statepoints instead of the shadow stack

Make.inc

@@ -687,9 +687,11 @@ endif
 ifeq ($(OS),WINNT)
 LIBUNWIND:=
 else
+# -lunwind must come _before_ -lunwind-generic to avoid
+# shadowing the actual _U_dyn_info_list_addr with a weak definition
 ifeq ($(USE_SYSTEM_LIBUNWIND), 1)
 ifneq ($(OS),Darwin)
-LIBUNWIND:=-lunwind-generic -lunwind
+LIBUNWIND:=-lunwind -lunwind-generic
 # Only for linux since we want to use not yet released libunwind features
 JCFLAGS+=-DSYSTEM_LIBUNWIND
 JCPPFLAGS+=-DSYSTEM_LIBUNWIND
@@ -699,7 +701,7 @@ ifeq ($(OS),Darwin)
 LIBUNWIND:=$(build_libdir)/libosxunwind.a
 JCPPFLAGS+=-DLIBOSXUNWIND
 else
-LIBUNWIND:=$(build_libdir)/libunwind-generic.a $(build_libdir)/libunwind.a
+LIBUNWIND:=$(build_libdir)/libunwind.a $(build_libdir)/libunwind-generic.a
 endif
 endif
 endif

Makefile

@@ -178,6 +178,7 @@ $(build_sysconfdir)/julia/juliarc.jl: $(JULIAHOME)/contrib/windows/juliarc.jl
 endif
 
 $(build_private_libdir)/%.$(SHLIB_EXT): $(build_private_libdir)/%.o
+	objcopy --globalize-symbol __LLVM_StackMaps $<
 	@$(call PRINT_LINK, $(CXX) $(LDFLAGS) -shared $(fPIC) -L$(build_private_libdir) -L$(build_libdir) -L$(build_shlibdir) -o $@ $< \
 		$(if $(findstring -debug.$(SHLIB_EXT),$(notdir $@)),-ljulia-debug,-ljulia) \
 		$$([ $(OS) = WINNT ] && echo '' -lssp))

deps/libunwind-remote-unw-hack.diff

@@ -0,0 +1,16 @@
+diff -ur libunwind-1.1-julia2/src/mi/Gfind_dynamic_proc_info.c libunwind-1.1-julia2.MINE/src/mi/Gfind_dynamic_proc_info.c
+--- libunwind-1.1-julia2/src/mi/Gfind_dynamic_proc_info.c	2016-01-06 07:48:23.000000000 -0500
++++ libunwind-1.1-julia2.MINE/src/mi/Gfind_dynamic_proc_info.c	2016-06-06 08:37:55.364416906 -0400
+@@ -84,8 +84,8 @@
+                              unw_proc_info_t *pi, int need_unwind_info,
+                              void *arg)
+ {
+-  if (as == unw_local_addr_space)
++    //if (as == unw_local_addr_space)
+     return local_find_proc_info (as, ip, pi, need_unwind_info, arg);
+-  else
+-    return remote_find_proc_info (as, ip, pi, need_unwind_info, arg);
++    /*else
++      return remote_find_proc_info (as, ip, pi, need_unwind_info, arg);*/
+ }
+Only in libunwind-1.1-julia2.MINE/src/mi: Gfind_dynamic_proc_info.c~

deps/llvm.mk

@@ -451,6 +451,8 @@ $(eval $(call LLVM_PATCH,llvm-D17712))
 $(eval $(call LLVM_PATCH,llvm-PR26180))
 $(eval $(call LLVM_PATCH,llvm-PR27046))
 $(eval $(call LLVM_PATCH,llvm-3.8.0_ppc64_SUBFC8))
+else ifeq ($(LLVM_VER),svn)
+$(eval $(call LLVM_PATCH,llvm-3.9_statepoint))
 endif # LLVM_VER
 
 ifeq ($(LLVM_VER),3.7.1)

deps/patches/llvm-3.9_statepoint.patch

@@ -0,0 +1,41 @@
+diff --git a/lib/Analysis/AliasAnalysis.cpp b/lib/Analysis/AliasAnalysis.cpp
+index 117f8cb..263dd48 100644
+--- a/lib/Analysis/AliasAnalysis.cpp
++++ b/lib/Analysis/AliasAnalysis.cpp
+@@ -445,7 +445,7 @@ ModRefInfo AAResults::callCapturesBefore(const Instruction *I,
+
+   unsigned ArgNo = 0;
+   ModRefInfo R = MRI_NoModRef;
+-  for (ImmutableCallSite::arg_iterator CI = CS.arg_begin(), CE = CS.arg_end();
++  for (auto CI = CS.data_operands_begin(), CE = CS.data_operands_end();
+        CI != CE; ++CI, ++ArgNo) {
+     // Only look at the no-capture or byval pointer arguments.  If this
+     // pointer were passed to arguments that were neither of these, then it
+diff --git a/lib/CodeGen/PrologEpilogInserter.cpp b/lib/CodeGen/PrologEpilogInserter.cpp
+index be3e5e2..6064e6f 100644
+--- a/lib/CodeGen/PrologEpilogInserter.cpp
++++ b/lib/CodeGen/PrologEpilogInserter.cpp
+@@ -1094,7 +1094,7 @@ void PEI::replaceFrameIndices(MachineBasicBlock *BB, MachineFunction &Fn,
+         unsigned Reg;
+         MachineOperand &Offset = MI->getOperand(i + 1);
+         const unsigned refOffset =
+-          TFI->getFrameIndexReferenceFromSP(Fn, MI->getOperand(i).getIndex(),
++          TFI->getFrameIndexReference(Fn, MI->getOperand(i).getIndex(),
+                                             Reg);
+
+         Offset.setImm(Offset.getImm() + refOffset);
+diff --git a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+index 6a727a1..5defd28 100644
+--- a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
++++ b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+@@ -7180,8 +7180,8 @@ void SelectionDAGBuilder::populateCallLoweringInfo(
+
+   // Populate the argument list.
+   // Attributes for args start at offset 1, after the return attribute.
+-  for (unsigned ArgI = ArgIdx, ArgE = ArgIdx + NumArgs, AttrI = ArgIdx + 1;
+-       ArgI != ArgE; ++ArgI) {
++  for (unsigned ArgI = ArgIdx, ArgE = ArgIdx + NumArgs, AttrI = ArgIdx+1;
++       ArgI != ArgE; ++ArgI, ++AttrI) {
+     const Value *V = CS->getOperand(ArgI);
+
+     assert(!V->getType()->isEmptyTy() && "Empty type passed to intrinsic.");

deps/unwind.mk

@@ -10,6 +10,7 @@ $(SRCDIR)/srccache/libunwind-$(UNWIND_VER).tar.gz: | $(SRCDIR)/srccache
 $(SRCDIR)/srccache/libunwind-$(UNWIND_VER)/configure: $(SRCDIR)/srccache/libunwind-$(UNWIND_VER).tar.gz
 	$(JLCHECKSUM) $<
 	cd $(dir $<) && $(TAR) xfz $<
+	patch -d $(SRCDIR)/srccache -p0 < $(SRCDIR)/libunwind-remote-unw-hack.diff
 	touch -c $@
 $(BUILDDIR)/libunwind-$(UNWIND_VER)/config.status: $(SRCDIR)/srccache/libunwind-$(UNWIND_VER)/configure
 	mkdir -p $(dir $@)

src/Makefile

@@ -139,7 +139,7 @@ $(BUILDDIR)/anticodegen.o $(BUILDDIR)/anticodegen.dbg.obj: $(SRCDIR)/intrinsics.
 $(BUILDDIR)/debuginfo.o $(BUILDDIR)/debuginfo.dbg.obj: $(SRCDIR)/codegen_internal.h
 $(BUILDDIR)/disasm.o $(BUILDDIR)/disasm.dbg.obj: $(SRCDIR)/codegen_internal.h
 $(BUILDDIR)/builtins.o $(BUILDDIR)/builtins.dbg.obj: $(SRCDIR)/table.c
-$(BUILDDIR)/gc.o $(BUILDDIR)/gc.dbg.obj: $(SRCDIR)/gc.h
+$(BUILDDIR)/gc.o $(BUILDDIR)/gc.dbg.obj: $(SRCDIR)/gc.h $(SRCDIR)/gc-stackmaps.c
 $(BUILDDIR)/gc-debug.o $(BUILDDIR)/gc-debug.dbg.obj: $(SRCDIR)/gc.h
 $(BUILDDIR)/gc-pages.o $(BUILDDIR)/gc-pages.dbg.obj: $(SRCDIR)/gc.h
 $(BUILDDIR)/signal-handling.o $(BUILDDIR)/signal-handling.dbg.obj: $(addprefix $(SRCDIR)/,signals-*.c)

src/codegen.cpp

@@ -5658,6 +5658,7 @@ static inline SmallVector<std::string,10> getTargetFeatures() {
     }
     return attr;
 }
+#include <llvm/CodeGen/LinkAllCodegenComponents.h>
 
 extern "C" void jl_init_codegen(void)
 {
@@ -5670,7 +5671,7 @@ extern "C" void jl_init_codegen(void)
 #ifdef JL_DEBUG_BUILD
     cl::ParseEnvironmentOptions("Julia", "JULIA_LLVM_ARGS");
 #endif
-
+    llvm::linkStatepointExampleGC();
     imaging_mode = jl_generating_output();
     jl_init_debuginfo();
     jl_init_runtime_ccall();
@@ -5778,8 +5779,8 @@ extern "C" void jl_init_codegen(void)
                                " Is the LLVM backend for this CPU enabled?");
 #if defined(USE_MCJIT) && (!defined(_CPU_ARM_) && !defined(_CPU_PPC64_))
     // FastISel seems to be buggy for ARM. Ref #13321
-    if (jl_options.opt_level < 3)
-        jl_TargetMachine->setFastISel(true);
+    //if (jl_options.opt_level < 3)
+    //    jl_TargetMachine->setFastISel(true);
 #endif
 
 #ifdef USE_ORCJIT

src/debuginfo.cpp

@@ -49,7 +49,7 @@ using namespace llvm;
 #include "julia_internal.h"
 #include "codegen_internal.h"
 #ifdef _OS_LINUX_
-#  define UNW_LOCAL_ONLY
+//#  define UNW_LOCAL_ONLY
 #  include <libunwind.h>
 #  include <link.h>
 #endif
@@ -233,7 +233,7 @@ void jl_callback_triggered_linfos(void)
             jl_call_tracer(jl_linfo_tracer, (jl_value_t*)linfo);
     }
 }
-
+#include <iostream>
 class JuliaJITEventListener: public JITEventListener
 {
 #ifndef USE_MCJIT
@@ -322,12 +322,22 @@ class JuliaJITEventListener: public JITEventListener
 
 #ifdef LLVM38
         std::map<StringRef,object::SectionRef,strrefcomp> loadedSections;
+        uint8_t *text_addr = NULL, *stackmaps_addr = NULL;
+        size_t text_size = 0;
         for (const object::SectionRef &lSection: obj.sections()) {
             StringRef sName;
             if (!lSection.getName(sName)) {
                 loadedSections[sName] = lSection;
+                if (sName == ".llvm_stackmaps") {
+                    stackmaps_addr = (uint8_t*)L.getSectionLoadAddress(lSection);
+                } else if (sName == ".text") {
+                    text_addr = (uint8_t*)L.getSectionLoadAddress(lSection);
+                    text_size = lSection.getSize();
+                }
             }
         }
+        if (stackmaps_addr)
+            jl_gc_register_stackmaps(stackmaps_addr);
         auto getLoadAddress = [&] (const StringRef &sName) -> uint64_t {
             auto search = loadedSections.find(sName);
             if (search == loadedSections.end())

src/disasm.cpp

@@ -344,7 +344,7 @@ void jl_dump_asm_internal(uintptr_t Fptr, size_t Fsize, int64_t slide,
     MCContext Ctx(*MAI, *MRI, MOFI.get(), &SrcMgr);
 #endif
 #ifdef LLVM39
-    MOFI->InitMCObjectFileInfo(TheTriple, /* PIC */ false,
+    MOFI->InitMCObjectFileInfo(TheTriple, false,
                                CodeModel::Default, Ctx);
 #elif defined(LLVM37)
     MOFI->InitMCObjectFileInfo(TheTriple, Reloc::Default, CodeModel::Default, Ctx);

src/dump.c

@@ -265,6 +265,7 @@ static int jl_load_sysimg_so(void)
         }
 #endif
     }
+    jl_gc_register_stackmaps((uint8_t*)jl_dlsym_e(jl_sysimg_handle, "__LLVM_StackMaps"));
     const char *sysimg_data = (const char*)jl_dlsym_e(jl_sysimg_handle, "jl_system_image_data");
     if (sysimg_data) {
         size_t len = *(size_t*)jl_dlsym(jl_sysimg_handle, "jl_system_image_size");