LLVM  7.0.0svn
MemorySanitizer.cpp
Go to the documentation of this file.
1 //===- MemorySanitizer.cpp - detector of uninitialized reads --------------===//
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 /// \file
11 /// This file is a part of MemorySanitizer, a detector of uninitialized
12 /// reads.
13 ///
14 /// The algorithm of the tool is similar to Memcheck
15 /// (http://goo.gl/QKbem). We associate a few shadow bits with every
16 /// byte of the application memory, poison the shadow of the malloc-ed
17 /// or alloca-ed memory, load the shadow bits on every memory read,
18 /// propagate the shadow bits through some of the arithmetic
19 /// instruction (including MOV), store the shadow bits on every memory
20 /// write, report a bug on some other instructions (e.g. JMP) if the
21 /// associated shadow is poisoned.
22 ///
23 /// But there are differences too. The first and the major one:
24 /// compiler instrumentation instead of binary instrumentation. This
25 /// gives us much better register allocation, possible compiler
26 /// optimizations and a fast start-up. But this brings the major issue
27 /// as well: msan needs to see all program events, including system
28 /// calls and reads/writes in system libraries, so we either need to
29 /// compile *everything* with msan or use a binary translation
30 /// component (e.g. DynamoRIO) to instrument pre-built libraries.
31 /// Another difference from Memcheck is that we use 8 shadow bits per
32 /// byte of application memory and use a direct shadow mapping. This
33 /// greatly simplifies the instrumentation code and avoids races on
34 /// shadow updates (Memcheck is single-threaded so races are not a
35 /// concern there. Memcheck uses 2 shadow bits per byte with a slow
36 /// path storage that uses 8 bits per byte).
37 ///
38 /// The default value of shadow is 0, which means "clean" (not poisoned).
39 ///
40 /// Every module initializer should call __msan_init to ensure that the
41 /// shadow memory is ready. On error, __msan_warning is called. Since
42 /// parameters and return values may be passed via registers, we have a
43 /// specialized thread-local shadow for return values
44 /// (__msan_retval_tls) and parameters (__msan_param_tls).
45 ///
46 /// Origin tracking.
47 ///
48 /// MemorySanitizer can track origins (allocation points) of all uninitialized
49 /// values. This behavior is controlled with a flag (msan-track-origins) and is
50 /// disabled by default.
51 ///
52 /// Origins are 4-byte values created and interpreted by the runtime library.
53 /// They are stored in a second shadow mapping, one 4-byte value for 4 bytes
54 /// of application memory. Propagation of origins is basically a bunch of
55 /// "select" instructions that pick the origin of a dirty argument, if an
56 /// instruction has one.
57 ///
58 /// Every 4 aligned, consecutive bytes of application memory have one origin
59 /// value associated with them. If these bytes contain uninitialized data
60 /// coming from 2 different allocations, the last store wins. Because of this,
61 /// MemorySanitizer reports can show unrelated origins, but this is unlikely in
62 /// practice.
63 ///
64 /// Origins are meaningless for fully initialized values, so MemorySanitizer
65 /// avoids storing origin to memory when a fully initialized value is stored.
66 /// This way it avoids needless overwritting origin of the 4-byte region on
67 /// a short (i.e. 1 byte) clean store, and it is also good for performance.
68 ///
69 /// Atomic handling.
70 ///
71 /// Ideally, every atomic store of application value should update the
72 /// corresponding shadow location in an atomic way. Unfortunately, atomic store
73 /// of two disjoint locations can not be done without severe slowdown.
74 ///
75 /// Therefore, we implement an approximation that may err on the safe side.
76 /// In this implementation, every atomically accessed location in the program
77 /// may only change from (partially) uninitialized to fully initialized, but
78 /// not the other way around. We load the shadow _after_ the application load,
79 /// and we store the shadow _before_ the app store. Also, we always store clean
80 /// shadow (if the application store is atomic). This way, if the store-load
81 /// pair constitutes a happens-before arc, shadow store and load are correctly
82 /// ordered such that the load will get either the value that was stored, or
83 /// some later value (which is always clean).
84 ///
85 /// This does not work very well with Compare-And-Swap (CAS) and
86 /// Read-Modify-Write (RMW) operations. To follow the above logic, CAS and RMW
87 /// must store the new shadow before the app operation, and load the shadow
88 /// after the app operation. Computers don't work this way. Current
89 /// implementation ignores the load aspect of CAS/RMW, always returning a clean
90 /// value. It implements the store part as a simple atomic store by storing a
91 /// clean shadow.
92 //
93 //===----------------------------------------------------------------------===//
94 
95 #include "llvm/ADT/APInt.h"
96 #include "llvm/ADT/ArrayRef.h"
98 #include "llvm/ADT/SmallString.h"
99 #include "llvm/ADT/SmallVector.h"
100 #include "llvm/ADT/StringExtras.h"
101 #include "llvm/ADT/StringRef.h"
102 #include "llvm/ADT/Triple.h"
104 #include "llvm/IR/Argument.h"
105 #include "llvm/IR/Attributes.h"
106 #include "llvm/IR/BasicBlock.h"
107 #include "llvm/IR/CallSite.h"
108 #include "llvm/IR/CallingConv.h"
109 #include "llvm/IR/Constant.h"
110 #include "llvm/IR/Constants.h"
111 #include "llvm/IR/DataLayout.h"
112 #include "llvm/IR/DerivedTypes.h"
113 #include "llvm/IR/Function.h"
114 #include "llvm/IR/GlobalValue.h"
115 #include "llvm/IR/GlobalVariable.h"
116 #include "llvm/IR/IRBuilder.h"
117 #include "llvm/IR/InlineAsm.h"
118 #include "llvm/IR/InstVisitor.h"
119 #include "llvm/IR/InstrTypes.h"
120 #include "llvm/IR/Instruction.h"
121 #include "llvm/IR/Instructions.h"
122 #include "llvm/IR/IntrinsicInst.h"
123 #include "llvm/IR/Intrinsics.h"
124 #include "llvm/IR/LLVMContext.h"
125 #include "llvm/IR/MDBuilder.h"
126 #include "llvm/IR/Module.h"
127 #include "llvm/IR/Type.h"
128 #include "llvm/IR/Value.h"
129 #include "llvm/IR/ValueMap.h"
130 #include "llvm/Pass.h"
132 #include "llvm/Support/Casting.h"
134 #include "llvm/Support/Compiler.h"
135 #include "llvm/Support/Debug.h"
137 #include "llvm/Support/MathExtras.h"
143 #include <algorithm>
144 #include <cassert>
145 #include <cstddef>
146 #include <cstdint>
147 #include <memory>
148 #include <string>
149 #include <tuple>
150 
151 using namespace llvm;
152 
153 #define DEBUG_TYPE "msan"
154 
155 static const unsigned kOriginSize = 4;
156 static const unsigned kMinOriginAlignment = 4;
157 static const unsigned kShadowTLSAlignment = 8;
158 
159 // These constants must be kept in sync with the ones in msan.h.
160 static const unsigned kParamTLSSize = 800;
161 static const unsigned kRetvalTLSSize = 800;
162 
163 // Accesses sizes are powers of two: 1, 2, 4, 8.
164 static const size_t kNumberOfAccessSizes = 4;
165 
166 /// \brief Track origins of uninitialized values.
167 ///
168 /// Adds a section to MemorySanitizer report that points to the allocation
169 /// (stack or heap) the uninitialized bits came from originally.
170 static cl::opt<int> ClTrackOrigins("msan-track-origins",
171  cl::desc("Track origins (allocation sites) of poisoned memory"),
172  cl::Hidden, cl::init(0));
173 
174 static cl::opt<bool> ClKeepGoing("msan-keep-going",
175  cl::desc("keep going after reporting a UMR"),
176  cl::Hidden, cl::init(false));
177 
178 static cl::opt<bool> ClPoisonStack("msan-poison-stack",
179  cl::desc("poison uninitialized stack variables"),
180  cl::Hidden, cl::init(true));
181 
182 static cl::opt<bool> ClPoisonStackWithCall("msan-poison-stack-with-call",
183  cl::desc("poison uninitialized stack variables with a call"),
184  cl::Hidden, cl::init(false));
185 
186 static cl::opt<int> ClPoisonStackPattern("msan-poison-stack-pattern",
187  cl::desc("poison uninitialized stack variables with the given pattern"),
188  cl::Hidden, cl::init(0xff));
189 
190 static cl::opt<bool> ClPoisonUndef("msan-poison-undef",
191  cl::desc("poison undef temps"),
192  cl::Hidden, cl::init(true));
193 
194 static cl::opt<bool> ClHandleICmp("msan-handle-icmp",
195  cl::desc("propagate shadow through ICmpEQ and ICmpNE"),
196  cl::Hidden, cl::init(true));
197 
198 static cl::opt<bool> ClHandleICmpExact("msan-handle-icmp-exact",
199  cl::desc("exact handling of relational integer ICmp"),
200  cl::Hidden, cl::init(false));
201 
202 // This flag controls whether we check the shadow of the address
203 // operand of load or store. Such bugs are very rare, since load from
204 // a garbage address typically results in SEGV, but still happen
205 // (e.g. only lower bits of address are garbage, or the access happens
206 // early at program startup where malloc-ed memory is more likely to
207 // be zeroed. As of 2012-08-28 this flag adds 20% slowdown.
208 static cl::opt<bool> ClCheckAccessAddress("msan-check-access-address",
209  cl::desc("report accesses through a pointer which has poisoned shadow"),
210  cl::Hidden, cl::init(true));
211 
212 static cl::opt<bool> ClDumpStrictInstructions("msan-dump-strict-instructions",
213  cl::desc("print out instructions with default strict semantics"),
214  cl::Hidden, cl::init(false));
215 
217  "msan-instrumentation-with-call-threshold",
218  cl::desc(
219  "If the function being instrumented requires more than "
220  "this number of checks and origin stores, use callbacks instead of "
221  "inline checks (-1 means never use callbacks)."),
222  cl::Hidden, cl::init(3500));
223 
224 // This is an experiment to enable handling of cases where shadow is a non-zero
225 // compile-time constant. For some unexplainable reason they were silently
226 // ignored in the instrumentation.
227 static cl::opt<bool> ClCheckConstantShadow("msan-check-constant-shadow",
228  cl::desc("Insert checks for constant shadow values"),
229  cl::Hidden, cl::init(false));
230 
231 // This is off by default because of a bug in gold:
232 // https://sourceware.org/bugzilla/show_bug.cgi?id=19002
233 static cl::opt<bool> ClWithComdat("msan-with-comdat",
234  cl::desc("Place MSan constructors in comdat sections"),
235  cl::Hidden, cl::init(false));
236 
237 static const char *const kMsanModuleCtorName = "msan.module_ctor";
238 static const char *const kMsanInitName = "__msan_init";
239 
240 namespace {
241 
242 // Memory map parameters used in application-to-shadow address calculation.
243 // Offset = (Addr & ~AndMask) ^ XorMask
244 // Shadow = ShadowBase + Offset
245 // Origin = OriginBase + Offset
246 struct MemoryMapParams {
247  uint64_t AndMask;
248  uint64_t XorMask;
249  uint64_t ShadowBase;
250  uint64_t OriginBase;
251 };
252 
253 struct PlatformMemoryMapParams {
254  const MemoryMapParams *bits32;
255  const MemoryMapParams *bits64;
256 };
257 
258 } // end anonymous namespace
259 
260 // i386 Linux
261 static const MemoryMapParams Linux_I386_MemoryMapParams = {
262  0x000080000000, // AndMask
263  0, // XorMask (not used)
264  0, // ShadowBase (not used)
265  0x000040000000, // OriginBase
266 };
267 
268 // x86_64 Linux
269 static const MemoryMapParams Linux_X86_64_MemoryMapParams = {
270 #ifdef MSAN_LINUX_X86_64_OLD_MAPPING
271  0x400000000000, // AndMask
272  0, // XorMask (not used)
273  0, // ShadowBase (not used)
274  0x200000000000, // OriginBase
275 #else
276  0, // AndMask (not used)
277  0x500000000000, // XorMask
278  0, // ShadowBase (not used)
279  0x100000000000, // OriginBase
280 #endif
281 };
282 
283 // mips64 Linux
284 static const MemoryMapParams Linux_MIPS64_MemoryMapParams = {
285  0, // AndMask (not used)
286  0x008000000000, // XorMask
287  0, // ShadowBase (not used)
288  0x002000000000, // OriginBase
289 };
290 
291 // ppc64 Linux
292 static const MemoryMapParams Linux_PowerPC64_MemoryMapParams = {
293  0xE00000000000, // AndMask
294  0x100000000000, // XorMask
295  0x080000000000, // ShadowBase
296  0x1C0000000000, // OriginBase
297 };
298 
299 // aarch64 Linux
300 static const MemoryMapParams Linux_AArch64_MemoryMapParams = {
301  0, // AndMask (not used)
302  0x06000000000, // XorMask
303  0, // ShadowBase (not used)
304  0x01000000000, // OriginBase
305 };
306 
307 // i386 FreeBSD
308 static const MemoryMapParams FreeBSD_I386_MemoryMapParams = {
309  0x000180000000, // AndMask
310  0x000040000000, // XorMask
311  0x000020000000, // ShadowBase
312  0x000700000000, // OriginBase
313 };
314 
315 // x86_64 FreeBSD
316 static const MemoryMapParams FreeBSD_X86_64_MemoryMapParams = {
317  0xc00000000000, // AndMask
318  0x200000000000, // XorMask
319  0x100000000000, // ShadowBase
320  0x380000000000, // OriginBase
321 };
322 
323 // x86_64 NetBSD
324 static const MemoryMapParams NetBSD_X86_64_MemoryMapParams = {
325  0, // AndMask
326  0x500000000000, // XorMask
327  0, // ShadowBase
328  0x100000000000, // OriginBase
329 };
330 
331 static const PlatformMemoryMapParams Linux_X86_MemoryMapParams = {
334 };
335 
336 static const PlatformMemoryMapParams Linux_MIPS_MemoryMapParams = {
337  nullptr,
339 };
340 
341 static const PlatformMemoryMapParams Linux_PowerPC_MemoryMapParams = {
342  nullptr,
344 };
345 
346 static const PlatformMemoryMapParams Linux_ARM_MemoryMapParams = {
347  nullptr,
349 };
350 
351 static const PlatformMemoryMapParams FreeBSD_X86_MemoryMapParams = {
354 };
355 
356 static const PlatformMemoryMapParams NetBSD_X86_MemoryMapParams = {
357  nullptr,
359 };
360 
361 namespace {
362 
363 /// \brief An instrumentation pass implementing detection of uninitialized
364 /// reads.
365 ///
366 /// MemorySanitizer: instrument the code in module to find
367 /// uninitialized reads.
368 class MemorySanitizer : public FunctionPass {
369 public:
370  // Pass identification, replacement for typeid.
371  static char ID;
372 
373  MemorySanitizer(int TrackOrigins = 0, bool Recover = false)
374  : FunctionPass(ID),
375  TrackOrigins(std::max(TrackOrigins, (int)ClTrackOrigins)),
376  Recover(Recover || ClKeepGoing) {}
377 
378  StringRef getPassName() const override { return "MemorySanitizer"; }
379 
380  void getAnalysisUsage(AnalysisUsage &AU) const override {
382  }
383 
384  bool runOnFunction(Function &F) override;
385  bool doInitialization(Module &M) override;
386 
387 private:
388  friend struct MemorySanitizerVisitor;
389  friend struct VarArgAMD64Helper;
390  friend struct VarArgMIPS64Helper;
391  friend struct VarArgAArch64Helper;
392  friend struct VarArgPowerPC64Helper;
393 
394  void initializeCallbacks(Module &M);
395 
396  /// \brief Track origins (allocation points) of uninitialized values.
397  int TrackOrigins;
398  bool Recover;
399 
400  LLVMContext *C;
401  Type *IntptrTy;
402  Type *OriginTy;
403 
404  /// \brief Thread-local shadow storage for function parameters.
405  GlobalVariable *ParamTLS;
406 
407  /// \brief Thread-local origin storage for function parameters.
408  GlobalVariable *ParamOriginTLS;
409 
410  /// \brief Thread-local shadow storage for function return value.
411  GlobalVariable *RetvalTLS;
412 
413  /// \brief Thread-local origin storage for function return value.
414  GlobalVariable *RetvalOriginTLS;
415 
416  /// \brief Thread-local shadow storage for in-register va_arg function
417  /// parameters (x86_64-specific).
418  GlobalVariable *VAArgTLS;
419 
420  /// \brief Thread-local shadow storage for va_arg overflow area
421  /// (x86_64-specific).
422  GlobalVariable *VAArgOverflowSizeTLS;
423 
424  /// \brief Thread-local space used to pass origin value to the UMR reporting
425  /// function.
426  GlobalVariable *OriginTLS;
427 
428  /// \brief The run-time callback to print a warning.
429  Value *WarningFn = nullptr;
430 
431  // These arrays are indexed by log2(AccessSize).
432  Value *MaybeWarningFn[kNumberOfAccessSizes];
433  Value *MaybeStoreOriginFn[kNumberOfAccessSizes];
434 
435  /// \brief Run-time helper that generates a new origin value for a stack
436  /// allocation.
437  Value *MsanSetAllocaOrigin4Fn;
438 
439  /// \brief Run-time helper that poisons stack on function entry.
440  Value *MsanPoisonStackFn;
441 
442  /// \brief Run-time helper that records a store (or any event) of an
443  /// uninitialized value and returns an updated origin id encoding this info.
444  Value *MsanChainOriginFn;
445 
446  /// \brief MSan runtime replacements for memmove, memcpy and memset.
447  Value *MemmoveFn, *MemcpyFn, *MemsetFn;
448 
449  /// \brief Memory map parameters used in application-to-shadow calculation.
450  const MemoryMapParams *MapParams;
451 
452  MDNode *ColdCallWeights;
453 
454  /// \brief Branch weights for origin store.
455  MDNode *OriginStoreWeights;
456 
457  /// \brief An empty volatile inline asm that prevents callback merge.
458  InlineAsm *EmptyAsm;
459 
460  Function *MsanCtorFunction;
461 };
462 
463 } // end anonymous namespace
464 
465 char MemorySanitizer::ID = 0;
466 
468  MemorySanitizer, "msan",
469  "MemorySanitizer: detects uninitialized reads.", false, false)
472  MemorySanitizer, "msan",
473  "MemorySanitizer: detects uninitialized reads.", false, false)
474 
475 FunctionPass *llvm::createMemorySanitizerPass(int TrackOrigins, bool Recover) {
476  return new MemorySanitizer(TrackOrigins, Recover);
477 }
478 
479 /// \brief Create a non-const global initialized with the given string.
480 ///
481 /// Creates a writable global for Str so that we can pass it to the
482 /// run-time lib. Runtime uses first 4 bytes of the string to store the
483 /// frame ID, so the string needs to be mutable.
485  StringRef Str) {
486  Constant *StrConst = ConstantDataArray::getString(M.getContext(), Str);
487  return new GlobalVariable(M, StrConst->getType(), /*isConstant=*/false,
488  GlobalValue::PrivateLinkage, StrConst, "");
489 }
490 
491 /// \brief Insert extern declaration of runtime-provided functions and globals.
492 void MemorySanitizer::initializeCallbacks(Module &M) {
493  // Only do this once.
494  if (WarningFn)
495  return;
496 
497  IRBuilder<> IRB(*C);
498  // Create the callback.
499  // FIXME: this function should have "Cold" calling conv,
500  // which is not yet implemented.
501  StringRef WarningFnName = Recover ? "__msan_warning"
502  : "__msan_warning_noreturn";
503  WarningFn = M.getOrInsertFunction(WarningFnName, IRB.getVoidTy());
504 
505  for (size_t AccessSizeIndex = 0; AccessSizeIndex < kNumberOfAccessSizes;
506  AccessSizeIndex++) {
507  unsigned AccessSize = 1 << AccessSizeIndex;
508  std::string FunctionName = "__msan_maybe_warning_" + itostr(AccessSize);
509  MaybeWarningFn[AccessSizeIndex] = M.getOrInsertFunction(
510  FunctionName, IRB.getVoidTy(), IRB.getIntNTy(AccessSize * 8),
511  IRB.getInt32Ty());
512 
513  FunctionName = "__msan_maybe_store_origin_" + itostr(AccessSize);
514  MaybeStoreOriginFn[AccessSizeIndex] = M.getOrInsertFunction(
515  FunctionName, IRB.getVoidTy(), IRB.getIntNTy(AccessSize * 8),
516  IRB.getInt8PtrTy(), IRB.getInt32Ty());
517  }
518 
519  MsanSetAllocaOrigin4Fn = M.getOrInsertFunction(
520  "__msan_set_alloca_origin4", IRB.getVoidTy(), IRB.getInt8PtrTy(), IntptrTy,
521  IRB.getInt8PtrTy(), IntptrTy);
522  MsanPoisonStackFn =
523  M.getOrInsertFunction("__msan_poison_stack", IRB.getVoidTy(),
524  IRB.getInt8PtrTy(), IntptrTy);
525  MsanChainOriginFn = M.getOrInsertFunction(
526  "__msan_chain_origin", IRB.getInt32Ty(), IRB.getInt32Ty());
527  MemmoveFn = M.getOrInsertFunction(
528  "__msan_memmove", IRB.getInt8PtrTy(), IRB.getInt8PtrTy(),
529  IRB.getInt8PtrTy(), IntptrTy);
530  MemcpyFn = M.getOrInsertFunction(
531  "__msan_memcpy", IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IRB.getInt8PtrTy(),
532  IntptrTy);
533  MemsetFn = M.getOrInsertFunction(
534  "__msan_memset", IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IRB.getInt32Ty(),
535  IntptrTy);
536 
537  // Create globals.
538  RetvalTLS = new GlobalVariable(
539  M, ArrayType::get(IRB.getInt64Ty(), kRetvalTLSSize / 8), false,
540  GlobalVariable::ExternalLinkage, nullptr, "__msan_retval_tls", nullptr,
542  RetvalOriginTLS = new GlobalVariable(
543  M, OriginTy, false, GlobalVariable::ExternalLinkage, nullptr,
544  "__msan_retval_origin_tls", nullptr, GlobalVariable::InitialExecTLSModel);
545 
546  ParamTLS = new GlobalVariable(
547  M, ArrayType::get(IRB.getInt64Ty(), kParamTLSSize / 8), false,
548  GlobalVariable::ExternalLinkage, nullptr, "__msan_param_tls", nullptr,
550  ParamOriginTLS = new GlobalVariable(
551  M, ArrayType::get(OriginTy, kParamTLSSize / 4), false,
552  GlobalVariable::ExternalLinkage, nullptr, "__msan_param_origin_tls",
554 
555  VAArgTLS = new GlobalVariable(
556  M, ArrayType::get(IRB.getInt64Ty(), kParamTLSSize / 8), false,
557  GlobalVariable::ExternalLinkage, nullptr, "__msan_va_arg_tls", nullptr,
559  VAArgOverflowSizeTLS = new GlobalVariable(
560  M, IRB.getInt64Ty(), false, GlobalVariable::ExternalLinkage, nullptr,
561  "__msan_va_arg_overflow_size_tls", nullptr,
563  OriginTLS = new GlobalVariable(
564  M, IRB.getInt32Ty(), false, GlobalVariable::ExternalLinkage, nullptr,
565  "__msan_origin_tls", nullptr, GlobalVariable::InitialExecTLSModel);
566 
567  // We insert an empty inline asm after __msan_report* to avoid callback merge.
568  EmptyAsm = InlineAsm::get(FunctionType::get(IRB.getVoidTy(), false),
569  StringRef(""), StringRef(""),
570  /*hasSideEffects=*/true);
571 }
572 
573 /// \brief Module-level initialization.
574 ///
575 /// inserts a call to __msan_init to the module's constructor list.
576 bool MemorySanitizer::doInitialization(Module &M) {
577  auto &DL = M.getDataLayout();
578 
579  Triple TargetTriple(M.getTargetTriple());
580  switch (TargetTriple.getOS()) {
581  case Triple::FreeBSD:
582  switch (TargetTriple.getArch()) {
583  case Triple::x86_64:
584  MapParams = FreeBSD_X86_MemoryMapParams.bits64;
585  break;
586  case Triple::x86:
587  MapParams = FreeBSD_X86_MemoryMapParams.bits32;
588  break;
589  default:
590  report_fatal_error("unsupported architecture");
591  }
592  break;
593  case Triple::NetBSD:
594  switch (TargetTriple.getArch()) {
595  case Triple::x86_64:
596  MapParams = NetBSD_X86_MemoryMapParams.bits64;
597  break;
598  default:
599  report_fatal_error("unsupported architecture");
600  }
601  break;
602  case Triple::Linux:
603  switch (TargetTriple.getArch()) {
604  case Triple::x86_64:
605  MapParams = Linux_X86_MemoryMapParams.bits64;
606  break;
607  case Triple::x86:
608  MapParams = Linux_X86_MemoryMapParams.bits32;
609  break;
610  case Triple::mips64:
611  case Triple::mips64el:
612  MapParams = Linux_MIPS_MemoryMapParams.bits64;
613  break;
614  case Triple::ppc64:
615  case Triple::ppc64le:
616  MapParams = Linux_PowerPC_MemoryMapParams.bits64;
617  break;
618  case Triple::aarch64:
619  case Triple::aarch64_be:
620  MapParams = Linux_ARM_MemoryMapParams.bits64;
621  break;
622  default:
623  report_fatal_error("unsupported architecture");
624  }
625  break;
626  default:
627  report_fatal_error("unsupported operating system");
628  }
629 
630  C = &(M.getContext());
631  IRBuilder<> IRB(*C);
632  IntptrTy = IRB.getIntPtrTy(DL);
633  OriginTy = IRB.getInt32Ty();
634 
635  ColdCallWeights = MDBuilder(*C).createBranchWeights(1, 1000);
636  OriginStoreWeights = MDBuilder(*C).createBranchWeights(1, 1000);
637 
638  std::tie(MsanCtorFunction, std::ignore) =
640  /*InitArgTypes=*/{},
641  /*InitArgs=*/{});
642  if (ClWithComdat) {
643  Comdat *MsanCtorComdat = M.getOrInsertComdat(kMsanModuleCtorName);
644  MsanCtorFunction->setComdat(MsanCtorComdat);
645  appendToGlobalCtors(M, MsanCtorFunction, 0, MsanCtorFunction);
646  } else {
647  appendToGlobalCtors(M, MsanCtorFunction, 0);
648  }
649 
650 
651  if (TrackOrigins)
653  IRB.getInt32(TrackOrigins), "__msan_track_origins");
654 
655  if (Recover)
657  IRB.getInt32(Recover), "__msan_keep_going");
658 
659  return true;
660 }
661 
662 namespace {
663 
664 /// \brief A helper class that handles instrumentation of VarArg
665 /// functions on a particular platform.
666 ///
667 /// Implementations are expected to insert the instrumentation
668 /// necessary to propagate argument shadow through VarArg function
669 /// calls. Visit* methods are called during an InstVisitor pass over
670 /// the function, and should avoid creating new basic blocks. A new
671 /// instance of this class is created for each instrumented function.
672 struct VarArgHelper {
673  virtual ~VarArgHelper() = default;
674 
675  /// \brief Visit a CallSite.
676  virtual void visitCallSite(CallSite &CS, IRBuilder<> &IRB) = 0;
677 
678  /// \brief Visit a va_start call.
679  virtual void visitVAStartInst(VAStartInst &I) = 0;
680 
681  /// \brief Visit a va_copy call.
682  virtual void visitVACopyInst(VACopyInst &I) = 0;
683 
684  /// \brief Finalize function instrumentation.
685  ///
686  /// This method is called after visiting all interesting (see above)
687  /// instructions in a function.
688  virtual void finalizeInstrumentation() = 0;
689 };
690 
691 struct MemorySanitizerVisitor;
692 
693 } // end anonymous namespace
694 
695 static VarArgHelper *CreateVarArgHelper(Function &Func, MemorySanitizer &Msan,
696  MemorySanitizerVisitor &Visitor);
697 
698 static unsigned TypeSizeToSizeIndex(unsigned TypeSize) {
699  if (TypeSize <= 8) return 0;
700  return Log2_32_Ceil((TypeSize + 7) / 8);
701 }
702 
703 namespace {
704 
705 /// This class does all the work for a given function. Store and Load
706 /// instructions store and load corresponding shadow and origin
707 /// values. Most instructions propagate shadow from arguments to their
708 /// return values. Certain instructions (most importantly, BranchInst)
709 /// test their argument shadow and print reports (with a runtime call) if it's
710 /// non-zero.
711 struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
712  Function &F;
713  MemorySanitizer &MS;
714  SmallVector<PHINode *, 16> ShadowPHINodes, OriginPHINodes;
715  ValueMap<Value*, Value*> ShadowMap, OriginMap;
716  std::unique_ptr<VarArgHelper> VAHelper;
717  const TargetLibraryInfo *TLI;
718 
719  // The following flags disable parts of MSan instrumentation based on
720  // blacklist contents and command-line options.
721  bool InsertChecks;
722  bool PropagateShadow;
723  bool PoisonStack;
724  bool PoisonUndef;
725  bool CheckReturnValue;
726 
727  struct ShadowOriginAndInsertPoint {
728  Value *Shadow;
729  Value *Origin;
730  Instruction *OrigIns;
731 
732  ShadowOriginAndInsertPoint(Value *S, Value *O, Instruction *I)
733  : Shadow(S), Origin(O), OrigIns(I) {}
734  };
737 
738  MemorySanitizerVisitor(Function &F, MemorySanitizer &MS)
739  : F(F), MS(MS), VAHelper(CreateVarArgHelper(F, MS, *this)) {
740  bool SanitizeFunction = F.hasFnAttribute(Attribute::SanitizeMemory);
741  InsertChecks = SanitizeFunction;
742  PropagateShadow = SanitizeFunction;
743  PoisonStack = SanitizeFunction && ClPoisonStack;
744  PoisonUndef = SanitizeFunction && ClPoisonUndef;
745  // FIXME: Consider using SpecialCaseList to specify a list of functions that
746  // must always return fully initialized values. For now, we hardcode "main".
747  CheckReturnValue = SanitizeFunction && (F.getName() == "main");
748  TLI = &MS.getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
749 
750  DEBUG(if (!InsertChecks)
751  dbgs() << "MemorySanitizer is not inserting checks into '"
752  << F.getName() << "'\n");
753  }
754 
755  Value *updateOrigin(Value *V, IRBuilder<> &IRB) {
756  if (MS.TrackOrigins <= 1) return V;
757  return IRB.CreateCall(MS.MsanChainOriginFn, V);
758  }
759 
760  Value *originToIntptr(IRBuilder<> &IRB, Value *Origin) {
761  const DataLayout &DL = F.getParent()->getDataLayout();
762  unsigned IntptrSize = DL.getTypeStoreSize(MS.IntptrTy);
763  if (IntptrSize == kOriginSize) return Origin;
764  assert(IntptrSize == kOriginSize * 2);
765  Origin = IRB.CreateIntCast(Origin, MS.IntptrTy, /* isSigned */ false);
766  return IRB.CreateOr(Origin, IRB.CreateShl(Origin, kOriginSize * 8));
767  }
768 
769  /// \brief Fill memory range with the given origin value.
770  void paintOrigin(IRBuilder<> &IRB, Value *Origin, Value *OriginPtr,
771  unsigned Size, unsigned Alignment) {
772  const DataLayout &DL = F.getParent()->getDataLayout();
773  unsigned IntptrAlignment = DL.getABITypeAlignment(MS.IntptrTy);
774  unsigned IntptrSize = DL.getTypeStoreSize(MS.IntptrTy);
775  assert(IntptrAlignment >= kMinOriginAlignment);
776  assert(IntptrSize >= kOriginSize);
777 
778  unsigned Ofs = 0;
779  unsigned CurrentAlignment = Alignment;
780  if (Alignment >= IntptrAlignment && IntptrSize > kOriginSize) {
781  Value *IntptrOrigin = originToIntptr(IRB, Origin);
782  Value *IntptrOriginPtr =
783  IRB.CreatePointerCast(OriginPtr, PointerType::get(MS.IntptrTy, 0));
784  for (unsigned i = 0; i < Size / IntptrSize; ++i) {
785  Value *Ptr = i ? IRB.CreateConstGEP1_32(MS.IntptrTy, IntptrOriginPtr, i)
786  : IntptrOriginPtr;
787  IRB.CreateAlignedStore(IntptrOrigin, Ptr, CurrentAlignment);
788  Ofs += IntptrSize / kOriginSize;
789  CurrentAlignment = IntptrAlignment;
790  }
791  }
792 
793  for (unsigned i = Ofs; i < (Size + kOriginSize - 1) / kOriginSize; ++i) {
794  Value *GEP =
795  i ? IRB.CreateConstGEP1_32(nullptr, OriginPtr, i) : OriginPtr;
796  IRB.CreateAlignedStore(Origin, GEP, CurrentAlignment);
797  CurrentAlignment = kMinOriginAlignment;
798  }
799  }
800 
801  void storeOrigin(IRBuilder<> &IRB, Value *Addr, Value *Shadow, Value *Origin,
802  Value *OriginPtr, unsigned Alignment, bool AsCall) {
803  const DataLayout &DL = F.getParent()->getDataLayout();
804  unsigned OriginAlignment = std::max(kMinOriginAlignment, Alignment);
805  unsigned StoreSize = DL.getTypeStoreSize(Shadow->getType());
806  if (Shadow->getType()->isAggregateType()) {
807  paintOrigin(IRB, updateOrigin(Origin, IRB), OriginPtr, StoreSize,
808  OriginAlignment);
809  } else {
810  Value *ConvertedShadow = convertToShadowTyNoVec(Shadow, IRB);
811  Constant *ConstantShadow = dyn_cast_or_null<Constant>(ConvertedShadow);
812  if (ConstantShadow) {
813  if (ClCheckConstantShadow && !ConstantShadow->isZeroValue())
814  paintOrigin(IRB, updateOrigin(Origin, IRB), OriginPtr, StoreSize,
815  OriginAlignment);
816  return;
817  }
818 
819  unsigned TypeSizeInBits =
820  DL.getTypeSizeInBits(ConvertedShadow->getType());
821  unsigned SizeIndex = TypeSizeToSizeIndex(TypeSizeInBits);
822  if (AsCall && SizeIndex < kNumberOfAccessSizes) {
823  Value *Fn = MS.MaybeStoreOriginFn[SizeIndex];
824  Value *ConvertedShadow2 = IRB.CreateZExt(
825  ConvertedShadow, IRB.getIntNTy(8 * (1 << SizeIndex)));
826  IRB.CreateCall(Fn, {ConvertedShadow2,
827  IRB.CreatePointerCast(Addr, IRB.getInt8PtrTy()),
828  Origin});
829  } else {
830  Value *Cmp = IRB.CreateICmpNE(
831  ConvertedShadow, getCleanShadow(ConvertedShadow), "_mscmp");
833  Cmp, &*IRB.GetInsertPoint(), false, MS.OriginStoreWeights);
834  IRBuilder<> IRBNew(CheckTerm);
835  paintOrigin(IRBNew, updateOrigin(Origin, IRBNew), OriginPtr, StoreSize,
836  OriginAlignment);
837  }
838  }
839  }
840 
841  void materializeStores(bool InstrumentWithCalls) {
842  for (StoreInst *SI : StoreList) {
843  IRBuilder<> IRB(SI);
844  Value *Val = SI->getValueOperand();
845  Value *Addr = SI->getPointerOperand();
846  Value *Shadow = SI->isAtomic() ? getCleanShadow(Val) : getShadow(Val);
847  Value *ShadowPtr, *OriginPtr;
848  Type *ShadowTy = Shadow->getType();
849  unsigned Alignment = SI->getAlignment();
850  unsigned OriginAlignment = std::max(kMinOriginAlignment, Alignment);
851  std::tie(ShadowPtr, OriginPtr) =
852  getShadowOriginPtr(Addr, IRB, ShadowTy, Alignment);
853 
854  StoreInst *NewSI = IRB.CreateAlignedStore(Shadow, ShadowPtr, Alignment);
855  DEBUG(dbgs() << " STORE: " << *NewSI << "\n");
856 
858  insertShadowCheck(Addr, NewSI);
859 
860  if (SI->isAtomic())
861  SI->setOrdering(addReleaseOrdering(SI->getOrdering()));
862 
863  if (MS.TrackOrigins && !SI->isAtomic())
864  storeOrigin(IRB, Addr, Shadow, getOrigin(Val), OriginPtr,
865  OriginAlignment, InstrumentWithCalls);
866  }
867  }
868 
869  void materializeOneCheck(Instruction *OrigIns, Value *Shadow, Value *Origin,
870  bool AsCall) {
871  IRBuilder<> IRB(OrigIns);
872  DEBUG(dbgs() << " SHAD0 : " << *Shadow << "\n");
873  Value *ConvertedShadow = convertToShadowTyNoVec(Shadow, IRB);
874  DEBUG(dbgs() << " SHAD1 : " << *ConvertedShadow << "\n");
875 
876  Constant *ConstantShadow = dyn_cast_or_null<Constant>(ConvertedShadow);
877  if (ConstantShadow) {
878  if (ClCheckConstantShadow && !ConstantShadow->isZeroValue()) {
879  if (MS.TrackOrigins) {
880  IRB.CreateStore(Origin ? (Value *)Origin : (Value *)IRB.getInt32(0),
881  MS.OriginTLS);
882  }
883  IRB.CreateCall(MS.WarningFn, {});
884  IRB.CreateCall(MS.EmptyAsm, {});
885  // FIXME: Insert UnreachableInst if !MS.Recover?
886  // This may invalidate some of the following checks and needs to be done
887  // at the very end.
888  }
889  return;
890  }
891 
892  const DataLayout &DL = OrigIns->getModule()->getDataLayout();
893 
894  unsigned TypeSizeInBits = DL.getTypeSizeInBits(ConvertedShadow->getType());
895  unsigned SizeIndex = TypeSizeToSizeIndex(TypeSizeInBits);
896  if (AsCall && SizeIndex < kNumberOfAccessSizes) {
897  Value *Fn = MS.MaybeWarningFn[SizeIndex];
898  Value *ConvertedShadow2 =
899  IRB.CreateZExt(ConvertedShadow, IRB.getIntNTy(8 * (1 << SizeIndex)));
900  IRB.CreateCall(Fn, {ConvertedShadow2, MS.TrackOrigins && Origin
901  ? Origin
902  : (Value *)IRB.getInt32(0)});
903  } else {
904  Value *Cmp = IRB.CreateICmpNE(ConvertedShadow,
905  getCleanShadow(ConvertedShadow), "_mscmp");
907  Cmp, OrigIns,
908  /* Unreachable */ !MS.Recover, MS.ColdCallWeights);
909 
910  IRB.SetInsertPoint(CheckTerm);
911  if (MS.TrackOrigins) {
912  IRB.CreateStore(Origin ? (Value *)Origin : (Value *)IRB.getInt32(0),
913  MS.OriginTLS);
914  }
915  IRB.CreateCall(MS.WarningFn, {});
916  IRB.CreateCall(MS.EmptyAsm, {});
917  DEBUG(dbgs() << " CHECK: " << *Cmp << "\n");
918  }
919  }
920 
921  void materializeChecks(bool InstrumentWithCalls) {
922  for (const auto &ShadowData : InstrumentationList) {
923  Instruction *OrigIns = ShadowData.OrigIns;
924  Value *Shadow = ShadowData.Shadow;
925  Value *Origin = ShadowData.Origin;
926  materializeOneCheck(OrigIns, Shadow, Origin, InstrumentWithCalls);
927  }
928  DEBUG(dbgs() << "DONE:\n" << F);
929  }
930 
931  /// \brief Add MemorySanitizer instrumentation to a function.
932  bool runOnFunction() {
933  MS.initializeCallbacks(*F.getParent());
934 
935  // In the presence of unreachable blocks, we may see Phi nodes with
936  // incoming nodes from such blocks. Since InstVisitor skips unreachable
937  // blocks, such nodes will not have any shadow value associated with them.
938  // It's easier to remove unreachable blocks than deal with missing shadow.
940 
941  // Iterate all BBs in depth-first order and create shadow instructions
942  // for all instructions (where applicable).
943  // For PHI nodes we create dummy shadow PHIs which will be finalized later.
944  for (BasicBlock *BB : depth_first(&F.getEntryBlock()))
945  visit(*BB);
946 
947  // Finalize PHI nodes.
948  for (PHINode *PN : ShadowPHINodes) {
949  PHINode *PNS = cast<PHINode>(getShadow(PN));
950  PHINode *PNO = MS.TrackOrigins ? cast<PHINode>(getOrigin(PN)) : nullptr;
951  size_t NumValues = PN->getNumIncomingValues();
952  for (size_t v = 0; v < NumValues; v++) {
953  PNS->addIncoming(getShadow(PN, v), PN->getIncomingBlock(v));
954  if (PNO) PNO->addIncoming(getOrigin(PN, v), PN->getIncomingBlock(v));
955  }
956  }
957 
958  VAHelper->finalizeInstrumentation();
959 
960  bool InstrumentWithCalls = ClInstrumentationWithCallThreshold >= 0 &&
961  InstrumentationList.size() + StoreList.size() >
963 
964  // Delayed instrumentation of StoreInst.
965  // This may add new checks to be inserted later.
966  materializeStores(InstrumentWithCalls);
967 
968  // Insert shadow value checks.
969  materializeChecks(InstrumentWithCalls);
970 
971  return true;
972  }
973 
974  /// \brief Compute the shadow type that corresponds to a given Value.
975  Type *getShadowTy(Value *V) {
976  return getShadowTy(V->getType());
977  }
978 
979  /// \brief Compute the shadow type that corresponds to a given Type.
980  Type *getShadowTy(Type *OrigTy) {
981  if (!OrigTy->isSized()) {
982  return nullptr;
983  }
984  // For integer type, shadow is the same as the original type.
985  // This may return weird-sized types like i1.
986  if (IntegerType *IT = dyn_cast<IntegerType>(OrigTy))
987  return IT;
988  const DataLayout &DL = F.getParent()->getDataLayout();
989  if (VectorType *VT = dyn_cast<VectorType>(OrigTy)) {
990  uint32_t EltSize = DL.getTypeSizeInBits(VT->getElementType());
991  return VectorType::get(IntegerType::get(*MS.C, EltSize),
992  VT->getNumElements());
993  }
994  if (ArrayType *AT = dyn_cast<ArrayType>(OrigTy)) {
995  return ArrayType::get(getShadowTy(AT->getElementType()),
996  AT->getNumElements());
997  }
998  if (StructType *ST = dyn_cast<StructType>(OrigTy)) {
999  SmallVector<Type*, 4> Elements;
1000  for (unsigned i = 0, n = ST->getNumElements(); i < n; i++)
1001  Elements.push_back(getShadowTy(ST->getElementType(i)));
1002  StructType *Res = StructType::get(*MS.C, Elements, ST->isPacked());
1003  DEBUG(dbgs() << "getShadowTy: " << *ST << " ===> " << *Res << "\n");
1004  return Res;
1005  }
1006  uint32_t TypeSize = DL.getTypeSizeInBits(OrigTy);
1007  return IntegerType::get(*MS.C, TypeSize);
1008  }
1009 
1010  /// \brief Flatten a vector type.
1011  Type *getShadowTyNoVec(Type *ty) {
1012  if (VectorType *vt = dyn_cast<VectorType>(ty))
1013  return IntegerType::get(*MS.C, vt->getBitWidth());
1014  return ty;
1015  }
1016 
1017  /// \brief Convert a shadow value to it's flattened variant.
1018  Value *convertToShadowTyNoVec(Value *V, IRBuilder<> &IRB) {
1019  Type *Ty = V->getType();
1020  Type *NoVecTy = getShadowTyNoVec(Ty);
1021  if (Ty == NoVecTy) return V;
1022  return IRB.CreateBitCast(V, NoVecTy);
1023  }
1024 
1025  /// \brief Compute the integer shadow offset that corresponds to a given
1026  /// application address.
1027  ///
1028  /// Offset = (Addr & ~AndMask) ^ XorMask
1029  Value *getShadowPtrOffset(Value *Addr, IRBuilder<> &IRB) {
1030  Value *OffsetLong = IRB.CreatePointerCast(Addr, MS.IntptrTy);
1031 
1032  uint64_t AndMask = MS.MapParams->AndMask;
1033  if (AndMask)
1034  OffsetLong =
1035  IRB.CreateAnd(OffsetLong, ConstantInt::get(MS.IntptrTy, ~AndMask));
1036 
1037  uint64_t XorMask = MS.MapParams->XorMask;
1038  if (XorMask)
1039  OffsetLong =
1040  IRB.CreateXor(OffsetLong, ConstantInt::get(MS.IntptrTy, XorMask));
1041  return OffsetLong;
1042  }
1043 
1044  /// \brief Compute the shadow and origin addresses corresponding to a given
1045  /// application address.
1046  ///
1047  /// Shadow = ShadowBase + Offset
1048  /// Origin = (OriginBase + Offset) & ~3ULL
1049  std::pair<Value *, Value *> getShadowOriginPtrUserspace(
1050  Value *Addr, IRBuilder<> &IRB, Type *ShadowTy, unsigned Alignment,
1051  Instruction **FirstInsn) {
1052  Value *ShadowOffset = getShadowPtrOffset(Addr, IRB);
1053  Value *ShadowLong = ShadowOffset;
1054  uint64_t ShadowBase = MS.MapParams->ShadowBase;
1055  *FirstInsn = dyn_cast<Instruction>(ShadowLong);
1056  if (ShadowBase != 0) {
1057  ShadowLong =
1058  IRB.CreateAdd(ShadowLong,
1059  ConstantInt::get(MS.IntptrTy, ShadowBase));
1060  }
1061  Value *ShadowPtr =
1062  IRB.CreateIntToPtr(ShadowLong, PointerType::get(ShadowTy, 0));
1063  Value *OriginPtr = nullptr;
1064  if (MS.TrackOrigins) {
1065  Value *OriginLong = ShadowOffset;
1066  uint64_t OriginBase = MS.MapParams->OriginBase;
1067  if (OriginBase != 0)
1068  OriginLong = IRB.CreateAdd(OriginLong,
1069  ConstantInt::get(MS.IntptrTy, OriginBase));
1070  if (Alignment < kMinOriginAlignment) {
1071  uint64_t Mask = kMinOriginAlignment - 1;
1072  OriginLong =
1073  IRB.CreateAnd(OriginLong, ConstantInt::get(MS.IntptrTy, ~Mask));
1074  }
1075  OriginPtr =
1076  IRB.CreateIntToPtr(OriginLong, PointerType::get(IRB.getInt32Ty(), 0));
1077  }
1078  return std::make_pair(ShadowPtr, OriginPtr);
1079  }
1080 
1081  std::pair<Value *, Value *> getShadowOriginPtr(Value *Addr, IRBuilder<> &IRB,
1082  Type *ShadowTy,
1083  unsigned Alignment) {
1084  Instruction *FirstInsn = nullptr;
1085  std::pair<Value *, Value *> ret =
1086  getShadowOriginPtrUserspace(Addr, IRB, ShadowTy, Alignment, &FirstInsn);
1087  return ret;
1088  }
1089 
1090  /// \brief Compute the shadow address for a given function argument.
1091  ///
1092  /// Shadow = ParamTLS+ArgOffset.
1093  Value *getShadowPtrForArgument(Value *A, IRBuilder<> &IRB,
1094  int ArgOffset) {
1095  Value *Base = IRB.CreatePointerCast(MS.ParamTLS, MS.IntptrTy);
1096  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
1097  return IRB.CreateIntToPtr(Base, PointerType::get(getShadowTy(A), 0),
1098  "_msarg");
1099  }
1100 
1101  /// \brief Compute the origin address for a given function argument.
1102  Value *getOriginPtrForArgument(Value *A, IRBuilder<> &IRB,
1103  int ArgOffset) {
1104  if (!MS.TrackOrigins) return nullptr;
1105  Value *Base = IRB.CreatePointerCast(MS.ParamOriginTLS, MS.IntptrTy);
1106  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
1107  return IRB.CreateIntToPtr(Base, PointerType::get(MS.OriginTy, 0),
1108  "_msarg_o");
1109  }
1110 
1111  /// \brief Compute the shadow address for a retval.
1112  Value *getShadowPtrForRetval(Value *A, IRBuilder<> &IRB) {
1113  return IRB.CreatePointerCast(MS.RetvalTLS,
1114  PointerType::get(getShadowTy(A), 0),
1115  "_msret");
1116  }
1117 
1118  /// \brief Compute the origin address for a retval.
1119  Value *getOriginPtrForRetval(IRBuilder<> &IRB) {
1120  // We keep a single origin for the entire retval. Might be too optimistic.
1121  return MS.RetvalOriginTLS;
1122  }
1123 
1124  /// \brief Set SV to be the shadow value for V.
1125  void setShadow(Value *V, Value *SV) {
1126  assert(!ShadowMap.count(V) && "Values may only have one shadow");
1127  ShadowMap[V] = PropagateShadow ? SV : getCleanShadow(V);
1128  }
1129 
1130  /// \brief Set Origin to be the origin value for V.
1131  void setOrigin(Value *V, Value *Origin) {
1132  if (!MS.TrackOrigins) return;
1133  assert(!OriginMap.count(V) && "Values may only have one origin");
1134  DEBUG(dbgs() << "ORIGIN: " << *V << " ==> " << *Origin << "\n");
1135  OriginMap[V] = Origin;
1136  }
1137 
1138  Constant *getCleanShadow(Type *OrigTy) {
1139  Type *ShadowTy = getShadowTy(OrigTy);
1140  if (!ShadowTy)
1141  return nullptr;
1142  return Constant::getNullValue(ShadowTy);
1143  }
1144 
1145  /// \brief Create a clean shadow value for a given value.
1146  ///
1147  /// Clean shadow (all zeroes) means all bits of the value are defined
1148  /// (initialized).
1149  Constant *getCleanShadow(Value *V) {
1150  return getCleanShadow(V->getType());
1151  }
1152 
1153  /// \brief Create a dirty shadow of a given shadow type.
1154  Constant *getPoisonedShadow(Type *ShadowTy) {
1155  assert(ShadowTy);
1156  if (isa<IntegerType>(ShadowTy) || isa<VectorType>(ShadowTy))
1157  return Constant::getAllOnesValue(ShadowTy);
1158  if (ArrayType *AT = dyn_cast<ArrayType>(ShadowTy)) {
1159  SmallVector<Constant *, 4> Vals(AT->getNumElements(),
1160  getPoisonedShadow(AT->getElementType()));
1161  return ConstantArray::get(AT, Vals);
1162  }
1163  if (StructType *ST = dyn_cast<StructType>(ShadowTy)) {
1165  for (unsigned i = 0, n = ST->getNumElements(); i < n; i++)
1166  Vals.push_back(getPoisonedShadow(ST->getElementType(i)));
1167  return ConstantStruct::get(ST, Vals);
1168  }
1169  llvm_unreachable("Unexpected shadow type");
1170  }
1171 
1172  /// \brief Create a dirty shadow for a given value.
1173  Constant *getPoisonedShadow(Value *V) {
1174  Type *ShadowTy = getShadowTy(V);
1175  if (!ShadowTy)
1176  return nullptr;
1177  return getPoisonedShadow(ShadowTy);
1178  }
1179 
1180  /// \brief Create a clean (zero) origin.
1181  Value *getCleanOrigin() {
1182  return Constant::getNullValue(MS.OriginTy);
1183  }
1184 
1185  /// \brief Get the shadow value for a given Value.
1186  ///
1187  /// This function either returns the value set earlier with setShadow,
1188  /// or extracts if from ParamTLS (for function arguments).
1189  Value *getShadow(Value *V) {
1190  if (!PropagateShadow) return getCleanShadow(V);
1191  if (Instruction *I = dyn_cast<Instruction>(V)) {
1192  if (I->getMetadata("nosanitize"))
1193  return getCleanShadow(V);
1194  // For instructions the shadow is already stored in the map.
1195  Value *Shadow = ShadowMap[V];
1196  if (!Shadow) {
1197  DEBUG(dbgs() << "No shadow: " << *V << "\n" << *(I->getParent()));
1198  (void)I;
1199  assert(Shadow && "No shadow for a value");
1200  }
1201  return Shadow;
1202  }
1203  if (UndefValue *U = dyn_cast<UndefValue>(V)) {
1204  Value *AllOnes = PoisonUndef ? getPoisonedShadow(V) : getCleanShadow(V);
1205  DEBUG(dbgs() << "Undef: " << *U << " ==> " << *AllOnes << "\n");
1206  (void)U;
1207  return AllOnes;
1208  }
1209  if (Argument *A = dyn_cast<Argument>(V)) {
1210  // For arguments we compute the shadow on demand and store it in the map.
1211  Value **ShadowPtr = &ShadowMap[V];
1212  if (*ShadowPtr)
1213  return *ShadowPtr;
1214  Function *F = A->getParent();
1215  IRBuilder<> EntryIRB(F->getEntryBlock().getFirstNonPHI());
1216  unsigned ArgOffset = 0;
1217  const DataLayout &DL = F->getParent()->getDataLayout();
1218  for (auto &FArg : F->args()) {
1219  if (!FArg.getType()->isSized()) {
1220  DEBUG(dbgs() << "Arg is not sized\n");
1221  continue;
1222  }
1223  unsigned Size =
1224  FArg.hasByValAttr()
1225  ? DL.getTypeAllocSize(FArg.getType()->getPointerElementType())
1226  : DL.getTypeAllocSize(FArg.getType());
1227  if (A == &FArg) {
1228  bool Overflow = ArgOffset + Size > kParamTLSSize;
1229  Value *Base = getShadowPtrForArgument(&FArg, EntryIRB, ArgOffset);
1230  if (FArg.hasByValAttr()) {
1231  // ByVal pointer itself has clean shadow. We copy the actual
1232  // argument shadow to the underlying memory.
1233  // Figure out maximal valid memcpy alignment.
1234  unsigned ArgAlign = FArg.getParamAlignment();
1235  if (ArgAlign == 0) {
1236  Type *EltType = A->getType()->getPointerElementType();
1237  ArgAlign = DL.getABITypeAlignment(EltType);
1238  }
1239  Value *CpShadowPtr =
1240  getShadowOriginPtr(V, EntryIRB, EntryIRB.getInt8Ty(), ArgAlign)
1241  .first;
1242  if (Overflow) {
1243  // ParamTLS overflow.
1244  EntryIRB.CreateMemSet(
1245  CpShadowPtr, Constant::getNullValue(EntryIRB.getInt8Ty()),
1246  Size, ArgAlign);
1247  } else {
1248  unsigned CopyAlign = std::min(ArgAlign, kShadowTLSAlignment);
1249  Value *Cpy =
1250  EntryIRB.CreateMemCpy(CpShadowPtr, Base, Size, CopyAlign);
1251  DEBUG(dbgs() << " ByValCpy: " << *Cpy << "\n");
1252  (void)Cpy;
1253  }
1254  *ShadowPtr = getCleanShadow(V);
1255  } else {
1256  if (Overflow) {
1257  // ParamTLS overflow.
1258  *ShadowPtr = getCleanShadow(V);
1259  } else {
1260  *ShadowPtr =
1261  EntryIRB.CreateAlignedLoad(Base, kShadowTLSAlignment);
1262  }
1263  }
1264  DEBUG(dbgs() << " ARG: " << FArg << " ==> " <<
1265  **ShadowPtr << "\n");
1266  if (MS.TrackOrigins && !Overflow) {
1267  Value *OriginPtr =
1268  getOriginPtrForArgument(&FArg, EntryIRB, ArgOffset);
1269  setOrigin(A, EntryIRB.CreateLoad(OriginPtr));
1270  } else {
1271  setOrigin(A, getCleanOrigin());
1272  }
1273  }
1274  ArgOffset += alignTo(Size, kShadowTLSAlignment);
1275  }
1276  assert(*ShadowPtr && "Could not find shadow for an argument");
1277  return *ShadowPtr;
1278  }
1279  // For everything else the shadow is zero.
1280  return getCleanShadow(V);
1281  }
1282 
1283  /// \brief Get the shadow for i-th argument of the instruction I.
1284  Value *getShadow(Instruction *I, int i) {
1285  return getShadow(I->getOperand(i));
1286  }
1287 
1288  /// \brief Get the origin for a value.
1289  Value *getOrigin(Value *V) {
1290  if (!MS.TrackOrigins) return nullptr;
1291  if (!PropagateShadow) return getCleanOrigin();
1292  if (isa<Constant>(V)) return getCleanOrigin();
1293  assert((isa<Instruction>(V) || isa<Argument>(V)) &&
1294  "Unexpected value type in getOrigin()");
1295  if (Instruction *I = dyn_cast<Instruction>(V)) {
1296  if (I->getMetadata("nosanitize"))
1297  return getCleanOrigin();
1298  }
1299  Value *Origin = OriginMap[V];
1300  assert(Origin && "Missing origin");
1301  return Origin;
1302  }
1303 
1304  /// \brief Get the origin for i-th argument of the instruction I.
1305  Value *getOrigin(Instruction *I, int i) {
1306  return getOrigin(I->getOperand(i));
1307  }
1308 
1309  /// \brief Remember the place where a shadow check should be inserted.
1310  ///
1311  /// This location will be later instrumented with a check that will print a
1312  /// UMR warning in runtime if the shadow value is not 0.
1313  void insertShadowCheck(Value *Shadow, Value *Origin, Instruction *OrigIns) {
1314  assert(Shadow);
1315  if (!InsertChecks) return;
1316 #ifndef NDEBUG
1317  Type *ShadowTy = Shadow->getType();
1318  assert((isa<IntegerType>(ShadowTy) || isa<VectorType>(ShadowTy)) &&
1319  "Can only insert checks for integer and vector shadow types");
1320 #endif
1321  InstrumentationList.push_back(
1322  ShadowOriginAndInsertPoint(Shadow, Origin, OrigIns));
1323  }
1324 
1325  /// \brief Remember the place where a shadow check should be inserted.
1326  ///
1327  /// This location will be later instrumented with a check that will print a
1328  /// UMR warning in runtime if the value is not fully defined.
1329  void insertShadowCheck(Value *Val, Instruction *OrigIns) {
1330  assert(Val);
1331  Value *Shadow, *Origin;
1332  if (ClCheckConstantShadow) {
1333  Shadow = getShadow(Val);
1334  if (!Shadow) return;
1335  Origin = getOrigin(Val);
1336  } else {
1337  Shadow = dyn_cast_or_null<Instruction>(getShadow(Val));
1338  if (!Shadow) return;
1339  Origin = dyn_cast_or_null<Instruction>(getOrigin(Val));
1340  }
1341  insertShadowCheck(Shadow, Origin, OrigIns);
1342  }
1343 
1344  AtomicOrdering addReleaseOrdering(AtomicOrdering a) {
1345  switch (a) {
1351  return AtomicOrdering::Release;
1357  }
1358  llvm_unreachable("Unknown ordering");
1359  }
1360 
1361  AtomicOrdering addAcquireOrdering(AtomicOrdering a) {
1362  switch (a) {
1368  return AtomicOrdering::Acquire;
1374  }
1375  llvm_unreachable("Unknown ordering");
1376  }
1377 
1378  // ------------------- Visitors.
1380  void visit(Instruction &I) {
1381  if (!I.getMetadata("nosanitize"))
1383  }
1384 
1385  /// \brief Instrument LoadInst
1386  ///
1387  /// Loads the corresponding shadow and (optionally) origin.
1388  /// Optionally, checks that the load address is fully defined.
1389  void visitLoadInst(LoadInst &I) {
1390  assert(I.getType()->isSized() && "Load type must have size");
1391  assert(!I.getMetadata("nosanitize"));
1392  IRBuilder<> IRB(I.getNextNode());
1393  Type *ShadowTy = getShadowTy(&I);
1394  Value *Addr = I.getPointerOperand();
1395  Value *ShadowPtr, *OriginPtr;
1396  unsigned Alignment = I.getAlignment();
1397  if (PropagateShadow) {
1398  std::tie(ShadowPtr, OriginPtr) =
1399  getShadowOriginPtr(Addr, IRB, ShadowTy, Alignment);
1400  setShadow(&I, IRB.CreateAlignedLoad(ShadowPtr, Alignment, "_msld"));
1401  } else {
1402  setShadow(&I, getCleanShadow(&I));
1403  }
1404 
1406  insertShadowCheck(I.getPointerOperand(), &I);
1407 
1408  if (I.isAtomic())
1409  I.setOrdering(addAcquireOrdering(I.getOrdering()));
1410 
1411  if (MS.TrackOrigins) {
1412  if (PropagateShadow) {
1413  unsigned OriginAlignment = std::max(kMinOriginAlignment, Alignment);
1414  setOrigin(&I, IRB.CreateAlignedLoad(OriginPtr, OriginAlignment));
1415  } else {
1416  setOrigin(&I, getCleanOrigin());
1417  }
1418  }
1419  }
1420 
1421  /// \brief Instrument StoreInst
1422  ///
1423  /// Stores the corresponding shadow and (optionally) origin.
1424  /// Optionally, checks that the store address is fully defined.
1425  void visitStoreInst(StoreInst &I) {
1426  StoreList.push_back(&I);
1427  }
1428 
1429  void handleCASOrRMW(Instruction &I) {
1430  assert(isa<AtomicRMWInst>(I) || isa<AtomicCmpXchgInst>(I));
1431 
1432  IRBuilder<> IRB(&I);
1433  Value *Addr = I.getOperand(0);
1434  Value *ShadowPtr =
1435  getShadowOriginPtr(Addr, IRB, I.getType(), /*Alignment*/ 1).first;
1436 
1438  insertShadowCheck(Addr, &I);
1439 
1440  // Only test the conditional argument of cmpxchg instruction.
1441  // The other argument can potentially be uninitialized, but we can not
1442  // detect this situation reliably without possible false positives.
1443  if (isa<AtomicCmpXchgInst>(I))
1444  insertShadowCheck(I.getOperand(1), &I);
1445 
1446  IRB.CreateStore(getCleanShadow(&I), ShadowPtr);
1447 
1448  setShadow(&I, getCleanShadow(&I));
1449  setOrigin(&I, getCleanOrigin());
1450  }
1451 
1452  void visitAtomicRMWInst(AtomicRMWInst &I) {
1453  handleCASOrRMW(I);
1454  I.setOrdering(addReleaseOrdering(I.getOrdering()));
1455  }
1456 
1457  void visitAtomicCmpXchgInst(AtomicCmpXchgInst &I) {
1458  handleCASOrRMW(I);
1459  I.setSuccessOrdering(addReleaseOrdering(I.getSuccessOrdering()));
1460  }
1461 
1462  // Vector manipulation.
1463  void visitExtractElementInst(ExtractElementInst &I) {
1464  insertShadowCheck(I.getOperand(1), &I);
1465  IRBuilder<> IRB(&I);
1466  setShadow(&I, IRB.CreateExtractElement(getShadow(&I, 0), I.getOperand(1),
1467  "_msprop"));
1468  setOrigin(&I, getOrigin(&I, 0));
1469  }
1470 
1471  void visitInsertElementInst(InsertElementInst &I) {
1472  insertShadowCheck(I.getOperand(2), &I);
1473  IRBuilder<> IRB(&I);
1474  setShadow(&I, IRB.CreateInsertElement(getShadow(&I, 0), getShadow(&I, 1),
1475  I.getOperand(2), "_msprop"));
1476  setOriginForNaryOp(I);
1477  }
1478 
1479  void visitShuffleVectorInst(ShuffleVectorInst &I) {
1480  insertShadowCheck(I.getOperand(2), &I);
1481  IRBuilder<> IRB(&I);
1482  setShadow(&I, IRB.CreateShuffleVector(getShadow(&I, 0), getShadow(&I, 1),
1483  I.getOperand(2), "_msprop"));
1484  setOriginForNaryOp(I);
1485  }
1486 
1487  // Casts.
1488  void visitSExtInst(SExtInst &I) {
1489  IRBuilder<> IRB(&I);
1490  setShadow(&I, IRB.CreateSExt(getShadow(&I, 0), I.getType(), "_msprop"));
1491  setOrigin(&I, getOrigin(&I, 0));
1492  }
1493 
1494  void visitZExtInst(ZExtInst &I) {
1495  IRBuilder<> IRB(&I);
1496  setShadow(&I, IRB.CreateZExt(getShadow(&I, 0), I.getType(), "_msprop"));
1497  setOrigin(&I, getOrigin(&I, 0));
1498  }
1499 
1500  void visitTruncInst(TruncInst &I) {
1501  IRBuilder<> IRB(&I);
1502  setShadow(&I, IRB.CreateTrunc(getShadow(&I, 0), I.getType(), "_msprop"));
1503  setOrigin(&I, getOrigin(&I, 0));
1504  }
1505 
1506  void visitBitCastInst(BitCastInst &I) {
1507  // Special case: if this is the bitcast (there is exactly 1 allowed) between
1508  // a musttail call and a ret, don't instrument. New instructions are not
1509  // allowed after a musttail call.
1510  if (auto *CI = dyn_cast<CallInst>(I.getOperand(0)))
1511  if (CI->isMustTailCall())
1512  return;
1513  IRBuilder<> IRB(&I);
1514  setShadow(&I, IRB.CreateBitCast(getShadow(&I, 0), getShadowTy(&I)));
1515  setOrigin(&I, getOrigin(&I, 0));
1516  }
1517 
1518  void visitPtrToIntInst(PtrToIntInst &I) {
1519  IRBuilder<> IRB(&I);
1520  setShadow(&I, IRB.CreateIntCast(getShadow(&I, 0), getShadowTy(&I), false,
1521  "_msprop_ptrtoint"));
1522  setOrigin(&I, getOrigin(&I, 0));
1523  }
1524 
1525  void visitIntToPtrInst(IntToPtrInst &I) {
1526  IRBuilder<> IRB(&I);
1527  setShadow(&I, IRB.CreateIntCast(getShadow(&I, 0), getShadowTy(&I), false,
1528  "_msprop_inttoptr"));
1529  setOrigin(&I, getOrigin(&I, 0));
1530  }
1531 
1532  void visitFPToSIInst(CastInst& I) { handleShadowOr(I); }
1533  void visitFPToUIInst(CastInst& I) { handleShadowOr(I); }
1534  void visitSIToFPInst(CastInst& I) { handleShadowOr(I); }
1535  void visitUIToFPInst(CastInst& I) { handleShadowOr(I); }
1536  void visitFPExtInst(CastInst& I) { handleShadowOr(I); }
1537  void visitFPTruncInst(CastInst& I) { handleShadowOr(I); }
1538 
1539  /// \brief Propagate shadow for bitwise AND.
1540  ///
1541  /// This code is exact, i.e. if, for example, a bit in the left argument
1542  /// is defined and 0, then neither the value not definedness of the
1543  /// corresponding bit in B don't affect the resulting shadow.
1544  void visitAnd(BinaryOperator &I) {
1545  IRBuilder<> IRB(&I);
1546  // "And" of 0 and a poisoned value results in unpoisoned value.
1547  // 1&1 => 1; 0&1 => 0; p&1 => p;
1548  // 1&0 => 0; 0&0 => 0; p&0 => 0;
1549  // 1&p => p; 0&p => 0; p&p => p;
1550  // S = (S1 & S2) | (V1 & S2) | (S1 & V2)
1551  Value *S1 = getShadow(&I, 0);
1552  Value *S2 = getShadow(&I, 1);
1553  Value *V1 = I.getOperand(0);
1554  Value *V2 = I.getOperand(1);
1555  if (V1->getType() != S1->getType()) {
1556  V1 = IRB.CreateIntCast(V1, S1->getType(), false);
1557  V2 = IRB.CreateIntCast(V2, S2->getType(), false);
1558  }
1559  Value *S1S2 = IRB.CreateAnd(S1, S2);
1560  Value *V1S2 = IRB.CreateAnd(V1, S2);
1561  Value *S1V2 = IRB.CreateAnd(S1, V2);
1562  setShadow(&I, IRB.CreateOr(S1S2, IRB.CreateOr(V1S2, S1V2)));
1563  setOriginForNaryOp(I);
1564  }
1565 
1566  void visitOr(BinaryOperator &I) {
1567  IRBuilder<> IRB(&I);
1568  // "Or" of 1 and a poisoned value results in unpoisoned value.
1569  // 1|1 => 1; 0|1 => 1; p|1 => 1;
1570  // 1|0 => 1; 0|0 => 0; p|0 => p;
1571  // 1|p => 1; 0|p => p; p|p => p;
1572  // S = (S1 & S2) | (~V1 & S2) | (S1 & ~V2)
1573  Value *S1 = getShadow(&I, 0);
1574  Value *S2 = getShadow(&I, 1);
1575  Value *V1 = IRB.CreateNot(I.getOperand(0));
1576  Value *V2 = IRB.CreateNot(I.getOperand(1));
1577  if (V1->getType() != S1->getType()) {
1578  V1 = IRB.CreateIntCast(V1, S1->getType(), false);
1579  V2 = IRB.CreateIntCast(V2, S2->getType(), false);
1580  }
1581  Value *S1S2 = IRB.CreateAnd(S1, S2);
1582  Value *V1S2 = IRB.CreateAnd(V1, S2);
1583  Value *S1V2 = IRB.CreateAnd(S1, V2);
1584  setShadow(&I, IRB.CreateOr(S1S2, IRB.CreateOr(V1S2, S1V2)));
1585  setOriginForNaryOp(I);
1586  }
1587 
1588  /// \brief Default propagation of shadow and/or origin.
1589  ///
1590  /// This class implements the general case of shadow propagation, used in all
1591  /// cases where we don't know and/or don't care about what the operation
1592  /// actually does. It converts all input shadow values to a common type
1593  /// (extending or truncating as necessary), and bitwise OR's them.
1594  ///
1595  /// This is much cheaper than inserting checks (i.e. requiring inputs to be
1596  /// fully initialized), and less prone to false positives.
1597  ///
1598  /// This class also implements the general case of origin propagation. For a
1599  /// Nary operation, result origin is set to the origin of an argument that is
1600  /// not entirely initialized. If there is more than one such arguments, the
1601  /// rightmost of them is picked. It does not matter which one is picked if all
1602  /// arguments are initialized.
1603  template <bool CombineShadow>
1604  class Combiner {
1605  Value *Shadow = nullptr;
1606  Value *Origin = nullptr;
1607  IRBuilder<> &IRB;
1608  MemorySanitizerVisitor *MSV;
1609 
1610  public:
1611  Combiner(MemorySanitizerVisitor *MSV, IRBuilder<> &IRB)
1612  : IRB(IRB), MSV(MSV) {}
1613 
1614  /// \brief Add a pair of shadow and origin values to the mix.
1615  Combiner &Add(Value *OpShadow, Value *OpOrigin) {
1616  if (CombineShadow) {
1617  assert(OpShadow);
1618  if (!Shadow)
1619  Shadow = OpShadow;
1620  else {
1621  OpShadow = MSV->CreateShadowCast(IRB, OpShadow, Shadow->getType());
1622  Shadow = IRB.CreateOr(Shadow, OpShadow, "_msprop");
1623  }
1624  }
1625 
1626  if (MSV->MS.TrackOrigins) {
1627  assert(OpOrigin);
1628  if (!Origin) {
1629  Origin = OpOrigin;
1630  } else {
1631  Constant *ConstOrigin = dyn_cast<Constant>(OpOrigin);
1632  // No point in adding something that might result in 0 origin value.
1633  if (!ConstOrigin || !ConstOrigin->isNullValue()) {
1634  Value *FlatShadow = MSV->convertToShadowTyNoVec(OpShadow, IRB);
1635  Value *Cond =
1636  IRB.CreateICmpNE(FlatShadow, MSV->getCleanShadow(FlatShadow));
1637  Origin = IRB.CreateSelect(Cond, OpOrigin, Origin);
1638  }
1639  }
1640  }
1641  return *this;
1642  }
1643 
1644  /// \brief Add an application value to the mix.
1645  Combiner &Add(Value *V) {
1646  Value *OpShadow = MSV->getShadow(V);
1647  Value *OpOrigin = MSV->MS.TrackOrigins ? MSV->getOrigin(V) : nullptr;
1648  return Add(OpShadow, OpOrigin);
1649  }
1650 
1651  /// \brief Set the current combined values as the given instruction's shadow
1652  /// and origin.
1653  void Done(Instruction *I) {
1654  if (CombineShadow) {
1655  assert(Shadow);
1656  Shadow = MSV->CreateShadowCast(IRB, Shadow, MSV->getShadowTy(I));
1657  MSV->setShadow(I, Shadow);
1658  }
1659  if (MSV->MS.TrackOrigins) {
1660  assert(Origin);
1661  MSV->setOrigin(I, Origin);
1662  }
1663  }
1664  };
1665 
1666  using ShadowAndOriginCombiner = Combiner<true>;
1667  using OriginCombiner = Combiner<false>;
1668 
1669  /// \brief Propagate origin for arbitrary operation.
1670  void setOriginForNaryOp(Instruction &I) {
1671  if (!MS.TrackOrigins) return;
1672  IRBuilder<> IRB(&I);
1673  OriginCombiner OC(this, IRB);
1674  for (Instruction::op_iterator OI = I.op_begin(); OI != I.op_end(); ++OI)
1675  OC.Add(OI->get());
1676  OC.Done(&I);
1677  }
1678 
1679  size_t VectorOrPrimitiveTypeSizeInBits(Type *Ty) {
1680  assert(!(Ty->isVectorTy() && Ty->getScalarType()->isPointerTy()) &&
1681  "Vector of pointers is not a valid shadow type");
1682  return Ty->isVectorTy() ?
1684  Ty->getPrimitiveSizeInBits();
1685  }
1686 
1687  /// \brief Cast between two shadow types, extending or truncating as
1688  /// necessary.
1689  Value *CreateShadowCast(IRBuilder<> &IRB, Value *V, Type *dstTy,
1690  bool Signed = false) {
1691  Type *srcTy = V->getType();
1692  size_t srcSizeInBits = VectorOrPrimitiveTypeSizeInBits(srcTy);
1693  size_t dstSizeInBits = VectorOrPrimitiveTypeSizeInBits(dstTy);
1694  if (srcSizeInBits > 1 && dstSizeInBits == 1)
1695  return IRB.CreateICmpNE(V, getCleanShadow(V));
1696 
1697  if (dstTy->isIntegerTy() && srcTy->isIntegerTy())
1698  return IRB.CreateIntCast(V, dstTy, Signed);
1699  if (dstTy->isVectorTy() && srcTy->isVectorTy() &&
1700  dstTy->getVectorNumElements() == srcTy->getVectorNumElements())
1701  return IRB.CreateIntCast(V, dstTy, Signed);
1702  Value *V1 = IRB.CreateBitCast(V, Type::getIntNTy(*MS.C, srcSizeInBits));
1703  Value *V2 =
1704  IRB.CreateIntCast(V1, Type::getIntNTy(*MS.C, dstSizeInBits), Signed);
1705  return IRB.CreateBitCast(V2, dstTy);
1706  // TODO: handle struct types.
1707  }
1708 
1709  /// \brief Cast an application value to the type of its own shadow.
1710  Value *CreateAppToShadowCast(IRBuilder<> &IRB, Value *V) {
1711  Type *ShadowTy = getShadowTy(V);
1712  if (V->getType() == ShadowTy)
1713  return V;
1714  if (V->getType()->isPtrOrPtrVectorTy())
1715  return IRB.CreatePtrToInt(V, ShadowTy);
1716  else
1717  return IRB.CreateBitCast(V, ShadowTy);
1718  }
1719 
1720  /// \brief Propagate shadow for arbitrary operation.
1721  void handleShadowOr(Instruction &I) {
1722  IRBuilder<> IRB(&I);
1723  ShadowAndOriginCombiner SC(this, IRB);
1724  for (Instruction::op_iterator OI = I.op_begin(); OI != I.op_end(); ++OI)
1725  SC.Add(OI->get());
1726  SC.Done(&I);
1727  }
1728 
1729  // \brief Handle multiplication by constant.
1730  //
1731  // Handle a special case of multiplication by constant that may have one or
1732  // more zeros in the lower bits. This makes corresponding number of lower bits
1733  // of the result zero as well. We model it by shifting the other operand
1734  // shadow left by the required number of bits. Effectively, we transform
1735  // (X * (A * 2**B)) to ((X << B) * A) and instrument (X << B) as (Sx << B).
1736  // We use multiplication by 2**N instead of shift to cover the case of
1737  // multiplication by 0, which may occur in some elements of a vector operand.
1738  void handleMulByConstant(BinaryOperator &I, Constant *ConstArg,
1739  Value *OtherArg) {
1740  Constant *ShadowMul;
1741  Type *Ty = ConstArg->getType();
1742  if (Ty->isVectorTy()) {
1743  unsigned NumElements = Ty->getVectorNumElements();
1744  Type *EltTy = Ty->getSequentialElementType();
1745  SmallVector<Constant *, 16> Elements;
1746  for (unsigned Idx = 0; Idx < NumElements; ++Idx) {
1747  if (ConstantInt *Elt =
1748  dyn_cast<ConstantInt>(ConstArg->getAggregateElement(Idx))) {
1749  const APInt &V = Elt->getValue();
1750  APInt V2 = APInt(V.getBitWidth(), 1) << V.countTrailingZeros();
1751  Elements.push_back(ConstantInt::get(EltTy, V2));
1752  } else {
1753  Elements.push_back(ConstantInt::get(EltTy, 1));
1754  }
1755  }
1756  ShadowMul = ConstantVector::get(Elements);
1757  } else {
1758  if (ConstantInt *Elt = dyn_cast<ConstantInt>(ConstArg)) {
1759  const APInt &V = Elt->getValue();
1760  APInt V2 = APInt(V.getBitWidth(), 1) << V.countTrailingZeros();
1761  ShadowMul = ConstantInt::get(Ty, V2);
1762  } else {
1763  ShadowMul = ConstantInt::get(Ty, 1);
1764  }
1765  }
1766 
1767  IRBuilder<> IRB(&I);
1768  setShadow(&I,
1769  IRB.CreateMul(getShadow(OtherArg), ShadowMul, "msprop_mul_cst"));
1770  setOrigin(&I, getOrigin(OtherArg));
1771  }
1772 
1773  void visitMul(BinaryOperator &I) {
1774  Constant *constOp0 = dyn_cast<Constant>(I.getOperand(0));
1775  Constant *constOp1 = dyn_cast<Constant>(I.getOperand(1));
1776  if (constOp0 && !constOp1)
1777  handleMulByConstant(I, constOp0, I.getOperand(1));
1778  else if (constOp1 && !constOp0)
1779  handleMulByConstant(I, constOp1, I.getOperand(0));
1780  else
1781  handleShadowOr(I);
1782  }
1783 
1784  void visitFAdd(BinaryOperator &I) { handleShadowOr(I); }
1785  void visitFSub(BinaryOperator &I) { handleShadowOr(I); }
1786  void visitFMul(BinaryOperator &I) { handleShadowOr(I); }
1787  void visitAdd(BinaryOperator &I) { handleShadowOr(I); }
1788  void visitSub(BinaryOperator &I) { handleShadowOr(I); }
1789  void visitXor(BinaryOperator &I) { handleShadowOr(I); }
1790 
1791  void handleDiv(Instruction &I) {
1792  IRBuilder<> IRB(&I);
1793  // Strict on the second argument.
1794  insertShadowCheck(I.getOperand(1), &I);
1795  setShadow(&I, getShadow(&I, 0));
1796  setOrigin(&I, getOrigin(&I, 0));
1797  }
1798 
1799  void visitUDiv(BinaryOperator &I) { handleDiv(I); }
1800  void visitSDiv(BinaryOperator &I) { handleDiv(I); }
1801  void visitFDiv(BinaryOperator &I) { handleDiv(I); }
1802  void visitURem(BinaryOperator &I) { handleDiv(I); }
1803  void visitSRem(BinaryOperator &I) { handleDiv(I); }
1804  void visitFRem(BinaryOperator &I) { handleDiv(I); }
1805 
1806  /// \brief Instrument == and != comparisons.
1807  ///
1808  /// Sometimes the comparison result is known even if some of the bits of the
1809  /// arguments are not.
1810  void handleEqualityComparison(ICmpInst &I) {
1811  IRBuilder<> IRB(&I);
1812  Value *A = I.getOperand(0);
1813  Value *B = I.getOperand(1);
1814  Value *Sa = getShadow(A);
1815  Value *Sb = getShadow(B);
1816 
1817  // Get rid of pointers and vectors of pointers.
1818  // For ints (and vectors of ints), types of A and Sa match,
1819  // and this is a no-op.
1820  A = IRB.CreatePointerCast(A, Sa->getType());
1821  B = IRB.CreatePointerCast(B, Sb->getType());
1822 
1823  // A == B <==> (C = A^B) == 0
1824  // A != B <==> (C = A^B) != 0
1825  // Sc = Sa | Sb
1826  Value *C = IRB.CreateXor(A, B);
1827  Value *Sc = IRB.CreateOr(Sa, Sb);
1828  // Now dealing with i = (C == 0) comparison (or C != 0, does not matter now)
1829  // Result is defined if one of the following is true
1830  // * there is a defined 1 bit in C
1831  // * C is fully defined
1832  // Si = !(C & ~Sc) && Sc
1833  Value *Zero = Constant::getNullValue(Sc->getType());
1834  Value *MinusOne = Constant::getAllOnesValue(Sc->getType());
1835  Value *Si =
1836  IRB.CreateAnd(IRB.CreateICmpNE(Sc, Zero),
1837  IRB.CreateICmpEQ(
1838  IRB.CreateAnd(IRB.CreateXor(Sc, MinusOne), C), Zero));
1839  Si->setName("_msprop_icmp");
1840  setShadow(&I, Si);
1841  setOriginForNaryOp(I);
1842  }
1843 
1844  /// \brief Build the lowest possible value of V, taking into account V's
1845  /// uninitialized bits.
1846  Value *getLowestPossibleValue(IRBuilder<> &IRB, Value *A, Value *Sa,
1847  bool isSigned) {
1848  if (isSigned) {
1849  // Split shadow into sign bit and other bits.
1850  Value *SaOtherBits = IRB.CreateLShr(IRB.CreateShl(Sa, 1), 1);
1851  Value *SaSignBit = IRB.CreateXor(Sa, SaOtherBits);
1852  // Maximise the undefined shadow bit, minimize other undefined bits.
1853  return
1854  IRB.CreateOr(IRB.CreateAnd(A, IRB.CreateNot(SaOtherBits)), SaSignBit);
1855  } else {
1856  // Minimize undefined bits.
1857  return IRB.CreateAnd(A, IRB.CreateNot(Sa));
1858  }
1859  }
1860 
1861  /// \brief Build the highest possible value of V, taking into account V's
1862  /// uninitialized bits.
1863  Value *getHighestPossibleValue(IRBuilder<> &IRB, Value *A, Value *Sa,
1864  bool isSigned) {
1865  if (isSigned) {
1866  // Split shadow into sign bit and other bits.
1867  Value *SaOtherBits = IRB.CreateLShr(IRB.CreateShl(Sa, 1), 1);
1868  Value *SaSignBit = IRB.CreateXor(Sa, SaOtherBits);
1869  // Minimise the undefined shadow bit, maximise other undefined bits.
1870  return
1871  IRB.CreateOr(IRB.CreateAnd(A, IRB.CreateNot(SaSignBit)), SaOtherBits);
1872  } else {
1873  // Maximize undefined bits.
1874  return IRB.CreateOr(A, Sa);
1875  }
1876  }
1877 
1878  /// \brief Instrument relational comparisons.
1879  ///
1880  /// This function does exact shadow propagation for all relational
1881  /// comparisons of integers, pointers and vectors of those.
1882  /// FIXME: output seems suboptimal when one of the operands is a constant
1883  void handleRelationalComparisonExact(ICmpInst &I) {
1884  IRBuilder<> IRB(&I);
1885  Value *A = I.getOperand(0);
1886  Value *B = I.getOperand(1);
1887  Value *Sa = getShadow(A);
1888  Value *Sb = getShadow(B);
1889 
1890  // Get rid of pointers and vectors of pointers.
1891  // For ints (and vectors of ints), types of A and Sa match,
1892  // and this is a no-op.
1893  A = IRB.CreatePointerCast(A, Sa->getType());
1894  B = IRB.CreatePointerCast(B, Sb->getType());
1895 
1896  // Let [a0, a1] be the interval of possible values of A, taking into account
1897  // its undefined bits. Let [b0, b1] be the interval of possible values of B.
1898  // Then (A cmp B) is defined iff (a0 cmp b1) == (a1 cmp b0).
1899  bool IsSigned = I.isSigned();
1900  Value *S1 = IRB.CreateICmp(I.getPredicate(),
1901  getLowestPossibleValue(IRB, A, Sa, IsSigned),
1902  getHighestPossibleValue(IRB, B, Sb, IsSigned));
1903  Value *S2 = IRB.CreateICmp(I.getPredicate(),
1904  getHighestPossibleValue(IRB, A, Sa, IsSigned),
1905  getLowestPossibleValue(IRB, B, Sb, IsSigned));
1906  Value *Si = IRB.CreateXor(S1, S2);
1907  setShadow(&I, Si);
1908  setOriginForNaryOp(I);
1909  }
1910 
1911  /// \brief Instrument signed relational comparisons.
1912  ///
1913  /// Handle sign bit tests: x<0, x>=0, x<=-1, x>-1 by propagating the highest
1914  /// bit of the shadow. Everything else is delegated to handleShadowOr().
1915  void handleSignedRelationalComparison(ICmpInst &I) {
1916  Constant *constOp;
1917  Value *op = nullptr;
1918  CmpInst::Predicate pre;
1919  if ((constOp = dyn_cast<Constant>(I.getOperand(1)))) {
1920  op = I.getOperand(0);
1921  pre = I.getPredicate();
1922  } else if ((constOp = dyn_cast<Constant>(I.getOperand(0)))) {
1923  op = I.getOperand(1);
1924  pre = I.getSwappedPredicate();
1925  } else {
1926  handleShadowOr(I);
1927  return;
1928  }
1929 
1930  if ((constOp->isNullValue() &&
1931  (pre == CmpInst::ICMP_SLT || pre == CmpInst::ICMP_SGE)) ||
1932  (constOp->isAllOnesValue() &&
1933  (pre == CmpInst::ICMP_SGT || pre == CmpInst::ICMP_SLE))) {
1934  IRBuilder<> IRB(&I);
1935  Value *Shadow = IRB.CreateICmpSLT(getShadow(op), getCleanShadow(op),
1936  "_msprop_icmp_s");
1937  setShadow(&I, Shadow);
1938  setOrigin(&I, getOrigin(op));
1939  } else {
1940  handleShadowOr(I);
1941  }
1942  }
1943 
1944  void visitICmpInst(ICmpInst &I) {
1945  if (!ClHandleICmp) {
1946  handleShadowOr(I);
1947  return;
1948  }
1949  if (I.isEquality()) {
1950  handleEqualityComparison(I);
1951  return;
1952  }
1953 
1954  assert(I.isRelational());
1955  if (ClHandleICmpExact) {
1956  handleRelationalComparisonExact(I);
1957  return;
1958  }
1959  if (I.isSigned()) {
1960  handleSignedRelationalComparison(I);
1961  return;
1962  }
1963 
1964  assert(I.isUnsigned());
1965  if ((isa<Constant>(I.getOperand(0)) || isa<Constant>(I.getOperand(1)))) {
1966  handleRelationalComparisonExact(I);
1967  return;
1968  }
1969 
1970  handleShadowOr(I);
1971  }
1972 
1973  void visitFCmpInst(FCmpInst &I) {
1974  handleShadowOr(I);
1975  }
1976 
1977  void handleShift(BinaryOperator &I) {
1978  IRBuilder<> IRB(&I);
1979  // If any of the S2 bits are poisoned, the whole thing is poisoned.
1980  // Otherwise perform the same shift on S1.
1981  Value *S1 = getShadow(&I, 0);
1982  Value *S2 = getShadow(&I, 1);
1983  Value *S2Conv = IRB.CreateSExt(IRB.CreateICmpNE(S2, getCleanShadow(S2)),
1984  S2->getType());
1985  Value *V2 = I.getOperand(1);
1986  Value *Shift = IRB.CreateBinOp(I.getOpcode(), S1, V2);
1987  setShadow(&I, IRB.CreateOr(Shift, S2Conv));
1988  setOriginForNaryOp(I);
1989  }
1990 
1991  void visitShl(BinaryOperator &I) { handleShift(I); }
1992  void visitAShr(BinaryOperator &I) { handleShift(I); }
1993  void visitLShr(BinaryOperator &I) { handleShift(I); }
1994 
1995  /// \brief Instrument llvm.memmove
1996  ///
1997  /// At this point we don't know if llvm.memmove will be inlined or not.
1998  /// If we don't instrument it and it gets inlined,
1999  /// our interceptor will not kick in and we will lose the memmove.
2000  /// If we instrument the call here, but it does not get inlined,
2001  /// we will memove the shadow twice: which is bad in case
2002  /// of overlapping regions. So, we simply lower the intrinsic to a call.
2003  ///
2004  /// Similar situation exists for memcpy and memset.
2005  void visitMemMoveInst(MemMoveInst &I) {
2006  IRBuilder<> IRB(&I);
2007  IRB.CreateCall(
2008  MS.MemmoveFn,
2009  {IRB.CreatePointerCast(I.getArgOperand(0), IRB.getInt8PtrTy()),
2010  IRB.CreatePointerCast(I.getArgOperand(1), IRB.getInt8PtrTy()),
2011  IRB.CreateIntCast(I.getArgOperand(2), MS.IntptrTy, false)});
2012  I.eraseFromParent();
2013  }
2014 
2015  // Similar to memmove: avoid copying shadow twice.
2016  // This is somewhat unfortunate as it may slowdown small constant memcpys.
2017  // FIXME: consider doing manual inline for small constant sizes and proper
2018  // alignment.
2019  void visitMemCpyInst(MemCpyInst &I) {
2020  IRBuilder<> IRB(&I);
2021  IRB.CreateCall(
2022  MS.MemcpyFn,
2023  {IRB.CreatePointerCast(I.getArgOperand(0), IRB.getInt8PtrTy()),
2024  IRB.CreatePointerCast(I.getArgOperand(1), IRB.getInt8PtrTy()),
2025  IRB.CreateIntCast(I.getArgOperand(2), MS.IntptrTy, false)});
2026  I.eraseFromParent();
2027  }
2028 
2029  // Same as memcpy.
2030  void visitMemSetInst(MemSetInst &I) {
2031  IRBuilder<> IRB(&I);
2032  IRB.CreateCall(
2033  MS.MemsetFn,
2034  {IRB.CreatePointerCast(I.getArgOperand(0), IRB.getInt8PtrTy()),
2035  IRB.CreateIntCast(I.getArgOperand(1), IRB.getInt32Ty(), false),
2036  IRB.CreateIntCast(I.getArgOperand(2), MS.IntptrTy, false)});
2037  I.eraseFromParent();
2038  }
2039 
2040  void visitVAStartInst(VAStartInst &I) {
2041  VAHelper->visitVAStartInst(I);
2042  }
2043 
2044  void visitVACopyInst(VACopyInst &I) {
2045  VAHelper->visitVACopyInst(I);
2046  }
2047 
2048  /// \brief Handle vector store-like intrinsics.
2049  ///
2050  /// Instrument intrinsics that look like a simple SIMD store: writes memory,
2051  /// has 1 pointer argument and 1 vector argument, returns void.
2052  bool handleVectorStoreIntrinsic(IntrinsicInst &I) {
2053  IRBuilder<> IRB(&I);
2054  Value* Addr = I.getArgOperand(0);
2055  Value *Shadow = getShadow(&I, 1);
2056  Value *ShadowPtr, *OriginPtr;
2057 
2058  // We don't know the pointer alignment (could be unaligned SSE store!).
2059  // Have to assume to worst case.
2060  std::tie(ShadowPtr, OriginPtr) =
2061  getShadowOriginPtr(Addr, IRB, Shadow->getType(), /*Alignment*/ 1);
2062  IRB.CreateAlignedStore(Shadow, ShadowPtr, 1);
2063 
2065  insertShadowCheck(Addr, &I);
2066 
2067  // FIXME: factor out common code from materializeStores
2068  if (MS.TrackOrigins) IRB.CreateStore(getOrigin(&I, 1), OriginPtr);
2069  return true;
2070  }
2071 
2072  /// \brief Handle vector load-like intrinsics.
2073  ///
2074  /// Instrument intrinsics that look like a simple SIMD load: reads memory,
2075  /// has 1 pointer argument, returns a vector.
2076  bool handleVectorLoadIntrinsic(IntrinsicInst &I) {
2077  IRBuilder<> IRB(&I);
2078  Value *Addr = I.getArgOperand(0);
2079 
2080  Type *ShadowTy = getShadowTy(&I);
2081  Value *ShadowPtr, *OriginPtr;
2082  if (PropagateShadow) {
2083  // We don't know the pointer alignment (could be unaligned SSE load!).
2084  // Have to assume to worst case.
2085  unsigned Alignment = 1;
2086  std::tie(ShadowPtr, OriginPtr) =
2087  getShadowOriginPtr(Addr, IRB, ShadowTy, Alignment);
2088  setShadow(&I, IRB.CreateAlignedLoad(ShadowPtr, Alignment, "_msld"));
2089  } else {
2090  setShadow(&I, getCleanShadow(&I));
2091  }
2092 
2094  insertShadowCheck(Addr, &I);
2095 
2096  if (MS.TrackOrigins) {
2097  if (PropagateShadow)
2098  setOrigin(&I, IRB.CreateLoad(OriginPtr));
2099  else
2100  setOrigin(&I, getCleanOrigin());
2101  }
2102  return true;
2103  }
2104 
2105  /// \brief Handle (SIMD arithmetic)-like intrinsics.
2106  ///
2107  /// Instrument intrinsics with any number of arguments of the same type,
2108  /// equal to the return type. The type should be simple (no aggregates or
2109  /// pointers; vectors are fine).
2110  /// Caller guarantees that this intrinsic does not access memory.
2111  bool maybeHandleSimpleNomemIntrinsic(IntrinsicInst &I) {
2112  Type *RetTy = I.getType();
2113  if (!(RetTy->isIntOrIntVectorTy() ||
2114  RetTy->isFPOrFPVectorTy() ||
2115  RetTy->isX86_MMXTy()))
2116  return false;
2117 
2118  unsigned NumArgOperands = I.getNumArgOperands();
2119 
2120  for (unsigned i = 0; i < NumArgOperands; ++i) {
2121  Type *Ty = I.getArgOperand(i)->getType();
2122  if (Ty != RetTy)
2123  return false;
2124  }
2125 
2126  IRBuilder<> IRB(&I);
2127  ShadowAndOriginCombiner SC(this, IRB);
2128  for (unsigned i = 0; i < NumArgOperands; ++i)
2129  SC.Add(I.getArgOperand(i));
2130  SC.Done(&I);
2131 
2132  return true;
2133  }
2134 
2135  /// \brief Heuristically instrument unknown intrinsics.
2136  ///
2137  /// The main purpose of this code is to do something reasonable with all
2138  /// random intrinsics we might encounter, most importantly - SIMD intrinsics.
2139  /// We recognize several classes of intrinsics by their argument types and
2140  /// ModRefBehaviour and apply special intrumentation when we are reasonably
2141  /// sure that we know what the intrinsic does.
2142  ///
2143  /// We special-case intrinsics where this approach fails. See llvm.bswap
2144  /// handling as an example of that.
2145  bool handleUnknownIntrinsic(IntrinsicInst &I) {
2146  unsigned NumArgOperands = I.getNumArgOperands();
2147  if (NumArgOperands == 0)
2148  return false;
2149 
2150  if (NumArgOperands == 2 &&
2151  I.getArgOperand(0)->getType()->isPointerTy() &&
2152  I.getArgOperand(1)->getType()->isVectorTy() &&
2153  I.getType()->isVoidTy() &&
2154  !I.onlyReadsMemory()) {
2155  // This looks like a vector store.
2156  return handleVectorStoreIntrinsic(I);
2157  }
2158 
2159  if (NumArgOperands == 1 &&
2160  I.getArgOperand(0)->getType()->isPointerTy() &&
2161  I.getType()->isVectorTy() &&
2162  I.onlyReadsMemory()) {
2163  // This looks like a vector load.
2164  return handleVectorLoadIntrinsic(I);
2165  }
2166 
2167  if (I.doesNotAccessMemory())
2168  if (maybeHandleSimpleNomemIntrinsic(I))
2169  return true;
2170 
2171  // FIXME: detect and handle SSE maskstore/maskload
2172  return false;
2173  }
2174 
2175  void handleBswap(IntrinsicInst &I) {
2176  IRBuilder<> IRB(&I);
2177  Value *Op = I.getArgOperand(0);
2178  Type *OpType = Op->getType();
2179  Function *BswapFunc = Intrinsic::getDeclaration(
2180  F.getParent(), Intrinsic::bswap, makeArrayRef(&OpType, 1));
2181  setShadow(&I, IRB.CreateCall(BswapFunc, getShadow(Op)));
2182  setOrigin(&I, getOrigin(Op));
2183  }
2184 
2185  // \brief Instrument vector convert instrinsic.
2186  //
2187  // This function instruments intrinsics like cvtsi2ss:
2188  // %Out = int_xxx_cvtyyy(%ConvertOp)
2189  // or
2190  // %Out = int_xxx_cvtyyy(%CopyOp, %ConvertOp)
2191  // Intrinsic converts \p NumUsedElements elements of \p ConvertOp to the same
2192  // number \p Out elements, and (if has 2 arguments) copies the rest of the
2193  // elements from \p CopyOp.
2194  // In most cases conversion involves floating-point value which may trigger a
2195  // hardware exception when not fully initialized. For this reason we require
2196  // \p ConvertOp[0:NumUsedElements] to be fully initialized and trap otherwise.
2197  // We copy the shadow of \p CopyOp[NumUsedElements:] to \p
2198  // Out[NumUsedElements:]. This means that intrinsics without \p CopyOp always
2199  // return a fully initialized value.
2200  void handleVectorConvertIntrinsic(IntrinsicInst &I, int NumUsedElements) {
2201  IRBuilder<> IRB(&I);
2202  Value *CopyOp, *ConvertOp;
2203 
2204  switch (I.getNumArgOperands()) {
2205  case 3:
2206  assert(isa<ConstantInt>(I.getArgOperand(2)) && "Invalid rounding mode");
2208  case 2:
2209  CopyOp = I.getArgOperand(0);
2210  ConvertOp = I.getArgOperand(1);
2211  break;
2212  case 1:
2213  ConvertOp = I.getArgOperand(0);
2214  CopyOp = nullptr;
2215  break;
2216  default:
2217  llvm_unreachable("Cvt intrinsic with unsupported number of arguments.");
2218  }
2219 
2220  // The first *NumUsedElements* elements of ConvertOp are converted to the
2221  // same number of output elements. The rest of the output is copied from
2222  // CopyOp, or (if not available) filled with zeroes.
2223  // Combine shadow for elements of ConvertOp that are used in this operation,
2224  // and insert a check.
2225  // FIXME: consider propagating shadow of ConvertOp, at least in the case of
2226  // int->any conversion.
2227  Value *ConvertShadow = getShadow(ConvertOp);
2228  Value *AggShadow = nullptr;
2229  if (ConvertOp->getType()->isVectorTy()) {
2230  AggShadow = IRB.CreateExtractElement(
2231  ConvertShadow, ConstantInt::get(IRB.getInt32Ty(), 0));
2232  for (int i = 1; i < NumUsedElements; ++i) {
2233  Value *MoreShadow = IRB.CreateExtractElement(
2234  ConvertShadow, ConstantInt::get(IRB.getInt32Ty(), i));
2235  AggShadow = IRB.CreateOr(AggShadow, MoreShadow);
2236  }
2237  } else {
2238  AggShadow = ConvertShadow;
2239  }
2240  assert(AggShadow->getType()->isIntegerTy());
2241  insertShadowCheck(AggShadow, getOrigin(ConvertOp), &I);
2242 
2243  // Build result shadow by zero-filling parts of CopyOp shadow that come from
2244  // ConvertOp.
2245  if (CopyOp) {
2246  assert(CopyOp->getType() == I.getType());
2247  assert(CopyOp->getType()->isVectorTy());
2248  Value *ResultShadow = getShadow(CopyOp);
2249  Type *EltTy = ResultShadow->getType()->getVectorElementType();
2250  for (int i = 0; i < NumUsedElements; ++i) {
2251  ResultShadow = IRB.CreateInsertElement(
2252  ResultShadow, ConstantInt::getNullValue(EltTy),
2253  ConstantInt::get(IRB.getInt32Ty(), i));
2254  }
2255  setShadow(&I, ResultShadow);
2256  setOrigin(&I, getOrigin(CopyOp));
2257  } else {
2258  setShadow(&I, getCleanShadow(&I));
2259  setOrigin(&I, getCleanOrigin());
2260  }
2261  }
2262 
2263  // Given a scalar or vector, extract lower 64 bits (or less), and return all
2264  // zeroes if it is zero, and all ones otherwise.
2265  Value *Lower64ShadowExtend(IRBuilder<> &IRB, Value *S, Type *T) {
2266  if (S->getType()->isVectorTy())
2267  S = CreateShadowCast(IRB, S, IRB.getInt64Ty(), /* Signed */ true);
2268  assert(S->getType()->getPrimitiveSizeInBits() <= 64);
2269  Value *S2 = IRB.CreateICmpNE(S, getCleanShadow(S));
2270  return CreateShadowCast(IRB, S2, T, /* Signed */ true);
2271  }
2272 
2273  // Given a vector, extract its first element, and return all
2274  // zeroes if it is zero, and all ones otherwise.
2275  Value *LowerElementShadowExtend(IRBuilder<> &IRB, Value *S, Type *T) {
2276  Value *S1 = IRB.CreateExtractElement(S, (uint64_t)0);
2277  Value *S2 = IRB.CreateICmpNE(S1, getCleanShadow(S1));
2278  return CreateShadowCast(IRB, S2, T, /* Signed */ true);
2279  }
2280 
2281  Value *VariableShadowExtend(IRBuilder<> &IRB, Value *S) {
2282  Type *T = S->getType();
2283  assert(T->isVectorTy());
2284  Value *S2 = IRB.CreateICmpNE(S, getCleanShadow(S));
2285  return IRB.CreateSExt(S2, T);
2286  }
2287 
2288  // \brief Instrument vector shift instrinsic.
2289  //
2290  // This function instruments intrinsics like int_x86_avx2_psll_w.
2291  // Intrinsic shifts %In by %ShiftSize bits.
2292  // %ShiftSize may be a vector. In that case the lower 64 bits determine shift
2293  // size, and the rest is ignored. Behavior is defined even if shift size is
2294  // greater than register (or field) width.
2295  void handleVectorShiftIntrinsic(IntrinsicInst &I, bool Variable) {
2296  assert(I.getNumArgOperands() == 2);
2297  IRBuilder<> IRB(&I);
2298  // If any of the S2 bits are poisoned, the whole thing is poisoned.
2299  // Otherwise perform the same shift on S1.
2300  Value *S1 = getShadow(&I, 0);
2301  Value *S2 = getShadow(&I, 1);
2302  Value *S2Conv = Variable ? VariableShadowExtend(IRB, S2)
2303  : Lower64ShadowExtend(IRB, S2, getShadowTy(&I));
2304  Value *V1 = I.getOperand(0);
2305  Value *V2 = I.getOperand(1);
2306  Value *Shift = IRB.CreateCall(I.getCalledValue(),
2307  {IRB.CreateBitCast(S1, V1->getType()), V2});
2308  Shift = IRB.CreateBitCast(Shift, getShadowTy(&I));
2309  setShadow(&I, IRB.CreateOr(Shift, S2Conv));
2310  setOriginForNaryOp(I);
2311  }
2312 
2313  // \brief Get an X86_MMX-sized vector type.
2314  Type *getMMXVectorTy(unsigned EltSizeInBits) {
2315  const unsigned X86_MMXSizeInBits = 64;
2316  return VectorType::get(IntegerType::get(*MS.C, EltSizeInBits),
2317  X86_MMXSizeInBits / EltSizeInBits);
2318  }
2319 
2320  // \brief Returns a signed counterpart for an (un)signed-saturate-and-pack
2321  // intrinsic.
2322  Intrinsic::ID getSignedPackIntrinsic(Intrinsic::ID id) {
2323  switch (id) {
2324  case Intrinsic::x86_sse2_packsswb_128:
2325  case Intrinsic::x86_sse2_packuswb_128:
2326  return Intrinsic::x86_sse2_packsswb_128;
2327 
2328  case Intrinsic::x86_sse2_packssdw_128:
2329  case Intrinsic::x86_sse41_packusdw:
2330  return Intrinsic::x86_sse2_packssdw_128;
2331 
2332  case Intrinsic::x86_avx2_packsswb:
2333  case Intrinsic::x86_avx2_packuswb:
2334  return Intrinsic::x86_avx2_packsswb;
2335 
2336  case Intrinsic::x86_avx2_packssdw:
2337  case Intrinsic::x86_avx2_packusdw:
2338  return Intrinsic::x86_avx2_packssdw;
2339 
2340  case Intrinsic::x86_mmx_packsswb:
2341  case Intrinsic::x86_mmx_packuswb:
2342  return Intrinsic::x86_mmx_packsswb;
2343 
2344  case Intrinsic::x86_mmx_packssdw:
2345  return Intrinsic::x86_mmx_packssdw;
2346  default:
2347  llvm_unreachable("unexpected intrinsic id");
2348  }
2349  }
2350 
2351  // \brief Instrument vector pack instrinsic.
2352  //
2353  // This function instruments intrinsics like x86_mmx_packsswb, that
2354  // packs elements of 2 input vectors into half as many bits with saturation.
2355  // Shadow is propagated with the signed variant of the same intrinsic applied
2356  // to sext(Sa != zeroinitializer), sext(Sb != zeroinitializer).
2357  // EltSizeInBits is used only for x86mmx arguments.
2358  void handleVectorPackIntrinsic(IntrinsicInst &I, unsigned EltSizeInBits = 0) {
2359  assert(I.getNumArgOperands() == 2);
2360  bool isX86_MMX = I.getOperand(0)->getType()->isX86_MMXTy();
2361  IRBuilder<> IRB(&I);
2362  Value *S1 = getShadow(&I, 0);
2363  Value *S2 = getShadow(&I, 1);
2364  assert(isX86_MMX || S1->getType()->isVectorTy());
2365 
2366  // SExt and ICmpNE below must apply to individual elements of input vectors.
2367  // In case of x86mmx arguments, cast them to appropriate vector types and
2368  // back.
2369  Type *T = isX86_MMX ? getMMXVectorTy(EltSizeInBits) : S1->getType();
2370  if (isX86_MMX) {
2371  S1 = IRB.CreateBitCast(S1, T);
2372  S2 = IRB.CreateBitCast(S2, T);
2373  }
2374  Value *S1_ext = IRB.CreateSExt(
2375  IRB.CreateICmpNE(S1, Constant::getNullValue(T)), T);
2376  Value *S2_ext = IRB.CreateSExt(
2377  IRB.CreateICmpNE(S2, Constant::getNullValue(T)), T);
2378  if (isX86_MMX) {
2379  Type *X86_MMXTy = Type::getX86_MMXTy(*MS.C);
2380  S1_ext = IRB.CreateBitCast(S1_ext, X86_MMXTy);
2381  S2_ext = IRB.CreateBitCast(S2_ext, X86_MMXTy);
2382  }
2383 
2384  Function *ShadowFn = Intrinsic::getDeclaration(
2385  F.getParent(), getSignedPackIntrinsic(I.getIntrinsicID()));
2386 
2387  Value *S =
2388  IRB.CreateCall(ShadowFn, {S1_ext, S2_ext}, "_msprop_vector_pack");
2389  if (isX86_MMX) S = IRB.CreateBitCast(S, getShadowTy(&I));
2390  setShadow(&I, S);
2391  setOriginForNaryOp(I);
2392  }
2393 
2394  // \brief Instrument sum-of-absolute-differencies intrinsic.
2395  void handleVectorSadIntrinsic(IntrinsicInst &I) {
2396  const unsigned SignificantBitsPerResultElement = 16;
2397  bool isX86_MMX = I.getOperand(0)->getType()->isX86_MMXTy();
2398  Type *ResTy = isX86_MMX ? IntegerType::get(*MS.C, 64) : I.getType();
2399  unsigned ZeroBitsPerResultElement =
2400  ResTy->getScalarSizeInBits() - SignificantBitsPerResultElement;
2401 
2402  IRBuilder<> IRB(&I);
2403  Value *S = IRB.CreateOr(getShadow(&I, 0), getShadow(&I, 1));
2404  S = IRB.CreateBitCast(S, ResTy);
2405  S = IRB.CreateSExt(IRB.CreateICmpNE(S, Constant::getNullValue(ResTy)),
2406  ResTy);
2407  S = IRB.CreateLShr(S, ZeroBitsPerResultElement);
2408  S = IRB.CreateBitCast(S, getShadowTy(&I));
2409  setShadow(&I, S);
2410  setOriginForNaryOp(I);
2411  }
2412 
2413  // \brief Instrument multiply-add intrinsic.
2414  void handleVectorPmaddIntrinsic(IntrinsicInst &I,
2415  unsigned EltSizeInBits = 0) {
2416  bool isX86_MMX = I.getOperand(0)->getType()->isX86_MMXTy();
2417  Type *ResTy = isX86_MMX ? getMMXVectorTy(EltSizeInBits * 2) : I.getType();
2418  IRBuilder<> IRB(&I);
2419  Value *S = IRB.CreateOr(getShadow(&I, 0), getShadow(&I, 1));
2420  S = IRB.CreateBitCast(S, ResTy);
2421  S = IRB.CreateSExt(IRB.CreateICmpNE(S, Constant::getNullValue(ResTy)),
2422  ResTy);
2423  S = IRB.CreateBitCast(S, getShadowTy(&I));
2424  setShadow(&I, S);
2425  setOriginForNaryOp(I);
2426  }
2427 
2428  // \brief Instrument compare-packed intrinsic.
2429  // Basically, an or followed by sext(icmp ne 0) to end up with all-zeros or
2430  // all-ones shadow.
2431  void handleVectorComparePackedIntrinsic(IntrinsicInst &I) {
2432  IRBuilder<> IRB(&I);
2433  Type *ResTy = getShadowTy(&I);
2434  Value *S0 = IRB.CreateOr(getShadow(&I, 0), getShadow(&I, 1));
2435  Value *S = IRB.CreateSExt(
2436  IRB.CreateICmpNE(S0, Constant::getNullValue(ResTy)), ResTy);
2437  setShadow(&I, S);
2438  setOriginForNaryOp(I);
2439  }
2440 
2441  // \brief Instrument compare-scalar intrinsic.
2442  // This handles both cmp* intrinsics which return the result in the first
2443  // element of a vector, and comi* which return the result as i32.
2444  void handleVectorCompareScalarIntrinsic(IntrinsicInst &I) {
2445  IRBuilder<> IRB(&I);
2446  Value *S0 = IRB.CreateOr(getShadow(&I, 0), getShadow(&I, 1));
2447  Value *S = LowerElementShadowExtend(IRB, S0, getShadowTy(&I));
2448  setShadow(&I, S);
2449  setOriginForNaryOp(I);
2450  }
2451 
2452  void handleStmxcsr(IntrinsicInst &I) {
2453  IRBuilder<> IRB(&I);
2454  Value* Addr = I.getArgOperand(0);
2455  Type *Ty = IRB.getInt32Ty();
2456  Value *ShadowPtr = getShadowOriginPtr(Addr, IRB, Ty, /*Alignment*/ 1).first;
2457 
2458  IRB.CreateStore(getCleanShadow(Ty),
2459  IRB.CreatePointerCast(ShadowPtr, Ty->getPointerTo()));
2460 
2462  insertShadowCheck(Addr, &I);
2463  }
2464 
2465  void handleLdmxcsr(IntrinsicInst &I) {
2466  if (!InsertChecks) return;
2467 
2468  IRBuilder<> IRB(&I);
2469  Value *Addr = I.getArgOperand(0);
2470  Type *Ty = IRB.getInt32Ty();
2471  unsigned Alignment = 1;
2472  Value *ShadowPtr, *OriginPtr;
2473  std::tie(ShadowPtr, OriginPtr) =
2474  getShadowOriginPtr(Addr, IRB, Ty, Alignment);
2475 
2477  insertShadowCheck(Addr, &I);
2478 
2479  Value *Shadow = IRB.CreateAlignedLoad(ShadowPtr, Alignment, "_ldmxcsr");
2480  Value *Origin =
2481  MS.TrackOrigins ? IRB.CreateLoad(OriginPtr) : getCleanOrigin();
2482  insertShadowCheck(Shadow, Origin, &I);
2483  }
2484 
2485  void visitIntrinsicInst(IntrinsicInst &I) {
2486  switch (I.getIntrinsicID()) {
2487  case Intrinsic::bswap:
2488  handleBswap(I);
2489  break;
2490  case Intrinsic::x86_sse_stmxcsr:
2491  handleStmxcsr(I);
2492  break;
2493  case Intrinsic::x86_sse_ldmxcsr:
2494  handleLdmxcsr(I);
2495  break;
2496  case Intrinsic::x86_avx512_vcvtsd2usi64:
2497  case Intrinsic::x86_avx512_vcvtsd2usi32:
2498  case Intrinsic::x86_avx512_vcvtss2usi64:
2499  case Intrinsic::x86_avx512_vcvtss2usi32:
2500  case Intrinsic::x86_avx512_cvttss2usi64:
2501  case Intrinsic::x86_avx512_cvttss2usi:
2502  case Intrinsic::x86_avx512_cvttsd2usi64:
2503  case Intrinsic::x86_avx512_cvttsd2usi:
2504  case Intrinsic::x86_avx512_cvtusi2sd:
2505  case Intrinsic::x86_avx512_cvtusi2ss:
2506  case Intrinsic::x86_avx512_cvtusi642sd:
2507  case Intrinsic::x86_avx512_cvtusi642ss:
2508  case Intrinsic::x86_sse2_cvtsd2si64:
2509  case Intrinsic::x86_sse2_cvtsd2si:
2510  case Intrinsic::x86_sse2_cvtsd2ss:
2511  case Intrinsic::x86_sse2_cvtsi2sd:
2512  case Intrinsic::x86_sse2_cvtsi642sd:
2513  case Intrinsic::x86_sse2_cvtss2sd:
2514  case Intrinsic::x86_sse2_cvttsd2si64:
2515  case Intrinsic::x86_sse2_cvttsd2si:
2516  case Intrinsic::x86_sse_cvtsi2ss:
2517  case Intrinsic::x86_sse_cvtsi642ss:
2518  case Intrinsic::x86_sse_cvtss2si64:
2519  case Intrinsic::x86_sse_cvtss2si:
2520  case Intrinsic::x86_sse_cvttss2si64:
2521  case Intrinsic::x86_sse_cvttss2si:
2522  handleVectorConvertIntrinsic(I, 1);
2523  break;
2524  case Intrinsic::x86_sse_cvtps2pi:
2525  case Intrinsic::x86_sse_cvttps2pi:
2526  handleVectorConvertIntrinsic(I, 2);
2527  break;
2528 
2529  case Intrinsic::x86_avx512_psll_w_512:
2530  case Intrinsic::x86_avx512_psll_d_512:
2531  case Intrinsic::x86_avx512_psll_q_512:
2532  case Intrinsic::x86_avx512_pslli_w_512:
2533  case Intrinsic::x86_avx512_pslli_d_512:
2534  case Intrinsic::x86_avx512_pslli_q_512:
2535  case Intrinsic::x86_avx512_psrl_w_512:
2536  case Intrinsic::x86_avx512_psrl_d_512:
2537  case Intrinsic::x86_avx512_psrl_q_512:
2538  case Intrinsic::x86_avx512_psra_w_512:
2539  case Intrinsic::x86_avx512_psra_d_512:
2540  case Intrinsic::x86_avx512_psra_q_512:
2541  case Intrinsic::x86_avx512_psrli_w_512:
2542  case Intrinsic::x86_avx512_psrli_d_512:
2543  case Intrinsic::x86_avx512_psrli_q_512:
2544  case Intrinsic::x86_avx512_psrai_w_512:
2545  case Intrinsic::x86_avx512_psrai_d_512:
2546  case Intrinsic::x86_avx512_psrai_q_512:
2547  case Intrinsic::x86_avx512_psra_q_256:
2548  case Intrinsic::x86_avx512_psra_q_128:
2549  case Intrinsic::x86_avx512_psrai_q_256:
2550  case Intrinsic::x86_avx512_psrai_q_128:
2551  case Intrinsic::x86_avx2_psll_w:
2552  case Intrinsic::x86_avx2_psll_d:
2553  case Intrinsic::x86_avx2_psll_q:
2554  case Intrinsic::x86_avx2_pslli_w:
2555  case Intrinsic::x86_avx2_pslli_d:
2556  case Intrinsic::x86_avx2_pslli_q:
2557  case Intrinsic::x86_avx2_psrl_w:
2558  case Intrinsic::x86_avx2_psrl_d:
2559  case Intrinsic::x86_avx2_psrl_q:
2560  case Intrinsic::x86_avx2_psra_w:
2561  case Intrinsic::x86_avx2_psra_d:
2562  case Intrinsic::x86_avx2_psrli_w:
2563  case Intrinsic::x86_avx2_psrli_d:
2564  case Intrinsic::x86_avx2_psrli_q:
2565  case Intrinsic::x86_avx2_psrai_w:
2566  case Intrinsic::x86_avx2_psrai_d:
2567  case Intrinsic::x86_sse2_psll_w:
2568  case Intrinsic::x86_sse2_psll_d:
2569  case Intrinsic::x86_sse2_psll_q:
2570  case Intrinsic::x86_sse2_pslli_w:
2571  case Intrinsic::x86_sse2_pslli_d:
2572  case Intrinsic::x86_sse2_pslli_q:
2573  case Intrinsic::x86_sse2_psrl_w:
2574  case Intrinsic::x86_sse2_psrl_d:
2575  case Intrinsic::x86_sse2_psrl_q:
2576  case Intrinsic::x86_sse2_psra_w:
2577  case Intrinsic::x86_sse2_psra_d:
2578  case Intrinsic::x86_sse2_psrli_w:
2579  case Intrinsic::x86_sse2_psrli_d:
2580  case Intrinsic::x86_sse2_psrli_q:
2581  case Intrinsic::x86_sse2_psrai_w:
2582  case Intrinsic::x86_sse2_psrai_d:
2583  case Intrinsic::x86_mmx_psll_w:
2584  case Intrinsic::x86_mmx_psll_d:
2585  case Intrinsic::x86_mmx_psll_q:
2586  case Intrinsic::x86_mmx_pslli_w:
2587  case Intrinsic::x86_mmx_pslli_d:
2588  case Intrinsic::x86_mmx_pslli_q:
2589  case Intrinsic::x86_mmx_psrl_w:
2590  case Intrinsic::x86_mmx_psrl_d:
2591  case Intrinsic::x86_mmx_psrl_q:
2592  case Intrinsic::x86_mmx_psra_w:
2593  case Intrinsic::x86_mmx_psra_d:
2594  case Intrinsic::x86_mmx_psrli_w:
2595  case Intrinsic::x86_mmx_psrli_d:
2596  case Intrinsic::x86_mmx_psrli_q:
2597  case Intrinsic::x86_mmx_psrai_w:
2598  case Intrinsic::x86_mmx_psrai_d:
2599  handleVectorShiftIntrinsic(I, /* Variable */ false);
2600  break;
2601  case Intrinsic::x86_avx2_psllv_d:
2602  case Intrinsic::x86_avx2_psllv_d_256:
2603  case Intrinsic::x86_avx512_psllv_d_512:
2604  case Intrinsic::x86_avx2_psllv_q:
2605  case Intrinsic::x86_avx2_psllv_q_256:
2606  case Intrinsic::x86_avx512_psllv_q_512:
2607  case Intrinsic::x86_avx2_psrlv_d:
2608  case Intrinsic::x86_avx2_psrlv_d_256:
2609  case Intrinsic::x86_avx512_psrlv_d_512:
2610  case Intrinsic::x86_avx2_psrlv_q:
2611  case Intrinsic::x86_avx2_psrlv_q_256:
2612  case Intrinsic::x86_avx512_psrlv_q_512:
2613  case Intrinsic::x86_avx2_psrav_d:
2614  case Intrinsic::x86_avx2_psrav_d_256:
2615  case Intrinsic::x86_avx512_psrav_d_512:
2616  case Intrinsic::x86_avx512_psrav_q_128:
2617  case Intrinsic::x86_avx512_psrav_q_256:
2618  case Intrinsic::x86_avx512_psrav_q_512:
2619  handleVectorShiftIntrinsic(I, /* Variable */ true);
2620  break;
2621 
2622  case Intrinsic::x86_sse2_packsswb_128:
2623  case Intrinsic::x86_sse2_packssdw_128:
2624  case Intrinsic::x86_sse2_packuswb_128:
2625  case Intrinsic::x86_sse41_packusdw:
2626  case Intrinsic::x86_avx2_packsswb:
2627  case Intrinsic::x86_avx2_packssdw:
2628  case Intrinsic::x86_avx2_packuswb:
2629  case Intrinsic::x86_avx2_packusdw:
2630  handleVectorPackIntrinsic(I);
2631  break;
2632 
2633  case Intrinsic::x86_mmx_packsswb:
2634  case Intrinsic::x86_mmx_packuswb:
2635  handleVectorPackIntrinsic(I, 16);
2636  break;
2637 
2638  case Intrinsic::x86_mmx_packssdw:
2639  handleVectorPackIntrinsic(I, 32);
2640  break;
2641 
2642  case Intrinsic::x86_mmx_psad_bw:
2643  case Intrinsic::x86_sse2_psad_bw:
2644  case Intrinsic::x86_avx2_psad_bw:
2645  handleVectorSadIntrinsic(I);
2646  break;
2647 
2648  case Intrinsic::x86_sse2_pmadd_wd:
2649  case Intrinsic::x86_avx2_pmadd_wd:
2650  case Intrinsic::x86_ssse3_pmadd_ub_sw_128:
2651  case Intrinsic::x86_avx2_pmadd_ub_sw:
2652  handleVectorPmaddIntrinsic(I);
2653  break;
2654 
2655  case Intrinsic::x86_ssse3_pmadd_ub_sw:
2656  handleVectorPmaddIntrinsic(I, 8);
2657  break;
2658 
2659  case Intrinsic::x86_mmx_pmadd_wd:
2660  handleVectorPmaddIntrinsic(I, 16);
2661  break;
2662 
2663  case Intrinsic::x86_sse_cmp_ss:
2664  case Intrinsic::x86_sse2_cmp_sd:
2665  case Intrinsic::x86_sse_comieq_ss:
2666  case Intrinsic::x86_sse_comilt_ss:
2667  case Intrinsic::x86_sse_comile_ss:
2668  case Intrinsic::x86_sse_comigt_ss:
2669  case Intrinsic::x86_sse_comige_ss:
2670  case Intrinsic::x86_sse_comineq_ss:
2671  case Intrinsic::x86_sse_ucomieq_ss:
2672  case Intrinsic::x86_sse_ucomilt_ss:
2673  case Intrinsic::x86_sse_ucomile_ss:
2674  case Intrinsic::x86_sse_ucomigt_ss:
2675  case Intrinsic::x86_sse_ucomige_ss:
2676  case Intrinsic::x86_sse_ucomineq_ss:
2677  case Intrinsic::x86_sse2_comieq_sd:
2678  case Intrinsic::x86_sse2_comilt_sd:
2679  case Intrinsic::x86_sse2_comile_sd:
2680  case Intrinsic::x86_sse2_comigt_sd:
2681  case Intrinsic::x86_sse2_comige_sd:
2682  case Intrinsic::x86_sse2_comineq_sd:
2683  case Intrinsic::x86_sse2_ucomieq_sd:
2684  case Intrinsic::x86_sse2_ucomilt_sd:
2685  case Intrinsic::x86_sse2_ucomile_sd:
2686  case Intrinsic::x86_sse2_ucomigt_sd:
2687  case Intrinsic::x86_sse2_ucomige_sd:
2688  case Intrinsic::x86_sse2_ucomineq_sd:
2689  handleVectorCompareScalarIntrinsic(I);
2690  break;
2691 
2692  case Intrinsic::x86_sse_cmp_ps:
2693  case Intrinsic::x86_sse2_cmp_pd:
2694  // FIXME: For x86_avx_cmp_pd_256 and x86_avx_cmp_ps_256 this function
2695  // generates reasonably looking IR that fails in the backend with "Do not
2696  // know how to split the result of this operator!".
2697  handleVectorComparePackedIntrinsic(I);
2698  break;
2699 
2700  default:
2701  if (!handleUnknownIntrinsic(I))
2702  visitInstruction(I);
2703  break;
2704  }
2705  }
2706 
2707  void visitCallSite(CallSite CS) {
2708  Instruction &I = *CS.getInstruction();
2709  assert(!I.getMetadata("nosanitize"));
2710  assert((CS.isCall() || CS.isInvoke()) && "Unknown type of CallSite");
2711  if (CS.isCall()) {
2712  CallInst *Call = cast<CallInst>(&I);
2713 
2714  // For inline asm, do the usual thing: check argument shadow and mark all
2715  // outputs as clean. Note that any side effects of the inline asm that are
2716  // not immediately visible in its constraints are not handled.
2717  if (Call->isInlineAsm()) {
2718  visitInstruction(I);
2719  return;
2720  }
2721 
2722  assert(!isa<IntrinsicInst>(&I) && "intrinsics are handled elsewhere");
2723 
2724  // We are going to insert code that relies on the fact that the callee
2725  // will become a non-readonly function after it is instrumented by us. To
2726  // prevent this code from being optimized out, mark that function
2727  // non-readonly in advance.
2728  if (Function *Func = Call->getCalledFunction()) {
2729  // Clear out readonly/readnone attributes.
2730  AttrBuilder B;
2731  B.addAttribute(Attribute::ReadOnly)
2732  .addAttribute(Attribute::ReadNone);
2734  }
2735 
2737  }
2738  IRBuilder<> IRB(&I);
2739 
2740  unsigned ArgOffset = 0;
2741  DEBUG(dbgs() << " CallSite: " << I << "\n");
2742  for (CallSite::arg_iterator ArgIt = CS.arg_begin(), End = CS.arg_end();
2743  ArgIt != End; ++ArgIt) {
2744  Value *A = *ArgIt;
2745  unsigned i = ArgIt - CS.arg_begin();
2746  if (!A->getType()->isSized()) {
2747  DEBUG(dbgs() << "Arg " << i << " is not sized: " << I << "\n");
2748  continue;
2749  }
2750  unsigned Size = 0;
2751  Value *Store = nullptr;
2752  // Compute the Shadow for arg even if it is ByVal, because
2753  // in that case getShadow() will copy the actual arg shadow to
2754  // __msan_param_tls.
2755  Value *ArgShadow = getShadow(A);
2756  Value *ArgShadowBase = getShadowPtrForArgument(A, IRB, ArgOffset);
2757  DEBUG(dbgs() << " Arg#" << i << ": " << *A <<
2758  " Shadow: " << *ArgShadow << "\n");
2759  bool ArgIsInitialized = false;
2760  const DataLayout &DL = F.getParent()->getDataLayout();
2761  if (CS.paramHasAttr(i, Attribute::ByVal)) {
2762  assert(A->getType()->isPointerTy() &&
2763  "ByVal argument is not a pointer!");
2764  Size = DL.getTypeAllocSize(A->getType()->getPointerElementType());
2765  if (ArgOffset + Size > kParamTLSSize) break;
2766  unsigned ParamAlignment = CS.getParamAlignment(i);
2767  unsigned Alignment = std::min(ParamAlignment, kShadowTLSAlignment);
2768  Value *AShadowPtr =
2769  getShadowOriginPtr(A, IRB, IRB.getInt8Ty(), Alignment).first;
2770 
2771  Store = IRB.CreateMemCpy(ArgShadowBase, AShadowPtr, Size, Alignment);
2772  } else {
2773  Size = DL.getTypeAllocSize(A->getType());
2774  if (ArgOffset + Size > kParamTLSSize) break;
2775  Store = IRB.CreateAlignedStore(ArgShadow, ArgShadowBase,
2776  kShadowTLSAlignment);
2777  Constant *Cst = dyn_cast<Constant>(ArgShadow);
2778  if (Cst && Cst->isNullValue()) ArgIsInitialized = true;
2779  }
2780  if (MS.TrackOrigins && !ArgIsInitialized)
2781  IRB.CreateStore(getOrigin(A),
2782  getOriginPtrForArgument(A, IRB, ArgOffset));
2783  (void)Store;
2784  assert(Size != 0 && Store != nullptr);
2785  DEBUG(dbgs() << " Param:" << *Store << "\n");
2786  ArgOffset += alignTo(Size, 8);
2787  }
2788  DEBUG(dbgs() << " done with call args\n");
2789 
2790  FunctionType *FT =
2791  cast<FunctionType>(CS.getCalledValue()->getType()->getContainedType(0));
2792  if (FT->isVarArg()) {
2793  VAHelper->visitCallSite(CS, IRB);
2794  }
2795 
2796  // Now, get the shadow for the RetVal.
2797  if (!I.getType()->isSized()) return;
2798  // Don't emit the epilogue for musttail call returns.
2799  if (CS.isCall() && cast<CallInst>(&I)->isMustTailCall()) return;
2800  IRBuilder<> IRBBefore(&I);
2801  // Until we have full dynamic coverage, make sure the retval shadow is 0.
2802  Value *Base = getShadowPtrForRetval(&I, IRBBefore);
2803  IRBBefore.CreateAlignedStore(getCleanShadow(&I), Base, kShadowTLSAlignment);
2804  BasicBlock::iterator NextInsn;
2805  if (CS.isCall()) {
2806  NextInsn = ++I.getIterator();
2807  assert(NextInsn != I.getParent()->end());
2808  } else {
2809  BasicBlock *NormalDest = cast<InvokeInst>(&I)->getNormalDest();
2810  if (!NormalDest->getSinglePredecessor()) {
2811  // FIXME: this case is tricky, so we are just conservative here.
2812  // Perhaps we need to split the edge between this BB and NormalDest,
2813  // but a naive attempt to use SplitEdge leads to a crash.
2814  setShadow(&I, getCleanShadow(&I));
2815  setOrigin(&I, getCleanOrigin());
2816  return;
2817  }
2818  // FIXME: NextInsn is likely in a basic block that has not been visited yet.
2819  // Anything inserted there will be instrumented by MSan later!
2820  NextInsn = NormalDest->getFirstInsertionPt();
2821  assert(NextInsn != NormalDest->end() &&
2822  "Could not find insertion point for retval shadow load");
2823  }
2824  IRBuilder<> IRBAfter(&*NextInsn);
2825  Value *RetvalShadow =
2826  IRBAfter.CreateAlignedLoad(getShadowPtrForRetval(&I, IRBAfter),
2827  kShadowTLSAlignment, "_msret");
2828  setShadow(&I, RetvalShadow);
2829  if (MS.TrackOrigins)
2830  setOrigin(&I, IRBAfter.CreateLoad(getOriginPtrForRetval(IRBAfter)));
2831  }
2832 
2833  bool isAMustTailRetVal(Value *RetVal) {
2834  if (auto *I = dyn_cast<BitCastInst>(RetVal)) {
2835  RetVal = I->getOperand(0);
2836  }
2837  if (auto *I = dyn_cast<CallInst>(RetVal)) {
2838  return I->isMustTailCall();
2839  }
2840  return false;
2841  }
2842 
2843  void visitReturnInst(ReturnInst &I) {
2844  IRBuilder<> IRB(&I);
2845  Value *RetVal = I.getReturnValue();
2846  if (!RetVal) return;
2847  // Don't emit the epilogue for musttail call returns.
2848  if (isAMustTailRetVal(RetVal)) return;
2849  Value *ShadowPtr = getShadowPtrForRetval(RetVal, IRB);
2850  if (CheckReturnValue) {
2851  insertShadowCheck(RetVal, &I);
2852  Value *Shadow = getCleanShadow(RetVal);
2853  IRB.CreateAlignedStore(Shadow, ShadowPtr, kShadowTLSAlignment);
2854  } else {
2855  Value *Shadow = getShadow(RetVal);
2856  IRB.CreateAlignedStore(Shadow, ShadowPtr, kShadowTLSAlignment);
2857  if (MS.TrackOrigins)
2858  IRB.CreateStore(getOrigin(RetVal), getOriginPtrForRetval(IRB));
2859  }
2860  }
2861 
2862  void visitPHINode(PHINode &I) {
2863  IRBuilder<> IRB(&I);
2864  if (!PropagateShadow) {
2865  setShadow(&I, getCleanShadow(&I));
2866  setOrigin(&I, getCleanOrigin());
2867  return;
2868  }
2869 
2870  ShadowPHINodes.push_back(&I);
2871  setShadow(&I, IRB.CreatePHI(getShadowTy(&I), I.getNumIncomingValues(),
2872  "_msphi_s"));
2873  if (MS.TrackOrigins)
2874  setOrigin(&I, IRB.CreatePHI(MS.OriginTy, I.getNumIncomingValues(),
2875  "_msphi_o"));
2876  }
2877 
2878  void visitAllocaInst(AllocaInst &I) {
2879  setShadow(&I, getCleanShadow(&I));
2880  setOrigin(&I, getCleanOrigin());
2881  IRBuilder<> IRB(I.getNextNode());
2882  const DataLayout &DL = F.getParent()->getDataLayout();
2883  uint64_t TypeSize = DL.getTypeAllocSize(I.getAllocatedType());
2884  Value *Len = ConstantInt::get(MS.IntptrTy, TypeSize);
2885  if (I.isArrayAllocation())
2886  Len = IRB.CreateMul(Len, I.getArraySize());
2887  if (PoisonStack && ClPoisonStackWithCall) {
2888  IRB.CreateCall(MS.MsanPoisonStackFn,
2889  {IRB.CreatePointerCast(&I, IRB.getInt8PtrTy()), Len});
2890  } else {
2891  Value *ShadowBase =
2892  getShadowOriginPtr(&I, IRB, IRB.getInt8Ty(), I.getAlignment()).first;
2893 
2894  Value *PoisonValue = IRB.getInt8(PoisonStack ? ClPoisonStackPattern : 0);
2895  IRB.CreateMemSet(ShadowBase, PoisonValue, Len, I.getAlignment());
2896  }
2897 
2898  if (PoisonStack && MS.TrackOrigins) {
2899  SmallString<2048> StackDescriptionStorage;
2900  raw_svector_ostream StackDescription(StackDescriptionStorage);
2901  // We create a string with a description of the stack allocation and
2902  // pass it into __msan_set_alloca_origin.
2903  // It will be printed by the run-time if stack-originated UMR is found.
2904  // The first 4 bytes of the string are set to '----' and will be replaced
2905  // by __msan_va_arg_overflow_size_tls at the first call.
2906  StackDescription << "----" << I.getName() << "@" << F.getName();
2907  Value *Descr =
2909  StackDescription.str());
2910 
2911  IRB.CreateCall(MS.MsanSetAllocaOrigin4Fn,
2912  {IRB.CreatePointerCast(&I, IRB.getInt8PtrTy()), Len,
2913  IRB.CreatePointerCast(Descr, IRB.getInt8PtrTy()),
2914  IRB.CreatePointerCast(&F, MS.IntptrTy)});
2915  }
2916  }
2917 
2918  void visitSelectInst(SelectInst& I) {
2919  IRBuilder<> IRB(&I);
2920  // a = select b, c, d
2921  Value *B = I.getCondition();
2922  Value *C = I.getTrueValue();
2923  Value *D = I.getFalseValue();
2924  Value *Sb = getShadow(B);
2925  Value *Sc = getShadow(C);
2926  Value *Sd = getShadow(D);
2927 
2928  // Result shadow if condition shadow is 0.
2929  Value *Sa0 = IRB.CreateSelect(B, Sc, Sd);
2930  Value *Sa1;
2931  if (I.getType()->isAggregateType()) {
2932  // To avoid "sign extending" i1 to an arbitrary aggregate type, we just do
2933  // an extra "select". This results in much more compact IR.
2934  // Sa = select Sb, poisoned, (select b, Sc, Sd)
2935  Sa1 = getPoisonedShadow(getShadowTy(I.getType()));
2936  } else {
2937  // Sa = select Sb, [ (c^d) | Sc | Sd ], [ b ? Sc : Sd ]
2938  // If Sb (condition is poisoned), look for bits in c and d that are equal
2939  // and both unpoisoned.
2940  // If !Sb (condition is unpoisoned), simply pick one of Sc and Sd.
2941 
2942  // Cast arguments to shadow-compatible type.
2943  C = CreateAppToShadowCast(IRB, C);
2944  D = CreateAppToShadowCast(IRB, D);
2945 
2946  // Result shadow if condition shadow is 1.
2947  Sa1 = IRB.CreateOr(IRB.CreateXor(C, D), IRB.CreateOr(Sc, Sd));
2948  }
2949  Value *Sa = IRB.CreateSelect(Sb, Sa1, Sa0, "_msprop_select");
2950  setShadow(&I, Sa);
2951  if (MS.TrackOrigins) {
2952  // Origins are always i32, so any vector conditions must be flattened.
2953  // FIXME: consider tracking vector origins for app vectors?
2954  if (B->getType()->isVectorTy()) {
2955  Type *FlatTy = getShadowTyNoVec(B->getType());
2956  B = IRB.CreateICmpNE(IRB.CreateBitCast(B, FlatTy),
2957  ConstantInt::getNullValue(FlatTy));
2958  Sb = IRB.CreateICmpNE(IRB.CreateBitCast(Sb, FlatTy),
2959  ConstantInt::getNullValue(FlatTy));
2960  }
2961  // a = select b, c, d
2962  // Oa = Sb ? Ob : (b ? Oc : Od)
2963  setOrigin(
2964  &I, IRB.CreateSelect(Sb, getOrigin(I.getCondition()),
2965  IRB.CreateSelect(B, getOrigin(I.getTrueValue()),
2966  getOrigin(I.getFalseValue()))));
2967  }
2968  }
2969 
2970  void visitLandingPadInst(LandingPadInst &I) {
2971  // Do nothing.
2972  // See https://github.com/google/sanitizers/issues/504
2973  setShadow(&I, getCleanShadow(&I));
2974  setOrigin(&I, getCleanOrigin());
2975  }
2976 
2977  void visitCatchSwitchInst(CatchSwitchInst &I) {
2978  setShadow(&I, getCleanShadow(&I));
2979  setOrigin(&I, getCleanOrigin());
2980  }
2981 
2982  void visitFuncletPadInst(FuncletPadInst &I) {
2983  setShadow(&I, getCleanShadow(&I));
2984  setOrigin(&I, getCleanOrigin());
2985  }
2986 
2987  void visitGetElementPtrInst(GetElementPtrInst &I) {
2988  handleShadowOr(I);
2989  }
2990 
2991  void visitExtractValueInst(ExtractValueInst &I) {
2992  IRBuilder<> IRB(&I);
2993  Value *Agg = I.getAggregateOperand();
2994  DEBUG(dbgs() << "ExtractValue: " << I << "\n");
2995  Value *AggShadow = getShadow(Agg);
2996  DEBUG(dbgs() << " AggShadow: " << *AggShadow << "\n");
2997  Value *ResShadow = IRB.CreateExtractValue(AggShadow, I.getIndices());
2998  DEBUG(dbgs() << " ResShadow: " << *ResShadow << "\n");
2999  setShadow(&I, ResShadow);
3000  setOriginForNaryOp(I);
3001  }
3002 
3003  void visitInsertValueInst(InsertValueInst &I) {
3004  IRBuilder<> IRB(&I);
3005  DEBUG(dbgs() << "InsertValue: " << I << "\n");
3006  Value *AggShadow = getShadow(I.getAggregateOperand());
3007  Value *InsShadow = getShadow(I.getInsertedValueOperand());
3008  DEBUG(dbgs() << " AggShadow: " << *AggShadow << "\n");
3009  DEBUG(dbgs() << " InsShadow: " << *InsShadow << "\n");
3010  Value *Res = IRB.CreateInsertValue(AggShadow, InsShadow, I.getIndices());
3011  DEBUG(dbgs() << " Res: " << *Res << "\n");
3012  setShadow(&I, Res);
3013  setOriginForNaryOp(I);
3014  }
3015 
3016  void dumpInst(Instruction &I) {
3017  if (CallInst *CI = dyn_cast<CallInst>(&I)) {
3018  errs() << "ZZZ call " << CI->getCalledFunction()->getName() << "\n";
3019  } else {
3020  errs() << "ZZZ " << I.getOpcodeName() << "\n";
3021  }
3022  errs() << "QQQ " << I << "\n";
3023  }
3024 
3025  void visitResumeInst(ResumeInst &I) {
3026  DEBUG(dbgs() << "Resume: " << I << "\n");
3027  // Nothing to do here.
3028  }
3029 
3030  void visitCleanupReturnInst(CleanupReturnInst &CRI) {
3031  DEBUG(dbgs() << "CleanupReturn: " << CRI << "\n");
3032  // Nothing to do here.
3033  }
3034 
3035  void visitCatchReturnInst(CatchReturnInst &CRI) {
3036  DEBUG(dbgs() << "CatchReturn: " << CRI << "\n");
3037  // Nothing to do here.
3038  }
3039 
3040  void visitInstruction(Instruction &I) {
3041  // Everything else: stop propagating and check for poisoned shadow.
3043  dumpInst(I);
3044  DEBUG(dbgs() << "DEFAULT: " << I << "\n");
3045  for (size_t i = 0, n = I.getNumOperands(); i < n; i++) {
3046  Value *Operand = I.getOperand(i);
3047  if (Operand->getType()->isSized())
3048  insertShadowCheck(Operand, &I);
3049  }
3050  setShadow(&I, getCleanShadow(&I));
3051  setOrigin(&I, getCleanOrigin());
3052  }
3053 };
3054 
3055 /// \brief AMD64-specific implementation of VarArgHelper.
3056 struct VarArgAMD64Helper : public VarArgHelper {
3057  // An unfortunate workaround for asymmetric lowering of va_arg stuff.
3058  // See a comment in visitCallSite for more details.
3059  static const unsigned AMD64GpEndOffset = 48; // AMD64 ABI Draft 0.99.6 p3.5.7
3060  static const unsigned AMD64FpEndOffset = 176;
3061 
3062  Function &F;
3063  MemorySanitizer &MS;
3064  MemorySanitizerVisitor &MSV;
3065  Value *VAArgTLSCopy = nullptr;
3066  Value *VAArgOverflowSize = nullptr;
3067 
3068  SmallVector<CallInst*, 16> VAStartInstrumentationList;
3069 
3070  enum ArgKind { AK_GeneralPurpose, AK_FloatingPoint, AK_Memory };
3071 
3072  VarArgAMD64Helper(Function &F, MemorySanitizer &MS,
3073  MemorySanitizerVisitor &MSV) : F(F), MS(MS), MSV(MSV) {}
3074 
3075  ArgKind classifyArgument(Value* arg) {
3076  // A very rough approximation of X86_64 argument classification rules.
3077  Type *T = arg->getType();
3078  if (T->isFPOrFPVectorTy() || T->isX86_MMXTy())
3079  return AK_FloatingPoint;
3080  if (T->isIntegerTy() && T->getPrimitiveSizeInBits() <= 64)
3081  return AK_GeneralPurpose;
3082  if (T->isPointerTy())
3083  return AK_GeneralPurpose;
3084  return AK_Memory;
3085  }
3086 
3087  // For VarArg functions, store the argument shadow in an ABI-specific format
3088  // that corresponds to va_list layout.
3089  // We do this because Clang lowers va_arg in the frontend, and this pass
3090  // only sees the low level code that deals with va_list internals.
3091  // A much easier alternative (provided that Clang emits va_arg instructions)
3092  // would have been to associate each live instance of va_list with a copy of
3093  // MSanParamTLS, and extract shadow on va_arg() call in the argument list
3094  // order.
3095  void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {
3096  unsigned GpOffset = 0;
3097  unsigned FpOffset = AMD64GpEndOffset;
3098  unsigned OverflowOffset = AMD64FpEndOffset;
3099  const DataLayout &DL = F.getParent()->getDataLayout();
3100  for (CallSite::arg_iterator ArgIt = CS.arg_begin(), End = CS.arg_end();
3101  ArgIt != End; ++ArgIt) {
3102  Value *A = *ArgIt;
3103  unsigned ArgNo = CS.getArgumentNo(ArgIt);
3104  bool IsFixed = ArgNo < CS.getFunctionType()->getNumParams();
3105  bool IsByVal = CS.paramHasAttr(ArgNo, Attribute::ByVal);
3106  if (IsByVal) {
3107  // ByVal arguments always go to the overflow area.
3108  // Fixed arguments passed through the overflow area will be stepped
3109  // over by va_start, so don't count them towards the offset.
3110  if (IsFixed)
3111  continue;
3112  assert(A->getType()->isPointerTy());
3113  Type *RealTy = A->getType()->getPointerElementType();
3114  uint64_t ArgSize = DL.getTypeAllocSize(RealTy);
3115  Value *ShadowBase =
3116  getShadowPtrForVAArgument(RealTy, IRB, OverflowOffset);
3117  OverflowOffset += alignTo(ArgSize, 8);
3118  Value *ShadowPtr, *OriginPtr;
3119  std::tie(ShadowPtr, OriginPtr) = MSV.getShadowOriginPtr(
3120  A, IRB, IRB.getInt8Ty(), kShadowTLSAlignment);
3121 
3122  IRB.CreateMemCpy(ShadowBase, ShadowPtr, ArgSize, kShadowTLSAlignment);
3123  } else {
3124  ArgKind AK = classifyArgument(A);
3125  if (AK == AK_GeneralPurpose && GpOffset >= AMD64GpEndOffset)
3126  AK = AK_Memory;
3127  if (AK == AK_FloatingPoint && FpOffset >= AMD64FpEndOffset)
3128  AK = AK_Memory;
3129  Value *ShadowBase;
3130  switch (AK) {
3131  case AK_GeneralPurpose:
3132  ShadowBase = getShadowPtrForVAArgument(A->getType(), IRB, GpOffset);
3133  GpOffset += 8;
3134  break;
3135  case AK_FloatingPoint:
3136  ShadowBase = getShadowPtrForVAArgument(A->getType(), IRB, FpOffset);
3137  FpOffset += 16;
3138  break;
3139  case AK_Memory:
3140  if (IsFixed)
3141  continue;
3142  uint64_t ArgSize = DL.getTypeAllocSize(A->getType());
3143  ShadowBase =
3144  getShadowPtrForVAArgument(A->getType(), IRB, OverflowOffset);
3145  OverflowOffset += alignTo(ArgSize, 8);
3146  }
3147  // Take fixed arguments into account for GpOffset and FpOffset,
3148  // but don't actually store shadows for them.
3149  if (IsFixed)
3150  continue;
3151  IRB.CreateAlignedStore(MSV.getShadow(A), ShadowBase,
3153  }
3154  }
3155  Constant *OverflowSize =
3156  ConstantInt::get(IRB.getInt64Ty(), OverflowOffset - AMD64FpEndOffset);
3157  IRB.CreateStore(OverflowSize, MS.VAArgOverflowSizeTLS);
3158  }
3159 
3160  /// \brief Compute the shadow address for a given va_arg.
3161  Value *getShadowPtrForVAArgument(Type *Ty, IRBuilder<> &IRB,
3162  int ArgOffset) {
3163  Value *Base = IRB.CreatePointerCast(MS.VAArgTLS, MS.IntptrTy);
3164  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
3165  return IRB.CreateIntToPtr(Base, PointerType::get(MSV.getShadowTy(Ty), 0),
3166  "_msarg");
3167  }
3168 
3169  void unpoisonVAListTagForInst(IntrinsicInst &I) {
3170  IRBuilder<> IRB(&I);
3171  Value *VAListTag = I.getArgOperand(0);
3172  Value *ShadowPtr, *OriginPtr;
3173  unsigned Alignment = 8;
3174  std::tie(ShadowPtr, OriginPtr) =
3175  MSV.getShadowOriginPtr(VAListTag, IRB, IRB.getInt8Ty(), Alignment);
3176 
3177  // Unpoison the whole __va_list_tag.
3178  // FIXME: magic ABI constants.
3179  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
3180  /* size */ 24, Alignment, false);
3181  // We shouldn't need to zero out the origins, as they're only checked for
3182  // nonzero shadow.
3183  }
3184 
3185  void visitVAStartInst(VAStartInst &I) override {
3187  return;
3188  VAStartInstrumentationList.push_back(&I);
3189  unpoisonVAListTagForInst(I);
3190  }
3191 
3192  void visitVACopyInst(VACopyInst &I) override {
3193  if (F.getCallingConv() == CallingConv::Win64) return;
3194  unpoisonVAListTagForInst(I);
3195  }
3196 
3197  void finalizeInstrumentation() override {
3198  assert(!VAArgOverflowSize && !VAArgTLSCopy &&
3199  "finalizeInstrumentation called twice");
3200  if (!VAStartInstrumentationList.empty()) {
3201  // If there is a va_start in this function, make a backup copy of
3202  // va_arg_tls somewhere in the function entry block.
3204  VAArgOverflowSize = IRB.CreateLoad(MS.VAArgOverflowSizeTLS);
3205  Value *CopySize =
3206  IRB.CreateAdd(ConstantInt::get(MS.IntptrTy, AMD64FpEndOffset),
3207  VAArgOverflowSize);
3208  VAArgTLSCopy = IRB.CreateAlloca(Type::getInt8Ty(*MS.C), CopySize);
3209  IRB.CreateMemCpy(VAArgTLSCopy, MS.VAArgTLS, CopySize, 8);
3210  }
3211 
3212  // Instrument va_start.
3213  // Copy va_list shadow from the backup copy of the TLS contents.
3214  for (size_t i = 0, n = VAStartInstrumentationList.size(); i < n; i++) {
3215  CallInst *OrigInst = VAStartInstrumentationList[i];
3216  IRBuilder<> IRB(OrigInst->getNextNode());
3217  Value *VAListTag = OrigInst->getArgOperand(0);
3218 
3219  Value *RegSaveAreaPtrPtr = IRB.CreateIntToPtr(
3220  IRB.CreateAdd(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
3221  ConstantInt::get(MS.IntptrTy, 16)),
3222  Type::getInt64PtrTy(*MS.C));
3223  Value *RegSaveAreaPtr = IRB.CreateLoad(RegSaveAreaPtrPtr);
3224  Value *RegSaveAreaShadowPtr, *RegSaveAreaOriginPtr;
3225  unsigned Alignment = 16;
3226  std::tie(RegSaveAreaShadowPtr, RegSaveAreaOriginPtr) =
3227  MSV.getShadowOriginPtr(RegSaveAreaPtr, IRB, IRB.getInt8Ty(),
3228  Alignment);
3229  IRB.CreateMemCpy(RegSaveAreaShadowPtr, VAArgTLSCopy, AMD64FpEndOffset,
3230  Alignment);
3231  Value *OverflowArgAreaPtrPtr = IRB.CreateIntToPtr(
3232  IRB.CreateAdd(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
3233  ConstantInt::get(MS.IntptrTy, 8)),
3234  Type::getInt64PtrTy(*MS.C));
3235  Value *OverflowArgAreaPtr = IRB.CreateLoad(OverflowArgAreaPtrPtr);
3236  Value *OverflowArgAreaShadowPtr, *OverflowArgAreaOriginPtr;
3237  std::tie(OverflowArgAreaShadowPtr, OverflowArgAreaOriginPtr) =
3238  MSV.getShadowOriginPtr(OverflowArgAreaPtr, IRB, IRB.getInt8Ty(),
3239  Alignment);
3240  Value *SrcPtr = IRB.CreateConstGEP1_32(IRB.getInt8Ty(), VAArgTLSCopy,
3241  AMD64FpEndOffset);
3242  IRB.CreateMemCpy(OverflowArgAreaShadowPtr, SrcPtr, VAArgOverflowSize,
3243  Alignment);
3244  }
3245  }
3246 };
3247 
3248 /// \brief MIPS64-specific implementation of VarArgHelper.
3249 struct VarArgMIPS64Helper : public VarArgHelper {
3250  Function &F;
3251  MemorySanitizer &MS;
3252  MemorySanitizerVisitor &MSV;
3253  Value *VAArgTLSCopy = nullptr;
3254  Value *VAArgSize = nullptr;
3255 
3256  SmallVector<CallInst*, 16> VAStartInstrumentationList;
3257 
3258  VarArgMIPS64Helper(Function &F, MemorySanitizer &MS,
3259  MemorySanitizerVisitor &MSV) : F(F), MS(MS), MSV(MSV) {}
3260 
3261  void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {
3262  unsigned VAArgOffset = 0;
3263  const DataLayout &DL = F.getParent()->getDataLayout();
3264  for (CallSite::arg_iterator ArgIt = CS.arg_begin() +
3265  CS.getFunctionType()->getNumParams(), End = CS.arg_end();
3266  ArgIt != End; ++ArgIt) {
3267  Triple TargetTriple(F.getParent()->getTargetTriple());
3268  Value *A = *ArgIt;
3269  Value *Base;
3270  uint64_t ArgSize = DL.getTypeAllocSize(A->getType());
3271  if (TargetTriple.getArch() == Triple::mips64) {
3272  // Adjusting the shadow for argument with size < 8 to match the placement
3273  // of bits in big endian system
3274  if (ArgSize < 8)
3275  VAArgOffset += (8 - ArgSize);
3276  }
3277  Base = getShadowPtrForVAArgument(A->getType(), IRB, VAArgOffset);
3278  VAArgOffset += ArgSize;
3279  VAArgOffset = alignTo(VAArgOffset, 8);
3280  IRB.CreateAlignedStore(MSV.getShadow(A), Base, kShadowTLSAlignment);
3281  }
3282 
3283  Constant *TotalVAArgSize = ConstantInt::get(IRB.getInt64Ty(), VAArgOffset);
3284  // Here using VAArgOverflowSizeTLS as VAArgSizeTLS to avoid creation of
3285  // a new class member i.e. it is the total size of all VarArgs.
3286  IRB.CreateStore(TotalVAArgSize, MS.VAArgOverflowSizeTLS);
3287  }
3288 
3289  /// \brief Compute the shadow address for a given va_arg.
3290  Value *getShadowPtrForVAArgument(Type *Ty, IRBuilder<> &IRB,
3291  int ArgOffset) {
3292  Value *Base = IRB.CreatePointerCast(MS.VAArgTLS, MS.IntptrTy);
3293  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
3294  return IRB.CreateIntToPtr(Base, PointerType::get(MSV.getShadowTy(Ty), 0),
3295  "_msarg");
3296  }
3297 
3298  void visitVAStartInst(VAStartInst &I) override {
3299  IRBuilder<> IRB(&I);
3300  VAStartInstrumentationList.push_back(&I);
3301  Value *VAListTag = I.getArgOperand(0);
3302  Value *ShadowPtr, *OriginPtr;
3303  unsigned Alignment = 8;
3304  std::tie(ShadowPtr, OriginPtr) =
3305  MSV.getShadowOriginPtr(VAListTag, IRB, IRB.getInt8Ty(), Alignment);
3306  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
3307  /* size */ 8, Alignment, false);
3308  }
3309 
3310  void visitVACopyInst(VACopyInst &I) override {
3311  IRBuilder<> IRB(&I);
3312  VAStartInstrumentationList.push_back(&I);
3313  Value *VAListTag = I.getArgOperand(0);
3314  Value *ShadowPtr, *OriginPtr;
3315  unsigned Alignment = 8;
3316  std::tie(ShadowPtr, OriginPtr) =
3317  MSV.getShadowOriginPtr(VAListTag, IRB, IRB.getInt8Ty(), Alignment);
3318  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
3319  /* size */ 8, Alignment, false);
3320  }
3321 
3322  void finalizeInstrumentation() override {
3323  assert(!VAArgSize && !VAArgTLSCopy &&
3324  "finalizeInstrumentation called twice");
3326  VAArgSize = IRB.CreateLoad(MS.VAArgOverflowSizeTLS);
3327  Value *CopySize = IRB.CreateAdd(ConstantInt::get(MS.IntptrTy, 0),
3328  VAArgSize);
3329 
3330  if (!VAStartInstrumentationList.empty()) {
3331  // If there is a va_start in this function, make a backup copy of
3332  // va_arg_tls somewhere in the function entry block.
3333  VAArgTLSCopy = IRB.CreateAlloca(Type::getInt8Ty(*MS.C), CopySize);
3334  IRB.CreateMemCpy(VAArgTLSCopy, MS.VAArgTLS, CopySize, 8);
3335  }
3336 
3337  // Instrument va_start.
3338  // Copy va_list shadow from the backup copy of the TLS contents.
3339  for (size_t i = 0, n = VAStartInstrumentationList.size(); i < n; i++) {
3340  CallInst *OrigInst = VAStartInstrumentationList[i];
3341  IRBuilder<> IRB(OrigInst->getNextNode());
3342  Value *VAListTag = OrigInst->getArgOperand(0);
3343  Value *RegSaveAreaPtrPtr =
3344  IRB.CreateIntToPtr(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
3345  Type::getInt64PtrTy(*MS.C));
3346  Value *RegSaveAreaPtr = IRB.CreateLoad(RegSaveAreaPtrPtr);
3347  Value *RegSaveAreaShadowPtr, *RegSaveAreaOriginPtr;
3348  unsigned Alignment = 8;
3349  std::tie(RegSaveAreaShadowPtr, RegSaveAreaOriginPtr) =
3350  MSV.getShadowOriginPtr(RegSaveAreaPtr, IRB, IRB.getInt8Ty(),
3351  Alignment);
3352  IRB.CreateMemCpy(RegSaveAreaShadowPtr, VAArgTLSCopy, CopySize, Alignment);
3353  }
3354  }
3355 };
3356 
3357 /// \brief AArch64-specific implementation of VarArgHelper.
3358 struct VarArgAArch64Helper : public VarArgHelper {
3359  static const unsigned kAArch64GrArgSize = 64;
3360  static const unsigned kAArch64VrArgSize = 128;
3361 
3362  static const unsigned AArch64GrBegOffset = 0;
3363  static const unsigned AArch64GrEndOffset = kAArch64GrArgSize;
3364  // Make VR space aligned to 16 bytes.
3365  static const unsigned AArch64VrBegOffset = AArch64GrEndOffset;
3366  static const unsigned AArch64VrEndOffset = AArch64VrBegOffset
3367  + kAArch64VrArgSize;
3368  static const unsigned AArch64VAEndOffset = AArch64VrEndOffset;
3369 
3370  Function &F;
3371  MemorySanitizer &MS;
3372  MemorySanitizerVisitor &MSV;
3373  Value *VAArgTLSCopy = nullptr;
3374  Value *VAArgOverflowSize = nullptr;
3375 
3376  SmallVector<CallInst*, 16> VAStartInstrumentationList;
3377 
3378  enum ArgKind { AK_GeneralPurpose, AK_FloatingPoint, AK_Memory };
3379 
3380  VarArgAArch64Helper(Function &F, MemorySanitizer &MS,
3381  MemorySanitizerVisitor &MSV) : F(F), MS(MS), MSV(MSV) {}
3382 
3383  ArgKind classifyArgument(Value* arg) {
3384  Type *T = arg->getType();
3385  if (T->isFPOrFPVectorTy())
3386  return AK_FloatingPoint;
3387  if ((T->isIntegerTy() && T->getPrimitiveSizeInBits() <= 64)
3388  || (T->isPointerTy()))
3389  return AK_GeneralPurpose;
3390  return AK_Memory;
3391  }
3392 
3393  // The instrumentation stores the argument shadow in a non ABI-specific
3394  // format because it does not know which argument is named (since Clang,
3395  // like x86_64 case, lowers the va_args in the frontend and this pass only
3396  // sees the low level code that deals with va_list internals).
3397  // The first seven GR registers are saved in the first 56 bytes of the
3398  // va_arg tls arra, followers by the first 8 FP/SIMD registers, and then
3399  // the remaining arguments.
3400  // Using constant offset within the va_arg TLS array allows fast copy
3401  // in the finalize instrumentation.
3402  void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {
3403  unsigned GrOffset = AArch64GrBegOffset;
3404  unsigned VrOffset = AArch64VrBegOffset;
3405  unsigned OverflowOffset = AArch64VAEndOffset;
3406 
3407  const DataLayout &DL = F.getParent()->getDataLayout();
3408  for (CallSite::arg_iterator ArgIt = CS.arg_begin(), End = CS.arg_end();
3409  ArgIt != End; ++ArgIt) {
3410  Value *A = *ArgIt;
3411  unsigned ArgNo = CS.getArgumentNo(ArgIt);
3412  bool IsFixed = ArgNo < CS.getFunctionType()->getNumParams();
3413  ArgKind AK = classifyArgument(A);
3414  if (AK == AK_GeneralPurpose && GrOffset >= AArch64GrEndOffset)
3415  AK = AK_Memory;
3416  if (AK == AK_FloatingPoint && VrOffset >= AArch64VrEndOffset)
3417  AK = AK_Memory;
3418  Value *Base;
3419  switch (AK) {
3420  case AK_GeneralPurpose:
3421  Base = getShadowPtrForVAArgument(A->getType(), IRB, GrOffset);
3422  GrOffset += 8;
3423  break;
3424  case AK_FloatingPoint:
3425  Base = getShadowPtrForVAArgument(A->getType(), IRB, VrOffset);
3426  VrOffset += 16;
3427  break;
3428  case AK_Memory:
3429  // Don't count fixed arguments in the overflow area - va_start will
3430  // skip right over them.
3431  if (IsFixed)
3432  continue;
3433  uint64_t ArgSize = DL.getTypeAllocSize(A->getType());
3434  Base = getShadowPtrForVAArgument(A->getType(), IRB, OverflowOffset);
3435  OverflowOffset += alignTo(ArgSize, 8);
3436  break;
3437  }
3438  // Count Gp/Vr fixed arguments to their respective offsets, but don't
3439  // bother to actually store a shadow.
3440  if (IsFixed)
3441  continue;
3442  IRB.CreateAlignedStore(MSV.getShadow(A), Base, kShadowTLSAlignment);
3443  }
3444  Constant *OverflowSize =
3445  ConstantInt::get(IRB.getInt64Ty(), OverflowOffset - AArch64VAEndOffset);
3446  IRB.CreateStore(OverflowSize, MS.VAArgOverflowSizeTLS);
3447  }
3448 
3449  /// Compute the shadow address for a given va_arg.
3450  Value *getShadowPtrForVAArgument(Type *Ty, IRBuilder<> &IRB,
3451  int ArgOffset) {
3452  Value *Base = IRB.CreatePointerCast(MS.VAArgTLS, MS.IntptrTy);
3453  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
3454  return IRB.CreateIntToPtr(Base, PointerType::get(MSV.getShadowTy(Ty), 0),
3455  "_msarg");
3456  }
3457 
3458  void visitVAStartInst(VAStartInst &I) override {
3459  IRBuilder<> IRB(&I);
3460  VAStartInstrumentationList.push_back(&I);
3461  Value *VAListTag = I.getArgOperand(0);
3462  Value *ShadowPtr, *OriginPtr;
3463  unsigned Alignment = 8;
3464  std::tie(ShadowPtr, OriginPtr) =
3465  MSV.getShadowOriginPtr(VAListTag, IRB, IRB.getInt8Ty(), Alignment);
3466  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
3467  /* size */ 32, Alignment, false);
3468  }
3469 
3470  void visitVACopyInst(VACopyInst &I) override {
3471  IRBuilder<> IRB(&I);
3472  VAStartInstrumentationList.push_back(&I);
3473  Value *VAListTag = I.getArgOperand(0);
3474  Value *ShadowPtr, *OriginPtr;
3475  unsigned Alignment = 8;
3476  std::tie(ShadowPtr, OriginPtr) =
3477  MSV.getShadowOriginPtr(VAListTag, IRB, IRB.getInt8Ty(), Alignment);
3478  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
3479  /* size */ 32, Alignment, false);
3480  }
3481 
3482  // Retrieve a va_list field of 'void*' size.
3483  Value* getVAField64(IRBuilder<> &IRB, Value *VAListTag, int offset) {
3484  Value *SaveAreaPtrPtr =
3485  IRB.CreateIntToPtr(
3486  IRB.CreateAdd(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
3487  ConstantInt::get(MS.IntptrTy, offset)),
3488  Type::getInt64PtrTy(*MS.C));
3489  return IRB.CreateLoad(SaveAreaPtrPtr);
3490  }
3491 
3492  // Retrieve a va_list field of 'int' size.
3493  Value* getVAField32(IRBuilder<> &IRB, Value *VAListTag, int offset) {
3494  Value *SaveAreaPtr =
3495  IRB.CreateIntToPtr(
3496  IRB.CreateAdd(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
3497  ConstantInt::get(MS.IntptrTy, offset)),
3498  Type::getInt32PtrTy(*MS.C));
3499  Value *SaveArea32 = IRB.CreateLoad(SaveAreaPtr);
3500  return IRB.CreateSExt(SaveArea32, MS.IntptrTy);
3501  }
3502 
3503  void finalizeInstrumentation() override {
3504  assert(!VAArgOverflowSize && !VAArgTLSCopy &&
3505  "finalizeInstrumentation called twice");
3506  if (!VAStartInstrumentationList.empty()) {
3507  // If there is a va_start in this function, make a backup copy of
3508  // va_arg_tls somewhere in the function entry block.
3510  VAArgOverflowSize = IRB.CreateLoad(MS.VAArgOverflowSizeTLS);
3511  Value *CopySize =
3512  IRB.CreateAdd(ConstantInt::get(MS.IntptrTy, AArch64VAEndOffset),
3513  VAArgOverflowSize);
3514  VAArgTLSCopy = IRB.CreateAlloca(Type::getInt8Ty(*MS.C), CopySize);
3515  IRB.CreateMemCpy(VAArgTLSCopy, MS.VAArgTLS, CopySize, 8);
3516  }
3517 
3518  Value *GrArgSize = ConstantInt::get(MS.IntptrTy, kAArch64GrArgSize);
3519  Value *VrArgSize = ConstantInt::get(MS.IntptrTy, kAArch64VrArgSize);
3520 
3521  // Instrument va_start, copy va_list shadow from the backup copy of
3522  // the TLS contents.
3523  for (size_t i = 0, n = VAStartInstrumentationList.size(); i < n; i++) {
3524  CallInst *OrigInst = VAStartInstrumentationList[i];
3525  IRBuilder<> IRB(OrigInst->getNextNode());
3526 
3527  Value *VAListTag = OrigInst->getArgOperand(0);
3528 
3529  // The variadic ABI for AArch64 creates two areas to save the incoming
3530  // argument registers (one for 64-bit general register xn-x7 and another
3531  // for 128-bit FP/SIMD vn-v7).
3532  // We need then to propagate the shadow arguments on both regions
3533  // 'va::__gr_top + va::__gr_offs' and 'va::__vr_top + va::__vr_offs'.
3534  // The remaning arguments are saved on shadow for 'va::stack'.
3535  // One caveat is it requires only to propagate the non-named arguments,
3536  // however on the call site instrumentation 'all' the arguments are
3537  // saved. So to copy the shadow values from the va_arg TLS array
3538  // we need to adjust the offset for both GR and VR fields based on
3539  // the __{gr,vr}_offs value (since they are stores based on incoming
3540  // named arguments).
3541 
3542  // Read the stack pointer from the va_list.
3543  Value *StackSaveAreaPtr = getVAField64(IRB, VAListTag, 0);
3544 
3545  // Read both the __gr_top and __gr_off and add them up.
3546  Value *GrTopSaveAreaPtr = getVAField64(IRB, VAListTag, 8);
3547  Value *GrOffSaveArea = getVAField32(IRB, VAListTag, 24);
3548 
3549  Value *GrRegSaveAreaPtr = IRB.CreateAdd(GrTopSaveAreaPtr, GrOffSaveArea);
3550 
3551  // Read both the __vr_top and __vr_off and add them up.
3552  Value *VrTopSaveAreaPtr = getVAField64(IRB, VAListTag, 16);
3553  Value *VrOffSaveArea = getVAField32(IRB, VAListTag, 28);
3554 
3555  Value *VrRegSaveAreaPtr = IRB.CreateAdd(VrTopSaveAreaPtr, VrOffSaveArea);
3556 
3557  // It does not know how many named arguments is being used and, on the
3558  // callsite all the arguments were saved. Since __gr_off is defined as
3559  // '0 - ((8 - named_gr) * 8)', the idea is to just propagate the variadic
3560  // argument by ignoring the bytes of shadow from named arguments.
3561  Value *GrRegSaveAreaShadowPtrOff =
3562  IRB.CreateAdd(GrArgSize, GrOffSaveArea);
3563 
3564  Value *GrRegSaveAreaShadowPtr =
3565  MSV.getShadowOriginPtr(GrRegSaveAreaPtr, IRB, IRB.getInt8Ty(),
3566  /*Alignment*/ 8)
3567  .first;
3568 
3569  Value *GrSrcPtr = IRB.CreateInBoundsGEP(IRB.getInt8Ty(), VAArgTLSCopy,
3570  GrRegSaveAreaShadowPtrOff);
3571  Value *GrCopySize = IRB.CreateSub(GrArgSize, GrRegSaveAreaShadowPtrOff);
3572 
3573  IRB.CreateMemCpy(GrRegSaveAreaShadowPtr, GrSrcPtr, GrCopySize, 8);
3574 
3575  // Again, but for FP/SIMD values.
3576  Value *VrRegSaveAreaShadowPtrOff =
3577  IRB.CreateAdd(VrArgSize, VrOffSaveArea);
3578 
3579  Value *VrRegSaveAreaShadowPtr =
3580  MSV.getShadowOriginPtr(VrRegSaveAreaPtr, IRB, IRB.getInt8Ty(),
3581  /*Alignment*/ 8)
3582  .first;
3583 
3584  Value *VrSrcPtr = IRB.CreateInBoundsGEP(
3585  IRB.getInt8Ty(),
3586  IRB.CreateInBoundsGEP(IRB.getInt8Ty(), VAArgTLSCopy,
3587  IRB.getInt32(AArch64VrBegOffset)),
3588  VrRegSaveAreaShadowPtrOff);
3589  Value *VrCopySize = IRB.CreateSub(VrArgSize, VrRegSaveAreaShadowPtrOff);
3590 
3591  IRB.CreateMemCpy(VrRegSaveAreaShadowPtr, VrSrcPtr, VrCopySize, 8);
3592 
3593  // And finally for remaining arguments.
3594  Value *StackSaveAreaShadowPtr =
3595  MSV.getShadowOriginPtr(StackSaveAreaPtr, IRB, IRB.getInt8Ty(),
3596  /*Alignment*/ 16)
3597  .first;
3598 
3599  Value *StackSrcPtr =
3600  IRB.CreateInBoundsGEP(IRB.getInt8Ty(), VAArgTLSCopy,
3601  IRB.getInt32(AArch64VAEndOffset));
3602 
3603  IRB.CreateMemCpy(StackSaveAreaShadowPtr, StackSrcPtr,
3604  VAArgOverflowSize, 16);
3605  }
3606  }
3607 };
3608 
3609 /// \brief PowerPC64-specific implementation of VarArgHelper.
3610 struct VarArgPowerPC64Helper : public VarArgHelper {
3611  Function &F;
3612  MemorySanitizer &MS;
3613  MemorySanitizerVisitor &MSV;
3614  Value *VAArgTLSCopy = nullptr;
3615  Value *VAArgSize = nullptr;
3616 
3617  SmallVector<CallInst*, 16> VAStartInstrumentationList;
3618 
3619  VarArgPowerPC64Helper(Function &F, MemorySanitizer &MS,
3620  MemorySanitizerVisitor &MSV) : F(F), MS(MS), MSV(MSV) {}
3621 
3622  void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {
3623  // For PowerPC, we need to deal with alignment of stack arguments -
3624  // they are mostly aligned to 8 bytes, but vectors and i128 arrays
3625  // are aligned to 16 bytes, byvals can be aligned to 8 or 16 bytes,
3626  // and QPX vectors are aligned to 32 bytes. For that reason, we
3627  // compute current offset from stack pointer (which is always properly
3628  // aligned), and offset for the first vararg, then subtract them.
3629  unsigned VAArgBase;
3630  Triple TargetTriple(F.getParent()->getTargetTriple());
3631  // Parameter save area starts at 48 bytes from frame pointer for ABIv1,
3632  // and 32 bytes for ABIv2. This is usually determined by target
3633  // endianness, but in theory could be overriden by function attribute.
3634  // For simplicity, we ignore it here (it'd only matter for QPX vectors).
3635  if (TargetTriple.getArch() == Triple::ppc64)
3636  VAArgBase = 48;
3637  else
3638  VAArgBase = 32;
3639  unsigned VAArgOffset = VAArgBase;
3640  const DataLayout &DL = F.getParent()->getDataLayout();
3641  for (CallSite::arg_iterator ArgIt = CS.arg_begin(), End = CS.arg_end();
3642  ArgIt != End; ++ArgIt) {
3643  Value *A = *ArgIt;
3644  unsigned ArgNo = CS.getArgumentNo(ArgIt);
3645  bool IsFixed = ArgNo < CS.getFunctionType()->getNumParams();
3646  bool IsByVal = CS.paramHasAttr(ArgNo, Attribute::ByVal);
3647  if (IsByVal) {
3648  assert(A->getType()->isPointerTy());
3649  Type *RealTy = A->getType()->getPointerElementType();
3650  uint64_t ArgSize = DL.getTypeAllocSize(RealTy);
3651  uint64_t ArgAlign = CS.getParamAlignment(ArgNo);
3652  if (ArgAlign < 8)
3653  ArgAlign = 8;
3654  VAArgOffset = alignTo(VAArgOffset, ArgAlign);
3655  if (!IsFixed) {
3656  Value *Base = getShadowPtrForVAArgument(RealTy, IRB,
3657  VAArgOffset - VAArgBase);
3658  Value *AShadowPtr, *AOriginPtr;
3659  std::tie(AShadowPtr, AOriginPtr) = MSV.getShadowOriginPtr(
3660  A, IRB, IRB.getInt8Ty(), kShadowTLSAlignment);
3661 
3662  IRB.CreateMemCpy(Base, AShadowPtr, ArgSize, kShadowTLSAlignment);
3663  }
3664  VAArgOffset += alignTo(ArgSize, 8);
3665  } else {
3666  Value *Base;
3667  uint64_t ArgSize = DL.getTypeAllocSize(A->getType());
3668  uint64_t ArgAlign = 8;
3669  if (A->getType()->isArrayTy()) {
3670  // Arrays are aligned to element size, except for long double
3671  // arrays, which are aligned to 8 bytes.
3672  Type *ElementTy = A->getType()->getArrayElementType();
3673  if (!ElementTy->isPPC_FP128Ty())
3674  ArgAlign = DL.getTypeAllocSize(ElementTy);
3675  } else if (A->getType()->isVectorTy()) {
3676  // Vectors are naturally aligned.
3677  ArgAlign = DL.getTypeAllocSize(A->getType());
3678  }
3679  if (ArgAlign < 8)
3680  ArgAlign = 8;
3681  VAArgOffset = alignTo(VAArgOffset, ArgAlign);
3682  if (DL.isBigEndian()) {
3683  // Adjusting the shadow for argument with size < 8 to match the placement
3684  // of bits in big endian system
3685  if (ArgSize < 8)
3686  VAArgOffset += (8 - ArgSize);
3687  }
3688  if (!IsFixed) {
3689  Base = getShadowPtrForVAArgument(A->getType(), IRB,
3690  VAArgOffset - VAArgBase);
3691  IRB.CreateAlignedStore(MSV.getShadow(A), Base, kShadowTLSAlignment);
3692  }
3693  VAArgOffset += ArgSize;
3694  VAArgOffset = alignTo(VAArgOffset, 8);
3695  }
3696  if (IsFixed)
3697  VAArgBase = VAArgOffset;
3698  }
3699 
3700  Constant *TotalVAArgSize = ConstantInt::get(IRB.getInt64Ty(),
3701  VAArgOffset - VAArgBase);
3702  // Here using VAArgOverflowSizeTLS as VAArgSizeTLS to avoid creation of
3703  // a new class member i.e. it is the total size of all VarArgs.
3704  IRB.CreateStore(TotalVAArgSize, MS.VAArgOverflowSizeTLS);
3705  }
3706 
3707  /// \brief Compute the shadow address for a given va_arg.
3708  Value *getShadowPtrForVAArgument(Type *Ty, IRBuilder<> &IRB,
3709  int ArgOffset) {
3710  Value *Base = IRB.CreatePointerCast(MS.VAArgTLS, MS.IntptrTy);
3711  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
3712  return IRB.CreateIntToPtr(Base, PointerType::get(MSV.getShadowTy(Ty), 0),
3713  "_msarg");
3714  }
3715 
3716  void visitVAStartInst(VAStartInst &I) override {
3717  IRBuilder<> IRB(&I);
3718  VAStartInstrumentationList.push_back(&I);
3719  Value *VAListTag = I.getArgOperand(0);
3720  Value *ShadowPtr, *OriginPtr;
3721  unsigned Alignment = 8;
3722  std::tie(ShadowPtr, OriginPtr) =
3723  MSV.getShadowOriginPtr(VAListTag, IRB, IRB.getInt8Ty(), Alignment);
3724  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
3725  /* size */ 8, Alignment, false);
3726  }
3727 
3728  void visitVACopyInst(VACopyInst &I) override {
3729  IRBuilder<> IRB(&I);
3730  Value *VAListTag = I.getArgOperand(0);
3731  Value *ShadowPtr, *OriginPtr;
3732  unsigned Alignment = 8;
3733  std::tie(ShadowPtr, OriginPtr) =
3734  MSV.getShadowOriginPtr(VAListTag, IRB, IRB.getInt8Ty(), Alignment);
3735  // Unpoison the whole __va_list_tag.
3736  // FIXME: magic ABI constants.
3737  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
3738  /* size */ 8, Alignment, false);
3739  }
3740 
3741  void finalizeInstrumentation() override {
3742  assert(!VAArgSize && !VAArgTLSCopy &&
3743  "finalizeInstrumentation called twice");
3745  VAArgSize = IRB.CreateLoad(MS.VAArgOverflowSizeTLS);
3746  Value *CopySize = IRB.CreateAdd(ConstantInt::get(MS.IntptrTy, 0),
3747  VAArgSize);
3748 
3749  if (!VAStartInstrumentationList.empty()) {
3750  // If there is a va_start in this function, make a backup copy of
3751  // va_arg_tls somewhere in the function entry block.
3752  VAArgTLSCopy = IRB.CreateAlloca(Type::getInt8Ty(*MS.C), CopySize);
3753  IRB.CreateMemCpy(VAArgTLSCopy, MS.VAArgTLS, CopySize, 8);
3754  }
3755 
3756  // Instrument va_start.
3757  // Copy va_list shadow from the backup copy of the TLS contents.
3758  for (size_t i = 0, n = VAStartInstrumentationList.size(); i < n; i++) {
3759  CallInst *OrigInst = VAStartInstrumentationList[i];
3760  IRBuilder<> IRB(OrigInst->getNextNode());
3761  Value *VAListTag = OrigInst->getArgOperand(0);
3762  Value *RegSaveAreaPtrPtr =
3763  IRB.CreateIntToPtr(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
3764  Type::getInt64PtrTy(*MS.C));
3765  Value *RegSaveAreaPtr = IRB.CreateLoad(RegSaveAreaPtrPtr);
3766  Value *RegSaveAreaShadowPtr, *RegSaveAreaOriginPtr;
3767  unsigned Alignment = 8;
3768  std::tie(RegSaveAreaShadowPtr, RegSaveAreaOriginPtr) =
3769  MSV.getShadowOriginPtr(RegSaveAreaPtr, IRB, IRB.getInt8Ty(),
3770  Alignment);
3771  IRB.CreateMemCpy(RegSaveAreaShadowPtr, VAArgTLSCopy, CopySize, Alignment);
3772  }
3773  }
3774 };
3775 
3776 /// \brief A no-op implementation of VarArgHelper.
3777 struct VarArgNoOpHelper : public VarArgHelper {
3778  VarArgNoOpHelper(Function &F, MemorySanitizer &MS,
3779  MemorySanitizerVisitor &MSV) {}
3780 
3781  void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {}
3782 
3783  void visitVAStartInst(VAStartInst &I) override {}
3784 
3785  void visitVACopyInst(VACopyInst &I) override {}
3786 
3787  void finalizeInstrumentation() override {}
3788 };
3789 
3790 } // end anonymous namespace
3791 
3792 static VarArgHelper *CreateVarArgHelper(Function &Func, MemorySanitizer &Msan,
3793  MemorySanitizerVisitor &Visitor) {
3794  // VarArg handling is only implemented on AMD64. False positives are possible
3795  // on other platforms.
3796  Triple TargetTriple(Func.getParent()->getTargetTriple());
3797  if (TargetTriple.getArch() == Triple::x86_64)
3798  return new VarArgAMD64Helper(Func, Msan, Visitor);
3799  else if (TargetTriple.getArch() == Triple::mips64 ||
3800  TargetTriple.getArch() == Triple::mips64el)
3801  return new VarArgMIPS64Helper(Func, Msan, Visitor);
3802  else if (TargetTriple.getArch() == Triple::aarch64)
3803  return new VarArgAArch64Helper(Func, Msan, Visitor);
3804  else if (TargetTriple.getArch() == Triple::ppc64 ||
3805  TargetTriple.getArch() == Triple::ppc64le)
3806  return new VarArgPowerPC64Helper(Func, Msan, Visitor);
3807  else
3808  return new VarArgNoOpHelper(Func, Msan, Visitor);
3809 }
3810 
3812  if (&F == MsanCtorFunction)
3813  return false;
3814  MemorySanitizerVisitor Visitor(F, *this);
3815 
3816  // Clear out readonly/readnone attributes.
3817  AttrBuilder B;
3818  B.addAttribute(Attribute::ReadOnly)
3819  .addAttribute(Attribute::ReadNone);
3821 
3822  return Visitor.runOnFunction();
3823 }
Value * CreateInBoundsGEP(Value *Ptr, ArrayRef< Value *> IdxList, const Twine &Name="")
Definition: IRBuilder.h:1244
Type * getVectorElementType() const
Definition: Type.h:368
uint64_t CallInst * C
Return a value (possibly void), from a function.
User::op_iterator arg_iterator
The type of iterator to use when looping over actual arguments at this call site. ...
Definition: CallSite.h:213
SymbolTableList< Instruction >::iterator eraseFromParent()
This method unlinks &#39;this&#39; from the containing basic block and deletes it.
Definition: Instruction.cpp:67
unsigned Log2_32_Ceil(uint32_t Value)
Return the ceil log base 2 of the specified value, 32 if the value is zero.
Definition: MathExtras.h:544
A parsed version of the target data layout string in and methods for querying it. ...
Definition: DataLayout.h:109
const std::string & getTargetTriple() const
Get the target triple which is a string describing the target host.
Definition: Module.h:233
static const MemoryMapParams Linux_PowerPC64_MemoryMapParams
Value * CreateICmp(CmpInst::Predicate P, Value *LHS, Value *RHS, const Twine &Name="")
Definition: IRBuilder.h:1638
Value * CreateConstGEP1_32(Value *Ptr, unsigned Idx0, const Twine &Name="")
Definition: IRBuilder.h:1278
static Constant * getString(LLVMContext &Context, StringRef Initializer, bool AddNull=true)
This method constructs a CDS and initializes it with a text string.
Definition: Constants.cpp:2450
bool isAllOnesValue() const
Return true if this is the value that would be returned by getAllOnesValue.
Definition: Constants.cpp:100
raw_ostream & errs()
This returns a reference to a raw_ostream for standard error.
void addIncoming(Value *V, BasicBlock *BB)
Add an incoming value to the end of the PHI list.
This instruction extracts a struct member or array element value from an aggregate value...
Value * CreateBinOp(Instruction::BinaryOps Opc, Value *LHS, Value *RHS, const Twine &Name="", MDNode *FPMathTag=nullptr)
Definition: IRBuilder.h:1112
GCNRegPressure max(const GCNRegPressure &P1, const GCNRegPressure &P2)
This class represents an incoming formal argument to a Function.
Definition: Argument.h:30
static const PlatformMemoryMapParams Linux_PowerPC_MemoryMapParams
AllocaInst * CreateAlloca(Type *Ty, unsigned AddrSpace, Value *ArraySize=nullptr, const Twine &Name="")
Definition: IRBuilder.h:1156
The C convention as implemented on Windows/x86-64 and AArch64.
Definition: CallingConv.h:154
Base class for instruction visitors.
Definition: InstVisitor.h:81
Value * getAggregateOperand()
Value * CreateICmpNE(Value *LHS, Value *RHS, const Twine &Name="")
Definition: IRBuilder.h:1553
Atomic ordering constants.
NodeTy * getNextNode()
Get the next node, or nullptr for the list tail.
Definition: ilist_node.h:289
LLVM_ATTRIBUTE_NORETURN void report_fatal_error(Error Err, bool gen_crash_diag=true)
Report a serious error, calling any installed error handler.
Definition: Error.cpp:115
Compute iterated dominance frontiers using a linear time algorithm.
Definition: AllocatorList.h:24
BinaryOps getOpcode() const
Definition: InstrTypes.h:523
unsigned getParamAlignment(unsigned ArgNo) const
Extract the alignment for a call or parameter (0=unknown).
Definition: CallSite.h:406
bool isAtomic() const
Return true if this instruction has an AtomicOrdering of unordered or higher.
Constant * getOrInsertFunction(StringRef Name, FunctionType *T, AttributeList AttributeList)
Look up the specified function in the module symbol table.
Definition: Module.cpp:142
Value * CreateXor(Value *LHS, Value *RHS, const Twine &Name="")
Definition: IRBuilder.h:1099
A Module instance is used to store all the information related to an LLVM module. ...
Definition: Module.h:63
bool isSized(SmallPtrSetImpl< Type *> *Visited=nullptr) const
Return true if it makes sense to take the size of this type.
Definition: Type.h:262
LLVM_ATTRIBUTE_ALWAYS_INLINE size_type size() const
Definition: SmallVector.h:136
void setOrdering(AtomicOrdering Ordering)
Sets the ordering constraint of this rmw instruction.
Definition: Instructions.h:762
static const MemoryMapParams Linux_I386_MemoryMapParams
Same, but only replaced by something equivalent.
Definition: GlobalValue.h:54
an instruction that atomically checks whether a specified value is in a memory location, and, if it is, stores a new value there.
Definition: Instructions.h:514
static const MemoryMapParams NetBSD_X86_64_MemoryMapParams
static cl::opt< bool > ClPoisonStackWithCall("msan-poison-stack-with-call", cl::desc("poison uninitialized stack variables with a call"), cl::Hidden, cl::init(false))
This class represents zero extension of integer types.
static const unsigned kRetvalTLSSize
Value * CreateICmpSLT(Value *LHS, Value *RHS, const Twine &Name="")
Definition: IRBuilder.h:1574
This class represents a function call, abstracting a target machine&#39;s calling convention.
static PointerType * getInt32PtrTy(LLVMContext &C, unsigned AS=0)
Definition: Type.cpp:228
void setOrdering(AtomicOrdering Ordering)
Sets the ordering constraint of this load instruction.
Definition: Instructions.h:239
static PointerType * get(Type *ElementType, unsigned AddressSpace)
This constructs a pointer to an object of the specified type in a numbered address space...
Definition: Type.cpp:617
const Value * getTrueValue() const
AtomicOrdering getOrdering() const
Returns the ordering constraint of this load instruction.
Definition: Instructions.h:233
Like Internal, but omit from symbol table.
Definition: GlobalValue.h:57
static VarArgHelper * CreateVarArgHelper(Function &Func, MemorySanitizer &Msan, MemorySanitizerVisitor &Visitor)
This instruction constructs a fixed permutation of two input vectors.
static cl::opt< bool > ClWithComdat("msan-with-comdat", cl::desc("Place MSan constructors in comdat sections"), cl::Hidden, cl::init(false))
Externally visible function.
Definition: GlobalValue.h:49
bool hasFnAttribute(Attribute::AttrKind Kind) const
Return true if the function has the attribute.
Definition: Function.h:302
A raw_ostream that writes to an SmallVector or SmallString.
Definition: raw_ostream.h:489
This class wraps the llvm.memset intrinsic.
static bool isEquality(Predicate P)
Return true if this predicate is either EQ or NE.
Value * CreateSExt(Value *V, Type *DestTy, const Twine &Name="")
Definition: IRBuilder.h:1390
Metadata node.
Definition: Metadata.h:862
static unsigned TypeSizeToSizeIndex(unsigned TypeSize)
F(f)
This class represents a sign extension of integer types.
uint64_t alignTo(uint64_t Value, uint64_t Align, uint64_t Skew=0)
Returns the next integer (mod 2**64) that is greater than or equal to Value and is a multiple of Alig...
Definition: MathExtras.h:677
CallInst * CreateMemSet(Value *Ptr, Value *Val, uint64_t Size, unsigned Align, bool isVolatile=false, MDNode *TBAATag=nullptr, MDNode *ScopeTag=nullptr, MDNode *NoAliasTag=nullptr)
Create and insert a memset to the specified pointer and the specified value.
Definition: IRBuilder.h:405
An instruction for reading from memory.
Definition: Instructions.h:164
AttrBuilder & addAttribute(Attribute::AttrKind Val)
Add an attribute to the builder.
an instruction that atomically reads a memory location, combines it with another value, and then stores the result back.
Definition: Instructions.h:677
Hexagon Common GEP
bool isVectorTy() const
True if this is an instance of VectorType.
Definition: Type.h:227
#define op(i)
bool isMustTailCall() const
static Type * getX86_MMXTy(LLVMContext &C)
Definition: Type.cpp:171
Use * op_iterator
Definition: User.h:209
bool isPPC_FP128Ty() const
Return true if this is powerpc long double.
Definition: Type.h:159
static const MemoryMapParams Linux_AArch64_MemoryMapParams
static cl::opt< bool > ClHandleICmp("msan-handle-icmp", cl::desc("propagate shadow through ICmpEQ and ICmpNE"), cl::Hidden, cl::init(true))
op_iterator op_begin()
Definition: User.h:214
static PointerType * getInt64PtrTy(LLVMContext &C, unsigned AS=0)
Definition: Type.cpp:232
static Constant * get(ArrayType *T, ArrayRef< Constant *> V)
Definition: Constants.cpp:887
unsigned getBitWidth() const
Return the number of bits in the APInt.
Definition: APInt.h:1488
static Constant * getNullValue(Type *Ty)
Constructor to create a &#39;0&#39; constant of arbitrary type.
Definition: Constants.cpp:206
Value * CreateNot(Value *V, const Twine &Name="")
Definition: IRBuilder.h:1146
StoreInst * CreateAlignedStore(Value *Val, Value *Ptr, unsigned Align, bool isVolatile=false)
Definition: IRBuilder.h:1203
IntegerType * getInt32Ty()
Fetch the type representing a 32-bit integer.
Definition: IRBuilder.h:348
unsigned countTrailingZeros() const
Count the number of trailing zero bits.
Definition: APInt.h:1611
static cl::opt< int > ClPoisonStackPattern("msan-poison-stack-pattern", cl::desc("poison uninitialized stack variables with the given pattern"), cl::Hidden, cl::init(0xff))
ArrayRef< unsigned > getIndices() const
AnalysisUsage & addRequired()
#define INITIALIZE_PASS_DEPENDENCY(depName)
Definition: PassSupport.h:51
bool isSigned() const
Determine if this instruction is using a signed comparison.
Definition: InstrTypes.h:994
static cl::opt< bool > ClDumpStrictInstructions("msan-dump-strict-instructions", cl::desc("print out instructions with default strict semantics"), cl::Hidden, cl::init(false))
This class represents the LLVM &#39;select&#39; instruction.
Type * getPointerElementType() const
Definition: Type.h:373
const DataLayout & getDataLayout() const
Get the data layout for the module&#39;s target platform.
Definition: Module.cpp:361
unsigned getAlignment() const
Return the alignment of the memory that is being allocated by the instruction.
Definition: Instructions.h:109
unsigned getNumArgOperands() const
Return the number of call arguments.
IntegerType * getInt64Ty()
Fetch the type representing a 64-bit integer.
Definition: IRBuilder.h:353
This is the base class for all instructions that perform data casts.
Definition: InstrTypes.h:560
ArrayRef< T > makeArrayRef(const T &OneElt)
Construct an ArrayRef from a single element.
Definition: ArrayRef.h:451
&#39;undef&#39; values are things that do not have specified contents.
Definition: Constants.h:1247
This class wraps the llvm.memmove intrinsic.
Class to represent struct types.
Definition: DerivedTypes.h:201
LLVMContext & getContext() const
Get the global data context.
Definition: Module.h:237
static cl::opt< bool > ClCheckAccessAddress("msan-check-access-address", cl::desc("report accesses through a pointer which has poisoned shadow"), cl::Hidden, cl::init(true))
PointerType * getPointerTo(unsigned AddrSpace=0) const
Return a pointer to the current type.
Definition: Type.cpp:639
IterTy arg_end() const
Definition: CallSite.h:575
bool isUnsigned() const
Determine if this instruction is using an unsigned comparison.
Definition: InstrTypes.h:1000
bool isIntegerTy() const
True if this is an instance of IntegerType.
Definition: Type.h:197
This provides a uniform API for creating instructions and inserting them into a basic block: either a...
Definition: IRBuilder.h:668
IntegerType * getIntPtrTy(const DataLayout &DL, unsigned AddrSpace=0)
Fetch the type representing a pointer to an integer value.
Definition: IRBuilder.h:391
This file contains the simple types necessary to represent the attributes associated with functions a...
InstrTy * getInstruction() const
Definition: CallSite.h:92
Value * CreateAdd(Value *LHS, Value *RHS, const Twine &Name="", bool HasNUW=false, bool HasNSW=false)
Definition: IRBuilder.h:893
void setName(const Twine &Name)
Change the name of the value.
Definition: Value.cpp:295
static StructType * get(LLVMContext &Context, ArrayRef< Type *> Elements, bool isPacked=false)
This static method is the primary way to create a literal StructType.
Definition: Type.cpp:336
This file implements a class to represent arbitrary precision integral constant values and operations...
Type * getVoidTy()
Fetch the type representing void.
Definition: IRBuilder.h:381
This class represents a cast from a pointer to an integer.
AtomicOrdering
Atomic ordering for LLVM&#39;s memory model.
StoreInst * CreateStore(Value *Val, Value *Ptr, bool isVolatile=false)
Definition: IRBuilder.h:1180
Value * CreateIntToPtr(Value *V, Type *DestTy, const Twine &Name="")
Definition: IRBuilder.h:1444
ValTy * getCalledValue() const
Return the pointer to function that is being called.
Definition: CallSite.h:100
bool isNullValue() const
Return true if this is the value that would be returned by getNullValue.
Definition: Constants.cpp:85
Class to represent function types.
Definition: DerivedTypes.h:103
Value * CreateBitCast(Value *V, Type *DestTy, const Twine &Name="")
Definition: IRBuilder.h:1448
Type * getType() const
All values are typed, get the type of this value.
Definition: Value.h:245
This represents the llvm.va_start intrinsic.
static const char *const kMsanInitName
std::string itostr(int64_t X)
Definition: StringExtras.h:208
AtomicOrdering getSuccessOrdering() const
Returns the success ordering constraint of this cmpxchg instruction.
Definition: Instructions.h:568
#define T
static cl::opt< int > ClTrackOrigins("msan-track-origins", cl::desc("Track origins (allocation sites) of poisoned memory"), cl::Hidden, cl::init(0))
Track origins of uninitialized values.
Class to represent array types.
Definition: DerivedTypes.h:369
This instruction compares its operands according to the predicate given to the constructor.
bool isVarArg() const
Definition: DerivedTypes.h:123
This class represents a no-op cast from one type to another.
bool paramHasAttr(unsigned ArgNo, Attribute::AttrKind Kind) const
Return true if the call or the callee has the given attribute.
Definition: CallSite.h:377
MDNode * getMetadata(unsigned KindID) const
Get the metadata of given kind attached to this Instruction.
Definition: Instruction.h:195
Value * getInsertedValueOperand()
SmallString - A SmallString is just a SmallVector with methods and accessors that make it work better...
Definition: SmallString.h:26
Value * CreateSub(Value *LHS, Value *RHS, const Twine &Name="", bool HasNUW=false, bool HasNSW=false)
Definition: IRBuilder.h:915
An instruction for storing to memory.
Definition: Instructions.h:306
bool removeUnreachableBlocks(Function &F, LazyValueInfo *LVI=nullptr, DeferredDominance *DDT=nullptr)
Remove all blocks that can not be reached from the function&#39;s entry.
Definition: Local.cpp:1826
bool isIntOrIntVectorTy() const
Return true if this is an integer type or a vector of integer types.
Definition: Type.h:203
static const unsigned kParamTLSSize
Value * CreateZExt(Value *V, Type *DestTy, const Twine &Name="")
Definition: IRBuilder.h:1387
static cl::opt< bool > ClPoisonStack("msan-poison-stack", cl::desc("poison uninitialized stack variables"), cl::Hidden, cl::init(true))
Function * getDeclaration(Module *M, ID id, ArrayRef< Type *> Tys=None)
Create or insert an LLVM Function declaration for an intrinsic, and return it.
Definition: Function.cpp:980
This class represents a truncation of integer types.
void SetInsertPoint(BasicBlock *TheBB)
This specifies that created instructions should be appended to the end of the specified block...
Definition: IRBuilder.h:128
bool onlyReadsMemory() const
Determine if the call does not access or only reads memory.
Value * getOperand(unsigned i) const
Definition: User.h:154
static const PlatformMemoryMapParams Linux_MIPS_MemoryMapParams
bool isRelational() const
Return true if the predicate is relational (not EQ or NE).
bool isCall() const
Return true if a CallInst is enclosed.
Definition: CallSite.h:87
Value * CreateOr(Value *LHS, Value *RHS, const Twine &Name="")
Definition: IRBuilder.h:1083
static const unsigned kMinOriginAlignment
Constant * getAggregateElement(unsigned Elt) const
For aggregates (struct/array/vector) return the constant that corresponds to the specified element if...
Definition: Constants.cpp:276
Type * getScalarType() const
If this is a vector type, return the element type, otherwise return &#39;this&#39;.
Definition: Type.h:301
bool isZeroValue() const
Return true if the value is negative zero or null value.
Definition: Constants.cpp:65
bool isVoidTy() const
Return true if this is &#39;void&#39;.
Definition: Type.h:141
const BasicBlock & getEntryBlock() const
Definition: Function.h:618
an instruction for type-safe pointer arithmetic to access elements of arrays and structs ...
Definition: Instructions.h:837
LoadInst * CreateLoad(Value *Ptr, const char *Name)
Definition: IRBuilder.h:1168
static bool runOnFunction(Function &F, bool PostInlining)
initializer< Ty > init(const Ty &Val)
Definition: CommandLine.h:406
This instruction inserts a single (scalar) element into a VectorType value.
The landingpad instruction holds all of the information necessary to generate correct exception handl...
const Instruction * getFirstNonPHI() const
Returns a pointer to the first instruction in this block that is not a PHINode instruction.
Definition: BasicBlock.cpp:171
const Value * getCalledValue() const
Get a pointer to the function that is invoked by this instruction.
static GCRegistry::Add< OcamlGC > B("ocaml", "ocaml 3.10-compatible GC")
const_iterator getFirstInsertionPt() const
Returns an iterator to the first instruction in this block that is suitable for inserting a non-PHI i...
Definition: BasicBlock.cpp:200
const BasicBlock * getSinglePredecessor() const
Return the predecessor of this block if it has a single predecessor block.
Definition: BasicBlock.cpp:217
INITIALIZE_PASS_BEGIN(MemorySanitizer, "msan", "MemorySanitizer: detects uninitialized reads.", false, false) INITIALIZE_PASS_END(MemorySanitizer
LLVM Basic Block Representation.
Definition: BasicBlock.h:59
The instances of the Type class are immutable: once they are created, they are never changed...
Definition: Type.h:46
std::pair< Function *, Function * > createSanitizerCtorAndInitFunctions(Module &M, StringRef CtorName, StringRef InitName, ArrayRef< Type *> InitArgTypes, ArrayRef< Value *> InitArgs, StringRef VersionCheckName=StringRef())
Creates sanitizer constructor function, and calls sanitizer&#39;s init function from it.
This is an important class for using LLVM in a threaded context.
Definition: LLVMContext.h:69
const char * getOpcodeName() const
Definition: Instruction.h:128
This is an important base class in LLVM.
Definition: Constant.h:42
Resume the propagation of an exception.
This file contains the declarations for the subclasses of Constant, which represent the different fla...
Value * CreateSelect(Value *C, Value *True, Value *False, const Twine &Name="", Instruction *MDFrom=nullptr)
Definition: IRBuilder.h:1693
bool isPointerTy() const
True if this is an instance of PointerType.
Definition: Type.h:221
unsigned getNumParams() const
Return the number of fixed parameters this function type requires.
Definition: DerivedTypes.h:139
Represent the analysis usage information of a pass.
op_iterator op_end()
Definition: User.h:216
static const PlatformMemoryMapParams Linux_X86_MemoryMapParams
This instruction compares its operands according to the predicate given to the constructor.
Predicate
This enumeration lists the possible predicates for CmpInst subclasses.
Definition: InstrTypes.h:853
static const unsigned End
FunctionPass class - This class is used to implement most global optimizations.
Definition: Pass.h:285
static const unsigned kShadowTLSAlignment
static FunctionType * get(Type *Result, ArrayRef< Type *> Params, bool isVarArg)
This static method is the primary way of constructing a FunctionType.
Definition: Type.cpp:297
static Constant * get(StructType *T, ArrayRef< Constant *> V)
Definition: Constants.cpp:948
Value * getPointerOperand()
Definition: Instructions.h:270
bool isX86_MMXTy() const
Return true if this is X86 MMX.
Definition: Type.h:182
Value * CreateICmpEQ(Value *LHS, Value *RHS, const Twine &Name="")
Definition: IRBuilder.h:1550
self_iterator getIterator()
Definition: ilist_node.h:82
Class to represent integer types.
Definition: DerivedTypes.h:40
IntegerType * getIntNTy(unsigned N)
Fetch the type representing an N-bit integer.
Definition: IRBuilder.h:361
Value * CreateExtractElement(Value *Vec, Value *Idx, const Twine &Name="")
Definition: IRBuilder.h:1713
This class represents a cast from an integer to a pointer.
const Value * getCondition() const
static Constant * getAllOnesValue(Type *Ty)
Get the all ones value.
Definition: Constants.cpp:260
static const PlatformMemoryMapParams Linux_ARM_MemoryMapParams
static const MemoryMapParams FreeBSD_X86_64_MemoryMapParams
Comdat * getOrInsertComdat(StringRef Name)
Return the Comdat in the module with the specified name.
Definition: Module.cpp:467
const Value * getArraySize() const
Get the number of elements allocated.
Definition: Instructions.h:93
Value * CreateExtractValue(Value *Agg, ArrayRef< unsigned > Idxs, const Twine &Name="")
Definition: IRBuilder.h:1755
PointerType * getInt8PtrTy(unsigned AddrSpace=0)
Fetch the type representing a pointer to an 8-bit integer value.
Definition: IRBuilder.h:386
MDNode * createBranchWeights(uint32_t TrueWeight, uint32_t FalseWeight)
Return metadata containing two branch weights.
Definition: MDBuilder.cpp:38
AtomicOrdering getOrdering() const
Returns the ordering constraint of this rmw instruction.
Definition: Instructions.h:757
INITIALIZE_PASS_END(RegBankSelect, DEBUG_TYPE, "Assign register bank of generic virtual registers", false, false) RegBankSelect
#define llvm_unreachable(msg)
Marks that the current location is not supposed to be reachable.
Value * CreateMul(Value *LHS, Value *RHS, const Twine &Name="", bool HasNUW=false, bool HasNSW=false)
Definition: IRBuilder.h:937
Value * CreateTrunc(Value *V, Type *DestTy, const Twine &Name="")
Definition: IRBuilder.h:1384
Type * getAllocatedType() const
Return the type that is being allocated by the instruction.
Definition: Instructions.h:102
Triple - Helper class for working with autoconf configuration names.
Definition: Triple.h:44
signed greater than
Definition: InstrTypes.h:880
unsigned first
bool isInvoke() const
Return true if a InvokeInst is enclosed.
Definition: CallSite.h:90
Intrinsic::ID getIntrinsicID() const
Return the intrinsic ID of this intrinsic.
Definition: IntrinsicInst.h:51
PHINode * CreatePHI(Type *Ty, unsigned NumReservedValues, const Twine &Name="")
Definition: IRBuilder.h:1658
bool isPtrOrPtrVectorTy() const
Return true if this is a pointer type or a vector of pointer types.
Definition: Type.h:224
static IntegerType * get(LLVMContext &C, unsigned NumBits)
This static method is the primary way of constructing an IntegerType.
Definition: Type.cpp:240
Type * getSequentialElementType() const
Definition: Type.h:355
Iterator for intrusive lists based on ilist_node.
unsigned getNumOperands() const
Definition: User.h:176
See the file comment.
Definition: ValueMap.h:86
This is the shared class of boolean and integer constants.
Definition: Constants.h:84
iterator end()
Definition: BasicBlock.h:254
unsigned getScalarSizeInBits() const LLVM_READONLY
If this is a vector type, return the getPrimitiveSizeInBits value for the element type...
Definition: Type.cpp:130
CallingConv::ID getCallingConv() const
getCallingConv()/setCallingConv(CC) - These method get and set the calling convention of this functio...
Definition: Function.h:194
IterTy arg_begin() const
Definition: CallSite.h:571
Value * CreateIntCast(Value *V, Type *DestTy, bool isSigned, const Twine &Name="")
Definition: IRBuilder.h:1511
This is a &#39;vector&#39; (really, a variable-sized array), optimized for the case when the array is small...
Definition: SmallVector.h:862
Module.h This file contains the declarations for the Module class.
Value * CreateInsertElement(Value *Vec, Value *NewElt, Value *Idx, const Twine &Name="")
Definition: IRBuilder.h:1726
Provides information about what library functions are available for the current target.
unsigned getABITypeAlignment(Type *Ty) const
Returns the minimum ABI-required alignment for the specified type.
Definition: DataLayout.cpp:682
bool isAggregateType() const
Return true if the type is an aggregate type.
Definition: Type.h:255
signed less than
Definition: InstrTypes.h:882
TerminatorInst * SplitBlockAndInsertIfThen(Value *Cond, Instruction *SplitBefore, bool Unreachable, MDNode *BranchWeights=nullptr, DominatorTree *DT=nullptr, LoopInfo *LI=nullptr)
Split the containing block at the specified instruction - everything before SplitBefore stays in the ...
CHAIN = SC CHAIN, Imm128 - System call.
ConstantInt * getInt32(uint32_t C)
Get a constant 32-bit value.
Definition: IRBuilder.h:308
static IntegerType * getIntNTy(LLVMContext &C, unsigned N)
Definition: Type.cpp:180
StringRef str()
Return a StringRef for the vector contents.
Definition: raw_ostream.h:514
static GCRegistry::Add< StatepointGC > D("statepoint-example", "an example strategy for statepoint")
This class wraps the llvm.memcpy intrinsic.
static Constant * get(Type *Ty, uint64_t V, bool isSigned=false)
If Ty is a vector type, return a Constant with a splat of the given value.
Definition: Constants.cpp:559
void appendToGlobalCtors(Module &M, Function *F, int Priority, Constant *Data=nullptr)
Append F to the list of global ctors of module M with the given Priority.
Definition: ModuleUtils.cpp:84
unsigned getNumIncomingValues() const
Return the number of incoming edges.
static const size_t kNumberOfAccessSizes
static GlobalVariable * createPrivateNonConstGlobalForString(Module &M, StringRef Str)
Create a non-const global initialized with the given string.
static cl::opt< bool > ClKeepGoing("msan-keep-going", cl::desc("keep going after reporting a UMR"), cl::Hidden, cl::init(false))
raw_ostream & dbgs()
dbgs() - This returns a reference to a raw_ostream for debugging messages.
Definition: Debug.cpp:132
unsigned getVectorNumElements() const
Definition: DerivedTypes.h:462
signed less or equal
Definition: InstrTypes.h:883
Class to represent vector types.
Definition: DerivedTypes.h:393
const Module * getModule() const
Return the module owning the function this instruction belongs to or nullptr it the function does not...
Definition: Instruction.cpp:55
Class for arbitrary precision integers.
Definition: APInt.h:69
IntegerType * getInt8Ty()
Fetch the type representing an 8-bit integer.
Definition: IRBuilder.h:338
Value * CreateShl(Value *LHS, Value *RHS, const Twine &Name="", bool HasNUW=false, bool HasNSW=false)
Definition: IRBuilder.h:1012
const Value * getFalseValue() const
Value * CreatePointerCast(Value *V, Type *DestTy, const Twine &Name="")
Definition: IRBuilder.h:1488
uint64_t getTypeSizeInBits(Type *Ty) const
Size examples:
Definition: DataLayout.h:530
static cl::opt< ITMode > IT(cl::desc("IT block support"), cl::Hidden, cl::init(DefaultIT), cl::ZeroOrMore, cl::values(clEnumValN(DefaultIT, "arm-default-it", "Generate IT block based on arch"), clEnumValN(RestrictedIT, "arm-restrict-it", "Disallow deprecated IT based on ARMv8"), clEnumValN(NoRestrictedIT, "arm-no-restrict-it", "Allow IT blocks based on ARMv7")))
uint64_t getTypeAllocSize(Type *Ty) const
Returns the offset in bytes between successive objects of the specified type, including alignment pad...
Definition: DataLayout.h:403
Function * getCalledFunction() const
Return the function called, or null if this is an indirect function invocation.
void removeAttributes(unsigned i, const AttrBuilder &Attrs)
removes the attributes from the list of attributes.
Definition: Function.cpp:395
Predicate getPredicate() const
Return the predicate for this instruction.
Definition: InstrTypes.h:927
bool isInlineAsm() const
Check if this call is an inline asm statement.
static cl::opt< bool > ClHandleICmpExact("msan-handle-icmp-exact", cl::desc("exact handling of relational integer ICmp"), cl::Hidden, cl::init(false))
CallInst * CreateMemCpy(Value *Dst, Value *Src, uint64_t Size, unsigned Align, bool isVolatile=false, MDNode *TBAATag=nullptr, MDNode *TBAAStructTag=nullptr, MDNode *ScopeTag=nullptr, MDNode *NoAliasTag=nullptr)
Create and insert a memcpy between the specified pointers.
Definition: IRBuilder.h:423
unsigned getAlignment() const
Return the alignment of the access that is being performed.
Definition: Instructions.h:226
LLVM_NODISCARD bool empty() const
Definition: SmallVector.h:61
Value * getArgOperand(unsigned i) const
getArgOperand/setArgOperand - Return/set the i-th call argument.
StringRef getName() const
Return a constant reference to the value&#39;s name.
Definition: Value.cpp:224
#define I(x, y, z)
Definition: MD5.cpp:58
static const MemoryMapParams Linux_X86_64_MemoryMapParams
static ArrayType * get(Type *ElementType, uint64_t NumElements)
This static method is the primary way to construct an ArrayType.
Definition: Type.cpp:568
LLVM_NODISCARD std::enable_if<!is_simple_type< Y >::value, typename cast_retty< X, const Y >::ret_type >::type dyn_cast(const Y &Val)
Definition: Casting.h:323
This instruction extracts a single (scalar) element from a VectorType value.
void maybeMarkSanitizerLibraryCallNoBuiltin(CallInst *CI, const TargetLibraryInfo *TLI)
Given a CallInst, check if it calls a string function known to CodeGen, and mark it with NoBuiltin if...
Definition: Local.cpp:2329
static const PlatformMemoryMapParams NetBSD_X86_MemoryMapParams
Value * getReturnValue() const
Convenience accessor. Returns null if there is no return value.
static InlineAsm * get(FunctionType *Ty, StringRef AsmString, StringRef Constraints, bool hasSideEffects, bool isAlignStack=false, AsmDialect asmDialect=AD_ATT)
InlineAsm::get - Return the specified uniqued inline asm string.
Definition: InlineAsm.cpp:43
static cl::opt< bool > ClCheckConstantShadow("msan-check-constant-shadow", cl::desc("Insert checks for constant shadow values"), cl::Hidden, cl::init(false))
static const MemoryMapParams Linux_MIPS64_MemoryMapParams
Value * CreateAnd(Value *LHS, Value *RHS, const Twine &Name="")
Definition: IRBuilder.h:1067
Value * CreatePtrToInt(Value *V, Type *DestTy, const Twine &Name="")
Definition: IRBuilder.h:1440
static const unsigned kOriginSize
static const MemoryMapParams FreeBSD_I386_MemoryMapParams
bool isFPOrFPVectorTy() const
Return true if this is a FP type or a vector of FP.
Definition: Type.h:185
iterator_range< df_iterator< T > > depth_first(const T &G)
This represents the llvm.va_copy intrinsic.
bool isArrayAllocation() const
Return true if there is an allocation size parameter to the allocation instruction that is not 1...
size_type count(const KeyT &Val) const
Return 1 if the specified key is in the map, 0 otherwise.
Definition: ValueMap.h:154
assert(ImpDefSCC.getReg()==AMDGPU::SCC &&ImpDefSCC.isDef())
LoadInst * CreateAlignedLoad(Value *Ptr, unsigned Align, const char *Name)
Definition: IRBuilder.h:1186
static cl::opt< int > ClInstrumentationWithCallThreshold("msan-instrumentation-with-call-threshold", cl::desc("If the function being instrumented requires more than " "this number of checks and origin stores, use callbacks instead of " "inline checks (-1 means never use callbacks)."), cl::Hidden, cl::init(3500))
unsigned getPrimitiveSizeInBits() const LLVM_READONLY
Return the basic size of this type if it is a primitive type.
Definition: Type.cpp:115
void setSuccessOrdering(AtomicOrdering Ordering)
Sets the success ordering constraint of this cmpxchg instruction.
Definition: Instructions.h:573
ArrayRef< unsigned > getIndices() const
Module * getParent()
Get the module that this global value is contained inside of...
Definition: GlobalValue.h:561
LLVM Value Representation.
Definition: Value.h:73
uint64_t getTypeStoreSize(Type *Ty) const
Returns the maximum number of bytes that may be overwritten by storing the specified type...
Definition: DataLayout.h:386
FunctionType * getFunctionType() const
Definition: CallSite.h:320
constexpr char Size[]
Key for Kernel::Arg::Metadata::mSize.
static VectorType * get(Type *ElementType, unsigned NumElements)
This static method is the primary way to construct an VectorType.
Definition: Type.cpp:593
#define LLVM_FALLTHROUGH
LLVM_FALLTHROUGH - Mark fallthrough cases in switch statements.
Definition: Compiler.h:235
std::underlying_type< E >::type Mask()
Get a bitmask with 1s in all places up to the high-order bit of E&#39;s largest value.
Definition: BitmaskEnum.h:81
unsigned getArgumentNo(Value::const_user_iterator I) const
Given a value use iterator, returns the argument that corresponds to it.
Definition: CallSite.h:199
BasicBlock::iterator GetInsertPoint() const
Definition: IRBuilder.h:123
bool doesNotAccessMemory() const
Determine if the call does not access memory.
FunctionPass * createMemorySanitizerPass(int TrackOrigins=0, bool Recover=false)
Value * CreateLShr(Value *LHS, Value *RHS, const Twine &Name="", bool isExact=false)
Definition: IRBuilder.h:1031
static cl::opt< bool > ClPoisonUndef("msan-poison-undef", cl::desc("poison undef temps"), cl::Hidden, cl::init(true))
#define DEBUG(X)
Definition: Debug.h:118
ConstantInt * getInt8(uint8_t C)
Get a constant 8-bit value.
Definition: IRBuilder.h:298
StringRef - Represent a constant reference to a string, i.e.
Definition: StringRef.h:49
Predicate getSwappedPredicate() const
For example, EQ->EQ, SLE->SGE, ULT->UGT, OEQ->OEQ, ULE->UGE, OLT->OGT, etc.
Definition: InstrTypes.h:967
static const PlatformMemoryMapParams FreeBSD_X86_MemoryMapParams
Type * getArrayElementType() const
Definition: Type.h:362
Value * CreateInsertValue(Value *Agg, Value *Val, ArrayRef< unsigned > Idxs, const Twine &Name="")
Definition: IRBuilder.h:1763
bool isBigEndian() const
Definition: DataLayout.h:217
static IntegerType * getInt8Ty(LLVMContext &C)
Definition: Type.cpp:174
static const char *const kMsanModuleCtorName
static Constant * get(ArrayRef< Constant *> V)
Definition: Constants.cpp:983
iterator_range< arg_iterator > args()
Definition: Function.h:667
signed greater or equal
Definition: InstrTypes.h:881
bool isArrayTy() const
True if this is an instance of ArrayType.
Definition: Type.h:218
Type * getContainedType(unsigned i) const
This method is used to implement the type iterator (defined at the end of the file).
Definition: Type.h:330
A wrapper class for inspecting calls to intrinsic functions.
Definition: IntrinsicInst.h:44
const BasicBlock * getParent() const
Definition: Instruction.h:67
an instruction to allocate memory on the stack
Definition: Instructions.h:60
This instruction inserts a struct field of array element value into an aggregate value.
CallInst * CreateCall(Value *Callee, ArrayRef< Value *> Args=None, const Twine &Name="", MDNode *FPMathTag=nullptr)
Definition: IRBuilder.h:1663