update deps

cmake.toml

@@ -2,7 +2,6 @@
 name = "vmemu"
 
 [subdir.deps]
-[subdir.tools]
 
 [target.vmemu]
 type = "static"

deps/vmprofiler

@@ -1 +1 @@
-Subproject commit 4de96fbaef64bbae7e85c37e32d287e6f29fe5f1
+Subproject commit f65c84050d39210c8057ab741b8cbaf9f1e2e17d

include/vmemu_t.hpp

@@ -26,7 +26,7 @@ class emu_t {
   ~emu_t();
   bool init();
   bool emulate(std::uint32_t vmenter_rva, vm::instrs::vrtn_t& vrtn);
-  const bool m_log_instructions = false;
+  const bool m_log_instructions = true;
 
  private:
   uc_engine* uc;

src/vmemu_t.cpp

@@ -65,10 +65,6 @@ bool emu_t::init() {
 }
 
 bool emu_t::emulate(std::uint32_t vmenter_rva, vm::instrs::vrtn_t& vrtn) {
-  static int vmenter_number = 0;
-  std::printf("\n[%.4d]  STARTING FROM NEW VMENTER AT %p (%p)\n\n", vmenter_number++,
-                                                              vmenter_rva + m_vm->m_module_base,
-                                                              vmenter_rva + m_vm->m_image_base);
   uc_err err;
   vrtn.m_rva = vmenter_rva;
 
@@ -128,6 +124,7 @@ bool emu_t::emulate(std::uint32_t vmenter_rva, vm::instrs::vrtn_t& vrtn) {
         auto& new_blk = vrtn.m_blks.emplace_back();
         new_blk.m_vip = {0ull, 0ull};
         new_blk.m_vm = {blk.m_jmp.m_vm.vip, blk.m_jmp.m_vm.vsp};
+        new_blk.is_branch = true;
         cc_blk = &new_blk;
 
         // emulate the branch...
@@ -272,20 +269,13 @@ bool emu_t::branch_pred_spec_exec(uc_engine* uc, uint64_t address,
     return false;
   }
 
-  if (instr.mnemonic == ZYDIS_MNEMONIC_INVALID) return false;
+  if (instr.mnemonic == ZYDIS_MNEMONIC_INVALID) {std::printf("> bad instruction\n"); return false;}
-
   uc_context* ctx;
   uct_context_alloc(uc, &ctx);
   uc_context_save(uc, ctx);
 
-  // if this is the first instruction of this handler then save the stack...
+  if (obj->cc_trace.m_instrs.empty())
-  if (!obj->cc_trace.m_instrs.size()) {
+    obj->cc_trace.m_begin = address;
-    obj->cc_trace.m_vip = obj->cc_blk->m_vm.vip;
-    obj->cc_trace.m_vsp = obj->cc_blk->m_vm.vsp;
-    obj->cc_trace.m_stack = new std::uint8_t[STACK_SIZE]; g_new_delete_tracker++;
-    uc_mem_read(uc, STACK_BASE, obj->cc_trace.m_stack, STACK_SIZE);
-  }
-
   obj->cc_trace.m_instrs.push_back({instr, ctx});
 
   // RET or JMP REG means the end of a vm handler...
@@ -295,8 +285,16 @@ bool emu_t::branch_pred_spec_exec(uc_engine* uc, uint64_t address,
     // deobfuscate the instruction stream before profiling...
     // makes it easier for profiles to be correct...
     vm::instrs::deobfuscate(obj->cc_trace);
-
+    std::printf("> vsp = %s, vip = %s, address = %p (%p)\n", ZydisRegisterGetString(obj->cc_trace.m_vsp), ZydisRegisterGetString(obj->cc_trace.m_vip),
-    // find the last MOV REG, DWORD PTR [VIP] in the instruction stream, then
+      obj->cc_trace.m_begin, obj->cc_trace.m_begin - obj->m_vm->m_module_base + obj->m_vm->m_image_base);
+    zydis_rtn_t inst_stream;
+    std::for_each(obj->cc_trace.m_instrs.begin(),
+                  obj->cc_trace.m_instrs.end(),
+                  [&](vm::instrs::emu_instr_t& instr) {
+                    inst_stream.push_back({instr.m_instr});
+                  });
+    vm::utils::print(inst_stream);
+/*     // find the last MOV REG, DWORD PTR [VIP] in the instruction stream, then
     // remove any instructions from this instruction to the JMP/RET...
     const auto rva_fetch = std::find_if(
         obj->cc_trace.m_instrs.rbegin(), obj->cc_trace.m_instrs.rend(),
@@ -317,27 +315,29 @@ bool emu_t::branch_pred_spec_exec(uc_engine* uc, uint64_t address,
       }
       obj->cc_trace.m_instrs.erase((rva_fetch + 1).base(),
                                    obj->cc_trace.m_instrs.end());
-    }
+    } */
 
     const auto vinstr = vm::instrs::determine(obj->cc_trace);
-
+    std::printf("instruction is: %d\n", vinstr.mnemonic);
     // -- free the trace since we will start a new one...
     std::for_each(obj->cc_trace.m_instrs.begin(), obj->cc_trace.m_instrs.end(),
                   [&](const vm::instrs::emu_instr_t& instr) {
                     uct_context_free(instr.m_cpu);
                   });
 
-    delete[] obj->cc_trace.m_stack; if (obj->cc_trace.m_stack) g_new_delete_tracker--;
     obj->cc_trace.m_instrs.clear();
 
     if (vinstr.mnemonic != vm::instrs::mnemonic_t::jmp) {
-      if (vinstr.mnemonic != vm::instrs::mnemonic_t::sreg) uc_emu_stop(uc);
+/*       if (vinstr.mnemonic != vm::instrs::mnemonic_t::sreg) uc_emu_stop(uc);
 
       if (!vinstr.imm.has_imm) uc_emu_stop(uc);
 
       if (vinstr.imm.size != 8 || vinstr.imm.val > 8 * VIRTUAL_REGISTER_COUNT)
-        uc_emu_stop(uc);
+        uc_emu_stop(uc); */
 
+      // stop if done or if instruction is invalid
+      if (vinstr.mnemonic == instrs::mnemonic_t::unknown)
+        uc_emu_stop(uc);
       // -- stop after 10 legit SREG's...
       if (++obj->m_sreg_cnt == 10) uc_emu_stop(uc);
     }
@@ -425,7 +425,7 @@ bool emu_t::code_exec_callback(uc_engine* uc, uint64_t address, uint32_t size,
       uc_context* backup;
       uct_context_alloc(uc, &backup);
       uc_context_save(uc, backup);
-      uc_context_restore(uc, (--vip_write)->m_cpu);
+        uc_context_restore(uc, vip_write->m_cpu);
       
       std::uintptr_t vip_addr = 0ull;
       uc_reg_read(uc, vm::instrs::reg_map[obj->cc_trace.m_vip], &vip_addr);
@@ -470,8 +470,8 @@ bool emu_t::code_exec_callback(uc_engine* uc, uint64_t address, uint32_t size,
         uc_emu_stop(uc);
         return false;
       }
-
+      //I don't think this if statement is neccesary
-      if (obj->cc_blk->m_vinstrs.size()) {
+      //if (obj->cc_blk->m_vinstrs.size()) {
         if (vinstr.mnemonic == vm::instrs::mnemonic_t::jmp) {
           uc_context *backup, *copy;
 
@@ -500,8 +500,7 @@ bool emu_t::code_exec_callback(uc_engine* uc, uint64_t address, uint32_t size,
         if (vinstr.mnemonic == vm::instrs::mnemonic_t::jmp ||
             vinstr.mnemonic == vm::instrs::mnemonic_t::vmexit)
           uc_emu_stop(obj->uc);
-      }
+      //}
-
       obj->cc_blk->m_vinstrs.push_back(vinstr);
     }
 
@@ -570,6 +569,10 @@ bool emu_t::legit_branch(vm::instrs::vblk_t& vblk, std::uintptr_t branch_addr) {
   uc_reg_read(uc, vm::instrs::reg_map[vblk.m_vm.vsp], &vsp);
   uc_mem_write(uc, vsp, &branch_addr, sizeof branch_addr);
   
+  // reset vsp and vip to their pre-jmp status in the current code trace
+  // this will be reset when the jmp instruction is emulated again
+  cc_trace.m_vsp = cc_blk->m_vm.vsp, cc_trace.m_vip = cc_blk->m_vm.vip;
+
   m_sreg_cnt = 0u;
   uc_emu_start(uc, rip, 0ull, 0ull, 0ull);
 
@@ -610,31 +613,32 @@ std::optional<std::pair<std::uintptr_t, std::uintptr_t>> emu_t::could_have_jcc(
     return v.mnemonic == vm::instrs::mnemonic_t::lconst && v.imm.size == 64;
   };
 
-  const auto lconst1 =
+  static const auto valid_mem = [=](uintptr_t loc)
-      std::find_if(vinstrs.rbegin(), vinstrs.rend(), lconst64_chk);
+  {
+      // check to see if the imm val is inside of the image...
+    if (loc > m_vm->m_image_base + m_vm->m_image_size ||
+        loc < m_vm->m_image_base ||
+        loc > m_vm->m_image_base + m_vm->m_image_size ||
+        loc < m_vm->m_image_base)
+      return false;
+    return true;
+  };
 
+  auto lconst1 = vinstrs.rbegin();
+  do {
+    lconst1 =
+      std::find_if(lconst1 + 1, vinstrs.rend(), lconst64_chk);
     if (lconst1 == vinstrs.rend()) return {};
+  } while (!valid_mem(lconst1->imm.val));
 
-  const auto lconst2 = std::find_if(lconst1 + 1, vinstrs.rend(), lconst64_chk);
+  auto lconst2 = lconst1;
-
+  do {
+    lconst2 =
+      std::find_if(lconst2 + 1, vinstrs.rend(), lconst64_chk);
   if (lconst2 == vinstrs.rend()) return {};
+  } while (!valid_mem(lconst1->imm.val));
 
-  // check to see if the imm val is inside of the image...
+  std::printf("possible jcc: %p, %p\n", lconst1->imm.val, lconst2->imm.val);
-  if (lconst1->imm.val > m_vm->m_image_base + m_vm->m_image_size ||
-      lconst1->imm.val < m_vm->m_image_base ||
-      lconst2->imm.val > m_vm->m_image_base + m_vm->m_image_size ||
-      lconst2->imm.val < m_vm->m_image_base)
-    return {};
-
-  // check to see if the imm's points to something inside of an executable
-  // section...
-  if (!vm::utils::scn::executable(
-          m_vm->m_module_base,
-          (lconst1->imm.val - m_vm->m_image_base) + m_vm->m_module_base) ||
-      !vm::utils::scn::executable(
-          m_vm->m_module_base,
-          (lconst2->imm.val - m_vm->m_image_base) + m_vm->m_module_base))
-    return {};
 
   return {{lconst1->imm.val, lconst2->imm.val}};
 }