ShadowStackGC.cpp revision 3ca2ad11567f83883ae2719c5fac5afc30c7b3d1 
1//===-- ShadowStackGC.cpp - GC support for uncooperative targets ----------===//
2//
3//                     The LLVM Compiler Infrastructure
4//
5// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
7//
8//===----------------------------------------------------------------------===//
9//
10// This file implements lowering for the llvm.gc* intrinsics for targets that do
11// not natively support them (which includes the C backend). Note that the code
12// generated is not quite as efficient as algorithms which generate stack maps
13// to identify roots.
14//
15// This pass implements the code transformation described in this paper:
16//   "Accurate Garbage Collection in an Uncooperative Environment"
17//   Fergus Henderson, ISMM, 2002
18//
19// In runtime/GC/SemiSpace.cpp is a prototype runtime which is compatible with
20// ShadowStackGC.
21//
22// In order to support this particular transformation, all stack roots are
23// coallocated in the stack. This allows a fully target-independent stack map
24// while introducing only minor runtime overhead.
25//
26//===----------------------------------------------------------------------===//
27
28#define DEBUG_TYPE "shadowstackgc"
29#include "llvm/CodeGen/GCs.h"
30#include "llvm/ADT/StringExtras.h"
31#include "llvm/CodeGen/GCStrategy.h"
32#include "llvm/IntrinsicInst.h"
33#include "llvm/Module.h"
34#include "llvm/Support/CallSite.h"
35#include "llvm/Support/IRBuilder.h"
36
37using namespace llvm;
38
39namespace {
40
41  class ShadowStackGC : public GCStrategy {
42    /// RootChain - This is the global linked-list that contains the chain of GC
43    /// roots.
44    GlobalVariable *Head;
45
46    /// StackEntryTy - Abstract type of a link in the shadow stack.
47    ///
48    StructType *StackEntryTy;
49    StructType *FrameMapTy;
50
51    /// Roots - GC roots in the current function. Each is a pair of the
52    /// intrinsic call and its corresponding alloca.
53    std::vector<std::pair<CallInst*,AllocaInst*> > Roots;
54
55  public:
56    ShadowStackGC();
57
58    bool initializeCustomLowering(Module &M);
59    bool performCustomLowering(Function &F);
60
61  private:
62    bool IsNullValue(Value *V);
63    Constant *GetFrameMap(Function &F);
64    Type* GetConcreteStackEntryType(Function &F);
65    void CollectRoots(Function &F);
66    static GetElementPtrInst *CreateGEP(LLVMContext &Context,
67                                        IRBuilder<> &B, Value *BasePtr,
68                                        int Idx1, const char *Name);
69    static GetElementPtrInst *CreateGEP(LLVMContext &Context,
70                                        IRBuilder<> &B, Value *BasePtr,
71                                        int Idx1, int Idx2, const char *Name);
72  };
73
74}
75
76static GCRegistry::Add<ShadowStackGC>
77X("shadow-stack", "Very portable GC for uncooperative code generators");
78
79namespace {
80  /// EscapeEnumerator - This is a little algorithm to find all escape points
81  /// from a function so that "finally"-style code can be inserted. In addition
82  /// to finding the existing return and unwind instructions, it also (if
83  /// necessary) transforms any call instructions into invokes and sends them to
84  /// a landing pad.
85  ///
86  /// It's wrapped up in a state machine using the same transform C# uses for
87  /// 'yield return' enumerators, This transform allows it to be non-allocating.
88  class EscapeEnumerator {
89    Function &F;
90    const char *CleanupBBName;
91
92    // State.
93    int State;
94    Function::iterator StateBB, StateE;
95    IRBuilder<> Builder;
96
97  public:
98    EscapeEnumerator(Function &F, const char *N = "cleanup")
99      : F(F), CleanupBBName(N), State(0), Builder(F.getContext()) {}
100
101    IRBuilder<> *Next() {
102      switch (State) {
103      default:
104        return 0;
105
106      case 0:
107        StateBB = F.begin();
108        StateE = F.end();
109        State = 1;
110
111      case 1:
112        // Find all 'return', 'resume', and 'unwind' instructions.
113        while (StateBB != StateE) {
114          BasicBlock *CurBB = StateBB++;
115
116          // Branches and invokes do not escape, only unwind, resume, and return
117          // do.
118          TerminatorInst *TI = CurBB->getTerminator();
119          if (!isa<UnwindInst>(TI) && !isa<ReturnInst>(TI) &&
120              !isa<ResumeInst>(TI))
121            continue;
122
123          Builder.SetInsertPoint(TI->getParent(), TI);
124          return &Builder;
125        }
126
127        State = 2;
128
129        // Find all 'call' instructions.
130        SmallVector<Instruction*,16> Calls;
131        for (Function::iterator BB = F.begin(),
132                                E = F.end(); BB != E; ++BB)
133          for (BasicBlock::iterator II = BB->begin(),
134                                    EE = BB->end(); II != EE; ++II)
135            if (CallInst *CI = dyn_cast<CallInst>(II))
136              if (!CI->getCalledFunction() ||
137                  !CI->getCalledFunction()->getIntrinsicID())
138                Calls.push_back(CI);
139
140        if (Calls.empty())
141          return 0;
142
143        // Create a cleanup block.
144        LLVMContext &C = F.getContext();
145        BasicBlock *CleanupBB = BasicBlock::Create(C, CleanupBBName, &F);
146        Type *ExnTy = StructType::get(Type::getInt8PtrTy(C),
147                                      Type::getInt32Ty(C), NULL);
148        // FIXME: Assuming the C++ personality function probably isn't the best
149        //        thing in the world.
150        Constant *PersFn =
151          F.getParent()->
152          getOrInsertFunction("__gxx_personality_v0",
153                              FunctionType::get(Type::getInt32Ty(C), true));
154        LandingPadInst *LPad = LandingPadInst::Create(ExnTy, PersFn, 1,
155                                                      "cleanup.lpad",
156                                                      CleanupBB);
157        LPad->setCleanup(true);
158        ResumeInst *RI = ResumeInst::Create(LPad, CleanupBB);
159
160        // Transform the 'call' instructions into 'invoke's branching to the
161        // cleanup block. Go in reverse order to make prettier BB names.
162        SmallVector<Value*,16> Args;
163        for (unsigned I = Calls.size(); I != 0; ) {
164          CallInst *CI = cast<CallInst>(Calls[--I]);
165
166          // Split the basic block containing the function call.
167          BasicBlock *CallBB = CI->getParent();
168          BasicBlock *NewBB =
169            CallBB->splitBasicBlock(CI, CallBB->getName() + ".cont");
170
171          // Remove the unconditional branch inserted at the end of CallBB.
172          CallBB->getInstList().pop_back();
173          NewBB->getInstList().remove(CI);
174
175          // Create a new invoke instruction.
176          Args.clear();
177          CallSite CS(CI);
178          Args.append(CS.arg_begin(), CS.arg_end());
179
180          InvokeInst *II = InvokeInst::Create(CI->getCalledValue(),
181                                              NewBB, CleanupBB,
182                                              Args, CI->getName(), CallBB);
183          II->setCallingConv(CI->getCallingConv());
184          II->setAttributes(CI->getAttributes());
185          CI->replaceAllUsesWith(II);
186          delete CI;
187        }
188
189        Builder.SetInsertPoint(RI->getParent(), RI);
190        return &Builder;
191      }
192    }
193  };
194}
195
196// -----------------------------------------------------------------------------
197
198void llvm::linkShadowStackGC() { }
199
200ShadowStackGC::ShadowStackGC() : Head(0), StackEntryTy(0) {
201  InitRoots = true;
202  CustomRoots = true;
203}
204
205Constant *ShadowStackGC::GetFrameMap(Function &F) {
206  // doInitialization creates the abstract type of this value.
207  Type *VoidPtr = Type::getInt8PtrTy(F.getContext());
208
209  // Truncate the ShadowStackDescriptor if some metadata is null.
210  unsigned NumMeta = 0;
211  SmallVector<Constant*, 16> Metadata;
212  for (unsigned I = 0; I != Roots.size(); ++I) {
213    Constant *C = cast<Constant>(Roots[I].first->getArgOperand(1));
214    if (!C->isNullValue())
215      NumMeta = I + 1;
216    Metadata.push_back(ConstantExpr::getBitCast(C, VoidPtr));
217  }
218  Metadata.resize(NumMeta);
219
220  Type *Int32Ty = Type::getInt32Ty(F.getContext());
221
222  Constant *BaseElts[] = {
223    ConstantInt::get(Int32Ty, Roots.size(), false),
224    ConstantInt::get(Int32Ty, NumMeta, false),
225  };
226
227  Constant *DescriptorElts[] = {
228    ConstantStruct::get(FrameMapTy, BaseElts),
229    ConstantArray::get(ArrayType::get(VoidPtr, NumMeta), Metadata)
230  };
231
232  Type *EltTys[] = { DescriptorElts[0]->getType(),DescriptorElts[1]->getType()};
233  StructType *STy = StructType::create(EltTys, "gc_map."+utostr(NumMeta));
234
235  Constant *FrameMap = ConstantStruct::get(STy, DescriptorElts);
236
237  // FIXME: Is this actually dangerous as WritingAnLLVMPass.html claims? Seems
238  //        that, short of multithreaded LLVM, it should be safe; all that is
239  //        necessary is that a simple Module::iterator loop not be invalidated.
240  //        Appending to the GlobalVariable list is safe in that sense.
241  //
242  //        All of the output passes emit globals last. The ExecutionEngine
243  //        explicitly supports adding globals to the module after
244  //        initialization.
245  //
246  //        Still, if it isn't deemed acceptable, then this transformation needs
247  //        to be a ModulePass (which means it cannot be in the 'llc' pipeline
248  //        (which uses a FunctionPassManager (which segfaults (not asserts) if
249  //        provided a ModulePass))).
250  Constant *GV = new GlobalVariable(*F.getParent(), FrameMap->getType(), true,
251                                    GlobalVariable::InternalLinkage,
252                                    FrameMap, "__gc_" + F.getName());
253
254  Constant *GEPIndices[2] = {
255                          ConstantInt::get(Type::getInt32Ty(F.getContext()), 0),
256                          ConstantInt::get(Type::getInt32Ty(F.getContext()), 0)
257                          };
258  return ConstantExpr::getGetElementPtr(GV, GEPIndices);
259}
260
261Type* ShadowStackGC::GetConcreteStackEntryType(Function &F) {
262  // doInitialization creates the generic version of this type.
263  std::vector<Type*> EltTys;
264  EltTys.push_back(StackEntryTy);
265  for (size_t I = 0; I != Roots.size(); I++)
266    EltTys.push_back(Roots[I].second->getAllocatedType());
267
268  return StructType::create(EltTys, "gc_stackentry."+F.getName().str());
269}
270
271/// doInitialization - If this module uses the GC intrinsics, find them now. If
272/// not, exit fast.
273bool ShadowStackGC::initializeCustomLowering(Module &M) {
274  // struct FrameMap {
275  //   int32_t NumRoots; // Number of roots in stack frame.
276  //   int32_t NumMeta;  // Number of metadata descriptors. May be < NumRoots.
277  //   void *Meta[];     // May be absent for roots without metadata.
278  // };
279  std::vector<Type*> EltTys;
280  // 32 bits is ok up to a 32GB stack frame. :)
281  EltTys.push_back(Type::getInt32Ty(M.getContext()));
282  // Specifies length of variable length array.
283  EltTys.push_back(Type::getInt32Ty(M.getContext()));
284  FrameMapTy = StructType::create(EltTys, "gc_map");
285  PointerType *FrameMapPtrTy = PointerType::getUnqual(FrameMapTy);
286
287  // struct StackEntry {
288  //   ShadowStackEntry *Next; // Caller's stack entry.
289  //   FrameMap *Map;          // Pointer to constant FrameMap.
290  //   void *Roots[];          // Stack roots (in-place array, so we pretend).
291  // };
292
293  StackEntryTy = StructType::create(M.getContext(), "gc_stackentry");
294
295  EltTys.clear();
296  EltTys.push_back(PointerType::getUnqual(StackEntryTy));
297  EltTys.push_back(FrameMapPtrTy);
298  StackEntryTy->setBody(EltTys);
299  PointerType *StackEntryPtrTy = PointerType::getUnqual(StackEntryTy);
300
301  // Get the root chain if it already exists.
302  Head = M.getGlobalVariable("llvm_gc_root_chain");
303  if (!Head) {
304    // If the root chain does not exist, insert a new one with linkonce
305    // linkage!
306    Head = new GlobalVariable(M, StackEntryPtrTy, false,
307                              GlobalValue::LinkOnceAnyLinkage,
308                              Constant::getNullValue(StackEntryPtrTy),
309                              "llvm_gc_root_chain");
310  } else if (Head->hasExternalLinkage() && Head->isDeclaration()) {
311    Head->setInitializer(Constant::getNullValue(StackEntryPtrTy));
312    Head->setLinkage(GlobalValue::LinkOnceAnyLinkage);
313  }
314
315  return true;
316}
317
318bool ShadowStackGC::IsNullValue(Value *V) {
319  if (Constant *C = dyn_cast<Constant>(V))
320    return C->isNullValue();
321  return false;
322}
323
324void ShadowStackGC::CollectRoots(Function &F) {
325  // FIXME: Account for original alignment. Could fragment the root array.
326  //   Approach 1: Null initialize empty slots at runtime. Yuck.
327  //   Approach 2: Emit a map of the array instead of just a count.
328
329  assert(Roots.empty() && "Not cleaned up?");
330
331  SmallVector<std::pair<CallInst*, AllocaInst*>, 16> MetaRoots;
332
333  for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
334    for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E;)
335      if (IntrinsicInst *CI = dyn_cast<IntrinsicInst>(II++))
336        if (Function *F = CI->getCalledFunction())
337          if (F->getIntrinsicID() == Intrinsic::gcroot) {
338            std::pair<CallInst*, AllocaInst*> Pair = std::make_pair(
339              CI, cast<AllocaInst>(CI->getArgOperand(0)->stripPointerCasts()));
340            if (IsNullValue(CI->getArgOperand(1)))
341              Roots.push_back(Pair);
342            else
343              MetaRoots.push_back(Pair);
344          }
345
346  // Number roots with metadata (usually empty) at the beginning, so that the
347  // FrameMap::Meta array can be elided.
348  Roots.insert(Roots.begin(), MetaRoots.begin(), MetaRoots.end());
349}
350
351GetElementPtrInst *
352ShadowStackGC::CreateGEP(LLVMContext &Context, IRBuilder<> &B, Value *BasePtr,
353                         int Idx, int Idx2, const char *Name) {
354  Value *Indices[] = { ConstantInt::get(Type::getInt32Ty(Context), 0),
355                       ConstantInt::get(Type::getInt32Ty(Context), Idx),
356                       ConstantInt::get(Type::getInt32Ty(Context), Idx2) };
357  Value* Val = B.CreateGEP(BasePtr, Indices, Name);
358
359  assert(isa<GetElementPtrInst>(Val) && "Unexpected folded constant");
360
361  return dyn_cast<GetElementPtrInst>(Val);
362}
363
364GetElementPtrInst *
365ShadowStackGC::CreateGEP(LLVMContext &Context, IRBuilder<> &B, Value *BasePtr,
366                         int Idx, const char *Name) {
367  Value *Indices[] = { ConstantInt::get(Type::getInt32Ty(Context), 0),
368                       ConstantInt::get(Type::getInt32Ty(Context), Idx) };
369  Value *Val = B.CreateGEP(BasePtr, Indices, Name);
370
371  assert(isa<GetElementPtrInst>(Val) && "Unexpected folded constant");
372
373  return dyn_cast<GetElementPtrInst>(Val);
374}
375
376/// runOnFunction - Insert code to maintain the shadow stack.
377bool ShadowStackGC::performCustomLowering(Function &F) {
378  LLVMContext &Context = F.getContext();
379
380  // Find calls to llvm.gcroot.
381  CollectRoots(F);
382
383  // If there are no roots in this function, then there is no need to add a
384  // stack map entry for it.
385  if (Roots.empty())
386    return false;
387
388  // Build the constant map and figure the type of the shadow stack entry.
389  Value *FrameMap = GetFrameMap(F);
390  Type *ConcreteStackEntryTy = GetConcreteStackEntryType(F);
391
392  // Build the shadow stack entry at the very start of the function.
393  BasicBlock::iterator IP = F.getEntryBlock().begin();
394  IRBuilder<> AtEntry(IP->getParent(), IP);
395
396  Instruction *StackEntry   = AtEntry.CreateAlloca(ConcreteStackEntryTy, 0,
397                                                   "gc_frame");
398
399  while (isa<AllocaInst>(IP)) ++IP;
400  AtEntry.SetInsertPoint(IP->getParent(), IP);
401
402  // Initialize the map pointer and load the current head of the shadow stack.
403  Instruction *CurrentHead  = AtEntry.CreateLoad(Head, "gc_currhead");
404  Instruction *EntryMapPtr  = CreateGEP(Context, AtEntry, StackEntry,
405                                        0,1,"gc_frame.map");
406  AtEntry.CreateStore(FrameMap, EntryMapPtr);
407
408  // After all the allocas...
409  for (unsigned I = 0, E = Roots.size(); I != E; ++I) {
410    // For each root, find the corresponding slot in the aggregate...
411    Value *SlotPtr = CreateGEP(Context, AtEntry, StackEntry, 1 + I, "gc_root");
412
413    // And use it in lieu of the alloca.
414    AllocaInst *OriginalAlloca = Roots[I].second;
415    SlotPtr->takeName(OriginalAlloca);
416    OriginalAlloca->replaceAllUsesWith(SlotPtr);
417  }
418
419  // Move past the original stores inserted by GCStrategy::InitRoots. This isn't
420  // really necessary (the collector would never see the intermediate state at
421  // runtime), but it's nicer not to push the half-initialized entry onto the
422  // shadow stack.
423  while (isa<StoreInst>(IP)) ++IP;
424  AtEntry.SetInsertPoint(IP->getParent(), IP);
425
426  // Push the entry onto the shadow stack.
427  Instruction *EntryNextPtr = CreateGEP(Context, AtEntry,
428                                        StackEntry,0,0,"gc_frame.next");
429  Instruction *NewHeadVal   = CreateGEP(Context, AtEntry,
430                                        StackEntry, 0, "gc_newhead");
431  AtEntry.CreateStore(CurrentHead, EntryNextPtr);
432  AtEntry.CreateStore(NewHeadVal, Head);
433
434  // For each instruction that escapes...
435  EscapeEnumerator EE(F, "gc_cleanup");
436  while (IRBuilder<> *AtExit = EE.Next()) {
437    // Pop the entry from the shadow stack. Don't reuse CurrentHead from
438    // AtEntry, since that would make the value live for the entire function.
439    Instruction *EntryNextPtr2 = CreateGEP(Context, *AtExit, StackEntry, 0, 0,
440                                           "gc_frame.next");
441    Value *SavedHead = AtExit->CreateLoad(EntryNextPtr2, "gc_savedhead");
442                       AtExit->CreateStore(SavedHead, Head);
443  }
444
445  // Delete the original allocas (which are no longer used) and the intrinsic
446  // calls (which are no longer valid). Doing this last avoids invalidating
447  // iterators.
448  for (unsigned I = 0, E = Roots.size(); I != E; ++I) {
449    Roots[I].first->eraseFromParent();
450    Roots[I].second->eraseFromParent();
451  }
452
453  Roots.clear();
454  return true;
455}
456