[x86/Linux] Stack align 16 bytes for JIT code

Change JIT code to align stack in 16 byte used in modern compiler

src/jit/codegenxarch.cpp

@@ -2631,16 +2631,14 @@ void CodeGen::genLclHeap(GenTreePtr tree)
  // Loop:
  genDefineTempLabel(loop);
 
-#if defined(_TARGET_AMD64_)
- // Push two 8-byte zeros. This matches the 16-byte STACK_ALIGN value.
- static_assert_no_msg(STACK_ALIGN == (REGSIZE_BYTES * 2));
- inst_IV(INS_push_hide, 0); // --- push 8-byte 0
- inst_IV(INS_push_hide, 0); // --- push 8-byte 0
-#elif defined(_TARGET_X86_)
- // Push a single 4-byte zero. This matches the 4-byte STACK_ALIGN value.
- static_assert_no_msg(STACK_ALIGN == REGSIZE_BYTES);
- inst_IV(INS_push_hide, 0); // --- push 4-byte 0
-#endif // _TARGET_X86_
+ static_assert_no_msg((STACK_ALIGN % REGSIZE_BYTES) == 0);
+ unsigned const count = (STACK_ALIGN / REGSIZE_BYTES);
+
+ for (unsigned i=0; i<count; i++)
+ {
+ inst_IV(INS_push_hide, 0); // --- push REG_SIZE bytes of 0
+ }
+ // Note that the stack must always be aligned to STACK_ALIGN bytes
 
  // Decrement the loop counter and loop if not done.
  inst_RV(INS_dec, regCnt, TYP_I_IMPL);

src/jit/lclvars.cpp

@@ -5750,6 +5750,7 @@ void Compiler::lvaAlignFrame()
 
 #elif defined(_TARGET_X86_)
 
+#if DOUBLE_ALIGN
  if (genDoubleAlign())
  {
  // Double Frame Alignement for x86 is handled in Compiler::lvaAssignVirtualFrameOffsetsToLocals()
@@ -5760,6 +5761,24 @@ void Compiler::lvaAlignFrame()
  lvaIncrementFrameSize(sizeof(void*));
  }
  }
+#endif
+
+ if (STACK_ALIGN > REGSIZE_BYTES)
+ {
+ if (lvaDoneFrameLayout != FINAL_FRAME_LAYOUT)
+ {
+ // If we are not doing final layout, we don't know the exact value of compLclFrameSize
+ // and thus do not know how much we will need to add in order to be aligned.
+ // We add the maximum pad that we could ever have (which is 12)
+ lvaIncrementFrameSize(STACK_ALIGN - REGSIZE_BYTES);
+ }
+
+ // The stack must always be 16 byte aligned.
+ if ((compLclFrameSize % STACK_ALIGN) != 0)
+ {
+ lvaIncrementFrameSize(STACK_ALIGN - (compLclFrameSize % STACK_ALIGN));
+ }
+ }
 
 #else
  NYI("TARGET specific lvaAlignFrame");

src/jit/target.h

@@ -484,9 +484,15 @@ typedef unsigned short regPairNoSmall; // arm: need 12 bits
  #define MIN_ARG_AREA_FOR_CALL 0 // Minimum required outgoing argument space for a call.
 
  #define CODE_ALIGN 1 // code alignment requirement
+#if !defined(FEATURE_PAL)
  #define STACK_ALIGN 4 // stack alignment requirement
  #define STACK_ALIGN_SHIFT 2 // Shift-right amount to convert stack size in bytes to size in DWORD_PTRs
  #define STACK_ALIGN_SHIFT_ALL 2 // Shift-right amount to convert stack size in bytes to size in STACK_ALIGN units
+#else
+ #define STACK_ALIGN 16 // stack alignment requirement
+ #define STACK_ALIGN_SHIFT 4 // Shift-right amount to convert stack size in bytes to size in DWORD_PTRs
+ #define STACK_ALIGN_SHIFT_ALL 4 // Shift-right amount to convert stack size in bytes to size in STACK_ALIGN units
+#endif // !FEATURE_PAL
 
  #define RBM_INT_CALLEE_SAVED (RBM_EBX|RBM_ESI|RBM_EDI)
  #define RBM_INT_CALLEE_TRASH (RBM_EAX|RBM_ECX|RBM_EDX)

src/vm/i386/asmhelpers.S

@@ -497,13 +497,14 @@ LEAF_ENTRY HelperMethodFrameRestoreState, _TEXT
 
  cmp dword ptr [eax+MachState__pRetAddr], 0
 
-#ifdef _DEBUG
+#ifdef _DEBUG_XX
  jnz LOCAL_LABEL(noConfirm)
  push ebp
  push ebx
  push edi
  push esi
  push ecx // HelperFrame*
+ CHECK_STACK_ALIGNMENT
  call C_FUNC(HelperMethodFrameConfirmState)
  // on return, eax = MachState*
  cmp DWORD PTR [eax + MachState__pRetAddr], 0
@@ -561,6 +562,7 @@ NESTED_ENTRY OnHijackTripThread, _TEXT, NoHandler
  sub esp,12
 
  push esp
+ CHECK_STACK_ALIGNMENT
  call C_FUNC(OnHijackWorker)
 
  // unused space for floating point state
@@ -596,6 +598,7 @@ NESTED_ENTRY OnHijackFPTripThread, _TEXT, NoHandler
  fstp QWORD PTR [esp]
 
  push esp
+ CHECK_STACK_ALIGNMENT
  call C_FUNC(OnHijackWorker)
 
  // restore top of the floating point stack
@@ -641,10 +644,20 @@ LEAF_ENTRY NDirectImportThunk, _TEXT
  push ecx
  push edx
 
+ // Compute padding size and adjust
+ mov esi, esp
+ lea ebx, [esp - 4] // for one push
+ and ebx, 15
+ sub esp, ebx
+
  // Invoke the function that does the real work.
  push eax
+ CHECK_STACK_ALIGNMENT
  call C_FUNC(NDirectImportWorker)
 
+ // Restore stack pointer
+ mov esp, esi
+
  // Restore argument registers
  pop edx
  pop ecx
@@ -686,6 +699,7 @@ NESTED_ENTRY UM2MThunk_WrapperHelper, _TEXT, NoHandler
  mov eax, [esp + 20] // pEntryThunk
  mov ecx, [esp + 24] // pThread
  mov ebx, [esp + 8] // pThunkArgs
+ CHECK_STACK_ALIGNMENT
  call [esp + 16] // pAddr
 
  pop ebx
@@ -699,6 +713,7 @@ NESTED_ENTRY UMThunkStubRareDisable, _TEXT, NoHandler
 
  push eax // Push the UMEntryThunk
  push ecx // Push thread
+ CHECK_STACK_ALIGNMENT
  call C_FUNC(UMThunkStubRareDisableWorker)
 
  pop ecx
@@ -756,6 +771,7 @@ LOCAL_LABEL(GoCallVarargWorker):
  push dword ptr [esi + 4*7] // pVaSigCookie
  push esi // pTransitionBlock
 
+ CHECK_STACK_ALIGNMENT
  call C_FUNC(VarargPInvokeStubWorker)
 
  // restore pMD
@@ -847,6 +863,7 @@ LOCAL_LABEL(GoCallCalliWorker):
  push dword ptr [esi + 4*7] // pVaSigCookie (first stack argument)
  push esi // pTransitionBlock
 
+ CHECK_STACK_ALIGNMENT
  call C_FUNC(GenericPInvokeCalliStubWorker)
 
  // restore target
@@ -872,6 +889,7 @@ NESTED_ENTRY StubDispatchFixupStub, _TEXT, NoHandler
  push eax // siteAddrForRegisterIndirect (for tailcalls)
  push esi // pTransitionBlock
 
+ CHECK_STACK_ALIGNMENT
  call C_FUNC(StubDispatchFixupWorker)
 
  STUB_EPILOG
@@ -906,6 +924,7 @@ NESTED_ENTRY ExternalMethodFixupStub, _TEXT_ NoHandler
  // pTransitionBlock
  push esi
 
+ CHECK_STACK_ALIGNMENT
  call C_FUNC(ExternalMethodFixupWorker)
 
  // eax now contains replacement stub. PreStubWorker will never return
@@ -939,6 +958,7 @@ NESTED_ENTRY DelayLoad_MethodCall, _TEXT, NoHandler
  // pTransitionBlock
  push esi
 
+ CHECK_STACK_ALIGNMENT
  call C_FUNC(ExternalMethodFixupWorker)
 
  // eax now contains replacement stub. PreStubWorker will never return
@@ -981,6 +1001,8 @@ NESTED_ENTRY VirtualMethodFixupStub, _TEXT, NoHandler
 
  push eax // address of the thunk
  push ecx // this ptr
+
+ CHECK_STACK_ALIGNMENT
  call C_FUNC(VirtualMethodFixupWorker)
 
  // Restore stack pointer
@@ -1070,6 +1092,7 @@ NESTED_ENTRY DelayLoad_Helper\suffix, _TEXT, NoHandler
  push eax // indirection cell address.
  push esi // pTransitionBlock
 
+ CHECK_STACK_ALIGNMENT
  call C_FUNC(DynamicHelperWorker)
  test eax,eax
  jnz LOCAL_LABEL(TailCallDelayLoad_Helper\suffix)
@@ -1121,6 +1144,7 @@ NESTED_ENTRY ResolveWorkerAsmStub, _TEXT, NoHandler
  lea ebp, [esi + 5*4]
 
  // Make the call
+ //CHECK_STACK_ALIGNMENT
  call C_FUNC(VSD_ResolveWorker)
 
  // From here on, mustn't trash eax
@@ -1228,6 +1252,7 @@ LOCAL_LABEL(main_loop):
  // promote the entry to the beginning of the chain
  mov ecx, eax
  // call C_FUNC(VirtualCallStubManager::PromoteChainEntry)
+ CHECK_STACK_ALIGNMENT
  call C_FUNC(_ZN22VirtualCallStubManager17PromoteChainEntryEP16ResolveCacheElem)
 
 LOCAL_LABEL(nopromote):