LLVM  9.0.0svn
MemorySanitizer.cpp
Go to the documentation of this file.
1 //===- MemorySanitizer.cpp - detector of uninitialized reads --------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 /// \file
10 /// This file is a part of MemorySanitizer, a detector of uninitialized
11 /// reads.
12 ///
13 /// The algorithm of the tool is similar to Memcheck
14 /// (http://goo.gl/QKbem). We associate a few shadow bits with every
15 /// byte of the application memory, poison the shadow of the malloc-ed
16 /// or alloca-ed memory, load the shadow bits on every memory read,
17 /// propagate the shadow bits through some of the arithmetic
18 /// instruction (including MOV), store the shadow bits on every memory
19 /// write, report a bug on some other instructions (e.g. JMP) if the
20 /// associated shadow is poisoned.
21 ///
22 /// But there are differences too. The first and the major one:
23 /// compiler instrumentation instead of binary instrumentation. This
24 /// gives us much better register allocation, possible compiler
25 /// optimizations and a fast start-up. But this brings the major issue
26 /// as well: msan needs to see all program events, including system
27 /// calls and reads/writes in system libraries, so we either need to
28 /// compile *everything* with msan or use a binary translation
29 /// component (e.g. DynamoRIO) to instrument pre-built libraries.
30 /// Another difference from Memcheck is that we use 8 shadow bits per
31 /// byte of application memory and use a direct shadow mapping. This
32 /// greatly simplifies the instrumentation code and avoids races on
33 /// shadow updates (Memcheck is single-threaded so races are not a
34 /// concern there. Memcheck uses 2 shadow bits per byte with a slow
35 /// path storage that uses 8 bits per byte).
36 ///
37 /// The default value of shadow is 0, which means "clean" (not poisoned).
38 ///
39 /// Every module initializer should call __msan_init to ensure that the
40 /// shadow memory is ready. On error, __msan_warning is called. Since
41 /// parameters and return values may be passed via registers, we have a
42 /// specialized thread-local shadow for return values
43 /// (__msan_retval_tls) and parameters (__msan_param_tls).
44 ///
45 /// Origin tracking.
46 ///
47 /// MemorySanitizer can track origins (allocation points) of all uninitialized
48 /// values. This behavior is controlled with a flag (msan-track-origins) and is
49 /// disabled by default.
50 ///
51 /// Origins are 4-byte values created and interpreted by the runtime library.
52 /// They are stored in a second shadow mapping, one 4-byte value for 4 bytes
53 /// of application memory. Propagation of origins is basically a bunch of
54 /// "select" instructions that pick the origin of a dirty argument, if an
55 /// instruction has one.
56 ///
57 /// Every 4 aligned, consecutive bytes of application memory have one origin
58 /// value associated with them. If these bytes contain uninitialized data
59 /// coming from 2 different allocations, the last store wins. Because of this,
60 /// MemorySanitizer reports can show unrelated origins, but this is unlikely in
61 /// practice.
62 ///
63 /// Origins are meaningless for fully initialized values, so MemorySanitizer
64 /// avoids storing origin to memory when a fully initialized value is stored.
65 /// This way it avoids needless overwritting origin of the 4-byte region on
66 /// a short (i.e. 1 byte) clean store, and it is also good for performance.
67 ///
68 /// Atomic handling.
69 ///
70 /// Ideally, every atomic store of application value should update the
71 /// corresponding shadow location in an atomic way. Unfortunately, atomic store
72 /// of two disjoint locations can not be done without severe slowdown.
73 ///
74 /// Therefore, we implement an approximation that may err on the safe side.
75 /// In this implementation, every atomically accessed location in the program
76 /// may only change from (partially) uninitialized to fully initialized, but
77 /// not the other way around. We load the shadow _after_ the application load,
78 /// and we store the shadow _before_ the app store. Also, we always store clean
79 /// shadow (if the application store is atomic). This way, if the store-load
80 /// pair constitutes a happens-before arc, shadow store and load are correctly
81 /// ordered such that the load will get either the value that was stored, or
82 /// some later value (which is always clean).
83 ///
84 /// This does not work very well with Compare-And-Swap (CAS) and
85 /// Read-Modify-Write (RMW) operations. To follow the above logic, CAS and RMW
86 /// must store the new shadow before the app operation, and load the shadow
87 /// after the app operation. Computers don't work this way. Current
88 /// implementation ignores the load aspect of CAS/RMW, always returning a clean
89 /// value. It implements the store part as a simple atomic store by storing a
90 /// clean shadow.
91 ///
92 /// Instrumenting inline assembly.
93 ///
94 /// For inline assembly code LLVM has little idea about which memory locations
95 /// become initialized depending on the arguments. It can be possible to figure
96 /// out which arguments are meant to point to inputs and outputs, but the
97 /// actual semantics can be only visible at runtime. In the Linux kernel it's
98 /// also possible that the arguments only indicate the offset for a base taken
99 /// from a segment register, so it's dangerous to treat any asm() arguments as
100 /// pointers. We take a conservative approach generating calls to
101 /// __msan_instrument_asm_store(ptr, size)
102 /// , which defer the memory unpoisoning to the runtime library.
103 /// The latter can perform more complex address checks to figure out whether
104 /// it's safe to touch the shadow memory.
105 /// Like with atomic operations, we call __msan_instrument_asm_store() before
106 /// the assembly call, so that changes to the shadow memory will be seen by
107 /// other threads together with main memory initialization.
108 ///
109 /// KernelMemorySanitizer (KMSAN) implementation.
110 ///
111 /// The major differences between KMSAN and MSan instrumentation are:
112 /// - KMSAN always tracks the origins and implies msan-keep-going=true;
113 /// - KMSAN allocates shadow and origin memory for each page separately, so
114 /// there are no explicit accesses to shadow and origin in the
115 /// instrumentation.
116 /// Shadow and origin values for a particular X-byte memory location
117 /// (X=1,2,4,8) are accessed through pointers obtained via the
118 /// __msan_metadata_ptr_for_load_X(ptr)
119 /// __msan_metadata_ptr_for_store_X(ptr)
120 /// functions. The corresponding functions check that the X-byte accesses
121 /// are possible and returns the pointers to shadow and origin memory.
122 /// Arbitrary sized accesses are handled with:
123 /// __msan_metadata_ptr_for_load_n(ptr, size)
124 /// __msan_metadata_ptr_for_store_n(ptr, size);
125 /// - TLS variables are stored in a single per-task struct. A call to a
126 /// function __msan_get_context_state() returning a pointer to that struct
127 /// is inserted into every instrumented function before the entry block;
128 /// - __msan_warning() takes a 32-bit origin parameter;
129 /// - local variables are poisoned with __msan_poison_alloca() upon function
130 /// entry and unpoisoned with __msan_unpoison_alloca() before leaving the
131 /// function;
132 /// - the pass doesn't declare any global variables or add global constructors
133 /// to the translation unit.
134 ///
135 /// Also, KMSAN currently ignores uninitialized memory passed into inline asm
136 /// calls, making sure we're on the safe side wrt. possible false positives.
137 ///
138 /// KernelMemorySanitizer only supports X86_64 at the moment.
139 ///
140 //===----------------------------------------------------------------------===//
141 
143 #include "llvm/ADT/APInt.h"
144 #include "llvm/ADT/ArrayRef.h"
146 #include "llvm/ADT/SmallString.h"
147 #include "llvm/ADT/SmallVector.h"
148 #include "llvm/ADT/StringExtras.h"
149 #include "llvm/ADT/StringRef.h"
150 #include "llvm/ADT/Triple.h"
152 #include "llvm/IR/Argument.h"
153 #include "llvm/IR/Attributes.h"
154 #include "llvm/IR/BasicBlock.h"
155 #include "llvm/IR/CallSite.h"
156 #include "llvm/IR/CallingConv.h"
157 #include "llvm/IR/Constant.h"
158 #include "llvm/IR/Constants.h"
159 #include "llvm/IR/DataLayout.h"
160 #include "llvm/IR/DerivedTypes.h"
161 #include "llvm/IR/Function.h"
162 #include "llvm/IR/GlobalValue.h"
163 #include "llvm/IR/GlobalVariable.h"
164 #include "llvm/IR/IRBuilder.h"
165 #include "llvm/IR/InlineAsm.h"
166 #include "llvm/IR/InstVisitor.h"
167 #include "llvm/IR/InstrTypes.h"
168 #include "llvm/IR/Instruction.h"
169 #include "llvm/IR/Instructions.h"
170 #include "llvm/IR/IntrinsicInst.h"
171 #include "llvm/IR/Intrinsics.h"
172 #include "llvm/IR/LLVMContext.h"
173 #include "llvm/IR/MDBuilder.h"
174 #include "llvm/IR/Module.h"
175 #include "llvm/IR/Type.h"
176 #include "llvm/IR/Value.h"
177 #include "llvm/IR/ValueMap.h"
178 #include "llvm/Pass.h"
180 #include "llvm/Support/Casting.h"
182 #include "llvm/Support/Compiler.h"
183 #include "llvm/Support/Debug.h"
185 #include "llvm/Support/MathExtras.h"
191 #include <algorithm>
192 #include <cassert>
193 #include <cstddef>
194 #include <cstdint>
195 #include <memory>
196 #include <string>
197 #include <tuple>
198 
199 using namespace llvm;
200 
201 #define DEBUG_TYPE "msan"
202 
203 static const unsigned kOriginSize = 4;
204 static const unsigned kMinOriginAlignment = 4;
205 static const unsigned kShadowTLSAlignment = 8;
206 
207 // These constants must be kept in sync with the ones in msan.h.
208 static const unsigned kParamTLSSize = 800;
209 static const unsigned kRetvalTLSSize = 800;
210 
211 // Accesses sizes are powers of two: 1, 2, 4, 8.
212 static const size_t kNumberOfAccessSizes = 4;
213 
214 /// Track origins of uninitialized values.
215 ///
216 /// Adds a section to MemorySanitizer report that points to the allocation
217 /// (stack or heap) the uninitialized bits came from originally.
218 static cl::opt<int> ClTrackOrigins("msan-track-origins",
219  cl::desc("Track origins (allocation sites) of poisoned memory"),
220  cl::Hidden, cl::init(0));
221 
222 static cl::opt<bool> ClKeepGoing("msan-keep-going",
223  cl::desc("keep going after reporting a UMR"),
224  cl::Hidden, cl::init(false));
225 
226 static cl::opt<bool> ClPoisonStack("msan-poison-stack",
227  cl::desc("poison uninitialized stack variables"),
228  cl::Hidden, cl::init(true));
229 
230 static cl::opt<bool> ClPoisonStackWithCall("msan-poison-stack-with-call",
231  cl::desc("poison uninitialized stack variables with a call"),
232  cl::Hidden, cl::init(false));
233 
234 static cl::opt<int> ClPoisonStackPattern("msan-poison-stack-pattern",
235  cl::desc("poison uninitialized stack variables with the given pattern"),
236  cl::Hidden, cl::init(0xff));
237 
238 static cl::opt<bool> ClPoisonUndef("msan-poison-undef",
239  cl::desc("poison undef temps"),
240  cl::Hidden, cl::init(true));
241 
242 static cl::opt<bool> ClHandleICmp("msan-handle-icmp",
243  cl::desc("propagate shadow through ICmpEQ and ICmpNE"),
244  cl::Hidden, cl::init(true));
245 
246 static cl::opt<bool> ClHandleICmpExact("msan-handle-icmp-exact",
247  cl::desc("exact handling of relational integer ICmp"),
248  cl::Hidden, cl::init(false));
249 
250 // When compiling the Linux kernel, we sometimes see false positives related to
251 // MSan being unable to understand that inline assembly calls may initialize
252 // local variables.
253 // This flag makes the compiler conservatively unpoison every memory location
254 // passed into an assembly call. Note that this may cause false positives.
255 // Because it's impossible to figure out the array sizes, we can only unpoison
256 // the first sizeof(type) bytes for each type* pointer.
257 // The instrumentation is only enabled in KMSAN builds, and only if
258 // -msan-handle-asm-conservative is on. This is done because we may want to
259 // quickly disable assembly instrumentation when it breaks.
261  "msan-handle-asm-conservative",
262  cl::desc("conservative handling of inline assembly"), cl::Hidden,
263  cl::init(true));
264 
265 // This flag controls whether we check the shadow of the address
266 // operand of load or store. Such bugs are very rare, since load from
267 // a garbage address typically results in SEGV, but still happen
268 // (e.g. only lower bits of address are garbage, or the access happens
269 // early at program startup where malloc-ed memory is more likely to
270 // be zeroed. As of 2012-08-28 this flag adds 20% slowdown.
271 static cl::opt<bool> ClCheckAccessAddress("msan-check-access-address",
272  cl::desc("report accesses through a pointer which has poisoned shadow"),
273  cl::Hidden, cl::init(true));
274 
275 static cl::opt<bool> ClDumpStrictInstructions("msan-dump-strict-instructions",
276  cl::desc("print out instructions with default strict semantics"),
277  cl::Hidden, cl::init(false));
278 
280  "msan-instrumentation-with-call-threshold",
281  cl::desc(
282  "If the function being instrumented requires more than "
283  "this number of checks and origin stores, use callbacks instead of "
284  "inline checks (-1 means never use callbacks)."),
285  cl::Hidden, cl::init(3500));
286 
287 static cl::opt<bool>
288  ClEnableKmsan("msan-kernel",
289  cl::desc("Enable KernelMemorySanitizer instrumentation"),
290  cl::Hidden, cl::init(false));
291 
292 // This is an experiment to enable handling of cases where shadow is a non-zero
293 // compile-time constant. For some unexplainable reason they were silently
294 // ignored in the instrumentation.
295 static cl::opt<bool> ClCheckConstantShadow("msan-check-constant-shadow",
296  cl::desc("Insert checks for constant shadow values"),
297  cl::Hidden, cl::init(false));
298 
299 // This is off by default because of a bug in gold:
300 // https://sourceware.org/bugzilla/show_bug.cgi?id=19002
301 static cl::opt<bool> ClWithComdat("msan-with-comdat",
302  cl::desc("Place MSan constructors in comdat sections"),
303  cl::Hidden, cl::init(false));
304 
305 // These options allow to specify custom memory map parameters
306 // See MemoryMapParams for details.
307 static cl::opt<unsigned long long> ClAndMask("msan-and-mask",
308  cl::desc("Define custom MSan AndMask"),
309  cl::Hidden, cl::init(0));
310 
311 static cl::opt<unsigned long long> ClXorMask("msan-xor-mask",
312  cl::desc("Define custom MSan XorMask"),
313  cl::Hidden, cl::init(0));
314 
315 static cl::opt<unsigned long long> ClShadowBase("msan-shadow-base",
316  cl::desc("Define custom MSan ShadowBase"),
317  cl::Hidden, cl::init(0));
318 
319 static cl::opt<unsigned long long> ClOriginBase("msan-origin-base",
320  cl::desc("Define custom MSan OriginBase"),
321  cl::Hidden, cl::init(0));
322 
323 static const char *const kMsanModuleCtorName = "msan.module_ctor";
324 static const char *const kMsanInitName = "__msan_init";
325 
326 namespace {
327 
328 // Memory map parameters used in application-to-shadow address calculation.
329 // Offset = (Addr & ~AndMask) ^ XorMask
330 // Shadow = ShadowBase + Offset
331 // Origin = OriginBase + Offset
332 struct MemoryMapParams {
333  uint64_t AndMask;
334  uint64_t XorMask;
335  uint64_t ShadowBase;
336  uint64_t OriginBase;
337 };
338 
339 struct PlatformMemoryMapParams {
340  const MemoryMapParams *bits32;
341  const MemoryMapParams *bits64;
342 };
343 
344 } // end anonymous namespace
345 
346 // i386 Linux
347 static const MemoryMapParams Linux_I386_MemoryMapParams = {
348  0x000080000000, // AndMask
349  0, // XorMask (not used)
350  0, // ShadowBase (not used)
351  0x000040000000, // OriginBase
352 };
353 
354 // x86_64 Linux
355 static const MemoryMapParams Linux_X86_64_MemoryMapParams = {
356 #ifdef MSAN_LINUX_X86_64_OLD_MAPPING
357  0x400000000000, // AndMask
358  0, // XorMask (not used)
359  0, // ShadowBase (not used)
360  0x200000000000, // OriginBase
361 #else
362  0, // AndMask (not used)
363  0x500000000000, // XorMask
364  0, // ShadowBase (not used)
365  0x100000000000, // OriginBase
366 #endif
367 };
368 
369 // mips64 Linux
370 static const MemoryMapParams Linux_MIPS64_MemoryMapParams = {
371  0, // AndMask (not used)
372  0x008000000000, // XorMask
373  0, // ShadowBase (not used)
374  0x002000000000, // OriginBase
375 };
376 
377 // ppc64 Linux
378 static const MemoryMapParams Linux_PowerPC64_MemoryMapParams = {
379  0xE00000000000, // AndMask
380  0x100000000000, // XorMask
381  0x080000000000, // ShadowBase
382  0x1C0000000000, // OriginBase
383 };
384 
385 // aarch64 Linux
386 static const MemoryMapParams Linux_AArch64_MemoryMapParams = {
387  0, // AndMask (not used)
388  0x06000000000, // XorMask
389  0, // ShadowBase (not used)
390  0x01000000000, // OriginBase
391 };
392 
393 // i386 FreeBSD
394 static const MemoryMapParams FreeBSD_I386_MemoryMapParams = {
395  0x000180000000, // AndMask
396  0x000040000000, // XorMask
397  0x000020000000, // ShadowBase
398  0x000700000000, // OriginBase
399 };
400 
401 // x86_64 FreeBSD
402 static const MemoryMapParams FreeBSD_X86_64_MemoryMapParams = {
403  0xc00000000000, // AndMask
404  0x200000000000, // XorMask
405  0x100000000000, // ShadowBase
406  0x380000000000, // OriginBase
407 };
408 
409 // x86_64 NetBSD
410 static const MemoryMapParams NetBSD_X86_64_MemoryMapParams = {
411  0, // AndMask
412  0x500000000000, // XorMask
413  0, // ShadowBase
414  0x100000000000, // OriginBase
415 };
416 
417 static const PlatformMemoryMapParams Linux_X86_MemoryMapParams = {
420 };
421 
422 static const PlatformMemoryMapParams Linux_MIPS_MemoryMapParams = {
423  nullptr,
425 };
426 
427 static const PlatformMemoryMapParams Linux_PowerPC_MemoryMapParams = {
428  nullptr,
430 };
431 
432 static const PlatformMemoryMapParams Linux_ARM_MemoryMapParams = {
433  nullptr,
435 };
436 
437 static const PlatformMemoryMapParams FreeBSD_X86_MemoryMapParams = {
440 };
441 
442 static const PlatformMemoryMapParams NetBSD_X86_MemoryMapParams = {
443  nullptr,
445 };
446 
447 namespace {
448 
449 /// Instrument functions of a module to detect uninitialized reads.
450 ///
451 /// Instantiating MemorySanitizer inserts the msan runtime library API function
452 /// declarations into the module if they don't exist already. Instantiating
453 /// ensures the __msan_init function is in the list of global constructors for
454 /// the module.
455 class MemorySanitizer {
456 public:
457  MemorySanitizer(Module &M, MemorySanitizerOptions Options) {
458  this->CompileKernel =
459  ClEnableKmsan.getNumOccurrences() > 0 ? ClEnableKmsan : Options.Kernel;
460  if (ClTrackOrigins.getNumOccurrences() > 0)
461  this->TrackOrigins = ClTrackOrigins;
462  else
463  this->TrackOrigins = this->CompileKernel ? 2 : Options.TrackOrigins;
464  this->Recover = ClKeepGoing.getNumOccurrences() > 0
465  ? ClKeepGoing
466  : (this->CompileKernel | Options.Recover);
467  initializeModule(M);
468  }
469 
470  // MSan cannot be moved or copied because of MapParams.
471  MemorySanitizer(MemorySanitizer &&) = delete;
472  MemorySanitizer &operator=(MemorySanitizer &&) = delete;
473  MemorySanitizer(const MemorySanitizer &) = delete;
474  MemorySanitizer &operator=(const MemorySanitizer &) = delete;
475 
476  bool sanitizeFunction(Function &F, TargetLibraryInfo &TLI);
477 
478 private:
479  friend struct MemorySanitizerVisitor;
480  friend struct VarArgAMD64Helper;
481  friend struct VarArgMIPS64Helper;
482  friend struct VarArgAArch64Helper;
483  friend struct VarArgPowerPC64Helper;
484 
485  void initializeModule(Module &M);
486  void initializeCallbacks(Module &M);
487  void createKernelApi(Module &M);
488  void createUserspaceApi(Module &M);
489 
490  /// True if we're compiling the Linux kernel.
491  bool CompileKernel;
492  /// Track origins (allocation points) of uninitialized values.
493  int TrackOrigins;
494  bool Recover;
495 
496  LLVMContext *C;
497  Type *IntptrTy;
498  Type *OriginTy;
499 
500  // XxxTLS variables represent the per-thread state in MSan and per-task state
501  // in KMSAN.
502  // For the userspace these point to thread-local globals. In the kernel land
503  // they point to the members of a per-task struct obtained via a call to
504  // __msan_get_context_state().
505 
506  /// Thread-local shadow storage for function parameters.
507  Value *ParamTLS;
508 
509  /// Thread-local origin storage for function parameters.
510  Value *ParamOriginTLS;
511 
512  /// Thread-local shadow storage for function return value.
513  Value *RetvalTLS;
514 
515  /// Thread-local origin storage for function return value.
516  Value *RetvalOriginTLS;
517 
518  /// Thread-local shadow storage for in-register va_arg function
519  /// parameters (x86_64-specific).
520  Value *VAArgTLS;
521 
522  /// Thread-local shadow storage for in-register va_arg function
523  /// parameters (x86_64-specific).
524  Value *VAArgOriginTLS;
525 
526  /// Thread-local shadow storage for va_arg overflow area
527  /// (x86_64-specific).
528  Value *VAArgOverflowSizeTLS;
529 
530  /// Thread-local space used to pass origin value to the UMR reporting
531  /// function.
532  Value *OriginTLS;
533 
534  /// Are the instrumentation callbacks set up?
535  bool CallbacksInitialized = false;
536 
537  /// The run-time callback to print a warning.
538  FunctionCallee WarningFn;
539 
540  // These arrays are indexed by log2(AccessSize).
541  FunctionCallee MaybeWarningFn[kNumberOfAccessSizes];
542  FunctionCallee MaybeStoreOriginFn[kNumberOfAccessSizes];
543 
544  /// Run-time helper that generates a new origin value for a stack
545  /// allocation.
546  FunctionCallee MsanSetAllocaOrigin4Fn;
547 
548  /// Run-time helper that poisons stack on function entry.
549  FunctionCallee MsanPoisonStackFn;
550 
551  /// Run-time helper that records a store (or any event) of an
552  /// uninitialized value and returns an updated origin id encoding this info.
553  FunctionCallee MsanChainOriginFn;
554 
555  /// MSan runtime replacements for memmove, memcpy and memset.
556  FunctionCallee MemmoveFn, MemcpyFn, MemsetFn;
557 
558  /// KMSAN callback for task-local function argument shadow.
559  StructType *MsanContextStateTy;
560  FunctionCallee MsanGetContextStateFn;
561 
562  /// Functions for poisoning/unpoisoning local variables
563  FunctionCallee MsanPoisonAllocaFn, MsanUnpoisonAllocaFn;
564 
565  /// Each of the MsanMetadataPtrXxx functions returns a pair of shadow/origin
566  /// pointers.
567  FunctionCallee MsanMetadataPtrForLoadN, MsanMetadataPtrForStoreN;
568  FunctionCallee MsanMetadataPtrForLoad_1_8[4];
569  FunctionCallee MsanMetadataPtrForStore_1_8[4];
570  FunctionCallee MsanInstrumentAsmStoreFn;
571 
572  /// Helper to choose between different MsanMetadataPtrXxx().
573  FunctionCallee getKmsanShadowOriginAccessFn(bool isStore, int size);
574 
575  /// Memory map parameters used in application-to-shadow calculation.
576  const MemoryMapParams *MapParams;
577 
578  /// Custom memory map parameters used when -msan-shadow-base or
579  // -msan-origin-base is provided.
580  MemoryMapParams CustomMapParams;
581 
582  MDNode *ColdCallWeights;
583 
584  /// Branch weights for origin store.
585  MDNode *OriginStoreWeights;
586 
587  /// An empty volatile inline asm that prevents callback merge.
588  InlineAsm *EmptyAsm;
589 
590  Function *MsanCtorFunction;
591 };
592 
593 /// A legacy function pass for msan instrumentation.
594 ///
595 /// Instruments functions to detect unitialized reads.
596 struct MemorySanitizerLegacyPass : public FunctionPass {
597  // Pass identification, replacement for typeid.
598  static char ID;
599 
600  MemorySanitizerLegacyPass(MemorySanitizerOptions Options = {})
601  : FunctionPass(ID), Options(Options) {}
602  StringRef getPassName() const override { return "MemorySanitizerLegacyPass"; }
603 
604  void getAnalysisUsage(AnalysisUsage &AU) const override {
606  }
607 
608  bool runOnFunction(Function &F) override {
609  return MSan->sanitizeFunction(
610  F, getAnalysis<TargetLibraryInfoWrapperPass>().getTLI());
611  }
612  bool doInitialization(Module &M) override;
613 
615  MemorySanitizerOptions Options;
616 };
617 
618 } // end anonymous namespace
619 
622  MemorySanitizer Msan(*F.getParent(), Options);
623  if (Msan.sanitizeFunction(F, FAM.getResult<TargetLibraryAnalysis>(F)))
624  return PreservedAnalyses::none();
625  return PreservedAnalyses::all();
626 }
627 
629 
630 INITIALIZE_PASS_BEGIN(MemorySanitizerLegacyPass, "msan",
631  "MemorySanitizer: detects uninitialized reads.", false,
632  false)
634 INITIALIZE_PASS_END(MemorySanitizerLegacyPass, "msan",
635  "MemorySanitizer: detects uninitialized reads.", false,
636  false)
637 
638 FunctionPass *
640  return new MemorySanitizerLegacyPass(Options);
641 }
642 
643 /// Create a non-const global initialized with the given string.
644 ///
645 /// Creates a writable global for Str so that we can pass it to the
646 /// run-time lib. Runtime uses first 4 bytes of the string to store the
647 /// frame ID, so the string needs to be mutable.
649  StringRef Str) {
650  Constant *StrConst = ConstantDataArray::getString(M.getContext(), Str);
651  return new GlobalVariable(M, StrConst->getType(), /*isConstant=*/false,
652  GlobalValue::PrivateLinkage, StrConst, "");
653 }
654 
655 /// Create KMSAN API callbacks.
656 void MemorySanitizer::createKernelApi(Module &M) {
657  IRBuilder<> IRB(*C);
658 
659  // These will be initialized in insertKmsanPrologue().
660  RetvalTLS = nullptr;
661  RetvalOriginTLS = nullptr;
662  ParamTLS = nullptr;
663  ParamOriginTLS = nullptr;
664  VAArgTLS = nullptr;
665  VAArgOriginTLS = nullptr;
666  VAArgOverflowSizeTLS = nullptr;
667  // OriginTLS is unused in the kernel.
668  OriginTLS = nullptr;
669 
670  // __msan_warning() in the kernel takes an origin.
671  WarningFn = M.getOrInsertFunction("__msan_warning", IRB.getVoidTy(),
672  IRB.getInt32Ty());
673  // Requests the per-task context state (kmsan_context_state*) from the
674  // runtime library.
675  MsanContextStateTy = StructType::get(
676  ArrayType::get(IRB.getInt64Ty(), kParamTLSSize / 8),
677  ArrayType::get(IRB.getInt64Ty(), kRetvalTLSSize / 8),
678  ArrayType::get(IRB.getInt64Ty(), kParamTLSSize / 8),
679  ArrayType::get(IRB.getInt64Ty(), kParamTLSSize / 8), /* va_arg_origin */
680  IRB.getInt64Ty(), ArrayType::get(OriginTy, kParamTLSSize / 4), OriginTy,
681  OriginTy);
682  MsanGetContextStateFn = M.getOrInsertFunction(
683  "__msan_get_context_state", PointerType::get(MsanContextStateTy, 0));
684 
685  Type *RetTy = StructType::get(PointerType::get(IRB.getInt8Ty(), 0),
686  PointerType::get(IRB.getInt32Ty(), 0));
687 
688  for (int ind = 0, size = 1; ind < 4; ind++, size <<= 1) {
689  std::string name_load =
690  "__msan_metadata_ptr_for_load_" + std::to_string(size);
691  std::string name_store =
692  "__msan_metadata_ptr_for_store_" + std::to_string(size);
693  MsanMetadataPtrForLoad_1_8[ind] = M.getOrInsertFunction(
694  name_load, RetTy, PointerType::get(IRB.getInt8Ty(), 0));
695  MsanMetadataPtrForStore_1_8[ind] = M.getOrInsertFunction(
696  name_store, RetTy, PointerType::get(IRB.getInt8Ty(), 0));
697  }
698 
699  MsanMetadataPtrForLoadN = M.getOrInsertFunction(
700  "__msan_metadata_ptr_for_load_n", RetTy,
701  PointerType::get(IRB.getInt8Ty(), 0), IRB.getInt64Ty());
702  MsanMetadataPtrForStoreN = M.getOrInsertFunction(
703  "__msan_metadata_ptr_for_store_n", RetTy,
704  PointerType::get(IRB.getInt8Ty(), 0), IRB.getInt64Ty());
705 
706  // Functions for poisoning and unpoisoning memory.
707  MsanPoisonAllocaFn =
708  M.getOrInsertFunction("__msan_poison_alloca", IRB.getVoidTy(),
709  IRB.getInt8PtrTy(), IntptrTy, IRB.getInt8PtrTy());
710  MsanUnpoisonAllocaFn = M.getOrInsertFunction(
711  "__msan_unpoison_alloca", IRB.getVoidTy(), IRB.getInt8PtrTy(), IntptrTy);
712 }
713 
715  return M.getOrInsertGlobal(Name, Ty, [&] {
716  return new GlobalVariable(M, Ty, false, GlobalVariable::ExternalLinkage,
717  nullptr, Name, nullptr,
719  });
720 }
721 
722 /// Insert declarations for userspace-specific functions and globals.
723 void MemorySanitizer::createUserspaceApi(Module &M) {
724  IRBuilder<> IRB(*C);
725  // Create the callback.
726  // FIXME: this function should have "Cold" calling conv,
727  // which is not yet implemented.
728  StringRef WarningFnName = Recover ? "__msan_warning"
729  : "__msan_warning_noreturn";
730  WarningFn = M.getOrInsertFunction(WarningFnName, IRB.getVoidTy());
731 
732  // Create the global TLS variables.
733  RetvalTLS =
734  getOrInsertGlobal(M, "__msan_retval_tls",
735  ArrayType::get(IRB.getInt64Ty(), kRetvalTLSSize / 8));
736 
737  RetvalOriginTLS = getOrInsertGlobal(M, "__msan_retval_origin_tls", OriginTy);
738 
739  ParamTLS =
740  getOrInsertGlobal(M, "__msan_param_tls",
741  ArrayType::get(IRB.getInt64Ty(), kParamTLSSize / 8));
742 
743  ParamOriginTLS =
744  getOrInsertGlobal(M, "__msan_param_origin_tls",
745  ArrayType::get(OriginTy, kParamTLSSize / 4));
746 
747  VAArgTLS =
748  getOrInsertGlobal(M, "__msan_va_arg_tls",
749  ArrayType::get(IRB.getInt64Ty(), kParamTLSSize / 8));
750 
751  VAArgOriginTLS =
752  getOrInsertGlobal(M, "__msan_va_arg_origin_tls",
753  ArrayType::get(OriginTy, kParamTLSSize / 4));
754 
755  VAArgOverflowSizeTLS =
756  getOrInsertGlobal(M, "__msan_va_arg_overflow_size_tls", IRB.getInt64Ty());
757  OriginTLS = getOrInsertGlobal(M, "__msan_origin_tls", IRB.getInt32Ty());
758 
759  for (size_t AccessSizeIndex = 0; AccessSizeIndex < kNumberOfAccessSizes;
760  AccessSizeIndex++) {
761  unsigned AccessSize = 1 << AccessSizeIndex;
762  std::string FunctionName = "__msan_maybe_warning_" + itostr(AccessSize);
763  MaybeWarningFn[AccessSizeIndex] = M.getOrInsertFunction(
764  FunctionName, IRB.getVoidTy(), IRB.getIntNTy(AccessSize * 8),
765  IRB.getInt32Ty());
766 
767  FunctionName = "__msan_maybe_store_origin_" + itostr(AccessSize);
768  MaybeStoreOriginFn[AccessSizeIndex] = M.getOrInsertFunction(
769  FunctionName, IRB.getVoidTy(), IRB.getIntNTy(AccessSize * 8),
770  IRB.getInt8PtrTy(), IRB.getInt32Ty());
771  }
772 
773  MsanSetAllocaOrigin4Fn = M.getOrInsertFunction(
774  "__msan_set_alloca_origin4", IRB.getVoidTy(), IRB.getInt8PtrTy(), IntptrTy,
775  IRB.getInt8PtrTy(), IntptrTy);
776  MsanPoisonStackFn =
777  M.getOrInsertFunction("__msan_poison_stack", IRB.getVoidTy(),
778  IRB.getInt8PtrTy(), IntptrTy);
779 }
780 
781 /// Insert extern declaration of runtime-provided functions and globals.
782 void MemorySanitizer::initializeCallbacks(Module &M) {
783  // Only do this once.
784  if (CallbacksInitialized)
785  return;
786 
787  IRBuilder<> IRB(*C);
788  // Initialize callbacks that are common for kernel and userspace
789  // instrumentation.
790  MsanChainOriginFn = M.getOrInsertFunction(
791  "__msan_chain_origin", IRB.getInt32Ty(), IRB.getInt32Ty());
792  MemmoveFn = M.getOrInsertFunction(
793  "__msan_memmove", IRB.getInt8PtrTy(), IRB.getInt8PtrTy(),
794  IRB.getInt8PtrTy(), IntptrTy);
795  MemcpyFn = M.getOrInsertFunction(
796  "__msan_memcpy", IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IRB.getInt8PtrTy(),
797  IntptrTy);
798  MemsetFn = M.getOrInsertFunction(
799  "__msan_memset", IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IRB.getInt32Ty(),
800  IntptrTy);
801  // We insert an empty inline asm after __msan_report* to avoid callback merge.
802  EmptyAsm = InlineAsm::get(FunctionType::get(IRB.getVoidTy(), false),
803  StringRef(""), StringRef(""),
804  /*hasSideEffects=*/true);
805 
806  MsanInstrumentAsmStoreFn =
807  M.getOrInsertFunction("__msan_instrument_asm_store", IRB.getVoidTy(),
808  PointerType::get(IRB.getInt8Ty(), 0), IntptrTy);
809 
810  if (CompileKernel) {
811  createKernelApi(M);
812  } else {
813  createUserspaceApi(M);
814  }
815  CallbacksInitialized = true;
816 }
817 
818 FunctionCallee MemorySanitizer::getKmsanShadowOriginAccessFn(bool isStore,
819  int size) {
820  FunctionCallee *Fns =
821  isStore ? MsanMetadataPtrForStore_1_8 : MsanMetadataPtrForLoad_1_8;
822  switch (size) {
823  case 1:
824  return Fns[0];
825  case 2:
826  return Fns[1];
827  case 4:
828  return Fns[2];
829  case 8:
830  return Fns[3];
831  default:
832  return nullptr;
833  }
834 }
835 
836 /// Module-level initialization.
837 ///
838 /// inserts a call to __msan_init to the module's constructor list.
839 void MemorySanitizer::initializeModule(Module &M) {
840  auto &DL = M.getDataLayout();
841 
842  bool ShadowPassed = ClShadowBase.getNumOccurrences() > 0;
843  bool OriginPassed = ClOriginBase.getNumOccurrences() > 0;
844  // Check the overrides first
845  if (ShadowPassed || OriginPassed) {
846  CustomMapParams.AndMask = ClAndMask;
847  CustomMapParams.XorMask = ClXorMask;
848  CustomMapParams.ShadowBase = ClShadowBase;
849  CustomMapParams.OriginBase = ClOriginBase;
850  MapParams = &CustomMapParams;
851  } else {
852  Triple TargetTriple(M.getTargetTriple());
853  switch (TargetTriple.getOS()) {
854  case Triple::FreeBSD:
855  switch (TargetTriple.getArch()) {
856  case Triple::x86_64:
857  MapParams = FreeBSD_X86_MemoryMapParams.bits64;
858  break;
859  case Triple::x86:
860  MapParams = FreeBSD_X86_MemoryMapParams.bits32;
861  break;
862  default:
863  report_fatal_error("unsupported architecture");
864  }
865  break;
866  case Triple::NetBSD:
867  switch (TargetTriple.getArch()) {
868  case Triple::x86_64:
869  MapParams = NetBSD_X86_MemoryMapParams.bits64;
870  break;
871  default:
872  report_fatal_error("unsupported architecture");
873  }
874  break;
875  case Triple::Linux:
876  switch (TargetTriple.getArch()) {
877  case Triple::x86_64:
878  MapParams = Linux_X86_MemoryMapParams.bits64;
879  break;
880  case Triple::x86:
881  MapParams = Linux_X86_MemoryMapParams.bits32;
882  break;
883  case Triple::mips64:
884  case Triple::mips64el:
885  MapParams = Linux_MIPS_MemoryMapParams.bits64;
886  break;
887  case Triple::ppc64:
888  case Triple::ppc64le:
889  MapParams = Linux_PowerPC_MemoryMapParams.bits64;
890  break;
891  case Triple::aarch64:
892  case Triple::aarch64_be:
893  MapParams = Linux_ARM_MemoryMapParams.bits64;
894  break;
895  default:
896  report_fatal_error("unsupported architecture");
897  }
898  break;
899  default:
900  report_fatal_error("unsupported operating system");
901  }
902  }
903 
904  C = &(M.getContext());
905  IRBuilder<> IRB(*C);
906  IntptrTy = IRB.getIntPtrTy(DL);
907  OriginTy = IRB.getInt32Ty();
908 
909  ColdCallWeights = MDBuilder(*C).createBranchWeights(1, 1000);
910  OriginStoreWeights = MDBuilder(*C).createBranchWeights(1, 1000);
911 
912  if (!CompileKernel) {
913  std::tie(MsanCtorFunction, std::ignore) =
916  /*InitArgTypes=*/{},
917  /*InitArgs=*/{},
918  // This callback is invoked when the functions are created the first
919  // time. Hook them into the global ctors list in that case:
920  [&](Function *Ctor, FunctionCallee) {
921  if (!ClWithComdat) {
922  appendToGlobalCtors(M, Ctor, 0);
923  return;
924  }
925  Comdat *MsanCtorComdat = M.getOrInsertComdat(kMsanModuleCtorName);
926  Ctor->setComdat(MsanCtorComdat);
927  appendToGlobalCtors(M, Ctor, 0, Ctor);
928  });
929 
930  if (TrackOrigins)
931  M.getOrInsertGlobal("__msan_track_origins", IRB.getInt32Ty(), [&] {
932  return new GlobalVariable(
933  M, IRB.getInt32Ty(), true, GlobalValue::WeakODRLinkage,
934  IRB.getInt32(TrackOrigins), "__msan_track_origins");
935  });
936 
937  if (Recover)
938  M.getOrInsertGlobal("__msan_keep_going", IRB.getInt32Ty(), [&] {
939  return new GlobalVariable(M, IRB.getInt32Ty(), true,
941  IRB.getInt32(Recover), "__msan_keep_going");
942  });
943 }
944 }
945 
946 bool MemorySanitizerLegacyPass::doInitialization(Module &M) {
947  MSan.emplace(M, Options);
948  return true;
949 }
950 
951 namespace {
952 
953 /// A helper class that handles instrumentation of VarArg
954 /// functions on a particular platform.
955 ///
956 /// Implementations are expected to insert the instrumentation
957 /// necessary to propagate argument shadow through VarArg function
958 /// calls. Visit* methods are called during an InstVisitor pass over
959 /// the function, and should avoid creating new basic blocks. A new
960 /// instance of this class is created for each instrumented function.
961 struct VarArgHelper {
962  virtual ~VarArgHelper() = default;
963 
964  /// Visit a CallSite.
965  virtual void visitCallSite(CallSite &CS, IRBuilder<> &IRB) = 0;
966 
967  /// Visit a va_start call.
968  virtual void visitVAStartInst(VAStartInst &I) = 0;
969 
970  /// Visit a va_copy call.
971  virtual void visitVACopyInst(VACopyInst &I) = 0;
972 
973  /// Finalize function instrumentation.
974  ///
975  /// This method is called after visiting all interesting (see above)
976  /// instructions in a function.
977  virtual void finalizeInstrumentation() = 0;
978 };
979 
980 struct MemorySanitizerVisitor;
981 
982 } // end anonymous namespace
983 
984 static VarArgHelper *CreateVarArgHelper(Function &Func, MemorySanitizer &Msan,
985  MemorySanitizerVisitor &Visitor);
986 
987 static unsigned TypeSizeToSizeIndex(unsigned TypeSize) {
988  if (TypeSize <= 8) return 0;
989  return Log2_32_Ceil((TypeSize + 7) / 8);
990 }
991 
992 namespace {
993 
994 /// This class does all the work for a given function. Store and Load
995 /// instructions store and load corresponding shadow and origin
996 /// values. Most instructions propagate shadow from arguments to their
997 /// return values. Certain instructions (most importantly, BranchInst)
998 /// test their argument shadow and print reports (with a runtime call) if it's
999 /// non-zero.
1000 struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
1001  Function &F;
1002  MemorySanitizer &MS;
1003  SmallVector<PHINode *, 16> ShadowPHINodes, OriginPHINodes;
1004  ValueMap<Value*, Value*> ShadowMap, OriginMap;
1005  std::unique_ptr<VarArgHelper> VAHelper;
1006  const TargetLibraryInfo *TLI;
1007  BasicBlock *ActualFnStart;
1008 
1009  // The following flags disable parts of MSan instrumentation based on
1010  // blacklist contents and command-line options.
1011  bool InsertChecks;
1012  bool PropagateShadow;
1013  bool PoisonStack;
1014  bool PoisonUndef;
1015  bool CheckReturnValue;
1016 
1017  struct ShadowOriginAndInsertPoint {
1018  Value *Shadow;
1019  Value *Origin;
1020  Instruction *OrigIns;
1021 
1022  ShadowOriginAndInsertPoint(Value *S, Value *O, Instruction *I)
1023  : Shadow(S), Origin(O), OrigIns(I) {}
1024  };
1025  SmallVector<ShadowOriginAndInsertPoint, 16> InstrumentationList;
1026  SmallVector<StoreInst *, 16> StoreList;
1027 
1028  MemorySanitizerVisitor(Function &F, MemorySanitizer &MS,
1029  const TargetLibraryInfo &TLI)
1030  : F(F), MS(MS), VAHelper(CreateVarArgHelper(F, MS, *this)), TLI(&TLI) {
1031  bool SanitizeFunction = F.hasFnAttribute(Attribute::SanitizeMemory);
1032  InsertChecks = SanitizeFunction;
1033  PropagateShadow = SanitizeFunction;
1034  PoisonStack = SanitizeFunction && ClPoisonStack;
1035  PoisonUndef = SanitizeFunction && ClPoisonUndef;
1036  // FIXME: Consider using SpecialCaseList to specify a list of functions that
1037  // must always return fully initialized values. For now, we hardcode "main".
1038  CheckReturnValue = SanitizeFunction && (F.getName() == "main");
1039 
1040  MS.initializeCallbacks(*F.getParent());
1041  if (MS.CompileKernel)
1042  ActualFnStart = insertKmsanPrologue(F);
1043  else
1044  ActualFnStart = &F.getEntryBlock();
1045 
1046  LLVM_DEBUG(if (!InsertChecks) dbgs()
1047  << "MemorySanitizer is not inserting checks into '"
1048  << F.getName() << "'\n");
1049  }
1050 
1051  Value *updateOrigin(Value *V, IRBuilder<> &IRB) {
1052  if (MS.TrackOrigins <= 1) return V;
1053  return IRB.CreateCall(MS.MsanChainOriginFn, V);
1054  }
1055 
1056  Value *originToIntptr(IRBuilder<> &IRB, Value *Origin) {
1057  const DataLayout &DL = F.getParent()->getDataLayout();
1058  unsigned IntptrSize = DL.getTypeStoreSize(MS.IntptrTy);
1059  if (IntptrSize == kOriginSize) return Origin;
1060  assert(IntptrSize == kOriginSize * 2);
1061  Origin = IRB.CreateIntCast(Origin, MS.IntptrTy, /* isSigned */ false);
1062  return IRB.CreateOr(Origin, IRB.CreateShl(Origin, kOriginSize * 8));
1063  }
1064 
1065  /// Fill memory range with the given origin value.
1066  void paintOrigin(IRBuilder<> &IRB, Value *Origin, Value *OriginPtr,
1067  unsigned Size, unsigned Alignment) {
1068  const DataLayout &DL = F.getParent()->getDataLayout();
1069  unsigned IntptrAlignment = DL.getABITypeAlignment(MS.IntptrTy);
1070  unsigned IntptrSize = DL.getTypeStoreSize(MS.IntptrTy);
1071  assert(IntptrAlignment >= kMinOriginAlignment);
1072  assert(IntptrSize >= kOriginSize);
1073 
1074  unsigned Ofs = 0;
1075  unsigned CurrentAlignment = Alignment;
1076  if (Alignment >= IntptrAlignment && IntptrSize > kOriginSize) {
1077  Value *IntptrOrigin = originToIntptr(IRB, Origin);
1078  Value *IntptrOriginPtr =
1079  IRB.CreatePointerCast(OriginPtr, PointerType::get(MS.IntptrTy, 0));
1080  for (unsigned i = 0; i < Size / IntptrSize; ++i) {
1081  Value *Ptr = i ? IRB.CreateConstGEP1_32(MS.IntptrTy, IntptrOriginPtr, i)
1082  : IntptrOriginPtr;
1083  IRB.CreateAlignedStore(IntptrOrigin, Ptr, CurrentAlignment);
1084  Ofs += IntptrSize / kOriginSize;
1085  CurrentAlignment = IntptrAlignment;
1086  }
1087  }
1088 
1089  for (unsigned i = Ofs; i < (Size + kOriginSize - 1) / kOriginSize; ++i) {
1090  Value *GEP =
1091  i ? IRB.CreateConstGEP1_32(MS.OriginTy, OriginPtr, i) : OriginPtr;
1092  IRB.CreateAlignedStore(Origin, GEP, CurrentAlignment);
1093  CurrentAlignment = kMinOriginAlignment;
1094  }
1095  }
1096 
1097  void storeOrigin(IRBuilder<> &IRB, Value *Addr, Value *Shadow, Value *Origin,
1098  Value *OriginPtr, unsigned Alignment, bool AsCall) {
1099  const DataLayout &DL = F.getParent()->getDataLayout();
1100  unsigned OriginAlignment = std::max(kMinOriginAlignment, Alignment);
1101  unsigned StoreSize = DL.getTypeStoreSize(Shadow->getType());
1102  if (Shadow->getType()->isAggregateType()) {
1103  paintOrigin(IRB, updateOrigin(Origin, IRB), OriginPtr, StoreSize,
1104  OriginAlignment);
1105  } else {
1106  Value *ConvertedShadow = convertToShadowTyNoVec(Shadow, IRB);
1107  Constant *ConstantShadow = dyn_cast_or_null<Constant>(ConvertedShadow);
1108  if (ConstantShadow) {
1109  if (ClCheckConstantShadow && !ConstantShadow->isZeroValue())
1110  paintOrigin(IRB, updateOrigin(Origin, IRB), OriginPtr, StoreSize,
1111  OriginAlignment);
1112  return;
1113  }
1114 
1115  unsigned TypeSizeInBits =
1116  DL.getTypeSizeInBits(ConvertedShadow->getType());
1117  unsigned SizeIndex = TypeSizeToSizeIndex(TypeSizeInBits);
1118  if (AsCall && SizeIndex < kNumberOfAccessSizes && !MS.CompileKernel) {
1119  FunctionCallee Fn = MS.MaybeStoreOriginFn[SizeIndex];
1120  Value *ConvertedShadow2 = IRB.CreateZExt(
1121  ConvertedShadow, IRB.getIntNTy(8 * (1 << SizeIndex)));
1122  IRB.CreateCall(Fn, {ConvertedShadow2,
1123  IRB.CreatePointerCast(Addr, IRB.getInt8PtrTy()),
1124  Origin});
1125  } else {
1126  Value *Cmp = IRB.CreateICmpNE(
1127  ConvertedShadow, getCleanShadow(ConvertedShadow), "_mscmp");
1129  Cmp, &*IRB.GetInsertPoint(), false, MS.OriginStoreWeights);
1130  IRBuilder<> IRBNew(CheckTerm);
1131  paintOrigin(IRBNew, updateOrigin(Origin, IRBNew), OriginPtr, StoreSize,
1132  OriginAlignment);
1133  }
1134  }
1135  }
1136 
1137  void materializeStores(bool InstrumentWithCalls) {
1138  for (StoreInst *SI : StoreList) {
1139  IRBuilder<> IRB(SI);
1140  Value *Val = SI->getValueOperand();
1141  Value *Addr = SI->getPointerOperand();
1142  Value *Shadow = SI->isAtomic() ? getCleanShadow(Val) : getShadow(Val);
1143  Value *ShadowPtr, *OriginPtr;
1144  Type *ShadowTy = Shadow->getType();
1145  unsigned Alignment = SI->getAlignment();
1146  unsigned OriginAlignment = std::max(kMinOriginAlignment, Alignment);
1147  std::tie(ShadowPtr, OriginPtr) =
1148  getShadowOriginPtr(Addr, IRB, ShadowTy, Alignment, /*isStore*/ true);
1149 
1150  StoreInst *NewSI = IRB.CreateAlignedStore(Shadow, ShadowPtr, Alignment);
1151  LLVM_DEBUG(dbgs() << " STORE: " << *NewSI << "\n");
1152  (void)NewSI;
1153 
1154  if (SI->isAtomic())
1155  SI->setOrdering(addReleaseOrdering(SI->getOrdering()));
1156 
1157  if (MS.TrackOrigins && !SI->isAtomic())
1158  storeOrigin(IRB, Addr, Shadow, getOrigin(Val), OriginPtr,
1159  OriginAlignment, InstrumentWithCalls);
1160  }
1161  }
1162 
1163  /// Helper function to insert a warning at IRB's current insert point.
1164  void insertWarningFn(IRBuilder<> &IRB, Value *Origin) {
1165  if (!Origin)
1166  Origin = (Value *)IRB.getInt32(0);
1167  if (MS.CompileKernel) {
1168  IRB.CreateCall(MS.WarningFn, Origin);
1169  } else {
1170  if (MS.TrackOrigins) {
1171  IRB.CreateStore(Origin, MS.OriginTLS);
1172  }
1173  IRB.CreateCall(MS.WarningFn, {});
1174  }
1175  IRB.CreateCall(MS.EmptyAsm, {});
1176  // FIXME: Insert UnreachableInst if !MS.Recover?
1177  // This may invalidate some of the following checks and needs to be done
1178  // at the very end.
1179  }
1180 
1181  void materializeOneCheck(Instruction *OrigIns, Value *Shadow, Value *Origin,
1182  bool AsCall) {
1183  IRBuilder<> IRB(OrigIns);
1184  LLVM_DEBUG(dbgs() << " SHAD0 : " << *Shadow << "\n");
1185  Value *ConvertedShadow = convertToShadowTyNoVec(Shadow, IRB);
1186  LLVM_DEBUG(dbgs() << " SHAD1 : " << *ConvertedShadow << "\n");
1187 
1188  Constant *ConstantShadow = dyn_cast_or_null<Constant>(ConvertedShadow);
1189  if (ConstantShadow) {
1190  if (ClCheckConstantShadow && !ConstantShadow->isZeroValue()) {
1191  insertWarningFn(IRB, Origin);
1192  }
1193  return;
1194  }
1195 
1196  const DataLayout &DL = OrigIns->getModule()->getDataLayout();
1197 
1198  unsigned TypeSizeInBits = DL.getTypeSizeInBits(ConvertedShadow->getType());
1199  unsigned SizeIndex = TypeSizeToSizeIndex(TypeSizeInBits);
1200  if (AsCall && SizeIndex < kNumberOfAccessSizes && !MS.CompileKernel) {
1201  FunctionCallee Fn = MS.MaybeWarningFn[SizeIndex];
1202  Value *ConvertedShadow2 =
1203  IRB.CreateZExt(ConvertedShadow, IRB.getIntNTy(8 * (1 << SizeIndex)));
1204  IRB.CreateCall(Fn, {ConvertedShadow2, MS.TrackOrigins && Origin
1205  ? Origin
1206  : (Value *)IRB.getInt32(0)});
1207  } else {
1208  Value *Cmp = IRB.CreateICmpNE(ConvertedShadow,
1209  getCleanShadow(ConvertedShadow), "_mscmp");
1211  Cmp, OrigIns,
1212  /* Unreachable */ !MS.Recover, MS.ColdCallWeights);
1213 
1214  IRB.SetInsertPoint(CheckTerm);
1215  insertWarningFn(IRB, Origin);
1216  LLVM_DEBUG(dbgs() << " CHECK: " << *Cmp << "\n");
1217  }
1218  }
1219 
1220  void materializeChecks(bool InstrumentWithCalls) {
1221  for (const auto &ShadowData : InstrumentationList) {
1222  Instruction *OrigIns = ShadowData.OrigIns;
1223  Value *Shadow = ShadowData.Shadow;
1224  Value *Origin = ShadowData.Origin;
1225  materializeOneCheck(OrigIns, Shadow, Origin, InstrumentWithCalls);
1226  }
1227  LLVM_DEBUG(dbgs() << "DONE:\n" << F);
1228  }
1229 
1230  BasicBlock *insertKmsanPrologue(Function &F) {
1231  BasicBlock *ret =
1234  Value *ContextState = IRB.CreateCall(MS.MsanGetContextStateFn, {});
1235  Constant *Zero = IRB.getInt32(0);
1236  MS.ParamTLS = IRB.CreateGEP(MS.MsanContextStateTy, ContextState,
1237  {Zero, IRB.getInt32(0)}, "param_shadow");
1238  MS.RetvalTLS = IRB.CreateGEP(MS.MsanContextStateTy, ContextState,
1239  {Zero, IRB.getInt32(1)}, "retval_shadow");
1240  MS.VAArgTLS = IRB.CreateGEP(MS.MsanContextStateTy, ContextState,
1241  {Zero, IRB.getInt32(2)}, "va_arg_shadow");
1242  MS.VAArgOriginTLS = IRB.CreateGEP(MS.MsanContextStateTy, ContextState,
1243  {Zero, IRB.getInt32(3)}, "va_arg_origin");
1244  MS.VAArgOverflowSizeTLS =
1245  IRB.CreateGEP(MS.MsanContextStateTy, ContextState,
1246  {Zero, IRB.getInt32(4)}, "va_arg_overflow_size");
1247  MS.ParamOriginTLS = IRB.CreateGEP(MS.MsanContextStateTy, ContextState,
1248  {Zero, IRB.getInt32(5)}, "param_origin");
1249  MS.RetvalOriginTLS =
1250  IRB.CreateGEP(MS.MsanContextStateTy, ContextState,
1251  {Zero, IRB.getInt32(6)}, "retval_origin");
1252  return ret;
1253  }
1254 
1255  /// Add MemorySanitizer instrumentation to a function.
1256  bool runOnFunction() {
1257  // In the presence of unreachable blocks, we may see Phi nodes with
1258  // incoming nodes from such blocks. Since InstVisitor skips unreachable
1259  // blocks, such nodes will not have any shadow value associated with them.
1260  // It's easier to remove unreachable blocks than deal with missing shadow.
1262 
1263  // Iterate all BBs in depth-first order and create shadow instructions
1264  // for all instructions (where applicable).
1265  // For PHI nodes we create dummy shadow PHIs which will be finalized later.
1266  for (BasicBlock *BB : depth_first(ActualFnStart))
1267  visit(*BB);
1268 
1269  // Finalize PHI nodes.
1270  for (PHINode *PN : ShadowPHINodes) {
1271  PHINode *PNS = cast<PHINode>(getShadow(PN));
1272  PHINode *PNO = MS.TrackOrigins ? cast<PHINode>(getOrigin(PN)) : nullptr;
1273  size_t NumValues = PN->getNumIncomingValues();
1274  for (size_t v = 0; v < NumValues; v++) {
1275  PNS->addIncoming(getShadow(PN, v), PN->getIncomingBlock(v));
1276  if (PNO) PNO->addIncoming(getOrigin(PN, v), PN->getIncomingBlock(v));
1277  }
1278  }
1279 
1280  VAHelper->finalizeInstrumentation();
1281 
1282  bool InstrumentWithCalls = ClInstrumentationWithCallThreshold >= 0 &&
1283  InstrumentationList.size() + StoreList.size() >
1285 
1286  // Insert shadow value checks.
1287  materializeChecks(InstrumentWithCalls);
1288 
1289  // Delayed instrumentation of StoreInst.
1290  // This may not add new address checks.
1291  materializeStores(InstrumentWithCalls);
1292 
1293  return true;
1294  }
1295 
1296  /// Compute the shadow type that corresponds to a given Value.
1297  Type *getShadowTy(Value *V) {
1298  return getShadowTy(V->getType());
1299  }
1300 
1301  /// Compute the shadow type that corresponds to a given Type.
1302  Type *getShadowTy(Type *OrigTy) {
1303  if (!OrigTy->isSized()) {
1304  return nullptr;
1305  }
1306  // For integer type, shadow is the same as the original type.
1307  // This may return weird-sized types like i1.
1308  if (IntegerType *IT = dyn_cast<IntegerType>(OrigTy))
1309  return IT;
1310  const DataLayout &DL = F.getParent()->getDataLayout();
1311  if (VectorType *VT = dyn_cast<VectorType>(OrigTy)) {
1312  uint32_t EltSize = DL.getTypeSizeInBits(VT->getElementType());
1313  return VectorType::get(IntegerType::get(*MS.C, EltSize),
1314  VT->getNumElements());
1315  }
1316  if (ArrayType *AT = dyn_cast<ArrayType>(OrigTy)) {
1317  return ArrayType::get(getShadowTy(AT->getElementType()),
1318  AT->getNumElements());
1319  }
1320  if (StructType *ST = dyn_cast<StructType>(OrigTy)) {
1321  SmallVector<Type*, 4> Elements;
1322  for (unsigned i = 0, n = ST->getNumElements(); i < n; i++)
1323  Elements.push_back(getShadowTy(ST->getElementType(i)));
1324  StructType *Res = StructType::get(*MS.C, Elements, ST->isPacked());
1325  LLVM_DEBUG(dbgs() << "getShadowTy: " << *ST << " ===> " << *Res << "\n");
1326  return Res;
1327  }
1328  uint32_t TypeSize = DL.getTypeSizeInBits(OrigTy);
1329  return IntegerType::get(*MS.C, TypeSize);
1330  }
1331 
1332  /// Flatten a vector type.
1333  Type *getShadowTyNoVec(Type *ty) {
1334  if (VectorType *vt = dyn_cast<VectorType>(ty))
1335  return IntegerType::get(*MS.C, vt->getBitWidth());
1336  return ty;
1337  }
1338 
1339  /// Convert a shadow value to it's flattened variant.
1340  Value *convertToShadowTyNoVec(Value *V, IRBuilder<> &IRB) {
1341  Type *Ty = V->getType();
1342  Type *NoVecTy = getShadowTyNoVec(Ty);
1343  if (Ty == NoVecTy) return V;
1344  return IRB.CreateBitCast(V, NoVecTy);
1345  }
1346 
1347  /// Compute the integer shadow offset that corresponds to a given
1348  /// application address.
1349  ///
1350  /// Offset = (Addr & ~AndMask) ^ XorMask
1351  Value *getShadowPtrOffset(Value *Addr, IRBuilder<> &IRB) {
1352  Value *OffsetLong = IRB.CreatePointerCast(Addr, MS.IntptrTy);
1353 
1354  uint64_t AndMask = MS.MapParams->AndMask;
1355  if (AndMask)
1356  OffsetLong =
1357  IRB.CreateAnd(OffsetLong, ConstantInt::get(MS.IntptrTy, ~AndMask));
1358 
1359  uint64_t XorMask = MS.MapParams->XorMask;
1360  if (XorMask)
1361  OffsetLong =
1362  IRB.CreateXor(OffsetLong, ConstantInt::get(MS.IntptrTy, XorMask));
1363  return OffsetLong;
1364  }
1365 
1366  /// Compute the shadow and origin addresses corresponding to a given
1367  /// application address.
1368  ///
1369  /// Shadow = ShadowBase + Offset
1370  /// Origin = (OriginBase + Offset) & ~3ULL
1371  std::pair<Value *, Value *> getShadowOriginPtrUserspace(Value *Addr,
1372  IRBuilder<> &IRB,
1373  Type *ShadowTy,
1374  unsigned Alignment) {
1375  Value *ShadowOffset = getShadowPtrOffset(Addr, IRB);
1376  Value *ShadowLong = ShadowOffset;
1377  uint64_t ShadowBase = MS.MapParams->ShadowBase;
1378  if (ShadowBase != 0) {
1379  ShadowLong =
1380  IRB.CreateAdd(ShadowLong,
1381  ConstantInt::get(MS.IntptrTy, ShadowBase));
1382  }
1383  Value *ShadowPtr =
1384  IRB.CreateIntToPtr(ShadowLong, PointerType::get(ShadowTy, 0));
1385  Value *OriginPtr = nullptr;
1386  if (MS.TrackOrigins) {
1387  Value *OriginLong = ShadowOffset;
1388  uint64_t OriginBase = MS.MapParams->OriginBase;
1389  if (OriginBase != 0)
1390  OriginLong = IRB.CreateAdd(OriginLong,
1391  ConstantInt::get(MS.IntptrTy, OriginBase));
1392  if (Alignment < kMinOriginAlignment) {
1393  uint64_t Mask = kMinOriginAlignment - 1;
1394  OriginLong =
1395  IRB.CreateAnd(OriginLong, ConstantInt::get(MS.IntptrTy, ~Mask));
1396  }
1397  OriginPtr =
1398  IRB.CreateIntToPtr(OriginLong, PointerType::get(MS.OriginTy, 0));
1399  }
1400  return std::make_pair(ShadowPtr, OriginPtr);
1401  }
1402 
1403  std::pair<Value *, Value *>
1404  getShadowOriginPtrKernel(Value *Addr, IRBuilder<> &IRB, Type *ShadowTy,
1405  unsigned Alignment, bool isStore) {
1406  Value *ShadowOriginPtrs;
1407  const DataLayout &DL = F.getParent()->getDataLayout();
1408  int Size = DL.getTypeStoreSize(ShadowTy);
1409 
1410  FunctionCallee Getter = MS.getKmsanShadowOriginAccessFn(isStore, Size);
1411  Value *AddrCast =
1412  IRB.CreatePointerCast(Addr, PointerType::get(IRB.getInt8Ty(), 0));
1413  if (Getter) {
1414  ShadowOriginPtrs = IRB.CreateCall(Getter, AddrCast);
1415  } else {
1416  Value *SizeVal = ConstantInt::get(MS.IntptrTy, Size);
1417  ShadowOriginPtrs = IRB.CreateCall(isStore ? MS.MsanMetadataPtrForStoreN
1418  : MS.MsanMetadataPtrForLoadN,
1419  {AddrCast, SizeVal});
1420  }
1421  Value *ShadowPtr = IRB.CreateExtractValue(ShadowOriginPtrs, 0);
1422  ShadowPtr = IRB.CreatePointerCast(ShadowPtr, PointerType::get(ShadowTy, 0));
1423  Value *OriginPtr = IRB.CreateExtractValue(ShadowOriginPtrs, 1);
1424 
1425  return std::make_pair(ShadowPtr, OriginPtr);
1426  }
1427 
1428  std::pair<Value *, Value *> getShadowOriginPtr(Value *Addr, IRBuilder<> &IRB,
1429  Type *ShadowTy,
1430  unsigned Alignment,
1431  bool isStore) {
1432  std::pair<Value *, Value *> ret;
1433  if (MS.CompileKernel)
1434  ret = getShadowOriginPtrKernel(Addr, IRB, ShadowTy, Alignment, isStore);
1435  else
1436  ret = getShadowOriginPtrUserspace(Addr, IRB, ShadowTy, Alignment);
1437  return ret;
1438  }
1439 
1440  /// Compute the shadow address for a given function argument.
1441  ///
1442  /// Shadow = ParamTLS+ArgOffset.
1443  Value *getShadowPtrForArgument(Value *A, IRBuilder<> &IRB,
1444  int ArgOffset) {
1445  Value *Base = IRB.CreatePointerCast(MS.ParamTLS, MS.IntptrTy);
1446  if (ArgOffset)
1447  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
1448  return IRB.CreateIntToPtr(Base, PointerType::get(getShadowTy(A), 0),
1449  "_msarg");
1450  }
1451 
1452  /// Compute the origin address for a given function argument.
1453  Value *getOriginPtrForArgument(Value *A, IRBuilder<> &IRB,
1454  int ArgOffset) {
1455  if (!MS.TrackOrigins)
1456  return nullptr;
1457  Value *Base = IRB.CreatePointerCast(MS.ParamOriginTLS, MS.IntptrTy);
1458  if (ArgOffset)
1459  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
1460  return IRB.CreateIntToPtr(Base, PointerType::get(MS.OriginTy, 0),
1461  "_msarg_o");
1462  }
1463 
1464  /// Compute the shadow address for a retval.
1465  Value *getShadowPtrForRetval(Value *A, IRBuilder<> &IRB) {
1466  return IRB.CreatePointerCast(MS.RetvalTLS,
1467  PointerType::get(getShadowTy(A), 0),
1468  "_msret");
1469  }
1470 
1471  /// Compute the origin address for a retval.
1472  Value *getOriginPtrForRetval(IRBuilder<> &IRB) {
1473  // We keep a single origin for the entire retval. Might be too optimistic.
1474  return MS.RetvalOriginTLS;
1475  }
1476 
1477  /// Set SV to be the shadow value for V.
1478  void setShadow(Value *V, Value *SV) {
1479  assert(!ShadowMap.count(V) && "Values may only have one shadow");
1480  ShadowMap[V] = PropagateShadow ? SV : getCleanShadow(V);
1481  }
1482 
1483  /// Set Origin to be the origin value for V.
1484  void setOrigin(Value *V, Value *Origin) {
1485  if (!MS.TrackOrigins) return;
1486  assert(!OriginMap.count(V) && "Values may only have one origin");
1487  LLVM_DEBUG(dbgs() << "ORIGIN: " << *V << " ==> " << *Origin << "\n");
1488  OriginMap[V] = Origin;
1489  }
1490 
1491  Constant *getCleanShadow(Type *OrigTy) {
1492  Type *ShadowTy = getShadowTy(OrigTy);
1493  if (!ShadowTy)
1494  return nullptr;
1495  return Constant::getNullValue(ShadowTy);
1496  }
1497 
1498  /// Create a clean shadow value for a given value.
1499  ///
1500  /// Clean shadow (all zeroes) means all bits of the value are defined
1501  /// (initialized).
1502  Constant *getCleanShadow(Value *V) {
1503  return getCleanShadow(V->getType());
1504  }
1505 
1506  /// Create a dirty shadow of a given shadow type.
1507  Constant *getPoisonedShadow(Type *ShadowTy) {
1508  assert(ShadowTy);
1509  if (isa<IntegerType>(ShadowTy) || isa<VectorType>(ShadowTy))
1510  return Constant::getAllOnesValue(ShadowTy);
1511  if (ArrayType *AT = dyn_cast<ArrayType>(ShadowTy)) {
1512  SmallVector<Constant *, 4> Vals(AT->getNumElements(),
1513  getPoisonedShadow(AT->getElementType()));
1514  return ConstantArray::get(AT, Vals);
1515  }
1516  if (StructType *ST = dyn_cast<StructType>(ShadowTy)) {
1518  for (unsigned i = 0, n = ST->getNumElements(); i < n; i++)
1519  Vals.push_back(getPoisonedShadow(ST->getElementType(i)));
1520  return ConstantStruct::get(ST, Vals);
1521  }
1522  llvm_unreachable("Unexpected shadow type");
1523  }
1524 
1525  /// Create a dirty shadow for a given value.
1526  Constant *getPoisonedShadow(Value *V) {
1527  Type *ShadowTy = getShadowTy(V);
1528  if (!ShadowTy)
1529  return nullptr;
1530  return getPoisonedShadow(ShadowTy);
1531  }
1532 
1533  /// Create a clean (zero) origin.
1534  Value *getCleanOrigin() {
1535  return Constant::getNullValue(MS.OriginTy);
1536  }
1537 
1538  /// Get the shadow value for a given Value.
1539  ///
1540  /// This function either returns the value set earlier with setShadow,
1541  /// or extracts if from ParamTLS (for function arguments).
1542  Value *getShadow(Value *V) {
1543  if (!PropagateShadow) return getCleanShadow(V);
1544  if (Instruction *I = dyn_cast<Instruction>(V)) {
1545  if (I->getMetadata("nosanitize"))
1546  return getCleanShadow(V);
1547  // For instructions the shadow is already stored in the map.
1548  Value *Shadow = ShadowMap[V];
1549  if (!Shadow) {
1550  LLVM_DEBUG(dbgs() << "No shadow: " << *V << "\n" << *(I->getParent()));
1551  (void)I;
1552  assert(Shadow && "No shadow for a value");
1553  }
1554  return Shadow;
1555  }
1556  if (UndefValue *U = dyn_cast<UndefValue>(V)) {
1557  Value *AllOnes = PoisonUndef ? getPoisonedShadow(V) : getCleanShadow(V);
1558  LLVM_DEBUG(dbgs() << "Undef: " << *U << " ==> " << *AllOnes << "\n");
1559  (void)U;
1560  return AllOnes;
1561  }
1562  if (Argument *A = dyn_cast<Argument>(V)) {
1563  // For arguments we compute the shadow on demand and store it in the map.
1564  Value **ShadowPtr = &ShadowMap[V];
1565  if (*ShadowPtr)
1566  return *ShadowPtr;
1567  Function *F = A->getParent();
1568  IRBuilder<> EntryIRB(ActualFnStart->getFirstNonPHI());
1569  unsigned ArgOffset = 0;
1570  const DataLayout &DL = F->getParent()->getDataLayout();
1571  for (auto &FArg : F->args()) {
1572  if (!FArg.getType()->isSized()) {
1573  LLVM_DEBUG(dbgs() << "Arg is not sized\n");
1574  continue;
1575  }
1576  unsigned Size =
1577  FArg.hasByValAttr()
1578  ? DL.getTypeAllocSize(FArg.getType()->getPointerElementType())
1579  : DL.getTypeAllocSize(FArg.getType());
1580  if (A == &FArg) {
1581  bool Overflow = ArgOffset + Size > kParamTLSSize;
1582  Value *Base = getShadowPtrForArgument(&FArg, EntryIRB, ArgOffset);
1583  if (FArg.hasByValAttr()) {
1584  // ByVal pointer itself has clean shadow. We copy the actual
1585  // argument shadow to the underlying memory.
1586  // Figure out maximal valid memcpy alignment.
1587  unsigned ArgAlign = FArg.getParamAlignment();
1588  if (ArgAlign == 0) {
1589  Type *EltType = A->getType()->getPointerElementType();
1590  ArgAlign = DL.getABITypeAlignment(EltType);
1591  }
1592  Value *CpShadowPtr =
1593  getShadowOriginPtr(V, EntryIRB, EntryIRB.getInt8Ty(), ArgAlign,
1594  /*isStore*/ true)
1595  .first;
1596  // TODO(glider): need to copy origins.
1597  if (Overflow) {
1598  // ParamTLS overflow.
1599  EntryIRB.CreateMemSet(
1600  CpShadowPtr, Constant::getNullValue(EntryIRB.getInt8Ty()),
1601  Size, ArgAlign);
1602  } else {
1603  unsigned CopyAlign = std::min(ArgAlign, kShadowTLSAlignment);
1604  Value *Cpy = EntryIRB.CreateMemCpy(CpShadowPtr, CopyAlign, Base,
1605  CopyAlign, Size);
1606  LLVM_DEBUG(dbgs() << " ByValCpy: " << *Cpy << "\n");
1607  (void)Cpy;
1608  }
1609  *ShadowPtr = getCleanShadow(V);
1610  } else {
1611  if (Overflow) {
1612  // ParamTLS overflow.
1613  *ShadowPtr = getCleanShadow(V);
1614  } else {
1615  *ShadowPtr = EntryIRB.CreateAlignedLoad(getShadowTy(&FArg), Base,
1616  kShadowTLSAlignment);
1617  }
1618  }
1619  LLVM_DEBUG(dbgs()
1620  << " ARG: " << FArg << " ==> " << **ShadowPtr << "\n");
1621  if (MS.TrackOrigins && !Overflow) {
1622  Value *OriginPtr =
1623  getOriginPtrForArgument(&FArg, EntryIRB, ArgOffset);
1624  setOrigin(A, EntryIRB.CreateLoad(MS.OriginTy, OriginPtr));
1625  } else {
1626  setOrigin(A, getCleanOrigin());
1627  }
1628  }
1629  ArgOffset += alignTo(Size, kShadowTLSAlignment);
1630  }
1631  assert(*ShadowPtr && "Could not find shadow for an argument");
1632  return *ShadowPtr;
1633  }
1634  // For everything else the shadow is zero.
1635  return getCleanShadow(V);
1636  }
1637 
1638  /// Get the shadow for i-th argument of the instruction I.
1639  Value *getShadow(Instruction *I, int i) {
1640  return getShadow(I->getOperand(i));
1641  }
1642 
1643  /// Get the origin for a value.
1644  Value *getOrigin(Value *V) {
1645  if (!MS.TrackOrigins) return nullptr;
1646  if (!PropagateShadow) return getCleanOrigin();
1647  if (isa<Constant>(V)) return getCleanOrigin();
1648  assert((isa<Instruction>(V) || isa<Argument>(V)) &&
1649  "Unexpected value type in getOrigin()");
1650  if (Instruction *I = dyn_cast<Instruction>(V)) {
1651  if (I->getMetadata("nosanitize"))
1652  return getCleanOrigin();
1653  }
1654  Value *Origin = OriginMap[V];
1655  assert(Origin && "Missing origin");
1656  return Origin;
1657  }
1658 
1659  /// Get the origin for i-th argument of the instruction I.
1660  Value *getOrigin(Instruction *I, int i) {
1661  return getOrigin(I->getOperand(i));
1662  }
1663 
1664  /// Remember the place where a shadow check should be inserted.
1665  ///
1666  /// This location will be later instrumented with a check that will print a
1667  /// UMR warning in runtime if the shadow value is not 0.
1668  void insertShadowCheck(Value *Shadow, Value *Origin, Instruction *OrigIns) {
1669  assert(Shadow);
1670  if (!InsertChecks) return;
1671 #ifndef NDEBUG
1672  Type *ShadowTy = Shadow->getType();
1673  assert((isa<IntegerType>(ShadowTy) || isa<VectorType>(ShadowTy)) &&
1674  "Can only insert checks for integer and vector shadow types");
1675 #endif
1676  InstrumentationList.push_back(
1677  ShadowOriginAndInsertPoint(Shadow, Origin, OrigIns));
1678  }
1679 
1680  /// Remember the place where a shadow check should be inserted.
1681  ///
1682  /// This location will be later instrumented with a check that will print a
1683  /// UMR warning in runtime if the value is not fully defined.
1684  void insertShadowCheck(Value *Val, Instruction *OrigIns) {
1685  assert(Val);
1686  Value *Shadow, *Origin;
1687  if (ClCheckConstantShadow) {
1688  Shadow = getShadow(Val);
1689  if (!Shadow) return;
1690  Origin = getOrigin(Val);
1691  } else {
1692  Shadow = dyn_cast_or_null<Instruction>(getShadow(Val));
1693  if (!Shadow) return;
1694  Origin = dyn_cast_or_null<Instruction>(getOrigin(Val));
1695  }
1696  insertShadowCheck(Shadow, Origin, OrigIns);
1697  }
1698 
1699  AtomicOrdering addReleaseOrdering(AtomicOrdering a) {
1700  switch (a) {
1706  return AtomicOrdering::Release;
1712  }
1713  llvm_unreachable("Unknown ordering");
1714  }
1715 
1716  AtomicOrdering addAcquireOrdering(AtomicOrdering a) {
1717  switch (a) {
1723  return AtomicOrdering::Acquire;
1729  }
1730  llvm_unreachable("Unknown ordering");
1731  }
1732 
1733  // ------------------- Visitors.
1735  void visit(Instruction &I) {
1736  if (!I.getMetadata("nosanitize"))
1738  }
1739 
1740  /// Instrument LoadInst
1741  ///
1742  /// Loads the corresponding shadow and (optionally) origin.
1743  /// Optionally, checks that the load address is fully defined.
1744  void visitLoadInst(LoadInst &I) {
1745  assert(I.getType()->isSized() && "Load type must have size");
1746  assert(!I.getMetadata("nosanitize"));
1747  IRBuilder<> IRB(I.getNextNode());
1748  Type *ShadowTy = getShadowTy(&I);
1749  Value *Addr = I.getPointerOperand();
1750  Value *ShadowPtr, *OriginPtr;
1751  unsigned Alignment = I.getAlignment();
1752  if (PropagateShadow) {
1753  std::tie(ShadowPtr, OriginPtr) =
1754  getShadowOriginPtr(Addr, IRB, ShadowTy, Alignment, /*isStore*/ false);
1755  setShadow(&I,
1756  IRB.CreateAlignedLoad(ShadowTy, ShadowPtr, Alignment, "_msld"));
1757  } else {
1758  setShadow(&I, getCleanShadow(&I));
1759  }
1760 
1762  insertShadowCheck(I.getPointerOperand(), &I);
1763 
1764  if (I.isAtomic())
1765  I.setOrdering(addAcquireOrdering(I.getOrdering()));
1766 
1767  if (MS.TrackOrigins) {
1768  if (PropagateShadow) {
1769  unsigned OriginAlignment = std::max(kMinOriginAlignment, Alignment);
1770  setOrigin(
1771  &I, IRB.CreateAlignedLoad(MS.OriginTy, OriginPtr, OriginAlignment));
1772  } else {
1773  setOrigin(&I, getCleanOrigin());
1774  }
1775  }
1776  }
1777 
1778  /// Instrument StoreInst
1779  ///
1780  /// Stores the corresponding shadow and (optionally) origin.
1781  /// Optionally, checks that the store address is fully defined.
1782  void visitStoreInst(StoreInst &I) {
1783  StoreList.push_back(&I);
1785  insertShadowCheck(I.getPointerOperand(), &I);
1786  }
1787 
1788  void handleCASOrRMW(Instruction &I) {
1789  assert(isa<AtomicRMWInst>(I) || isa<AtomicCmpXchgInst>(I));
1790 
1791  IRBuilder<> IRB(&I);
1792  Value *Addr = I.getOperand(0);
1793  Value *ShadowPtr = getShadowOriginPtr(Addr, IRB, I.getType(),
1794  /*Alignment*/ 1, /*isStore*/ true)
1795  .first;
1796 
1798  insertShadowCheck(Addr, &I);
1799 
1800  // Only test the conditional argument of cmpxchg instruction.
1801  // The other argument can potentially be uninitialized, but we can not
1802  // detect this situation reliably without possible false positives.
1803  if (isa<AtomicCmpXchgInst>(I))
1804  insertShadowCheck(I.getOperand(1), &I);
1805 
1806  IRB.CreateStore(getCleanShadow(&I), ShadowPtr);
1807 
1808  setShadow(&I, getCleanShadow(&I));
1809  setOrigin(&I, getCleanOrigin());
1810  }
1811 
1812  void visitAtomicRMWInst(AtomicRMWInst &I) {
1813  handleCASOrRMW(I);
1814  I.setOrdering(addReleaseOrdering(I.getOrdering()));
1815  }
1816 
1817  void visitAtomicCmpXchgInst(AtomicCmpXchgInst &I) {
1818  handleCASOrRMW(I);
1819  I.setSuccessOrdering(addReleaseOrdering(I.getSuccessOrdering()));
1820  }
1821 
1822  // Vector manipulation.
1823  void visitExtractElementInst(ExtractElementInst &I) {
1824  insertShadowCheck(I.getOperand(1), &I);
1825  IRBuilder<> IRB(&I);
1826  setShadow(&I, IRB.CreateExtractElement(getShadow(&I, 0), I.getOperand(1),
1827  "_msprop"));
1828  setOrigin(&I, getOrigin(&I, 0));
1829  }
1830 
1831  void visitInsertElementInst(InsertElementInst &I) {
1832  insertShadowCheck(I.getOperand(2), &I);
1833  IRBuilder<> IRB(&I);
1834  setShadow(&I, IRB.CreateInsertElement(getShadow(&I, 0), getShadow(&I, 1),
1835  I.getOperand(2), "_msprop"));
1836  setOriginForNaryOp(I);
1837  }
1838 
1839  void visitShuffleVectorInst(ShuffleVectorInst &I) {
1840  insertShadowCheck(I.getOperand(2), &I);
1841  IRBuilder<> IRB(&I);
1842  setShadow(&I, IRB.CreateShuffleVector(getShadow(&I, 0), getShadow(&I, 1),
1843  I.getOperand(2), "_msprop"));
1844  setOriginForNaryOp(I);
1845  }
1846 
1847  // Casts.
1848  void visitSExtInst(SExtInst &I) {
1849  IRBuilder<> IRB(&I);
1850  setShadow(&I, IRB.CreateSExt(getShadow(&I, 0), I.getType(), "_msprop"));
1851  setOrigin(&I, getOrigin(&I, 0));
1852  }
1853 
1854  void visitZExtInst(ZExtInst &I) {
1855  IRBuilder<> IRB(&I);
1856  setShadow(&I, IRB.CreateZExt(getShadow(&I, 0), I.getType(), "_msprop"));
1857  setOrigin(&I, getOrigin(&I, 0));
1858  }
1859 
1860  void visitTruncInst(TruncInst &I) {
1861  IRBuilder<> IRB(&I);
1862  setShadow(&I, IRB.CreateTrunc(getShadow(&I, 0), I.getType(), "_msprop"));
1863  setOrigin(&I, getOrigin(&I, 0));
1864  }
1865 
1866  void visitBitCastInst(BitCastInst &I) {
1867  // Special case: if this is the bitcast (there is exactly 1 allowed) between
1868  // a musttail call and a ret, don't instrument. New instructions are not
1869  // allowed after a musttail call.
1870  if (auto *CI = dyn_cast<CallInst>(I.getOperand(0)))
1871  if (CI->isMustTailCall())
1872  return;
1873  IRBuilder<> IRB(&I);
1874  setShadow(&I, IRB.CreateBitCast(getShadow(&I, 0), getShadowTy(&I)));
1875  setOrigin(&I, getOrigin(&I, 0));
1876  }
1877 
1878  void visitPtrToIntInst(PtrToIntInst &I) {
1879  IRBuilder<> IRB(&I);
1880  setShadow(&I, IRB.CreateIntCast(getShadow(&I, 0), getShadowTy(&I), false,
1881  "_msprop_ptrtoint"));
1882  setOrigin(&I, getOrigin(&I, 0));
1883  }
1884 
1885  void visitIntToPtrInst(IntToPtrInst &I) {
1886  IRBuilder<> IRB(&I);
1887  setShadow(&I, IRB.CreateIntCast(getShadow(&I, 0), getShadowTy(&I), false,
1888  "_msprop_inttoptr"));
1889  setOrigin(&I, getOrigin(&I, 0));
1890  }
1891 
1892  void visitFPToSIInst(CastInst& I) { handleShadowOr(I); }
1893  void visitFPToUIInst(CastInst& I) { handleShadowOr(I); }
1894  void visitSIToFPInst(CastInst& I) { handleShadowOr(I); }
1895  void visitUIToFPInst(CastInst& I) { handleShadowOr(I); }
1896  void visitFPExtInst(CastInst& I) { handleShadowOr(I); }
1897  void visitFPTruncInst(CastInst& I) { handleShadowOr(I); }
1898 
1899  /// Propagate shadow for bitwise AND.
1900  ///
1901  /// This code is exact, i.e. if, for example, a bit in the left argument
1902  /// is defined and 0, then neither the value not definedness of the
1903  /// corresponding bit in B don't affect the resulting shadow.
1904  void visitAnd(BinaryOperator &I) {
1905  IRBuilder<> IRB(&I);
1906  // "And" of 0 and a poisoned value results in unpoisoned value.
1907  // 1&1 => 1; 0&1 => 0; p&1 => p;
1908  // 1&0 => 0; 0&0 => 0; p&0 => 0;
1909  // 1&p => p; 0&p => 0; p&p => p;
1910  // S = (S1 & S2) | (V1 & S2) | (S1 & V2)
1911  Value *S1 = getShadow(&I, 0);
1912  Value *S2 = getShadow(&I, 1);
1913  Value *V1 = I.getOperand(0);
1914  Value *V2 = I.getOperand(1);
1915  if (V1->getType() != S1->getType()) {
1916  V1 = IRB.CreateIntCast(V1, S1->getType(), false);
1917  V2 = IRB.CreateIntCast(V2, S2->getType(), false);
1918  }
1919  Value *S1S2 = IRB.CreateAnd(S1, S2);
1920  Value *V1S2 = IRB.CreateAnd(V1, S2);
1921  Value *S1V2 = IRB.CreateAnd(S1, V2);
1922  setShadow(&I, IRB.CreateOr(S1S2, IRB.CreateOr(V1S2, S1V2)));
1923  setOriginForNaryOp(I);
1924  }
1925 
1926  void visitOr(BinaryOperator &I) {
1927  IRBuilder<> IRB(&I);
1928  // "Or" of 1 and a poisoned value results in unpoisoned value.
1929  // 1|1 => 1; 0|1 => 1; p|1 => 1;
1930  // 1|0 => 1; 0|0 => 0; p|0 => p;
1931  // 1|p => 1; 0|p => p; p|p => p;
1932  // S = (S1 & S2) | (~V1 & S2) | (S1 & ~V2)
1933  Value *S1 = getShadow(&I, 0);
1934  Value *S2 = getShadow(&I, 1);
1935  Value *V1 = IRB.CreateNot(I.getOperand(0));
1936  Value *V2 = IRB.CreateNot(I.getOperand(1));
1937  if (V1->getType() != S1->getType()) {
1938  V1 = IRB.CreateIntCast(V1, S1->getType(), false);
1939  V2 = IRB.CreateIntCast(V2, S2->getType(), false);
1940  }
1941  Value *S1S2 = IRB.CreateAnd(S1, S2);
1942  Value *V1S2 = IRB.CreateAnd(V1, S2);
1943  Value *S1V2 = IRB.CreateAnd(S1, V2);
1944  setShadow(&I, IRB.CreateOr(S1S2, IRB.CreateOr(V1S2, S1V2)));
1945  setOriginForNaryOp(I);
1946  }
1947 
1948  /// Default propagation of shadow and/or origin.
1949  ///
1950  /// This class implements the general case of shadow propagation, used in all
1951  /// cases where we don't know and/or don't care about what the operation
1952  /// actually does. It converts all input shadow values to a common type
1953  /// (extending or truncating as necessary), and bitwise OR's them.
1954  ///
1955  /// This is much cheaper than inserting checks (i.e. requiring inputs to be
1956  /// fully initialized), and less prone to false positives.
1957  ///
1958  /// This class also implements the general case of origin propagation. For a
1959  /// Nary operation, result origin is set to the origin of an argument that is
1960  /// not entirely initialized. If there is more than one such arguments, the
1961  /// rightmost of them is picked. It does not matter which one is picked if all
1962  /// arguments are initialized.
1963  template <bool CombineShadow>
1964  class Combiner {
1965  Value *Shadow = nullptr;
1966  Value *Origin = nullptr;
1967  IRBuilder<> &IRB;
1968  MemorySanitizerVisitor *MSV;
1969 
1970  public:
1971  Combiner(MemorySanitizerVisitor *MSV, IRBuilder<> &IRB)
1972  : IRB(IRB), MSV(MSV) {}
1973 
1974  /// Add a pair of shadow and origin values to the mix.
1975  Combiner &Add(Value *OpShadow, Value *OpOrigin) {
1976  if (CombineShadow) {
1977  assert(OpShadow);
1978  if (!Shadow)
1979  Shadow = OpShadow;
1980  else {
1981  OpShadow = MSV->CreateShadowCast(IRB, OpShadow, Shadow->getType());
1982  Shadow = IRB.CreateOr(Shadow, OpShadow, "_msprop");
1983  }
1984  }
1985 
1986  if (MSV->MS.TrackOrigins) {
1987  assert(OpOrigin);
1988  if (!Origin) {
1989  Origin = OpOrigin;
1990  } else {
1991  Constant *ConstOrigin = dyn_cast<Constant>(OpOrigin);
1992  // No point in adding something that might result in 0 origin value.
1993  if (!ConstOrigin || !ConstOrigin->isNullValue()) {
1994  Value *FlatShadow = MSV->convertToShadowTyNoVec(OpShadow, IRB);
1995  Value *Cond =
1996  IRB.CreateICmpNE(FlatShadow, MSV->getCleanShadow(FlatShadow));
1997  Origin = IRB.CreateSelect(Cond, OpOrigin, Origin);
1998  }
1999  }
2000  }
2001  return *this;
2002  }
2003 
2004  /// Add an application value to the mix.
2005  Combiner &Add(Value *V) {
2006  Value *OpShadow = MSV->getShadow(V);
2007  Value *OpOrigin = MSV->MS.TrackOrigins ? MSV->getOrigin(V) : nullptr;
2008  return Add(OpShadow, OpOrigin);
2009  }
2010 
2011  /// Set the current combined values as the given instruction's shadow
2012  /// and origin.
2013  void Done(Instruction *I) {
2014  if (CombineShadow) {
2015  assert(Shadow);
2016  Shadow = MSV->CreateShadowCast(IRB, Shadow, MSV->getShadowTy(I));
2017  MSV->setShadow(I, Shadow);
2018  }
2019  if (MSV->MS.TrackOrigins) {
2020  assert(Origin);
2021  MSV->setOrigin(I, Origin);
2022  }
2023  }
2024  };
2025 
2026  using ShadowAndOriginCombiner = Combiner<true>;
2027  using OriginCombiner = Combiner<false>;
2028 
2029  /// Propagate origin for arbitrary operation.
2030  void setOriginForNaryOp(Instruction &I) {
2031  if (!MS.TrackOrigins) return;
2032  IRBuilder<> IRB(&I);
2033  OriginCombiner OC(this, IRB);
2034  for (Instruction::op_iterator OI = I.op_begin(); OI != I.op_end(); ++OI)
2035  OC.Add(OI->get());
2036  OC.Done(&I);
2037  }
2038 
2039  size_t VectorOrPrimitiveTypeSizeInBits(Type *Ty) {
2040  assert(!(Ty->isVectorTy() && Ty->getScalarType()->isPointerTy()) &&
2041  "Vector of pointers is not a valid shadow type");
2042  return Ty->isVectorTy() ?
2044  Ty->getPrimitiveSizeInBits();
2045  }
2046 
2047  /// Cast between two shadow types, extending or truncating as
2048  /// necessary.
2049  Value *CreateShadowCast(IRBuilder<> &IRB, Value *V, Type *dstTy,
2050  bool Signed = false) {
2051  Type *srcTy = V->getType();
2052  size_t srcSizeInBits = VectorOrPrimitiveTypeSizeInBits(srcTy);
2053  size_t dstSizeInBits = VectorOrPrimitiveTypeSizeInBits(dstTy);
2054  if (srcSizeInBits > 1 && dstSizeInBits == 1)
2055  return IRB.CreateICmpNE(V, getCleanShadow(V));
2056 
2057  if (dstTy->isIntegerTy() && srcTy->isIntegerTy())
2058  return IRB.CreateIntCast(V, dstTy, Signed);
2059  if (dstTy->isVectorTy() && srcTy->isVectorTy() &&
2060  dstTy->getVectorNumElements() == srcTy->getVectorNumElements())
2061  return IRB.CreateIntCast(V, dstTy, Signed);
2062  Value *V1 = IRB.CreateBitCast(V, Type::getIntNTy(*MS.C, srcSizeInBits));
2063  Value *V2 =
2064  IRB.CreateIntCast(V1, Type::getIntNTy(*MS.C, dstSizeInBits), Signed);
2065  return IRB.CreateBitCast(V2, dstTy);
2066  // TODO: handle struct types.
2067  }
2068 
2069  /// Cast an application value to the type of its own shadow.
2070  Value *CreateAppToShadowCast(IRBuilder<> &IRB, Value *V) {
2071  Type *ShadowTy = getShadowTy(V);
2072  if (V->getType() == ShadowTy)
2073  return V;
2074  if (V->getType()->isPtrOrPtrVectorTy())
2075  return IRB.CreatePtrToInt(V, ShadowTy);
2076  else
2077  return IRB.CreateBitCast(V, ShadowTy);
2078  }
2079 
2080  /// Propagate shadow for arbitrary operation.
2081  void handleShadowOr(Instruction &I) {
2082  IRBuilder<> IRB(&I);
2083  ShadowAndOriginCombiner SC(this, IRB);
2084  for (Instruction::op_iterator OI = I.op_begin(); OI != I.op_end(); ++OI)
2085  SC.Add(OI->get());
2086  SC.Done(&I);
2087  }
2088 
2089  // Handle multiplication by constant.
2090  //
2091  // Handle a special case of multiplication by constant that may have one or
2092  // more zeros in the lower bits. This makes corresponding number of lower bits
2093  // of the result zero as well. We model it by shifting the other operand
2094  // shadow left by the required number of bits. Effectively, we transform
2095  // (X * (A * 2**B)) to ((X << B) * A) and instrument (X << B) as (Sx << B).
2096  // We use multiplication by 2**N instead of shift to cover the case of
2097  // multiplication by 0, which may occur in some elements of a vector operand.
2098  void handleMulByConstant(BinaryOperator &I, Constant *ConstArg,
2099  Value *OtherArg) {
2100  Constant *ShadowMul;
2101  Type *Ty = ConstArg->getType();
2102  if (Ty->isVectorTy()) {
2103  unsigned NumElements = Ty->getVectorNumElements();
2104  Type *EltTy = Ty->getSequentialElementType();
2105  SmallVector<Constant *, 16> Elements;
2106  for (unsigned Idx = 0; Idx < NumElements; ++Idx) {
2107  if (ConstantInt *Elt =
2108  dyn_cast<ConstantInt>(ConstArg->getAggregateElement(Idx))) {
2109  const APInt &V = Elt->getValue();
2110  APInt V2 = APInt(V.getBitWidth(), 1) << V.countTrailingZeros();
2111  Elements.push_back(ConstantInt::get(EltTy, V2));
2112  } else {
2113  Elements.push_back(ConstantInt::get(EltTy, 1));
2114  }
2115  }
2116  ShadowMul = ConstantVector::get(Elements);
2117  } else {
2118  if (ConstantInt *Elt = dyn_cast<ConstantInt>(ConstArg)) {
2119  const APInt &V = Elt->getValue();
2120  APInt V2 = APInt(V.getBitWidth(), 1) << V.countTrailingZeros();
2121  ShadowMul = ConstantInt::get(Ty, V2);
2122  } else {
2123  ShadowMul = ConstantInt::get(Ty, 1);
2124  }
2125  }
2126 
2127  IRBuilder<> IRB(&I);
2128  setShadow(&I,
2129  IRB.CreateMul(getShadow(OtherArg), ShadowMul, "msprop_mul_cst"));
2130  setOrigin(&I, getOrigin(OtherArg));
2131  }
2132 
2133  void visitMul(BinaryOperator &I) {
2134  Constant *constOp0 = dyn_cast<Constant>(I.getOperand(0));
2135  Constant *constOp1 = dyn_cast<Constant>(I.getOperand(1));
2136  if (constOp0 && !constOp1)
2137  handleMulByConstant(I, constOp0, I.getOperand(1));
2138  else if (constOp1 && !constOp0)
2139  handleMulByConstant(I, constOp1, I.getOperand(0));
2140  else
2141  handleShadowOr(I);
2142  }
2143 
2144  void visitFAdd(BinaryOperator &I) { handleShadowOr(I); }
2145  void visitFSub(BinaryOperator &I) { handleShadowOr(I); }
2146  void visitFMul(BinaryOperator &I) { handleShadowOr(I); }
2147  void visitAdd(BinaryOperator &I) { handleShadowOr(I); }
2148  void visitSub(BinaryOperator &I) { handleShadowOr(I); }
2149  void visitXor(BinaryOperator &I) { handleShadowOr(I); }
2150 
2151  void handleIntegerDiv(Instruction &I) {
2152  IRBuilder<> IRB(&I);
2153  // Strict on the second argument.
2154  insertShadowCheck(I.getOperand(1), &I);
2155  setShadow(&I, getShadow(&I, 0));
2156  setOrigin(&I, getOrigin(&I, 0));
2157  }
2158 
2159  void visitUDiv(BinaryOperator &I) { handleIntegerDiv(I); }
2160  void visitSDiv(BinaryOperator &I) { handleIntegerDiv(I); }
2161  void visitURem(BinaryOperator &I) { handleIntegerDiv(I); }
2162  void visitSRem(BinaryOperator &I) { handleIntegerDiv(I); }
2163 
2164  // Floating point division is side-effect free. We can not require that the
2165  // divisor is fully initialized and must propagate shadow. See PR37523.
2166  void visitFDiv(BinaryOperator &I) { handleShadowOr(I); }
2167  void visitFRem(BinaryOperator &I) { handleShadowOr(I); }
2168 
2169  /// Instrument == and != comparisons.
2170  ///
2171  /// Sometimes the comparison result is known even if some of the bits of the
2172  /// arguments are not.
2173  void handleEqualityComparison(ICmpInst &I) {
2174  IRBuilder<> IRB(&I);
2175  Value *A = I.getOperand(0);
2176  Value *B = I.getOperand(1);
2177  Value *Sa = getShadow(A);
2178  Value *Sb = getShadow(B);
2179 
2180  // Get rid of pointers and vectors of pointers.
2181  // For ints (and vectors of ints), types of A and Sa match,
2182  // and this is a no-op.
2183  A = IRB.CreatePointerCast(A, Sa->getType());
2184  B = IRB.CreatePointerCast(B, Sb->getType());
2185 
2186  // A == B <==> (C = A^B) == 0
2187  // A != B <==> (C = A^B) != 0
2188  // Sc = Sa | Sb
2189  Value *C = IRB.CreateXor(A, B);
2190  Value *Sc = IRB.CreateOr(Sa, Sb);
2191  // Now dealing with i = (C == 0) comparison (or C != 0, does not matter now)
2192  // Result is defined if one of the following is true
2193  // * there is a defined 1 bit in C
2194  // * C is fully defined
2195  // Si = !(C & ~Sc) && Sc
2196  Value *Zero = Constant::getNullValue(Sc->getType());
2197  Value *MinusOne = Constant::getAllOnesValue(Sc->getType());
2198  Value *Si =
2199  IRB.CreateAnd(IRB.CreateICmpNE(Sc, Zero),
2200  IRB.CreateICmpEQ(
2201  IRB.CreateAnd(IRB.CreateXor(Sc, MinusOne), C), Zero));
2202  Si->setName("_msprop_icmp");
2203  setShadow(&I, Si);
2204  setOriginForNaryOp(I);
2205  }
2206 
2207  /// Build the lowest possible value of V, taking into account V's
2208  /// uninitialized bits.
2209  Value *getLowestPossibleValue(IRBuilder<> &IRB, Value *A, Value *Sa,
2210  bool isSigned) {
2211  if (isSigned) {
2212  // Split shadow into sign bit and other bits.
2213  Value *SaOtherBits = IRB.CreateLShr(IRB.CreateShl(Sa, 1), 1);
2214  Value *SaSignBit = IRB.CreateXor(Sa, SaOtherBits);
2215  // Maximise the undefined shadow bit, minimize other undefined bits.
2216  return
2217  IRB.CreateOr(IRB.CreateAnd(A, IRB.CreateNot(SaOtherBits)), SaSignBit);
2218  } else {
2219  // Minimize undefined bits.
2220  return IRB.CreateAnd(A, IRB.CreateNot(Sa));
2221  }
2222  }
2223 
2224  /// Build the highest possible value of V, taking into account V's
2225  /// uninitialized bits.
2226  Value *getHighestPossibleValue(IRBuilder<> &IRB, Value *A, Value *Sa,
2227  bool isSigned) {
2228  if (isSigned) {
2229  // Split shadow into sign bit and other bits.
2230  Value *SaOtherBits = IRB.CreateLShr(IRB.CreateShl(Sa, 1), 1);
2231  Value *SaSignBit = IRB.CreateXor(Sa, SaOtherBits);
2232  // Minimise the undefined shadow bit, maximise other undefined bits.
2233  return
2234  IRB.CreateOr(IRB.CreateAnd(A, IRB.CreateNot(SaSignBit)), SaOtherBits);
2235  } else {
2236  // Maximize undefined bits.
2237  return IRB.CreateOr(A, Sa);
2238  }
2239  }
2240 
2241  /// Instrument relational comparisons.
2242  ///
2243  /// This function does exact shadow propagation for all relational
2244  /// comparisons of integers, pointers and vectors of those.
2245  /// FIXME: output seems suboptimal when one of the operands is a constant
2246  void handleRelationalComparisonExact(ICmpInst &I) {
2247  IRBuilder<> IRB(&I);
2248  Value *A = I.getOperand(0);
2249  Value *B = I.getOperand(1);
2250  Value *Sa = getShadow(A);
2251  Value *Sb = getShadow(B);
2252 
2253  // Get rid of pointers and vectors of pointers.
2254  // For ints (and vectors of ints), types of A and Sa match,
2255  // and this is a no-op.
2256  A = IRB.CreatePointerCast(A, Sa->getType());
2257  B = IRB.CreatePointerCast(B, Sb->getType());
2258 
2259  // Let [a0, a1] be the interval of possible values of A, taking into account
2260  // its undefined bits. Let [b0, b1] be the interval of possible values of B.
2261  // Then (A cmp B) is defined iff (a0 cmp b1) == (a1 cmp b0).
2262  bool IsSigned = I.isSigned();
2263  Value *S1 = IRB.CreateICmp(I.getPredicate(),
2264  getLowestPossibleValue(IRB, A, Sa, IsSigned),
2265  getHighestPossibleValue(IRB, B, Sb, IsSigned));
2266  Value *S2 = IRB.CreateICmp(I.getPredicate(),
2267  getHighestPossibleValue(IRB, A, Sa, IsSigned),
2268  getLowestPossibleValue(IRB, B, Sb, IsSigned));
2269  Value *Si = IRB.CreateXor(S1, S2);
2270  setShadow(&I, Si);
2271  setOriginForNaryOp(I);
2272  }
2273 
2274  /// Instrument signed relational comparisons.
2275  ///
2276  /// Handle sign bit tests: x<0, x>=0, x<=-1, x>-1 by propagating the highest
2277  /// bit of the shadow. Everything else is delegated to handleShadowOr().
2278  void handleSignedRelationalComparison(ICmpInst &I) {
2279  Constant *constOp;
2280  Value *op = nullptr;
2281  CmpInst::Predicate pre;
2282  if ((constOp = dyn_cast<Constant>(I.getOperand(1)))) {
2283  op = I.getOperand(0);
2284  pre = I.getPredicate();
2285  } else if ((constOp = dyn_cast<Constant>(I.getOperand(0)))) {
2286  op = I.getOperand(1);
2287  pre = I.getSwappedPredicate();
2288  } else {
2289  handleShadowOr(I);
2290  return;
2291  }
2292 
2293  if ((constOp->isNullValue() &&
2294  (pre == CmpInst::ICMP_SLT || pre == CmpInst::ICMP_SGE)) ||
2295  (constOp->isAllOnesValue() &&
2296  (pre == CmpInst::ICMP_SGT || pre == CmpInst::ICMP_SLE))) {
2297  IRBuilder<> IRB(&I);
2298  Value *Shadow = IRB.CreateICmpSLT(getShadow(op), getCleanShadow(op),
2299  "_msprop_icmp_s");
2300  setShadow(&I, Shadow);
2301  setOrigin(&I, getOrigin(op));
2302  } else {
2303  handleShadowOr(I);
2304  }
2305  }
2306 
2307  void visitICmpInst(ICmpInst &I) {
2308  if (!ClHandleICmp) {
2309  handleShadowOr(I);
2310  return;
2311  }
2312  if (I.isEquality()) {
2313  handleEqualityComparison(I);
2314  return;
2315  }
2316 
2317  assert(I.isRelational());
2318  if (ClHandleICmpExact) {
2319  handleRelationalComparisonExact(I);
2320  return;
2321  }
2322  if (I.isSigned()) {
2323  handleSignedRelationalComparison(I);
2324  return;
2325  }
2326 
2327  assert(I.isUnsigned());
2328  if ((isa<Constant>(I.getOperand(0)) || isa<Constant>(I.getOperand(1)))) {
2329  handleRelationalComparisonExact(I);
2330  return;
2331  }
2332 
2333  handleShadowOr(I);
2334  }
2335 
2336  void visitFCmpInst(FCmpInst &I) {
2337  handleShadowOr(I);
2338  }
2339 
2340  void handleShift(BinaryOperator &I) {
2341  IRBuilder<> IRB(&I);
2342  // If any of the S2 bits are poisoned, the whole thing is poisoned.
2343  // Otherwise perform the same shift on S1.
2344  Value *S1 = getShadow(&I, 0);
2345  Value *S2 = getShadow(&I, 1);
2346  Value *S2Conv = IRB.CreateSExt(IRB.CreateICmpNE(S2, getCleanShadow(S2)),
2347  S2->getType());
2348  Value *V2 = I.getOperand(1);
2349  Value *Shift = IRB.CreateBinOp(I.getOpcode(), S1, V2);
2350  setShadow(&I, IRB.CreateOr(Shift, S2Conv));
2351  setOriginForNaryOp(I);
2352  }
2353 
2354  void visitShl(BinaryOperator &I) { handleShift(I); }
2355  void visitAShr(BinaryOperator &I) { handleShift(I); }
2356  void visitLShr(BinaryOperator &I) { handleShift(I); }
2357 
2358  /// Instrument llvm.memmove
2359  ///
2360  /// At this point we don't know if llvm.memmove will be inlined or not.
2361  /// If we don't instrument it and it gets inlined,
2362  /// our interceptor will not kick in and we will lose the memmove.
2363  /// If we instrument the call here, but it does not get inlined,
2364  /// we will memove the shadow twice: which is bad in case
2365  /// of overlapping regions. So, we simply lower the intrinsic to a call.
2366  ///
2367  /// Similar situation exists for memcpy and memset.
2368  void visitMemMoveInst(MemMoveInst &I) {
2369  IRBuilder<> IRB(&I);
2370  IRB.CreateCall(
2371  MS.MemmoveFn,
2372  {IRB.CreatePointerCast(I.getArgOperand(0), IRB.getInt8PtrTy()),
2373  IRB.CreatePointerCast(I.getArgOperand(1), IRB.getInt8PtrTy()),
2374  IRB.CreateIntCast(I.getArgOperand(2), MS.IntptrTy, false)});
2375  I.eraseFromParent();
2376  }
2377 
2378  // Similar to memmove: avoid copying shadow twice.
2379  // This is somewhat unfortunate as it may slowdown small constant memcpys.
2380  // FIXME: consider doing manual inline for small constant sizes and proper
2381  // alignment.
2382  void visitMemCpyInst(MemCpyInst &I) {
2383  IRBuilder<> IRB(&I);
2384  IRB.CreateCall(
2385  MS.MemcpyFn,
2386  {IRB.CreatePointerCast(I.getArgOperand(0), IRB.getInt8PtrTy()),
2387  IRB.CreatePointerCast(I.getArgOperand(1), IRB.getInt8PtrTy()),
2388  IRB.CreateIntCast(I.getArgOperand(2), MS.IntptrTy, false)});
2389  I.eraseFromParent();
2390  }
2391 
2392  // Same as memcpy.
2393  void visitMemSetInst(MemSetInst &I) {
2394  IRBuilder<> IRB(&I);
2395  IRB.CreateCall(
2396  MS.MemsetFn,
2397  {IRB.CreatePointerCast(I.getArgOperand(0), IRB.getInt8PtrTy()),
2398  IRB.CreateIntCast(I.getArgOperand(1), IRB.getInt32Ty(), false),
2399  IRB.CreateIntCast(I.getArgOperand(2), MS.IntptrTy, false)});
2400  I.eraseFromParent();
2401  }
2402 
2403  void visitVAStartInst(VAStartInst &I) {
2404  VAHelper->visitVAStartInst(I);
2405  }
2406 
2407  void visitVACopyInst(VACopyInst &I) {
2408  VAHelper->visitVACopyInst(I);
2409  }
2410 
2411  /// Handle vector store-like intrinsics.
2412  ///
2413  /// Instrument intrinsics that look like a simple SIMD store: writes memory,
2414  /// has 1 pointer argument and 1 vector argument, returns void.
2415  bool handleVectorStoreIntrinsic(IntrinsicInst &I) {
2416  IRBuilder<> IRB(&I);
2417  Value* Addr = I.getArgOperand(0);
2418  Value *Shadow = getShadow(&I, 1);
2419  Value *ShadowPtr, *OriginPtr;
2420 
2421  // We don't know the pointer alignment (could be unaligned SSE store!).
2422  // Have to assume to worst case.
2423  std::tie(ShadowPtr, OriginPtr) = getShadowOriginPtr(
2424  Addr, IRB, Shadow->getType(), /*Alignment*/ 1, /*isStore*/ true);
2425  IRB.CreateAlignedStore(Shadow, ShadowPtr, 1);
2426 
2428  insertShadowCheck(Addr, &I);
2429 
2430  // FIXME: factor out common code from materializeStores
2431  if (MS.TrackOrigins) IRB.CreateStore(getOrigin(&I, 1), OriginPtr);
2432  return true;
2433  }
2434 
2435  /// Handle vector load-like intrinsics.
2436  ///
2437  /// Instrument intrinsics that look like a simple SIMD load: reads memory,
2438  /// has 1 pointer argument, returns a vector.
2439  bool handleVectorLoadIntrinsic(IntrinsicInst &I) {
2440  IRBuilder<> IRB(&I);
2441  Value *Addr = I.getArgOperand(0);
2442 
2443  Type *ShadowTy = getShadowTy(&I);
2444  Value *ShadowPtr, *OriginPtr;
2445  if (PropagateShadow) {
2446  // We don't know the pointer alignment (could be unaligned SSE load!).
2447  // Have to assume to worst case.
2448  unsigned Alignment = 1;
2449  std::tie(ShadowPtr, OriginPtr) =
2450  getShadowOriginPtr(Addr, IRB, ShadowTy, Alignment, /*isStore*/ false);
2451  setShadow(&I,
2452  IRB.CreateAlignedLoad(ShadowTy, ShadowPtr, Alignment, "_msld"));
2453  } else {
2454  setShadow(&I, getCleanShadow(&I));
2455  }
2456 
2458  insertShadowCheck(Addr, &I);
2459 
2460  if (MS.TrackOrigins) {
2461  if (PropagateShadow)
2462  setOrigin(&I, IRB.CreateLoad(MS.OriginTy, OriginPtr));
2463  else
2464  setOrigin(&I, getCleanOrigin());
2465  }
2466  return true;
2467  }
2468 
2469  /// Handle (SIMD arithmetic)-like intrinsics.
2470  ///
2471  /// Instrument intrinsics with any number of arguments of the same type,
2472  /// equal to the return type. The type should be simple (no aggregates or
2473  /// pointers; vectors are fine).
2474  /// Caller guarantees that this intrinsic does not access memory.
2475  bool maybeHandleSimpleNomemIntrinsic(IntrinsicInst &I) {
2476  Type *RetTy = I.getType();
2477  if (!(RetTy->isIntOrIntVectorTy() ||
2478  RetTy->isFPOrFPVectorTy() ||
2479  RetTy->isX86_MMXTy()))
2480  return false;
2481 
2482  unsigned NumArgOperands = I.getNumArgOperands();
2483 
2484  for (unsigned i = 0; i < NumArgOperands; ++i) {
2485  Type *Ty = I.getArgOperand(i)->getType();
2486  if (Ty != RetTy)
2487  return false;
2488  }
2489 
2490  IRBuilder<> IRB(&I);
2491  ShadowAndOriginCombiner SC(this, IRB);
2492  for (unsigned i = 0; i < NumArgOperands; ++i)
2493  SC.Add(I.getArgOperand(i));
2494  SC.Done(&I);
2495 
2496  return true;
2497  }
2498 
2499  /// Heuristically instrument unknown intrinsics.
2500  ///
2501  /// The main purpose of this code is to do something reasonable with all
2502  /// random intrinsics we might encounter, most importantly - SIMD intrinsics.
2503  /// We recognize several classes of intrinsics by their argument types and
2504  /// ModRefBehaviour and apply special intrumentation when we are reasonably
2505  /// sure that we know what the intrinsic does.
2506  ///
2507  /// We special-case intrinsics where this approach fails. See llvm.bswap
2508  /// handling as an example of that.
2509  bool handleUnknownIntrinsic(IntrinsicInst &I) {
2510  unsigned NumArgOperands = I.getNumArgOperands();
2511  if (NumArgOperands == 0)
2512  return false;
2513 
2514  if (NumArgOperands == 2 &&
2515  I.getArgOperand(0)->getType()->isPointerTy() &&
2516  I.getArgOperand(1)->getType()->isVectorTy() &&
2517  I.getType()->isVoidTy() &&
2518  !I.onlyReadsMemory()) {
2519  // This looks like a vector store.
2520  return handleVectorStoreIntrinsic(I);
2521  }
2522 
2523  if (NumArgOperands == 1 &&
2524  I.getArgOperand(0)->getType()->isPointerTy() &&
2525  I.getType()->isVectorTy() &&
2526  I.onlyReadsMemory()) {
2527  // This looks like a vector load.
2528  return handleVectorLoadIntrinsic(I);
2529  }
2530 
2531  if (I.doesNotAccessMemory())
2532  if (maybeHandleSimpleNomemIntrinsic(I))
2533  return true;
2534 
2535  // FIXME: detect and handle SSE maskstore/maskload
2536  return false;
2537  }
2538 
2539  void handleBswap(IntrinsicInst &I) {
2540  IRBuilder<> IRB(&I);
2541  Value *Op = I.getArgOperand(0);
2542  Type *OpType = Op->getType();
2543  Function *BswapFunc = Intrinsic::getDeclaration(
2544  F.getParent(), Intrinsic::bswap, makeArrayRef(&OpType, 1));
2545  setShadow(&I, IRB.CreateCall(BswapFunc, getShadow(Op)));
2546  setOrigin(&I, getOrigin(Op));
2547  }
2548 
2549  // Instrument vector convert instrinsic.
2550  //
2551  // This function instruments intrinsics like cvtsi2ss:
2552  // %Out = int_xxx_cvtyyy(%ConvertOp)
2553  // or
2554  // %Out = int_xxx_cvtyyy(%CopyOp, %ConvertOp)
2555  // Intrinsic converts \p NumUsedElements elements of \p ConvertOp to the same
2556  // number \p Out elements, and (if has 2 arguments) copies the rest of the
2557  // elements from \p CopyOp.
2558  // In most cases conversion involves floating-point value which may trigger a
2559  // hardware exception when not fully initialized. For this reason we require
2560  // \p ConvertOp[0:NumUsedElements] to be fully initialized and trap otherwise.
2561  // We copy the shadow of \p CopyOp[NumUsedElements:] to \p
2562  // Out[NumUsedElements:]. This means that intrinsics without \p CopyOp always
2563  // return a fully initialized value.
2564  void handleVectorConvertIntrinsic(IntrinsicInst &I, int NumUsedElements) {
2565  IRBuilder<> IRB(&I);
2566  Value *CopyOp, *ConvertOp;
2567 
2568  switch (I.getNumArgOperands()) {
2569  case 3:
2570  assert(isa<ConstantInt>(I.getArgOperand(2)) && "Invalid rounding mode");
2572  case 2:
2573  CopyOp = I.getArgOperand(0);
2574  ConvertOp = I.getArgOperand(1);
2575  break;
2576  case 1:
2577  ConvertOp = I.getArgOperand(0);
2578  CopyOp = nullptr;
2579  break;
2580  default:
2581  llvm_unreachable("Cvt intrinsic with unsupported number of arguments.");
2582  }
2583 
2584  // The first *NumUsedElements* elements of ConvertOp are converted to the
2585  // same number of output elements. The rest of the output is copied from
2586  // CopyOp, or (if not available) filled with zeroes.
2587  // Combine shadow for elements of ConvertOp that are used in this operation,
2588  // and insert a check.
2589  // FIXME: consider propagating shadow of ConvertOp, at least in the case of
2590  // int->any conversion.
2591  Value *ConvertShadow = getShadow(ConvertOp);
2592  Value *AggShadow = nullptr;
2593  if (ConvertOp->getType()->isVectorTy()) {
2594  AggShadow = IRB.CreateExtractElement(
2595  ConvertShadow, ConstantInt::get(IRB.getInt32Ty(), 0));
2596  for (int i = 1; i < NumUsedElements; ++i) {
2597  Value *MoreShadow = IRB.CreateExtractElement(
2598  ConvertShadow, ConstantInt::get(IRB.getInt32Ty(), i));
2599  AggShadow = IRB.CreateOr(AggShadow, MoreShadow);
2600  }
2601  } else {
2602  AggShadow = ConvertShadow;
2603  }
2604  assert(AggShadow->getType()->isIntegerTy());
2605  insertShadowCheck(AggShadow, getOrigin(ConvertOp), &I);
2606 
2607  // Build result shadow by zero-filling parts of CopyOp shadow that come from
2608  // ConvertOp.
2609  if (CopyOp) {
2610  assert(CopyOp->getType() == I.getType());
2611  assert(CopyOp->getType()->isVectorTy());
2612  Value *ResultShadow = getShadow(CopyOp);
2613  Type *EltTy = ResultShadow->getType()->getVectorElementType();
2614  for (int i = 0; i < NumUsedElements; ++i) {
2615  ResultShadow = IRB.CreateInsertElement(
2616  ResultShadow, ConstantInt::getNullValue(EltTy),
2617  ConstantInt::get(IRB.getInt32Ty(), i));
2618  }
2619  setShadow(&I, ResultShadow);
2620  setOrigin(&I, getOrigin(CopyOp));
2621  } else {
2622  setShadow(&I, getCleanShadow(&I));
2623  setOrigin(&I, getCleanOrigin());
2624  }
2625  }
2626 
2627  // Given a scalar or vector, extract lower 64 bits (or less), and return all
2628  // zeroes if it is zero, and all ones otherwise.
2629  Value *Lower64ShadowExtend(IRBuilder<> &IRB, Value *S, Type *T) {
2630  if (S->getType()->isVectorTy())
2631  S = CreateShadowCast(IRB, S, IRB.getInt64Ty(), /* Signed */ true);
2632  assert(S->getType()->getPrimitiveSizeInBits() <= 64);
2633  Value *S2 = IRB.CreateICmpNE(S, getCleanShadow(S));
2634  return CreateShadowCast(IRB, S2, T, /* Signed */ true);
2635  }
2636 
2637  // Given a vector, extract its first element, and return all
2638  // zeroes if it is zero, and all ones otherwise.
2639  Value *LowerElementShadowExtend(IRBuilder<> &IRB, Value *S, Type *T) {
2640  Value *S1 = IRB.CreateExtractElement(S, (uint64_t)0);
2641  Value *S2 = IRB.CreateICmpNE(S1, getCleanShadow(S1));
2642  return CreateShadowCast(IRB, S2, T, /* Signed */ true);
2643  }
2644 
2645  Value *VariableShadowExtend(IRBuilder<> &IRB, Value *S) {
2646  Type *T = S->getType();
2647  assert(T->isVectorTy());
2648  Value *S2 = IRB.CreateICmpNE(S, getCleanShadow(S));
2649  return IRB.CreateSExt(S2, T);
2650  }
2651 
2652  // Instrument vector shift instrinsic.
2653  //
2654  // This function instruments intrinsics like int_x86_avx2_psll_w.
2655  // Intrinsic shifts %In by %ShiftSize bits.
2656  // %ShiftSize may be a vector. In that case the lower 64 bits determine shift
2657  // size, and the rest is ignored. Behavior is defined even if shift size is
2658  // greater than register (or field) width.
2659  void handleVectorShiftIntrinsic(IntrinsicInst &I, bool Variable) {
2660  assert(I.getNumArgOperands() == 2);
2661  IRBuilder<> IRB(&I);
2662  // If any of the S2 bits are poisoned, the whole thing is poisoned.
2663  // Otherwise perform the same shift on S1.
2664  Value *S1 = getShadow(&I, 0);
2665  Value *S2 = getShadow(&I, 1);
2666  Value *S2Conv = Variable ? VariableShadowExtend(IRB, S2)
2667  : Lower64ShadowExtend(IRB, S2, getShadowTy(&I));
2668  Value *V1 = I.getOperand(0);
2669  Value *V2 = I.getOperand(1);
2670  Value *Shift = IRB.CreateCall(I.getFunctionType(), I.getCalledValue(),
2671  {IRB.CreateBitCast(S1, V1->getType()), V2});
2672  Shift = IRB.CreateBitCast(Shift, getShadowTy(&I));
2673  setShadow(&I, IRB.CreateOr(Shift, S2Conv));
2674  setOriginForNaryOp(I);
2675  }
2676 
2677  // Get an X86_MMX-sized vector type.
2678  Type *getMMXVectorTy(unsigned EltSizeInBits) {
2679  const unsigned X86_MMXSizeInBits = 64;
2680  return VectorType::get(IntegerType::get(*MS.C, EltSizeInBits),
2681  X86_MMXSizeInBits / EltSizeInBits);
2682  }
2683 
2684  // Returns a signed counterpart for an (un)signed-saturate-and-pack
2685  // intrinsic.
2686  Intrinsic::ID getSignedPackIntrinsic(Intrinsic::ID id) {
2687  switch (id) {
2688  case Intrinsic::x86_sse2_packsswb_128:
2689  case Intrinsic::x86_sse2_packuswb_128:
2690  return Intrinsic::x86_sse2_packsswb_128;
2691 
2692  case Intrinsic::x86_sse2_packssdw_128:
2693  case Intrinsic::x86_sse41_packusdw:
2694  return Intrinsic::x86_sse2_packssdw_128;
2695 
2696  case Intrinsic::x86_avx2_packsswb:
2697  case Intrinsic::x86_avx2_packuswb:
2698  return Intrinsic::x86_avx2_packsswb;
2699 
2700  case Intrinsic::x86_avx2_packssdw:
2701  case Intrinsic::x86_avx2_packusdw:
2702  return Intrinsic::x86_avx2_packssdw;
2703 
2704  case Intrinsic::x86_mmx_packsswb:
2705  case Intrinsic::x86_mmx_packuswb:
2706  return Intrinsic::x86_mmx_packsswb;
2707 
2708  case Intrinsic::x86_mmx_packssdw:
2709  return Intrinsic::x86_mmx_packssdw;
2710  default:
2711  llvm_unreachable("unexpected intrinsic id");
2712  }
2713  }
2714 
2715  // Instrument vector pack instrinsic.
2716  //
2717  // This function instruments intrinsics like x86_mmx_packsswb, that
2718  // packs elements of 2 input vectors into half as many bits with saturation.
2719  // Shadow is propagated with the signed variant of the same intrinsic applied
2720  // to sext(Sa != zeroinitializer), sext(Sb != zeroinitializer).
2721  // EltSizeInBits is used only for x86mmx arguments.
2722  void handleVectorPackIntrinsic(IntrinsicInst &I, unsigned EltSizeInBits = 0) {
2723  assert(I.getNumArgOperands() == 2);
2724  bool isX86_MMX = I.getOperand(0)->getType()->isX86_MMXTy();
2725  IRBuilder<> IRB(&I);
2726  Value *S1 = getShadow(&I, 0);
2727  Value *S2 = getShadow(&I, 1);
2728  assert(isX86_MMX || S1->getType()->isVectorTy());
2729 
2730  // SExt and ICmpNE below must apply to individual elements of input vectors.
2731  // In case of x86mmx arguments, cast them to appropriate vector types and
2732  // back.
2733  Type *T = isX86_MMX ? getMMXVectorTy(EltSizeInBits) : S1->getType();
2734  if (isX86_MMX) {
2735  S1 = IRB.CreateBitCast(S1, T);
2736  S2 = IRB.CreateBitCast(S2, T);
2737  }
2738  Value *S1_ext = IRB.CreateSExt(
2739  IRB.CreateICmpNE(S1, Constant::getNullValue(T)), T);
2740  Value *S2_ext = IRB.CreateSExt(
2741  IRB.CreateICmpNE(S2, Constant::getNullValue(T)), T);
2742  if (isX86_MMX) {
2743  Type *X86_MMXTy = Type::getX86_MMXTy(*MS.C);
2744  S1_ext = IRB.CreateBitCast(S1_ext, X86_MMXTy);
2745  S2_ext = IRB.CreateBitCast(S2_ext, X86_MMXTy);
2746  }
2747 
2748  Function *ShadowFn = Intrinsic::getDeclaration(
2749  F.getParent(), getSignedPackIntrinsic(I.getIntrinsicID()));
2750 
2751  Value *S =
2752  IRB.CreateCall(ShadowFn, {S1_ext, S2_ext}, "_msprop_vector_pack");
2753  if (isX86_MMX) S = IRB.CreateBitCast(S, getShadowTy(&I));
2754  setShadow(&I, S);
2755  setOriginForNaryOp(I);
2756  }
2757 
2758  // Instrument sum-of-absolute-differencies intrinsic.
2759  void handleVectorSadIntrinsic(IntrinsicInst &I) {
2760  const unsigned SignificantBitsPerResultElement = 16;
2761  bool isX86_MMX = I.getOperand(0)->getType()->isX86_MMXTy();
2762  Type *ResTy = isX86_MMX ? IntegerType::get(*MS.C, 64) : I.getType();
2763  unsigned ZeroBitsPerResultElement =
2764  ResTy->getScalarSizeInBits() - SignificantBitsPerResultElement;
2765 
2766  IRBuilder<> IRB(&I);
2767  Value *S = IRB.CreateOr(getShadow(&I, 0), getShadow(&I, 1));
2768  S = IRB.CreateBitCast(S, ResTy);
2769  S = IRB.CreateSExt(IRB.CreateICmpNE(S, Constant::getNullValue(ResTy)),
2770  ResTy);
2771  S = IRB.CreateLShr(S, ZeroBitsPerResultElement);
2772  S = IRB.CreateBitCast(S, getShadowTy(&I));
2773  setShadow(&I, S);
2774  setOriginForNaryOp(I);
2775  }
2776 
2777  // Instrument multiply-add intrinsic.
2778  void handleVectorPmaddIntrinsic(IntrinsicInst &I,
2779  unsigned EltSizeInBits = 0) {
2780  bool isX86_MMX = I.getOperand(0)->getType()->isX86_MMXTy();
2781  Type *ResTy = isX86_MMX ? getMMXVectorTy(EltSizeInBits * 2) : I.getType();
2782  IRBuilder<> IRB(&I);
2783  Value *S = IRB.CreateOr(getShadow(&I, 0), getShadow(&I, 1));
2784  S = IRB.CreateBitCast(S, ResTy);
2785  S = IRB.CreateSExt(IRB.CreateICmpNE(S, Constant::getNullValue(ResTy)),
2786  ResTy);
2787  S = IRB.CreateBitCast(S, getShadowTy(&I));
2788  setShadow(&I, S);
2789  setOriginForNaryOp(I);
2790  }
2791 
2792  // Instrument compare-packed intrinsic.
2793  // Basically, an or followed by sext(icmp ne 0) to end up with all-zeros or
2794  // all-ones shadow.
2795  void handleVectorComparePackedIntrinsic(IntrinsicInst &I) {
2796  IRBuilder<> IRB(&I);
2797  Type *ResTy = getShadowTy(&I);
2798  Value *S0 = IRB.CreateOr(getShadow(&I, 0), getShadow(&I, 1));
2799  Value *S = IRB.CreateSExt(
2800  IRB.CreateICmpNE(S0, Constant::getNullValue(ResTy)), ResTy);
2801  setShadow(&I, S);
2802  setOriginForNaryOp(I);
2803  }
2804 
2805  // Instrument compare-scalar intrinsic.
2806  // This handles both cmp* intrinsics which return the result in the first
2807  // element of a vector, and comi* which return the result as i32.
2808  void handleVectorCompareScalarIntrinsic(IntrinsicInst &I) {
2809  IRBuilder<> IRB(&I);
2810  Value *S0 = IRB.CreateOr(getShadow(&I, 0), getShadow(&I, 1));
2811  Value *S = LowerElementShadowExtend(IRB, S0, getShadowTy(&I));
2812  setShadow(&I, S);
2813  setOriginForNaryOp(I);
2814  }
2815 
2816  void handleStmxcsr(IntrinsicInst &I) {
2817  IRBuilder<> IRB(&I);
2818  Value* Addr = I.getArgOperand(0);
2819  Type *Ty = IRB.getInt32Ty();
2820  Value *ShadowPtr =
2821  getShadowOriginPtr(Addr, IRB, Ty, /*Alignment*/ 1, /*isStore*/ true)
2822  .first;
2823 
2824  IRB.CreateStore(getCleanShadow(Ty),
2825  IRB.CreatePointerCast(ShadowPtr, Ty->getPointerTo()));
2826 
2828  insertShadowCheck(Addr, &I);
2829  }
2830 
2831  void handleLdmxcsr(IntrinsicInst &I) {
2832  if (!InsertChecks) return;
2833 
2834  IRBuilder<> IRB(&I);
2835  Value *Addr = I.getArgOperand(0);
2836  Type *Ty = IRB.getInt32Ty();
2837  unsigned Alignment = 1;
2838  Value *ShadowPtr, *OriginPtr;
2839  std::tie(ShadowPtr, OriginPtr) =
2840  getShadowOriginPtr(Addr, IRB, Ty, Alignment, /*isStore*/ false);
2841 
2843  insertShadowCheck(Addr, &I);
2844 
2845  Value *Shadow = IRB.CreateAlignedLoad(Ty, ShadowPtr, Alignment, "_ldmxcsr");
2846  Value *Origin = MS.TrackOrigins ? IRB.CreateLoad(MS.OriginTy, OriginPtr)
2847  : getCleanOrigin();
2848  insertShadowCheck(Shadow, Origin, &I);
2849  }
2850 
2851  void handleMaskedStore(IntrinsicInst &I) {
2852  IRBuilder<> IRB(&I);
2853  Value *V = I.getArgOperand(0);
2854  Value *Addr = I.getArgOperand(1);
2855  unsigned Align = cast<ConstantInt>(I.getArgOperand(2))->getZExtValue();
2856  Value *Mask = I.getArgOperand(3);
2857  Value *Shadow = getShadow(V);
2858 
2859  Value *ShadowPtr;
2860  Value *OriginPtr;
2861  std::tie(ShadowPtr, OriginPtr) = getShadowOriginPtr(
2862  Addr, IRB, Shadow->getType(), Align, /*isStore*/ true);
2863 
2864  if (ClCheckAccessAddress) {
2865  insertShadowCheck(Addr, &I);
2866  // Uninitialized mask is kind of like uninitialized address, but not as
2867  // scary.
2868  insertShadowCheck(Mask, &I);
2869  }
2870 
2871  IRB.CreateMaskedStore(Shadow, ShadowPtr, Align, Mask);
2872 
2873  if (MS.TrackOrigins) {
2874  auto &DL = F.getParent()->getDataLayout();
2875  paintOrigin(IRB, getOrigin(V), OriginPtr,
2876  DL.getTypeStoreSize(Shadow->getType()),
2877  std::max(Align, kMinOriginAlignment));
2878  }
2879  }
2880 
2881  bool handleMaskedLoad(IntrinsicInst &I) {
2882  IRBuilder<> IRB(&I);
2883  Value *Addr = I.getArgOperand(0);
2884  unsigned Align = cast<ConstantInt>(I.getArgOperand(1))->getZExtValue();
2885  Value *Mask = I.getArgOperand(2);
2886  Value *PassThru = I.getArgOperand(3);
2887 
2888  Type *ShadowTy = getShadowTy(&I);
2889  Value *ShadowPtr, *OriginPtr;
2890  if (PropagateShadow) {
2891  std::tie(ShadowPtr, OriginPtr) =
2892  getShadowOriginPtr(Addr, IRB, ShadowTy, Align, /*isStore*/ false);
2893  setShadow(&I, IRB.CreateMaskedLoad(ShadowPtr, Align, Mask,
2894  getShadow(PassThru), "_msmaskedld"));
2895  } else {
2896  setShadow(&I, getCleanShadow(&I));
2897  }
2898 
2899  if (ClCheckAccessAddress) {
2900  insertShadowCheck(Addr, &I);
2901  insertShadowCheck(Mask, &I);
2902  }
2903 
2904  if (MS.TrackOrigins) {
2905  if (PropagateShadow) {
2906  // Choose between PassThru's and the loaded value's origins.
2907  Value *MaskedPassThruShadow = IRB.CreateAnd(
2908  getShadow(PassThru), IRB.CreateSExt(IRB.CreateNeg(Mask), ShadowTy));
2909 
2910  Value *Acc = IRB.CreateExtractElement(
2911  MaskedPassThruShadow, ConstantInt::get(IRB.getInt32Ty(), 0));
2912  for (int i = 1, N = PassThru->getType()->getVectorNumElements(); i < N;
2913  ++i) {
2914  Value *More = IRB.CreateExtractElement(
2915  MaskedPassThruShadow, ConstantInt::get(IRB.getInt32Ty(), i));
2916  Acc = IRB.CreateOr(Acc, More);
2917  }
2918 
2919  Value *Origin = IRB.CreateSelect(
2920  IRB.CreateICmpNE(Acc, Constant::getNullValue(Acc->getType())),
2921  getOrigin(PassThru), IRB.CreateLoad(MS.OriginTy, OriginPtr));
2922 
2923  setOrigin(&I, Origin);
2924  } else {
2925  setOrigin(&I, getCleanOrigin());
2926  }
2927  }
2928  return true;
2929  }
2930 
2931  // Instrument BMI / BMI2 intrinsics.
2932  // All of these intrinsics are Z = I(X, Y)
2933  // where the types of all operands and the result match, and are either i32 or i64.
2934  // The following instrumentation happens to work for all of them:
2935  // Sz = I(Sx, Y) | (sext (Sy != 0))
2936  void handleBmiIntrinsic(IntrinsicInst &I) {
2937  IRBuilder<> IRB(&I);
2938  Type *ShadowTy = getShadowTy(&I);
2939 
2940  // If any bit of the mask operand is poisoned, then the whole thing is.
2941  Value *SMask = getShadow(&I, 1);
2942  SMask = IRB.CreateSExt(IRB.CreateICmpNE(SMask, getCleanShadow(ShadowTy)),
2943  ShadowTy);
2944  // Apply the same intrinsic to the shadow of the first operand.
2945  Value *S = IRB.CreateCall(I.getCalledFunction(),
2946  {getShadow(&I, 0), I.getOperand(1)});
2947  S = IRB.CreateOr(SMask, S);
2948  setShadow(&I, S);
2949  setOriginForNaryOp(I);
2950  }
2951 
2952  void visitIntrinsicInst(IntrinsicInst &I) {
2953  switch (I.getIntrinsicID()) {
2954  case Intrinsic::bswap:
2955  handleBswap(I);
2956  break;
2957  case Intrinsic::masked_store:
2958  handleMaskedStore(I);
2959  break;
2960  case Intrinsic::masked_load:
2961  handleMaskedLoad(I);
2962  break;
2963  case Intrinsic::x86_sse_stmxcsr:
2964  handleStmxcsr(I);
2965  break;
2966  case Intrinsic::x86_sse_ldmxcsr:
2967  handleLdmxcsr(I);
2968  break;
2969  case Intrinsic::x86_avx512_vcvtsd2usi64:
2970  case Intrinsic::x86_avx512_vcvtsd2usi32:
2971  case Intrinsic::x86_avx512_vcvtss2usi64:
2972  case Intrinsic::x86_avx512_vcvtss2usi32:
2973  case Intrinsic::x86_avx512_cvttss2usi64:
2974  case Intrinsic::x86_avx512_cvttss2usi:
2975  case Intrinsic::x86_avx512_cvttsd2usi64:
2976  case Intrinsic::x86_avx512_cvttsd2usi:
2977  case Intrinsic::x86_avx512_cvtusi2ss:
2978  case Intrinsic::x86_avx512_cvtusi642sd:
2979  case Intrinsic::x86_avx512_cvtusi642ss:
2980  case Intrinsic::x86_sse2_cvtsd2si64:
2981  case Intrinsic::x86_sse2_cvtsd2si:
2982  case Intrinsic::x86_sse2_cvtsd2ss:
2983  case Intrinsic::x86_sse2_cvttsd2si64:
2984  case Intrinsic::x86_sse2_cvttsd2si:
2985  case Intrinsic::x86_sse_cvtss2si64:
2986  case Intrinsic::x86_sse_cvtss2si:
2987  case Intrinsic::x86_sse_cvttss2si64:
2988  case Intrinsic::x86_sse_cvttss2si:
2989  handleVectorConvertIntrinsic(I, 1);
2990  break;
2991  case Intrinsic::x86_sse_cvtps2pi:
2992  case Intrinsic::x86_sse_cvttps2pi:
2993  handleVectorConvertIntrinsic(I, 2);
2994  break;
2995 
2996  case Intrinsic::x86_avx512_psll_w_512:
2997  case Intrinsic::x86_avx512_psll_d_512:
2998  case Intrinsic::x86_avx512_psll_q_512:
2999  case Intrinsic::x86_avx512_pslli_w_512:
3000  case Intrinsic::x86_avx512_pslli_d_512:
3001  case Intrinsic::x86_avx512_pslli_q_512:
3002  case Intrinsic::x86_avx512_psrl_w_512:
3003  case Intrinsic::x86_avx512_psrl_d_512:
3004  case Intrinsic::x86_avx512_psrl_q_512:
3005  case Intrinsic::x86_avx512_psra_w_512:
3006  case Intrinsic::x86_avx512_psra_d_512:
3007  case Intrinsic::x86_avx512_psra_q_512:
3008  case Intrinsic::x86_avx512_psrli_w_512:
3009  case Intrinsic::x86_avx512_psrli_d_512:
3010  case Intrinsic::x86_avx512_psrli_q_512:
3011  case Intrinsic::x86_avx512_psrai_w_512:
3012  case Intrinsic::x86_avx512_psrai_d_512:
3013  case Intrinsic::x86_avx512_psrai_q_512:
3014  case Intrinsic::x86_avx512_psra_q_256:
3015  case Intrinsic::x86_avx512_psra_q_128:
3016  case Intrinsic::x86_avx512_psrai_q_256:
3017  case Intrinsic::x86_avx512_psrai_q_128:
3018  case Intrinsic::x86_avx2_psll_w:
3019  case Intrinsic::x86_avx2_psll_d:
3020  case Intrinsic::x86_avx2_psll_q:
3021  case Intrinsic::x86_avx2_pslli_w:
3022  case Intrinsic::x86_avx2_pslli_d:
3023  case Intrinsic::x86_avx2_pslli_q:
3024  case Intrinsic::x86_avx2_psrl_w:
3025  case Intrinsic::x86_avx2_psrl_d:
3026  case Intrinsic::x86_avx2_psrl_q:
3027  case Intrinsic::x86_avx2_psra_w:
3028  case Intrinsic::x86_avx2_psra_d:
3029  case Intrinsic::x86_avx2_psrli_w:
3030  case Intrinsic::x86_avx2_psrli_d:
3031  case Intrinsic::x86_avx2_psrli_q:
3032  case Intrinsic::x86_avx2_psrai_w:
3033  case Intrinsic::x86_avx2_psrai_d:
3034  case Intrinsic::x86_sse2_psll_w:
3035  case Intrinsic::x86_sse2_psll_d:
3036  case Intrinsic::x86_sse2_psll_q:
3037  case Intrinsic::x86_sse2_pslli_w:
3038  case Intrinsic::x86_sse2_pslli_d:
3039  case Intrinsic::x86_sse2_pslli_q:
3040  case Intrinsic::x86_sse2_psrl_w:
3041  case Intrinsic::x86_sse2_psrl_d:
3042  case Intrinsic::x86_sse2_psrl_q:
3043  case Intrinsic::x86_sse2_psra_w:
3044  case Intrinsic::x86_sse2_psra_d:
3045  case Intrinsic::x86_sse2_psrli_w:
3046  case Intrinsic::x86_sse2_psrli_d:
3047  case Intrinsic::x86_sse2_psrli_q:
3048  case Intrinsic::x86_sse2_psrai_w:
3049  case Intrinsic::x86_sse2_psrai_d:
3050  case Intrinsic::x86_mmx_psll_w:
3051  case Intrinsic::x86_mmx_psll_d:
3052  case Intrinsic::x86_mmx_psll_q:
3053  case Intrinsic::x86_mmx_pslli_w:
3054  case Intrinsic::x86_mmx_pslli_d:
3055  case Intrinsic::x86_mmx_pslli_q:
3056  case Intrinsic::x86_mmx_psrl_w:
3057  case Intrinsic::x86_mmx_psrl_d:
3058  case Intrinsic::x86_mmx_psrl_q:
3059  case Intrinsic::x86_mmx_psra_w:
3060  case Intrinsic::x86_mmx_psra_d:
3061  case Intrinsic::x86_mmx_psrli_w:
3062  case Intrinsic::x86_mmx_psrli_d:
3063  case Intrinsic::x86_mmx_psrli_q:
3064  case Intrinsic::x86_mmx_psrai_w:
3065  case Intrinsic::x86_mmx_psrai_d:
3066  handleVectorShiftIntrinsic(I, /* Variable */ false);
3067  break;
3068  case Intrinsic::x86_avx2_psllv_d:
3069  case Intrinsic::x86_avx2_psllv_d_256:
3070  case Intrinsic::x86_avx512_psllv_d_512:
3071  case Intrinsic::x86_avx2_psllv_q:
3072  case Intrinsic::x86_avx2_psllv_q_256:
3073  case Intrinsic::x86_avx512_psllv_q_512:
3074  case Intrinsic::x86_avx2_psrlv_d:
3075  case Intrinsic::x86_avx2_psrlv_d_256:
3076  case Intrinsic::x86_avx512_psrlv_d_512:
3077  case Intrinsic::x86_avx2_psrlv_q:
3078  case Intrinsic::x86_avx2_psrlv_q_256:
3079  case Intrinsic::x86_avx512_psrlv_q_512:
3080  case Intrinsic::x86_avx2_psrav_d:
3081  case Intrinsic::x86_avx2_psrav_d_256:
3082  case Intrinsic::x86_avx512_psrav_d_512:
3083  case Intrinsic::x86_avx512_psrav_q_128:
3084  case Intrinsic::x86_avx512_psrav_q_256:
3085  case Intrinsic::x86_avx512_psrav_q_512:
3086  handleVectorShiftIntrinsic(I, /* Variable */ true);
3087  break;
3088 
3089  case Intrinsic::x86_sse2_packsswb_128:
3090  case Intrinsic::x86_sse2_packssdw_128:
3091  case Intrinsic::x86_sse2_packuswb_128:
3092  case Intrinsic::x86_sse41_packusdw:
3093  case Intrinsic::x86_avx2_packsswb:
3094  case Intrinsic::x86_avx2_packssdw:
3095  case Intrinsic::x86_avx2_packuswb:
3096  case Intrinsic::x86_avx2_packusdw:
3097  handleVectorPackIntrinsic(I);
3098  break;
3099 
3100  case Intrinsic::x86_mmx_packsswb:
3101  case Intrinsic::x86_mmx_packuswb:
3102  handleVectorPackIntrinsic(I, 16);
3103  break;
3104 
3105  case Intrinsic::x86_mmx_packssdw:
3106  handleVectorPackIntrinsic(I, 32);
3107  break;
3108 
3109  case Intrinsic::x86_mmx_psad_bw:
3110  case Intrinsic::x86_sse2_psad_bw:
3111  case Intrinsic::x86_avx2_psad_bw:
3112  handleVectorSadIntrinsic(I);
3113  break;
3114 
3115  case Intrinsic::x86_sse2_pmadd_wd:
3116  case Intrinsic::x86_avx2_pmadd_wd:
3117  case Intrinsic::x86_ssse3_pmadd_ub_sw_128:
3118  case Intrinsic::x86_avx2_pmadd_ub_sw:
3119  handleVectorPmaddIntrinsic(I);
3120  break;
3121 
3122  case Intrinsic::x86_ssse3_pmadd_ub_sw:
3123  handleVectorPmaddIntrinsic(I, 8);
3124  break;
3125 
3126  case Intrinsic::x86_mmx_pmadd_wd:
3127  handleVectorPmaddIntrinsic(I, 16);
3128  break;
3129 
3130  case Intrinsic::x86_sse_cmp_ss:
3131  case Intrinsic::x86_sse2_cmp_sd:
3132  case Intrinsic::x86_sse_comieq_ss:
3133  case Intrinsic::x86_sse_comilt_ss:
3134  case Intrinsic::x86_sse_comile_ss:
3135  case Intrinsic::x86_sse_comigt_ss:
3136  case Intrinsic::x86_sse_comige_ss:
3137  case Intrinsic::x86_sse_comineq_ss:
3138  case Intrinsic::x86_sse_ucomieq_ss:
3139  case Intrinsic::x86_sse_ucomilt_ss:
3140  case Intrinsic::x86_sse_ucomile_ss:
3141  case Intrinsic::x86_sse_ucomigt_ss:
3142  case Intrinsic::x86_sse_ucomige_ss:
3143  case Intrinsic::x86_sse_ucomineq_ss:
3144  case Intrinsic::x86_sse2_comieq_sd:
3145  case Intrinsic::x86_sse2_comilt_sd:
3146  case Intrinsic::x86_sse2_comile_sd:
3147  case Intrinsic::x86_sse2_comigt_sd:
3148  case Intrinsic::x86_sse2_comige_sd:
3149  case Intrinsic::x86_sse2_comineq_sd:
3150  case Intrinsic::x86_sse2_ucomieq_sd:
3151  case Intrinsic::x86_sse2_ucomilt_sd:
3152  case Intrinsic::x86_sse2_ucomile_sd:
3153  case Intrinsic::x86_sse2_ucomigt_sd:
3154  case Intrinsic::x86_sse2_ucomige_sd:
3155  case Intrinsic::x86_sse2_ucomineq_sd:
3156  handleVectorCompareScalarIntrinsic(I);
3157  break;
3158 
3159  case Intrinsic::x86_sse_cmp_ps:
3160  case Intrinsic::x86_sse2_cmp_pd:
3161  // FIXME: For x86_avx_cmp_pd_256 and x86_avx_cmp_ps_256 this function
3162  // generates reasonably looking IR that fails in the backend with "Do not
3163  // know how to split the result of this operator!".
3164  handleVectorComparePackedIntrinsic(I);
3165  break;
3166 
3167  case Intrinsic::x86_bmi_bextr_32:
3168  case Intrinsic::x86_bmi_bextr_64:
3169  case Intrinsic::x86_bmi_bzhi_32:
3170  case Intrinsic::x86_bmi_bzhi_64:
3171  case Intrinsic::x86_bmi_pdep_32:
3172  case Intrinsic::x86_bmi_pdep_64:
3173  case Intrinsic::x86_bmi_pext_32:
3174  case Intrinsic::x86_bmi_pext_64:
3175  handleBmiIntrinsic(I);
3176  break;
3177 
3178  case Intrinsic::is_constant:
3179  // The result of llvm.is.constant() is always defined.
3180  setShadow(&I, getCleanShadow(&I));
3181  setOrigin(&I, getCleanOrigin());
3182  break;
3183 
3184  default:
3185  if (!handleUnknownIntrinsic(I))
3186  visitInstruction(I);
3187  break;
3188  }
3189  }
3190 
3191  void visitCallSite(CallSite CS) {
3192  Instruction &I = *CS.getInstruction();
3193  assert(!I.getMetadata("nosanitize"));
3194  assert((CS.isCall() || CS.isInvoke()) && "Unknown type of CallSite");
3195  if (CS.isCall()) {
3196  CallInst *Call = cast<CallInst>(&I);
3197 
3198  // For inline asm, do the usual thing: check argument shadow and mark all
3199  // outputs as clean. Note that any side effects of the inline asm that are
3200  // not immediately visible in its constraints are not handled.
3201  if (Call->isInlineAsm()) {
3202  if (ClHandleAsmConservative && MS.CompileKernel)
3203  visitAsmInstruction(I);
3204  else
3205  visitInstruction(I);
3206  return;
3207  }
3208 
3209  assert(!isa<IntrinsicInst>(&I) && "intrinsics are handled elsewhere");
3210 
3211  // We are going to insert code that relies on the fact that the callee
3212  // will become a non-readonly function after it is instrumented by us. To
3213  // prevent this code from being optimized out, mark that function
3214  // non-readonly in advance.
3215  if (Function *Func = Call->getCalledFunction()) {
3216  // Clear out readonly/readnone attributes.
3217  AttrBuilder B;
3218  B.addAttribute(Attribute::ReadOnly)
3219  .addAttribute(Attribute::ReadNone);
3221  }
3222 
3224  }
3225  IRBuilder<> IRB(&I);
3226 
3227  unsigned ArgOffset = 0;
3228  LLVM_DEBUG(dbgs() << " CallSite: " << I << "\n");
3229  for (CallSite::arg_iterator ArgIt = CS.arg_begin(), End = CS.arg_end();
3230  ArgIt != End; ++ArgIt) {
3231  Value *A = *ArgIt;
3232  unsigned i = ArgIt - CS.arg_begin();
3233  if (!A->getType()->isSized()) {
3234  LLVM_DEBUG(dbgs() << "Arg " << i << " is not sized: " << I << "\n");
3235  continue;
3236  }
3237  unsigned Size = 0;
3238  Value *Store = nullptr;
3239  // Compute the Shadow for arg even if it is ByVal, because
3240  // in that case getShadow() will copy the actual arg shadow to
3241  // __msan_param_tls.
3242  Value *ArgShadow = getShadow(A);
3243  Value *ArgShadowBase = getShadowPtrForArgument(A, IRB, ArgOffset);
3244  LLVM_DEBUG(dbgs() << " Arg#" << i << ": " << *A
3245  << " Shadow: " << *ArgShadow << "\n");
3246  bool ArgIsInitialized = false;
3247  const DataLayout &DL = F.getParent()->getDataLayout();
3248  if (CS.paramHasAttr(i, Attribute::ByVal)) {
3249  assert(A->getType()->isPointerTy() &&
3250  "ByVal argument is not a pointer!");
3251  Size = DL.getTypeAllocSize(A->getType()->getPointerElementType());
3252  if (ArgOffset + Size > kParamTLSSize) break;
3253  unsigned ParamAlignment = CS.getParamAlignment(i);
3254  unsigned Alignment = std::min(ParamAlignment, kShadowTLSAlignment);
3255  Value *AShadowPtr =
3256  getShadowOriginPtr(A, IRB, IRB.getInt8Ty(), Alignment,
3257  /*isStore*/ false)
3258  .first;
3259 
3260  Store = IRB.CreateMemCpy(ArgShadowBase, Alignment, AShadowPtr,
3261  Alignment, Size);
3262  // TODO(glider): need to copy origins.
3263  } else {
3264  Size = DL.getTypeAllocSize(A->getType());
3265  if (ArgOffset + Size > kParamTLSSize) break;
3266  Store = IRB.CreateAlignedStore(ArgShadow, ArgShadowBase,
3267  kShadowTLSAlignment);
3268  Constant *Cst = dyn_cast<Constant>(ArgShadow);
3269  if (Cst && Cst->isNullValue()) ArgIsInitialized = true;
3270  }
3271  if (MS.TrackOrigins && !ArgIsInitialized)
3272  IRB.CreateStore(getOrigin(A),
3273  getOriginPtrForArgument(A, IRB, ArgOffset));
3274  (void)Store;
3275  assert(Size != 0 && Store != nullptr);
3276  LLVM_DEBUG(dbgs() << " Param:" << *Store << "\n");
3277  ArgOffset += alignTo(Size, 8);
3278  }
3279  LLVM_DEBUG(dbgs() << " done with call args\n");
3280 
3281  FunctionType *FT = CS.getFunctionType();
3282  if (FT->isVarArg()) {
3283  VAHelper->visitCallSite(CS, IRB);
3284  }
3285 
3286  // Now, get the shadow for the RetVal.
3287  if (!I.getType()->isSized()) return;
3288  // Don't emit the epilogue for musttail call returns.
3289  if (CS.isCall() && cast<CallInst>(&I)->isMustTailCall()) return;
3290  IRBuilder<> IRBBefore(&I);
3291  // Until we have full dynamic coverage, make sure the retval shadow is 0.
3292  Value *Base = getShadowPtrForRetval(&I, IRBBefore);
3293  IRBBefore.CreateAlignedStore(getCleanShadow(&I), Base, kShadowTLSAlignment);
3294  BasicBlock::iterator NextInsn;
3295  if (CS.isCall()) {
3296  NextInsn = ++I.getIterator();
3297  assert(NextInsn != I.getParent()->end());
3298  } else {
3299  BasicBlock *NormalDest = cast<InvokeInst>(&I)->getNormalDest();
3300  if (!NormalDest->getSinglePredecessor()) {
3301  // FIXME: this case is tricky, so we are just conservative here.
3302  // Perhaps we need to split the edge between this BB and NormalDest,
3303  // but a naive attempt to use SplitEdge leads to a crash.
3304  setShadow(&I, getCleanShadow(&I));
3305  setOrigin(&I, getCleanOrigin());
3306  return;
3307  }
3308  // FIXME: NextInsn is likely in a basic block that has not been visited yet.
3309  // Anything inserted there will be instrumented by MSan later!
3310  NextInsn = NormalDest->getFirstInsertionPt();
3311  assert(NextInsn != NormalDest->end() &&
3312  "Could not find insertion point for retval shadow load");
3313  }
3314  IRBuilder<> IRBAfter(&*NextInsn);
3315  Value *RetvalShadow = IRBAfter.CreateAlignedLoad(
3316  getShadowTy(&I), getShadowPtrForRetval(&I, IRBAfter),
3317  kShadowTLSAlignment, "_msret");
3318  setShadow(&I, RetvalShadow);
3319  if (MS.TrackOrigins)
3320  setOrigin(&I, IRBAfter.CreateLoad(MS.OriginTy,
3321  getOriginPtrForRetval(IRBAfter)));
3322  }
3323 
3324  bool isAMustTailRetVal(Value *RetVal) {
3325  if (auto *I = dyn_cast<BitCastInst>(RetVal)) {
3326  RetVal = I->getOperand(0);
3327  }
3328  if (auto *I = dyn_cast<CallInst>(RetVal)) {
3329  return I->isMustTailCall();
3330  }
3331  return false;
3332  }
3333 
3334  void visitReturnInst(ReturnInst &I) {
3335  IRBuilder<> IRB(&I);
3336  Value *RetVal = I.getReturnValue();
3337  if (!RetVal) return;
3338  // Don't emit the epilogue for musttail call returns.
3339  if (isAMustTailRetVal(RetVal)) return;
3340  Value *ShadowPtr = getShadowPtrForRetval(RetVal, IRB);
3341  if (CheckReturnValue) {
3342  insertShadowCheck(RetVal, &I);
3343  Value *Shadow = getCleanShadow(RetVal);
3344  IRB.CreateAlignedStore(Shadow, ShadowPtr, kShadowTLSAlignment);
3345  } else {
3346  Value *Shadow = getShadow(RetVal);
3347  IRB.CreateAlignedStore(Shadow, ShadowPtr, kShadowTLSAlignment);
3348  if (MS.TrackOrigins)
3349  IRB.CreateStore(getOrigin(RetVal), getOriginPtrForRetval(IRB));
3350  }
3351  }
3352 
3353  void visitPHINode(PHINode &I) {
3354  IRBuilder<> IRB(&I);
3355  if (!PropagateShadow) {
3356  setShadow(&I, getCleanShadow(&I));
3357  setOrigin(&I, getCleanOrigin());
3358  return;
3359  }
3360 
3361  ShadowPHINodes.push_back(&I);
3362  setShadow(&I, IRB.CreatePHI(getShadowTy(&I), I.getNumIncomingValues(),
3363  "_msphi_s"));
3364  if (MS.TrackOrigins)
3365  setOrigin(&I, IRB.CreatePHI(MS.OriginTy, I.getNumIncomingValues(),
3366  "_msphi_o"));
3367  }
3368 
3369  Value *getLocalVarDescription(AllocaInst &I) {
3370  SmallString<2048> StackDescriptionStorage;
3371  raw_svector_ostream StackDescription(StackDescriptionStorage);
3372  // We create a string with a description of the stack allocation and
3373  // pass it into __msan_set_alloca_origin.
3374  // It will be printed by the run-time if stack-originated UMR is found.
3375  // The first 4 bytes of the string are set to '----' and will be replaced
3376  // by __msan_va_arg_overflow_size_tls at the first call.
3377  StackDescription << "----" << I.getName() << "@" << F.getName();
3379  StackDescription.str());
3380  }
3381 
3382  void instrumentAllocaUserspace(AllocaInst &I, IRBuilder<> &IRB, Value *Len) {
3383  if (PoisonStack && ClPoisonStackWithCall) {
3384  IRB.CreateCall(MS.MsanPoisonStackFn,
3385  {IRB.CreatePointerCast(&I, IRB.getInt8PtrTy()), Len});
3386  } else {
3387  Value *ShadowBase, *OriginBase;
3388  std::tie(ShadowBase, OriginBase) =
3389  getShadowOriginPtr(&I, IRB, IRB.getInt8Ty(), 1, /*isStore*/ true);
3390 
3391  Value *PoisonValue = IRB.getInt8(PoisonStack ? ClPoisonStackPattern : 0);
3392  IRB.CreateMemSet(ShadowBase, PoisonValue, Len, I.getAlignment());
3393  }
3394 
3395  if (PoisonStack && MS.TrackOrigins) {
3396  Value *Descr = getLocalVarDescription(I);
3397  IRB.CreateCall(MS.MsanSetAllocaOrigin4Fn,
3398  {IRB.CreatePointerCast(&I, IRB.getInt8PtrTy()), Len,
3399  IRB.CreatePointerCast(Descr, IRB.getInt8PtrTy()),
3400  IRB.CreatePointerCast(&F, MS.IntptrTy)});
3401  }
3402  }
3403 
3404  void instrumentAllocaKmsan(AllocaInst &I, IRBuilder<> &IRB, Value *Len) {
3405  Value *Descr = getLocalVarDescription(I);
3406  if (PoisonStack) {
3407  IRB.CreateCall(MS.MsanPoisonAllocaFn,
3408  {IRB.CreatePointerCast(&I, IRB.getInt8PtrTy()), Len,
3409  IRB.CreatePointerCast(Descr, IRB.getInt8PtrTy())});
3410  } else {
3411  IRB.CreateCall(MS.MsanUnpoisonAllocaFn,
3412  {IRB.CreatePointerCast(&I, IRB.getInt8PtrTy()), Len});
3413  }
3414  }
3415 
3416  void visitAllocaInst(AllocaInst &I) {
3417  setShadow(&I, getCleanShadow(&I));
3418  setOrigin(&I, getCleanOrigin());
3419  IRBuilder<> IRB(I.getNextNode());
3420  const DataLayout &DL = F.getParent()->getDataLayout();
3421  uint64_t TypeSize = DL.getTypeAllocSize(I.getAllocatedType());
3422  Value *Len = ConstantInt::get(MS.IntptrTy, TypeSize);
3423  if (I.isArrayAllocation())
3424  Len = IRB.CreateMul(Len, I.getArraySize());
3425 
3426  if (MS.CompileKernel)
3427  instrumentAllocaKmsan(I, IRB, Len);
3428  else
3429  instrumentAllocaUserspace(I, IRB, Len);
3430  }
3431 
3432  void visitSelectInst(SelectInst& I) {
3433  IRBuilder<> IRB(&I);
3434  // a = select b, c, d
3435  Value *B = I.getCondition();
3436  Value *C = I.getTrueValue();
3437  Value *D = I.getFalseValue();
3438  Value *Sb = getShadow(B);
3439  Value *Sc = getShadow(C);
3440  Value *Sd = getShadow(D);
3441 
3442  // Result shadow if condition shadow is 0.
3443  Value *Sa0 = IRB.CreateSelect(B, Sc, Sd);
3444  Value *Sa1;
3445  if (I.getType()->isAggregateType()) {
3446  // To avoid "sign extending" i1 to an arbitrary aggregate type, we just do
3447  // an extra "select". This results in much more compact IR.
3448  // Sa = select Sb, poisoned, (select b, Sc, Sd)
3449  Sa1 = getPoisonedShadow(getShadowTy(I.getType()));
3450  } else {
3451  // Sa = select Sb, [ (c^d) | Sc | Sd ], [ b ? Sc : Sd ]
3452  // If Sb (condition is poisoned), look for bits in c and d that are equal
3453  // and both unpoisoned.
3454  // If !Sb (condition is unpoisoned), simply pick one of Sc and Sd.
3455 
3456  // Cast arguments to shadow-compatible type.
3457  C = CreateAppToShadowCast(IRB, C);
3458  D = CreateAppToShadowCast(IRB, D);
3459 
3460  // Result shadow if condition shadow is 1.
3461  Sa1 = IRB.CreateOr(IRB.CreateXor(C, D), IRB.CreateOr(Sc, Sd));
3462  }
3463  Value *Sa = IRB.CreateSelect(Sb, Sa1, Sa0, "_msprop_select");
3464  setShadow(&I, Sa);
3465  if (MS.TrackOrigins) {
3466  // Origins are always i32, so any vector conditions must be flattened.
3467  // FIXME: consider tracking vector origins for app vectors?
3468  if (B->getType()->isVectorTy()) {
3469  Type *FlatTy = getShadowTyNoVec(B->getType());
3470  B = IRB.CreateICmpNE(IRB.CreateBitCast(B, FlatTy),
3471  ConstantInt::getNullValue(FlatTy));
3472  Sb = IRB.CreateICmpNE(IRB.CreateBitCast(Sb, FlatTy),
3473  ConstantInt::getNullValue(FlatTy));
3474  }
3475  // a = select b, c, d
3476  // Oa = Sb ? Ob : (b ? Oc : Od)
3477  setOrigin(
3478  &I, IRB.CreateSelect(Sb, getOrigin(I.getCondition()),
3479  IRB.CreateSelect(B, getOrigin(I.getTrueValue()),
3480  getOrigin(I.getFalseValue()))));
3481  }
3482  }
3483 
3484  void visitLandingPadInst(LandingPadInst &I) {
3485  // Do nothing.
3486  // See https://github.com/google/sanitizers/issues/504
3487  setShadow(&I, getCleanShadow(&I));
3488  setOrigin(&I, getCleanOrigin());
3489  }
3490 
3491  void visitCatchSwitchInst(CatchSwitchInst &I) {
3492  setShadow(&I, getCleanShadow(&I));
3493  setOrigin(&I, getCleanOrigin());
3494  }
3495 
3496  void visitFuncletPadInst(FuncletPadInst &I) {
3497  setShadow(&I, getCleanShadow(&I));
3498  setOrigin(&I, getCleanOrigin());
3499  }
3500 
3501  void visitGetElementPtrInst(GetElementPtrInst &I) {
3502  handleShadowOr(I);
3503  }
3504 
3505  void visitExtractValueInst(ExtractValueInst &I) {
3506  IRBuilder<> IRB(&I);
3507  Value *Agg = I.getAggregateOperand();
3508  LLVM_DEBUG(dbgs() << "ExtractValue: " << I << "\n");
3509  Value *AggShadow = getShadow(Agg);
3510  LLVM_DEBUG(dbgs() << " AggShadow: " << *AggShadow << "\n");
3511  Value *ResShadow = IRB.CreateExtractValue(AggShadow, I.getIndices());
3512  LLVM_DEBUG(dbgs() << " ResShadow: " << *ResShadow << "\n");
3513  setShadow(&I, ResShadow);
3514  setOriginForNaryOp(I);
3515  }
3516 
3517  void visitInsertValueInst(InsertValueInst &I) {
3518  IRBuilder<> IRB(&I);
3519  LLVM_DEBUG(dbgs() << "InsertValue: " << I << "\n");
3520  Value *AggShadow = getShadow(I.getAggregateOperand());
3521  Value *InsShadow = getShadow(I.getInsertedValueOperand());
3522  LLVM_DEBUG(dbgs() << " AggShadow: " << *AggShadow << "\n");
3523  LLVM_DEBUG(dbgs() << " InsShadow: " << *InsShadow << "\n");
3524  Value *Res = IRB.CreateInsertValue(AggShadow, InsShadow, I.getIndices());
3525  LLVM_DEBUG(dbgs() << " Res: " << *Res << "\n");
3526  setShadow(&I, Res);
3527  setOriginForNaryOp(I);
3528  }
3529 
3530  void dumpInst(Instruction &I) {
3531  if (CallInst *CI = dyn_cast<CallInst>(&I)) {
3532  errs() << "ZZZ call " << CI->getCalledFunction()->getName() << "\n";
3533  } else {
3534  errs() << "ZZZ " << I.getOpcodeName() << "\n";
3535  }
3536  errs() << "QQQ " << I << "\n";
3537  }
3538 
3539  void visitResumeInst(ResumeInst &I) {
3540  LLVM_DEBUG(dbgs() << "Resume: " << I << "\n");
3541  // Nothing to do here.
3542  }
3543 
3544  void visitCleanupReturnInst(CleanupReturnInst &CRI) {
3545  LLVM_DEBUG(dbgs() << "CleanupReturn: " << CRI << "\n");
3546  // Nothing to do here.
3547  }
3548 
3549  void visitCatchReturnInst(CatchReturnInst &CRI) {
3550  LLVM_DEBUG(dbgs() << "CatchReturn: " << CRI << "\n");
3551  // Nothing to do here.
3552  }
3553 
3554  void instrumentAsmArgument(Value *Operand, Instruction &I, IRBuilder<> &IRB,
3555  const DataLayout &DL, bool isOutput) {
3556  // For each assembly argument, we check its value for being initialized.
3557  // If the argument is a pointer, we assume it points to a single element
3558  // of the corresponding type (or to a 8-byte word, if the type is unsized).
3559  // Each such pointer is instrumented with a call to the runtime library.
3560  Type *OpType = Operand->getType();
3561  // Check the operand value itself.
3562  insertShadowCheck(Operand, &I);
3563  if (!OpType->isPointerTy() || !isOutput) {
3564  assert(!isOutput);
3565  return;
3566  }
3567  Type *ElType = OpType->getPointerElementType();
3568  if (!ElType->isSized())
3569  return;
3570  int Size = DL.getTypeStoreSize(ElType);
3571  Value *Ptr = IRB.CreatePointerCast(Operand, IRB.getInt8PtrTy());
3572  Value *SizeVal = ConstantInt::get(MS.IntptrTy, Size);
3573  IRB.CreateCall(MS.MsanInstrumentAsmStoreFn, {Ptr, SizeVal});
3574  }
3575 
3576  /// Get the number of output arguments returned by pointers.
3577  int getNumOutputArgs(InlineAsm *IA, CallInst *CI) {
3578  int NumRetOutputs = 0;
3579  int NumOutputs = 0;
3580  Type *RetTy = dyn_cast<Value>(CI)->getType();
3581  if (!RetTy->isVoidTy()) {
3582  // Register outputs are returned via the CallInst return value.
3583  StructType *ST = dyn_cast_or_null<StructType>(RetTy);
3584  if (ST)
3585  NumRetOutputs = ST->getNumElements();
3586  else
3587  NumRetOutputs = 1;
3588  }
3590  for (size_t i = 0, n = Constraints.size(); i < n; i++) {
3591  InlineAsm::ConstraintInfo Info = Constraints[i];
3592  switch (Info.Type) {
3593  case InlineAsm::isOutput:
3594  NumOutputs++;
3595  break;
3596  default:
3597  break;
3598  }
3599  }
3600  return NumOutputs - NumRetOutputs;
3601  }
3602 
3603  void visitAsmInstruction(Instruction &I) {
3604  // Conservative inline assembly handling: check for poisoned shadow of
3605  // asm() arguments, then unpoison the result and all the memory locations
3606  // pointed to by those arguments.
3607  // An inline asm() statement in C++ contains lists of input and output
3608  // arguments used by the assembly code. These are mapped to operands of the
3609  // CallInst as follows:
3610  // - nR register outputs ("=r) are returned by value in a single structure
3611  // (SSA value of the CallInst);
3612  // - nO other outputs ("=m" and others) are returned by pointer as first
3613  // nO operands of the CallInst;
3614  // - nI inputs ("r", "m" and others) are passed to CallInst as the
3615  // remaining nI operands.
3616  // The total number of asm() arguments in the source is nR+nO+nI, and the
3617  // corresponding CallInst has nO+nI+1 operands (the last operand is the
3618  // function to be called).
3619  const DataLayout &DL = F.getParent()->getDataLayout();
3620  CallInst *CI = dyn_cast<CallInst>(&I);
3621  IRBuilder<> IRB(&I);
3622  InlineAsm *IA = cast<InlineAsm>(CI->getCalledValue());
3623  int OutputArgs = getNumOutputArgs(IA, CI);
3624  // The last operand of a CallInst is the function itself.
3625  int NumOperands = CI->getNumOperands() - 1;
3626 
3627  // Check input arguments. Doing so before unpoisoning output arguments, so
3628  // that we won't overwrite uninit values before checking them.
3629  for (int i = OutputArgs; i < NumOperands; i++) {
3630  Value *Operand = CI->getOperand(i);
3631  instrumentAsmArgument(Operand, I, IRB, DL, /*isOutput*/ false);
3632  }
3633  // Unpoison output arguments. This must happen before the actual InlineAsm
3634  // call, so that the shadow for memory published in the asm() statement
3635  // remains valid.
3636  for (int i = 0; i < OutputArgs; i++) {
3637  Value *Operand = CI->getOperand(i);
3638  instrumentAsmArgument(Operand, I, IRB, DL, /*isOutput*/ true);
3639  }
3640 
3641  setShadow(&I, getCleanShadow(&I));
3642  setOrigin(&I, getCleanOrigin());
3643  }
3644 
3645  void visitInstruction(Instruction &I) {
3646  // Everything else: stop propagating and check for poisoned shadow.
3648  dumpInst(I);
3649  LLVM_DEBUG(dbgs() << "DEFAULT: " << I << "\n");
3650  for (size_t i = 0, n = I.getNumOperands(); i < n; i++) {
3651  Value *Operand = I.getOperand(i);
3652  if (Operand->getType()->isSized())
3653  insertShadowCheck(Operand, &I);
3654  }
3655  setShadow(&I, getCleanShadow(&I));
3656  setOrigin(&I, getCleanOrigin());
3657  }
3658 };
3659 
3660 /// AMD64-specific implementation of VarArgHelper.
3661 struct VarArgAMD64Helper : public VarArgHelper {
3662  // An unfortunate workaround for asymmetric lowering of va_arg stuff.
3663  // See a comment in visitCallSite for more details.
3664  static const unsigned AMD64GpEndOffset = 48; // AMD64 ABI Draft 0.99.6 p3.5.7
3665  static const unsigned AMD64FpEndOffsetSSE = 176;
3666  // If SSE is disabled, fp_offset in va_list is zero.
3667  static const unsigned AMD64FpEndOffsetNoSSE = AMD64GpEndOffset;
3668 
3669  unsigned AMD64FpEndOffset;
3670  Function &F;
3671  MemorySanitizer &MS;
3672  MemorySanitizerVisitor &MSV;
3673  Value *VAArgTLSCopy = nullptr;
3674  Value *VAArgTLSOriginCopy = nullptr;
3675  Value *VAArgOverflowSize = nullptr;
3676 
3677  SmallVector<CallInst*, 16> VAStartInstrumentationList;
3678 
3679  enum ArgKind { AK_GeneralPurpose, AK_FloatingPoint, AK_Memory };
3680 
3681  VarArgAMD64Helper(Function &F, MemorySanitizer &MS,
3682  MemorySanitizerVisitor &MSV)
3683  : F(F), MS(MS), MSV(MSV) {
3684  AMD64FpEndOffset = AMD64FpEndOffsetSSE;
3685  for (const auto &Attr : F.getAttributes().getFnAttributes()) {
3686  if (Attr.isStringAttribute() &&
3687  (Attr.getKindAsString() == "target-features")) {
3688  if (Attr.getValueAsString().contains("-sse"))
3689  AMD64FpEndOffset = AMD64FpEndOffsetNoSSE;
3690  break;
3691  }
3692  }
3693  }
3694 
3695  ArgKind classifyArgument(Value* arg) {
3696  // A very rough approximation of X86_64 argument classification rules.
3697  Type *T = arg->getType();
3698  if (T->isFPOrFPVectorTy() || T->isX86_MMXTy())
3699  return AK_FloatingPoint;
3700  if (T->isIntegerTy() && T->getPrimitiveSizeInBits() <= 64)
3701  return AK_GeneralPurpose;
3702  if (T->isPointerTy())
3703  return AK_GeneralPurpose;
3704  return AK_Memory;
3705  }
3706 
3707  // For VarArg functions, store the argument shadow in an ABI-specific format
3708  // that corresponds to va_list layout.
3709  // We do this because Clang lowers va_arg in the frontend, and this pass
3710  // only sees the low level code that deals with va_list internals.
3711  // A much easier alternative (provided that Clang emits va_arg instructions)
3712  // would have been to associate each live instance of va_list with a copy of
3713  // MSanParamTLS, and extract shadow on va_arg() call in the argument list
3714  // order.
3715  void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {
3716  unsigned GpOffset = 0;
3717  unsigned FpOffset = AMD64GpEndOffset;
3718  unsigned OverflowOffset = AMD64FpEndOffset;
3719  const DataLayout &DL = F.getParent()->getDataLayout();
3720  for (CallSite::arg_iterator ArgIt = CS.arg_begin(), End = CS.arg_end();
3721  ArgIt != End; ++ArgIt) {
3722  Value *A = *ArgIt;
3723  unsigned ArgNo = CS.getArgumentNo(ArgIt);
3724  bool IsFixed = ArgNo < CS.getFunctionType()->getNumParams();
3725  bool IsByVal = CS.paramHasAttr(ArgNo, Attribute::ByVal);
3726  if (IsByVal) {
3727  // ByVal arguments always go to the overflow area.
3728  // Fixed arguments passed through the overflow area will be stepped
3729  // over by va_start, so don't count them towards the offset.
3730  if (IsFixed)
3731  continue;
3732  assert(A->getType()->isPointerTy());
3733  Type *RealTy = A->getType()->getPointerElementType();
3734  uint64_t ArgSize = DL.getTypeAllocSize(RealTy);
3735  Value *ShadowBase = getShadowPtrForVAArgument(
3736  RealTy, IRB, OverflowOffset, alignTo(ArgSize, 8));
3737  Value *OriginBase = nullptr;
3738  if (MS.TrackOrigins)
3739  OriginBase = getOriginPtrForVAArgument(RealTy, IRB, OverflowOffset);
3740  OverflowOffset += alignTo(ArgSize, 8);
3741  if (!ShadowBase)
3742  continue;
3743  Value *ShadowPtr, *OriginPtr;
3744  std::tie(ShadowPtr, OriginPtr) =
3745  MSV.getShadowOriginPtr(A, IRB, IRB.getInt8Ty(), kShadowTLSAlignment,
3746  /*isStore*/ false);
3747 
3748  IRB.CreateMemCpy(ShadowBase, kShadowTLSAlignment, ShadowPtr,
3749  kShadowTLSAlignment, ArgSize);
3750  if (MS.TrackOrigins)
3751  IRB.CreateMemCpy(OriginBase, kShadowTLSAlignment, OriginPtr,
3752  kShadowTLSAlignment, ArgSize);
3753  } else {
3754  ArgKind AK = classifyArgument(A);
3755  if (AK == AK_GeneralPurpose && GpOffset >= AMD64GpEndOffset)
3756  AK = AK_Memory;
3757  if (AK == AK_FloatingPoint && FpOffset >= AMD64FpEndOffset)
3758  AK = AK_Memory;
3759  Value *ShadowBase, *OriginBase = nullptr;
3760  switch (AK) {
3761  case AK_GeneralPurpose:
3762  ShadowBase =
3763  getShadowPtrForVAArgument(A->getType(), IRB, GpOffset, 8);
3764  if (MS.TrackOrigins)
3765  OriginBase =
3766  getOriginPtrForVAArgument(A->getType(), IRB, GpOffset);
3767  GpOffset += 8;
3768  break;
3769  case AK_FloatingPoint:
3770  ShadowBase =
3771  getShadowPtrForVAArgument(A->getType(), IRB, FpOffset, 16);
3772  if (MS.TrackOrigins)
3773  OriginBase =
3774  getOriginPtrForVAArgument(A->getType(), IRB, FpOffset);
3775  FpOffset += 16;
3776  break;
3777  case AK_Memory:
3778  if (IsFixed)
3779  continue;
3780  uint64_t ArgSize = DL.getTypeAllocSize(A->getType());
3781  ShadowBase =
3782  getShadowPtrForVAArgument(A->getType(), IRB, OverflowOffset, 8);
3783  if (MS.TrackOrigins)
3784  OriginBase =
3785  getOriginPtrForVAArgument(A->getType(), IRB, OverflowOffset);
3786  OverflowOffset += alignTo(ArgSize, 8);
3787  }
3788  // Take fixed arguments into account for GpOffset and FpOffset,
3789  // but don't actually store shadows for them.
3790  // TODO(glider): don't call get*PtrForVAArgument() for them.
3791  if (IsFixed)
3792  continue;
3793  if (!ShadowBase)
3794  continue;
3795  Value *Shadow = MSV.getShadow(A);
3796  IRB.CreateAlignedStore(Shadow, ShadowBase, kShadowTLSAlignment);
3797  if (MS.TrackOrigins) {
3798  Value *Origin = MSV.getOrigin(A);
3799  unsigned StoreSize = DL.getTypeStoreSize(Shadow->getType());
3800  MSV.paintOrigin(IRB, Origin, OriginBase, StoreSize,
3801  std::max(kShadowTLSAlignment, kMinOriginAlignment));
3802  }
3803  }
3804  }
3805  Constant *OverflowSize =
3806  ConstantInt::get(IRB.getInt64Ty(), OverflowOffset - AMD64FpEndOffset);
3807  IRB.CreateStore(OverflowSize, MS.VAArgOverflowSizeTLS);
3808  }
3809 
3810  /// Compute the shadow address for a given va_arg.
3811  Value *getShadowPtrForVAArgument(Type *Ty, IRBuilder<> &IRB,
3812  unsigned ArgOffset, unsigned ArgSize) {
3813  // Make sure we don't overflow __msan_va_arg_tls.
3814  if (ArgOffset + ArgSize > kParamTLSSize)
3815  return nullptr;
3816  Value *Base = IRB.CreatePointerCast(MS.VAArgTLS, MS.IntptrTy);
3817  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
3818  return IRB.CreateIntToPtr(Base, PointerType::get(MSV.getShadowTy(Ty), 0),
3819  "_msarg_va_s");
3820  }
3821 
3822  /// Compute the origin address for a given va_arg.
3823  Value *getOriginPtrForVAArgument(Type *Ty, IRBuilder<> &IRB, int ArgOffset) {
3824  Value *Base = IRB.CreatePointerCast(MS.VAArgOriginTLS, MS.IntptrTy);
3825  // getOriginPtrForVAArgument() is always called after
3826  // getShadowPtrForVAArgument(), so __msan_va_arg_origin_tls can never
3827  // overflow.
3828  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
3829  return IRB.CreateIntToPtr(Base, PointerType::get(MS.OriginTy, 0),
3830  "_msarg_va_o");
3831  }
3832 
3833  void unpoisonVAListTagForInst(IntrinsicInst &I) {
3834  IRBuilder<> IRB(&I);
3835  Value *VAListTag = I.getArgOperand(0);
3836  Value *ShadowPtr, *OriginPtr;
3837  unsigned Alignment = 8;
3838  std::tie(ShadowPtr, OriginPtr) =
3839  MSV.getShadowOriginPtr(VAListTag, IRB, IRB.getInt8Ty(), Alignment,
3840  /*isStore*/ true);
3841 
3842  // Unpoison the whole __va_list_tag.
3843  // FIXME: magic ABI constants.
3844  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
3845  /* size */ 24, Alignment, false);
3846  // We shouldn't need to zero out the origins, as they're only checked for
3847  // nonzero shadow.
3848  }
3849 
3850  void visitVAStartInst(VAStartInst &I) override {
3852  return;
3853  VAStartInstrumentationList.push_back(&I);
3854  unpoisonVAListTagForInst(I);
3855  }
3856 
3857  void visitVACopyInst(VACopyInst &I) override {
3858  if (F.getCallingConv() == CallingConv::Win64) return;
3859  unpoisonVAListTagForInst(I);
3860  }
3861 
3862  void finalizeInstrumentation() override {
3863  assert(!VAArgOverflowSize && !VAArgTLSCopy &&
3864  "finalizeInstrumentation called twice");
3865  if (!VAStartInstrumentationList.empty()) {
3866  // If there is a va_start in this function, make a backup copy of
3867  // va_arg_tls somewhere in the function entry block.
3868  IRBuilder<> IRB(MSV.ActualFnStart->getFirstNonPHI());
3869  VAArgOverflowSize =
3870  IRB.CreateLoad(IRB.getInt64Ty(), MS.VAArgOverflowSizeTLS);
3871  Value *CopySize =
3872  IRB.CreateAdd(ConstantInt::get(MS.IntptrTy, AMD64FpEndOffset),
3873  VAArgOverflowSize);
3874  VAArgTLSCopy = IRB.CreateAlloca(Type::getInt8Ty(*MS.C), CopySize);
3875  IRB.CreateMemCpy(VAArgTLSCopy, 8, MS.VAArgTLS, 8, CopySize);
3876  if (MS.TrackOrigins) {
3877  VAArgTLSOriginCopy = IRB.CreateAlloca(Type::getInt8Ty(*MS.C), CopySize);
3878  IRB.CreateMemCpy(VAArgTLSOriginCopy, 8, MS.VAArgOriginTLS, 8, CopySize);
3879  }
3880  }
3881 
3882  // Instrument va_start.
3883  // Copy va_list shadow from the backup copy of the TLS contents.
3884  for (size_t i = 0, n = VAStartInstrumentationList.size(); i < n; i++) {
3885  CallInst *OrigInst = VAStartInstrumentationList[i];
3886  IRBuilder<> IRB(OrigInst->getNextNode());
3887  Value *VAListTag = OrigInst->getArgOperand(0);
3888 
3889  Type *RegSaveAreaPtrTy = Type::getInt64PtrTy(*MS.C);
3890  Value *RegSaveAreaPtrPtr = IRB.CreateIntToPtr(
3891  IRB.CreateAdd(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
3892  ConstantInt::get(MS.IntptrTy, 16)),
3893  PointerType::get(RegSaveAreaPtrTy, 0));
3894  Value *RegSaveAreaPtr =
3895  IRB.CreateLoad(RegSaveAreaPtrTy, RegSaveAreaPtrPtr);
3896  Value *RegSaveAreaShadowPtr, *RegSaveAreaOriginPtr;
3897  unsigned Alignment = 16;
3898  std::tie(RegSaveAreaShadowPtr, RegSaveAreaOriginPtr) =
3899  MSV.getShadowOriginPtr(RegSaveAreaPtr, IRB, IRB.getInt8Ty(),
3900  Alignment, /*isStore*/ true);
3901  IRB.CreateMemCpy(RegSaveAreaShadowPtr, Alignment, VAArgTLSCopy, Alignment,
3902  AMD64FpEndOffset);
3903  if (MS.TrackOrigins)
3904  IRB.CreateMemCpy(RegSaveAreaOriginPtr, Alignment, VAArgTLSOriginCopy,
3905  Alignment, AMD64FpEndOffset);
3906  Type *OverflowArgAreaPtrTy = Type::getInt64PtrTy(*MS.C);
3907  Value *OverflowArgAreaPtrPtr = IRB.CreateIntToPtr(
3908  IRB.CreateAdd(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
3909  ConstantInt::get(MS.IntptrTy, 8)),
3910  PointerType::get(OverflowArgAreaPtrTy, 0));
3911  Value *OverflowArgAreaPtr =
3912  IRB.CreateLoad(OverflowArgAreaPtrTy, OverflowArgAreaPtrPtr);
3913  Value *OverflowArgAreaShadowPtr, *OverflowArgAreaOriginPtr;
3914  std::tie(OverflowArgAreaShadowPtr, OverflowArgAreaOriginPtr) =
3915  MSV.getShadowOriginPtr(OverflowArgAreaPtr, IRB, IRB.getInt8Ty(),
3916  Alignment, /*isStore*/ true);
3917  Value *SrcPtr = IRB.CreateConstGEP1_32(IRB.getInt8Ty(), VAArgTLSCopy,
3918  AMD64FpEndOffset);
3919  IRB.CreateMemCpy(OverflowArgAreaShadowPtr, Alignment, SrcPtr, Alignment,
3920  VAArgOverflowSize);
3921  if (MS.TrackOrigins) {
3922  SrcPtr = IRB.CreateConstGEP1_32(IRB.getInt8Ty(), VAArgTLSOriginCopy,
3923  AMD64FpEndOffset);
3924  IRB.CreateMemCpy(OverflowArgAreaOriginPtr, Alignment, SrcPtr, Alignment,
3925  VAArgOverflowSize);
3926  }
3927  }
3928  }
3929 };
3930 
3931 /// MIPS64-specific implementation of VarArgHelper.
3932 struct VarArgMIPS64Helper : public VarArgHelper {
3933  Function &F;
3934  MemorySanitizer &MS;
3935  MemorySanitizerVisitor &MSV;
3936  Value *VAArgTLSCopy = nullptr;
3937  Value *VAArgSize = nullptr;
3938 
3939  SmallVector<CallInst*, 16> VAStartInstrumentationList;
3940 
3941  VarArgMIPS64Helper(Function &F, MemorySanitizer &MS,
3942  MemorySanitizerVisitor &MSV) : F(F), MS(MS), MSV(MSV) {}
3943 
3944  void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {
3945  unsigned VAArgOffset = 0;
3946  const DataLayout &DL = F.getParent()->getDataLayout();
3947  for (CallSite::arg_iterator ArgIt = CS.arg_begin() +
3948  CS.getFunctionType()->getNumParams(), End = CS.arg_end();
3949  ArgIt != End; ++ArgIt) {
3950  Triple TargetTriple(F.getParent()->getTargetTriple());
3951  Value *A = *ArgIt;
3952  Value *Base;
3953  uint64_t ArgSize = DL.getTypeAllocSize(A->getType());
3954  if (TargetTriple.getArch() == Triple::mips64) {
3955  // Adjusting the shadow for argument with size < 8 to match the placement
3956  // of bits in big endian system
3957  if (ArgSize < 8)
3958  VAArgOffset += (8 - ArgSize);
3959  }
3960  Base = getShadowPtrForVAArgument(A->getType(), IRB, VAArgOffset, ArgSize);
3961  VAArgOffset += ArgSize;
3962  VAArgOffset = alignTo(VAArgOffset, 8);
3963  if (!Base)
3964  continue;
3965  IRB.CreateAlignedStore(MSV.getShadow(A), Base, kShadowTLSAlignment);
3966  }
3967 
3968  Constant *TotalVAArgSize = ConstantInt::get(IRB.getInt64Ty(), VAArgOffset);
3969  // Here using VAArgOverflowSizeTLS as VAArgSizeTLS to avoid creation of
3970  // a new class member i.e. it is the total size of all VarArgs.
3971  IRB.CreateStore(TotalVAArgSize, MS.VAArgOverflowSizeTLS);
3972  }
3973 
3974  /// Compute the shadow address for a given va_arg.
3975  Value *getShadowPtrForVAArgument(Type *Ty, IRBuilder<> &IRB,
3976  unsigned ArgOffset, unsigned ArgSize) {
3977  // Make sure we don't overflow __msan_va_arg_tls.
3978  if (ArgOffset + ArgSize > kParamTLSSize)
3979  return nullptr;
3980  Value *Base = IRB.CreatePointerCast(MS.VAArgTLS, MS.IntptrTy);
3981  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
3982  return IRB.CreateIntToPtr(Base, PointerType::get(MSV.getShadowTy(Ty), 0),
3983  "_msarg");
3984  }
3985 
3986  void visitVAStartInst(VAStartInst &I) override {
3987  IRBuilder<> IRB(&I);
3988  VAStartInstrumentationList.push_back(&I);
3989  Value *VAListTag = I.getArgOperand(0);
3990  Value *ShadowPtr, *OriginPtr;
3991  unsigned Alignment = 8;
3992  std::tie(ShadowPtr, OriginPtr) = MSV.getShadowOriginPtr(
3993  VAListTag, IRB, IRB.getInt8Ty(), Alignment, /*isStore*/ true);
3994  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
3995  /* size */ 8, Alignment, false);
3996  }
3997 
3998  void visitVACopyInst(VACopyInst &I) override {
3999  IRBuilder<> IRB(&I);
4000  VAStartInstrumentationList.push_back(&I);
4001  Value *VAListTag = I.getArgOperand(0);
4002  Value *ShadowPtr, *OriginPtr;
4003  unsigned Alignment = 8;
4004  std::tie(ShadowPtr, OriginPtr) = MSV.getShadowOriginPtr(
4005  VAListTag, IRB, IRB.getInt8Ty(), Alignment, /*isStore*/ true);
4006  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
4007  /* size */ 8, Alignment, false);
4008  }
4009 
4010  void finalizeInstrumentation() override {
4011  assert(!VAArgSize && !VAArgTLSCopy &&
4012  "finalizeInstrumentation called twice");
4013  IRBuilder<> IRB(MSV.ActualFnStart->getFirstNonPHI());
4014  VAArgSize = IRB.CreateLoad(IRB.getInt64Ty(), MS.VAArgOverflowSizeTLS);
4015  Value *CopySize = IRB.CreateAdd(ConstantInt::get(MS.IntptrTy, 0),
4016  VAArgSize);
4017 
4018  if (!VAStartInstrumentationList.empty()) {
4019  // If there is a va_start in this function, make a backup copy of
4020  // va_arg_tls somewhere in the function entry block.
4021  VAArgTLSCopy = IRB.CreateAlloca(Type::getInt8Ty(*MS.C), CopySize);
4022  IRB.CreateMemCpy(VAArgTLSCopy, 8, MS.VAArgTLS, 8, CopySize);
4023  }
4024 
4025  // Instrument va_start.
4026  // Copy va_list shadow from the backup copy of the TLS contents.
4027  for (size_t i = 0, n = VAStartInstrumentationList.size(); i < n; i++) {
4028  CallInst *OrigInst = VAStartInstrumentationList[i];
4029  IRBuilder<> IRB(OrigInst->getNextNode());
4030  Value *VAListTag = OrigInst->getArgOperand(0);
4031  Type *RegSaveAreaPtrTy = Type::getInt64PtrTy(*MS.C);
4032  Value *RegSaveAreaPtrPtr =
4033  IRB.CreateIntToPtr(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
4034  PointerType::get(RegSaveAreaPtrTy, 0));
4035  Value *RegSaveAreaPtr =
4036  IRB.CreateLoad(RegSaveAreaPtrTy, RegSaveAreaPtrPtr);
4037  Value *RegSaveAreaShadowPtr, *RegSaveAreaOriginPtr;
4038  unsigned Alignment = 8;
4039  std::tie(RegSaveAreaShadowPtr, RegSaveAreaOriginPtr) =
4040  MSV.getShadowOriginPtr(RegSaveAreaPtr, IRB, IRB.getInt8Ty(),
4041  Alignment, /*isStore*/ true);
4042  IRB.CreateMemCpy(RegSaveAreaShadowPtr, Alignment, VAArgTLSCopy, Alignment,
4043  CopySize);
4044  }
4045  }
4046 };
4047 
4048 /// AArch64-specific implementation of VarArgHelper.
4049 struct VarArgAArch64Helper : public VarArgHelper {
4050  static const unsigned kAArch64GrArgSize = 64;
4051  static const unsigned kAArch64VrArgSize = 128;
4052 
4053  static const unsigned AArch64GrBegOffset = 0;
4054  static const unsigned AArch64GrEndOffset = kAArch64GrArgSize;
4055  // Make VR space aligned to 16 bytes.
4056  static const unsigned AArch64VrBegOffset = AArch64GrEndOffset;
4057  static const unsigned AArch64VrEndOffset = AArch64VrBegOffset
4058  + kAArch64VrArgSize;
4059  static const unsigned AArch64VAEndOffset = AArch64VrEndOffset;
4060 
4061  Function &F;
4062  MemorySanitizer &MS;
4063  MemorySanitizerVisitor &MSV;
4064  Value *VAArgTLSCopy = nullptr;
4065  Value *VAArgOverflowSize = nullptr;
4066 
4067  SmallVector<CallInst*, 16> VAStartInstrumentationList;
4068 
4069  enum ArgKind { AK_GeneralPurpose, AK_FloatingPoint, AK_Memory };
4070 
4071  VarArgAArch64Helper(Function &F, MemorySanitizer &MS,
4072  MemorySanitizerVisitor &MSV) : F(F), MS(MS), MSV(MSV) {}
4073 
4074  ArgKind classifyArgument(Value* arg) {
4075  Type *T = arg->getType();
4076  if (T->isFPOrFPVectorTy())
4077  return AK_FloatingPoint;
4078  if ((T->isIntegerTy() && T->getPrimitiveSizeInBits() <= 64)
4079  || (T->isPointerTy()))
4080  return AK_GeneralPurpose;
4081  return AK_Memory;
4082  }
4083 
4084  // The instrumentation stores the argument shadow in a non ABI-specific
4085  // format because it does not know which argument is named (since Clang,
4086  // like x86_64 case, lowers the va_args in the frontend and this pass only
4087  // sees the low level code that deals with va_list internals).
4088  // The first seven GR registers are saved in the first 56 bytes of the
4089  // va_arg tls arra, followers by the first 8 FP/SIMD registers, and then
4090  // the remaining arguments.
4091  // Using constant offset within the va_arg TLS array allows fast copy
4092  // in the finalize instrumentation.
4093  void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {
4094  unsigned GrOffset = AArch64GrBegOffset;
4095  unsigned VrOffset = AArch64VrBegOffset;
4096  unsigned OverflowOffset = AArch64VAEndOffset;
4097 
4098  const DataLayout &DL = F.getParent()->getDataLayout();
4099  for (CallSite::arg_iterator ArgIt = CS.arg_begin(), End = CS.arg_end();
4100  ArgIt != End; ++ArgIt) {
4101  Value *A = *ArgIt;
4102  unsigned ArgNo = CS.getArgumentNo(ArgIt);
4103  bool IsFixed = ArgNo < CS.getFunctionType()->getNumParams();
4104  ArgKind AK = classifyArgument(A);
4105  if (AK == AK_GeneralPurpose && GrOffset >= AArch64GrEndOffset)
4106  AK = AK_Memory;
4107  if (AK == AK_FloatingPoint && VrOffset >= AArch64VrEndOffset)
4108  AK = AK_Memory;
4109  Value *Base;
4110  switch (AK) {
4111  case AK_GeneralPurpose:
4112  Base = getShadowPtrForVAArgument(A->getType(), IRB, GrOffset, 8);
4113  GrOffset += 8;
4114  break;
4115  case AK_FloatingPoint:
4116  Base = getShadowPtrForVAArgument(A->getType(), IRB, VrOffset, 8);
4117  VrOffset += 16;
4118  break;
4119  case AK_Memory:
4120  // Don't count fixed arguments in the overflow area - va_start will
4121  // skip right over them.
4122  if (IsFixed)
4123  continue;
4124  uint64_t ArgSize = DL.getTypeAllocSize(A->getType());
4125  Base = getShadowPtrForVAArgument(A->getType(), IRB, OverflowOffset,
4126  alignTo(ArgSize, 8));
4127  OverflowOffset += alignTo(ArgSize, 8);
4128  break;
4129  }
4130  // Count Gp/Vr fixed arguments to their respective offsets, but don't
4131  // bother to actually store a shadow.
4132  if (IsFixed)
4133  continue;
4134  if (!Base)
4135  continue;
4136  IRB.CreateAlignedStore(MSV.getShadow(A), Base, kShadowTLSAlignment);
4137  }
4138  Constant *OverflowSize =
4139  ConstantInt::get(IRB.getInt64Ty(), OverflowOffset - AArch64VAEndOffset);
4140  IRB.CreateStore(OverflowSize, MS.VAArgOverflowSizeTLS);
4141  }
4142 
4143  /// Compute the shadow address for a given va_arg.
4144  Value *getShadowPtrForVAArgument(Type *Ty, IRBuilder<> &IRB,
4145  unsigned ArgOffset, unsigned ArgSize) {
4146  // Make sure we don't overflow __msan_va_arg_tls.
4147  if (ArgOffset + ArgSize > kParamTLSSize)
4148  return nullptr;
4149  Value *Base = IRB.CreatePointerCast(MS.VAArgTLS, MS.IntptrTy);
4150  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
4151  return IRB.CreateIntToPtr(Base, PointerType::get(MSV.getShadowTy(Ty), 0),
4152  "_msarg");
4153  }
4154 
4155  void visitVAStartInst(VAStartInst &I) override {
4156  IRBuilder<> IRB(&I);
4157  VAStartInstrumentationList.push_back(&I);
4158  Value *VAListTag = I.getArgOperand(0);
4159  Value *ShadowPtr, *OriginPtr;
4160  unsigned Alignment = 8;
4161  std::tie(ShadowPtr, OriginPtr) = MSV.getShadowOriginPtr(
4162  VAListTag, IRB, IRB.getInt8Ty(), Alignment, /*isStore*/ true);
4163  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
4164  /* size */ 32, Alignment, false);
4165  }
4166 
4167  void visitVACopyInst(VACopyInst &I) override {
4168  IRBuilder<> IRB(&I);
4169  VAStartInstrumentationList.push_back(&I);
4170  Value *VAListTag = I.getArgOperand(0);
4171  Value *ShadowPtr, *OriginPtr;
4172  unsigned Alignment = 8;
4173  std::tie(ShadowPtr, OriginPtr) = MSV.getShadowOriginPtr(
4174  VAListTag, IRB, IRB.getInt8Ty(), Alignment, /*isStore*/ true);
4175  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
4176  /* size */ 32, Alignment, false);
4177  }
4178 
4179  // Retrieve a va_list field of 'void*' size.
4180  Value* getVAField64(IRBuilder<> &IRB, Value *VAListTag, int offset) {
4181  Value *SaveAreaPtrPtr =
4182  IRB.CreateIntToPtr(
4183  IRB.CreateAdd(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
4184  ConstantInt::get(MS.IntptrTy, offset)),
4185  Type::getInt64PtrTy(*MS.C));
4186  return IRB.CreateLoad(Type::getInt64Ty(*MS.C), SaveAreaPtrPtr);
4187  }
4188 
4189  // Retrieve a va_list field of 'int' size.
4190  Value* getVAField32(IRBuilder<> &IRB, Value *VAListTag, int offset) {
4191  Value *SaveAreaPtr =
4192  IRB.CreateIntToPtr(
4193  IRB.CreateAdd(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
4194  ConstantInt::get(MS.IntptrTy, offset)),
4195  Type::getInt32PtrTy(*MS.C));
4196  Value *SaveArea32 = IRB.CreateLoad(IRB.getInt32Ty(), SaveAreaPtr);
4197  return IRB.CreateSExt(SaveArea32, MS.IntptrTy);
4198  }
4199 
4200  void finalizeInstrumentation() override {
4201  assert(!VAArgOverflowSize && !VAArgTLSCopy &&
4202  "finalizeInstrumentation called twice");
4203  if (!VAStartInstrumentationList.empty()) {
4204  // If there is a va_start in this function, make a backup copy of
4205  // va_arg_tls somewhere in the function entry block.
4206  IRBuilder<> IRB(MSV.ActualFnStart->getFirstNonPHI());
4207  VAArgOverflowSize =
4208  IRB.CreateLoad(IRB.getInt64Ty(), MS.VAArgOverflowSizeTLS);
4209  Value *CopySize =
4210  IRB.CreateAdd(ConstantInt::get(MS.IntptrTy, AArch64VAEndOffset),
4211  VAArgOverflowSize);
4212  VAArgTLSCopy = IRB.CreateAlloca(Type::getInt8Ty(*MS.C), CopySize);
4213  IRB.CreateMemCpy(VAArgTLSCopy, 8, MS.VAArgTLS, 8, CopySize);
4214  }
4215 
4216  Value *GrArgSize = ConstantInt::get(MS.IntptrTy, kAArch64GrArgSize);
4217  Value *VrArgSize = ConstantInt::get(MS.IntptrTy, kAArch64VrArgSize);
4218 
4219  // Instrument va_start, copy va_list shadow from the backup copy of
4220  // the TLS contents.
4221  for (size_t i = 0, n = VAStartInstrumentationList.size(); i < n; i++) {
4222  CallInst *OrigInst = VAStartInstrumentationList[i];
4223  IRBuilder<> IRB(OrigInst->getNextNode());
4224 
4225  Value *VAListTag = OrigInst->getArgOperand(0);
4226 
4227  // The variadic ABI for AArch64 creates two areas to save the incoming
4228  // argument registers (one for 64-bit general register xn-x7 and another
4229  // for 128-bit FP/SIMD vn-v7).
4230  // We need then to propagate the shadow arguments on both regions
4231  // 'va::__gr_top + va::__gr_offs' and 'va::__vr_top + va::__vr_offs'.
4232  // The remaning arguments are saved on shadow for 'va::stack'.
4233  // One caveat is it requires only to propagate the non-named arguments,
4234  // however on the call site instrumentation 'all' the arguments are
4235  // saved. So to copy the shadow values from the va_arg TLS array
4236  // we need to adjust the offset for both GR and VR fields based on
4237  // the __{gr,vr}_offs value (since they are stores based on incoming
4238  // named arguments).
4239 
4240  // Read the stack pointer from the va_list.
4241  Value *StackSaveAreaPtr = getVAField64(IRB, VAListTag, 0);
4242 
4243  // Read both the __gr_top and __gr_off and add them up.
4244  Value *GrTopSaveAreaPtr = getVAField64(IRB, VAListTag, 8);
4245  Value *GrOffSaveArea = getVAField32(IRB, VAListTag, 24);
4246 
4247  Value *GrRegSaveAreaPtr = IRB.CreateAdd(GrTopSaveAreaPtr, GrOffSaveArea);
4248 
4249  // Read both the __vr_top and __vr_off and add them up.
4250  Value *VrTopSaveAreaPtr = getVAField64(IRB, VAListTag, 16);
4251  Value *VrOffSaveArea = getVAField32(IRB, VAListTag, 28);
4252 
4253  Value *VrRegSaveAreaPtr = IRB.CreateAdd(VrTopSaveAreaPtr, VrOffSaveArea);
4254 
4255  // It does not know how many named arguments is being used and, on the
4256  // callsite all the arguments were saved. Since __gr_off is defined as
4257  // '0 - ((8 - named_gr) * 8)', the idea is to just propagate the variadic
4258  // argument by ignoring the bytes of shadow from named arguments.
4259  Value *GrRegSaveAreaShadowPtrOff =
4260  IRB.CreateAdd(GrArgSize, GrOffSaveArea);
4261 
4262  Value *GrRegSaveAreaShadowPtr =
4263  MSV.getShadowOriginPtr(GrRegSaveAreaPtr, IRB, IRB.getInt8Ty(),
4264  /*Alignment*/ 8, /*isStore*/ true)
4265  .first;
4266 
4267  Value *GrSrcPtr = IRB.CreateInBoundsGEP(IRB.getInt8Ty(), VAArgTLSCopy,
4268  GrRegSaveAreaShadowPtrOff);
4269  Value *GrCopySize = IRB.CreateSub(GrArgSize, GrRegSaveAreaShadowPtrOff);
4270 
4271  IRB.CreateMemCpy(GrRegSaveAreaShadowPtr, 8, GrSrcPtr, 8, GrCopySize);
4272 
4273  // Again, but for FP/SIMD values.
4274  Value *VrRegSaveAreaShadowPtrOff =
4275  IRB.CreateAdd(VrArgSize, VrOffSaveArea);
4276 
4277  Value *VrRegSaveAreaShadowPtr =
4278  MSV.getShadowOriginPtr(VrRegSaveAreaPtr, IRB, IRB.getInt8Ty(),
4279  /*Alignment*/ 8, /*isStore*/ true)
4280  .first;
4281 
4282  Value *VrSrcPtr = IRB.CreateInBoundsGEP(
4283  IRB.getInt8Ty(),
4284  IRB.CreateInBoundsGEP(IRB.getInt8Ty(), VAArgTLSCopy,
4285  IRB.getInt32(AArch64VrBegOffset)),
4286  VrRegSaveAreaShadowPtrOff);
4287  Value *VrCopySize = IRB.CreateSub(VrArgSize, VrRegSaveAreaShadowPtrOff);
4288 
4289  IRB.CreateMemCpy(VrRegSaveAreaShadowPtr, 8, VrSrcPtr, 8, VrCopySize);
4290 
4291  // And finally for remaining arguments.
4292  Value *StackSaveAreaShadowPtr =
4293  MSV.getShadowOriginPtr(StackSaveAreaPtr, IRB, IRB.getInt8Ty(),
4294  /*Alignment*/ 16, /*isStore*/ true)
4295  .first;
4296 
4297  Value *StackSrcPtr =
4298  IRB.CreateInBoundsGEP(IRB.getInt8Ty(), VAArgTLSCopy,
4299  IRB.getInt32(AArch64VAEndOffset));
4300 
4301  IRB.CreateMemCpy(StackSaveAreaShadowPtr, 16, StackSrcPtr, 16,
4302  VAArgOverflowSize);
4303  }
4304  }
4305 };
4306 
4307 /// PowerPC64-specific implementation of VarArgHelper.
4308 struct VarArgPowerPC64Helper : public VarArgHelper {
4309  Function &F;
4310  MemorySanitizer &MS;
4311  MemorySanitizerVisitor &MSV;
4312  Value *VAArgTLSCopy = nullptr;
4313  Value *VAArgSize = nullptr;
4314 
4315  SmallVector<CallInst*, 16> VAStartInstrumentationList;
4316 
4317  VarArgPowerPC64Helper(Function &F, MemorySanitizer &MS,
4318  MemorySanitizerVisitor &MSV) : F(F), MS(MS), MSV(MSV) {}
4319 
4320  void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {
4321  // For PowerPC, we need to deal with alignment of stack arguments -
4322  // they are mostly aligned to 8 bytes, but vectors and i128 arrays
4323  // are aligned to 16 bytes, byvals can be aligned to 8 or 16 bytes,
4324  // and QPX vectors are aligned to 32 bytes. For that reason, we
4325  // compute current offset from stack pointer (which is always properly
4326  // aligned), and offset for the first vararg, then subtract them.
4327  unsigned VAArgBase;
4328  Triple TargetTriple(F.getParent()->getTargetTriple());
4329  // Parameter save area starts at 48 bytes from frame pointer for ABIv1,
4330  // and 32 bytes for ABIv2. This is usually determined by target
4331  // endianness, but in theory could be overriden by function attribute.
4332  // For simplicity, we ignore it here (it'd only matter for QPX vectors).
4333  if (TargetTriple.getArch() == Triple::ppc64)
4334  VAArgBase = 48;
4335  else
4336  VAArgBase = 32;
4337  unsigned VAArgOffset = VAArgBase;
4338  const DataLayout &DL = F.getParent()->getDataLayout();
4339  for (CallSite::arg_iterator ArgIt = CS.arg_begin(), End = CS.arg_end();
4340  ArgIt != End; ++ArgIt) {
4341  Value *A = *ArgIt;
4342  unsigned ArgNo = CS.getArgumentNo(ArgIt);
4343  bool IsFixed = ArgNo < CS.getFunctionType()->getNumParams();
4344  bool IsByVal = CS.paramHasAttr(ArgNo, Attribute::ByVal);
4345  if (IsByVal) {
4346  assert(A->getType()->isPointerTy());
4347  Type *RealTy = A->getType()->getPointerElementType();
4348  uint64_t ArgSize = DL.getTypeAllocSize(RealTy);
4349  uint64_t ArgAlign = CS.getParamAlignment(ArgNo);
4350  if (ArgAlign < 8)
4351  ArgAlign = 8;
4352  VAArgOffset = alignTo(VAArgOffset, ArgAlign);
4353  if (!IsFixed) {
4354  Value *Base = getShadowPtrForVAArgument(
4355  RealTy, IRB, VAArgOffset - VAArgBase, ArgSize);
4356  if (Base) {
4357  Value *AShadowPtr, *AOriginPtr;
4358  std::tie(AShadowPtr, AOriginPtr) =
4359  MSV.getShadowOriginPtr(A, IRB, IRB.getInt8Ty(),
4360  kShadowTLSAlignment, /*isStore*/ false);
4361 
4362  IRB.CreateMemCpy(Base, kShadowTLSAlignment, AShadowPtr,
4363  kShadowTLSAlignment, ArgSize);
4364  }
4365  }
4366  VAArgOffset += alignTo(ArgSize, 8);
4367  } else {
4368  Value *Base;
4369  uint64_t ArgSize = DL.getTypeAllocSize(A->getType());
4370  uint64_t ArgAlign = 8;
4371  if (A->getType()->isArrayTy()) {
4372  // Arrays are aligned to element size, except for long double
4373  // arrays, which are aligned to 8 bytes.
4374  Type *ElementTy = A->getType()->getArrayElementType();
4375  if (!ElementTy->isPPC_FP128Ty())
4376  ArgAlign = DL.getTypeAllocSize(ElementTy);
4377  } else if (A->getType()->isVectorTy()) {
4378  // Vectors are naturally aligned.
4379  ArgAlign = DL.getTypeAllocSize(A->getType());
4380  }
4381  if (ArgAlign < 8)
4382  ArgAlign = 8;
4383  VAArgOffset = alignTo(VAArgOffset, ArgAlign);
4384  if (DL.isBigEndian()) {
4385  // Adjusting the shadow for argument with size < 8 to match the placement
4386  // of bits in big endian system
4387  if (ArgSize < 8)
4388  VAArgOffset += (8 - ArgSize);
4389  }
4390  if (!IsFixed) {
4391  Base = getShadowPtrForVAArgument(A->getType(), IRB,
4392  VAArgOffset - VAArgBase, ArgSize);
4393  if (Base)
4394  IRB.CreateAlignedStore(MSV.getShadow(A), Base, kShadowTLSAlignment);
4395  }
4396  VAArgOffset += ArgSize;
4397  VAArgOffset = alignTo(VAArgOffset, 8);
4398  }
4399  if (IsFixed)
4400  VAArgBase = VAArgOffset;
4401  }
4402 
4403  Constant *TotalVAArgSize = ConstantInt::get(IRB.getInt64Ty(),
4404  VAArgOffset - VAArgBase);
4405  // Here using VAArgOverflowSizeTLS as VAArgSizeTLS to avoid creation of
4406  // a new class member i.e. it is the total size of all VarArgs.
4407  IRB.CreateStore(TotalVAArgSize, MS.VAArgOverflowSizeTLS);
4408  }
4409 
4410  /// Compute the shadow address for a given va_arg.
4411  Value *getShadowPtrForVAArgument(Type *Ty, IRBuilder<> &IRB,
4412  unsigned ArgOffset, unsigned ArgSize) {
4413  // Make sure we don't overflow __msan_va_arg_tls.
4414  if (ArgOffset + ArgSize > kParamTLSSize)
4415  return nullptr;
4416  Value *Base = IRB.CreatePointerCast(MS.VAArgTLS, MS.IntptrTy);
4417  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
4418  return IRB.CreateIntToPtr(Base, PointerType::get(MSV.getShadowTy(Ty), 0),
4419  "_msarg");
4420  }
4421 
4422  void visitVAStartInst(VAStartInst &I) override {
4423  IRBuilder<> IRB(&I);
4424  VAStartInstrumentationList.push_back(&I);
4425  Value *VAListTag = I.getArgOperand(0);
4426  Value *ShadowPtr, *OriginPtr;
4427  unsigned Alignment = 8;
4428  std::tie(ShadowPtr, OriginPtr) = MSV.getShadowOriginPtr(
4429  VAListTag, IRB, IRB.getInt8Ty(), Alignment, /*isStore*/ true);
4430  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
4431  /* size */ 8, Alignment, false);
4432  }
4433 
4434  void visitVACopyInst(VACopyInst &I) override {
4435  IRBuilder<> IRB(&I);
4436  Value *VAListTag = I.getArgOperand(0);
4437  Value *ShadowPtr, *OriginPtr;
4438  unsigned Alignment = 8;
4439  std::tie(ShadowPtr, OriginPtr) = MSV.getShadowOriginPtr(
4440  VAListTag, IRB, IRB.getInt8Ty(), Alignment, /*isStore*/ true);
4441  // Unpoison the whole __va_list_tag.
4442  // FIXME: magic ABI constants.
4443  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
4444  /* size */ 8, Alignment, false);
4445  }
4446 
4447  void finalizeInstrumentation() override {
4448  assert(!VAArgSize && !VAArgTLSCopy &&
4449  "finalizeInstrumentation called twice");
4450  IRBuilder<> IRB(MSV.ActualFnStart->getFirstNonPHI());
4451  VAArgSize = IRB.CreateLoad(IRB.getInt64Ty(), MS.VAArgOverflowSizeTLS);
4452  Value *CopySize = IRB.CreateAdd(ConstantInt::get(MS.IntptrTy, 0),
4453  VAArgSize);
4454 
4455  if (!VAStartInstrumentationList.empty()) {
4456  // If there is a va_start in this function, make a backup copy of
4457  // va_arg_tls somewhere in the function entry block.
4458  VAArgTLSCopy = IRB.CreateAlloca(Type::getInt8Ty(*MS.C), CopySize);
4459  IRB.CreateMemCpy(VAArgTLSCopy, 8, MS.VAArgTLS, 8, CopySize);
4460  }
4461 
4462  // Instrument va_start.
4463  // Copy va_list shadow from the backup copy of the TLS contents.
4464  for (size_t i = 0, n = VAStartInstrumentationList.size(); i < n; i++) {
4465  CallInst *OrigInst = VAStartInstrumentationList[i];
4466  IRBuilder<> IRB(OrigInst->getNextNode());
4467  Value *VAListTag = OrigInst->getArgOperand(0);
4468  Type *RegSaveAreaPtrTy = Type::getInt64PtrTy(*MS.C);
4469  Value *RegSaveAreaPtrPtr =
4470  IRB.CreateIntToPtr(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
4471  PointerType::get(RegSaveAreaPtrTy, 0));
4472  Value *RegSaveAreaPtr =
4473  IRB.CreateLoad(RegSaveAreaPtrTy, RegSaveAreaPtrPtr);
4474  Value *RegSaveAreaShadowPtr, *RegSaveAreaOriginPtr;
4475  unsigned Alignment = 8;
4476  std::tie(RegSaveAreaShadowPtr, RegSaveAreaOriginPtr) =
4477  MSV.getShadowOriginPtr(RegSaveAreaPtr, IRB, IRB.getInt8Ty(),
4478  Alignment, /*isStore*/ true);
4479  IRB.CreateMemCpy(RegSaveAreaShadowPtr, Alignment, VAArgTLSCopy, Alignment,
4480  CopySize);
4481  }
4482  }
4483 };
4484 
4485 /// A no-op implementation of VarArgHelper.
4486 struct VarArgNoOpHelper : public VarArgHelper {
4487  VarArgNoOpHelper(Function &F, MemorySanitizer &MS,
4488  MemorySanitizerVisitor &MSV) {}
4489 
4490  void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {}
4491 
4492  void visitVAStartInst(VAStartInst &I) override {}
4493 
4494  void visitVACopyInst(VACopyInst &I) override {}
4495 
4496  void finalizeInstrumentation() override {}
4497 };
4498 
4499 } // end anonymous namespace
4500 
4501 static VarArgHelper *CreateVarArgHelper(Function &Func, MemorySanitizer &Msan,
4502  MemorySanitizerVisitor &Visitor) {
4503  // VarArg handling is only implemented on AMD64. False positives are possible
4504  // on other platforms.
4505  Triple TargetTriple(Func.getParent()->getTargetTriple());
4506  if (TargetTriple.getArch() == Triple::x86_64)
4507  return new VarArgAMD64Helper(Func, Msan, Visitor);
4508  else if (TargetTriple.isMIPS64())
4509  return new VarArgMIPS64Helper(Func, Msan, Visitor);
4510  else if (TargetTriple.getArch() == Triple::aarch64)
4511  return new VarArgAArch64Helper(Func, Msan, Visitor);
4512  else if (TargetTriple.getArch() == Triple::ppc64 ||
4513  TargetTriple.getArch() == Triple::ppc64le)
4514  return new VarArgPowerPC64Helper(Func, Msan, Visitor);
4515  else
4516  return new VarArgNoOpHelper(Func, Msan, Visitor);
4517 }
4518 
4519 bool MemorySanitizer::sanitizeFunction(Function &F, TargetLibraryInfo &TLI) {
4520  if (!CompileKernel && (&F == MsanCtorFunction))
4521  return false;
4522  MemorySanitizerVisitor Visitor(F, *this, TLI);
4523 
4524  // Clear out readonly/readnone attributes.
4525  AttrBuilder B;
4526  B.addAttribute(Attribute::ReadOnly)
4527  .addAttribute(Attribute::ReadNone);
4529 
4530  return Visitor.runOnFunction();
4531 }
Value * CreateInBoundsGEP(Value *Ptr, ArrayRef< Value *> IdxList, const Twine &Name="")
Definition: IRBuilder.h:1512
Type * getVectorElementType() const
Definition: Type.h:370
IterTy arg_end() const
Definition: CallSite.h:583
uint64_t CallInst * C
Return a value (possibly void), from a function.
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:551
A parsed version of the target data layout string in and methods for querying it. ...
Definition: DataLayout.h:110
constexpr char Align[]
Key for Kernel::Arg::Metadata::mAlign.
const std::string & getTargetTriple() const
Get the target triple which is a string describing the target host.
Definition: Module.h:240
static const MemoryMapParams Linux_PowerPC64_MemoryMapParams
Value * CreateICmp(CmpInst::Predicate P, Value *LHS, Value *RHS, const Twine &Name="")
Definition: IRBuilder.h:1984
Value * CreateConstGEP1_32(Value *Ptr, unsigned Idx0, const Twine &Name="")
Definition: IRBuilder.h:1551
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:2576
bool isAllOnesValue() const
Return true if this is the value that would be returned by getAllOnesValue.
Definition: Constants.cpp:99
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:1333
GCNRegPressure max(const GCNRegPressure &P1, const GCNRegPressure &P2)
This class represents an incoming formal argument to a Function.
Definition: Argument.h:29
static const PlatformMemoryMapParams Linux_PowerPC_MemoryMapParams
AllocaInst * CreateAlloca(Type *Ty, unsigned AddrSpace, Value *ArraySize=nullptr, const Twine &Name="")
Definition: IRBuilder.h:1379
bool doesNotAccessMemory(unsigned OpNo) const
Definition: InstrTypes.h:1465
Base class for instruction visitors.
Definition: InstVisitor.h:80
Value * getAggregateOperand()
Value * CreateICmpNE(Value *LHS, Value *RHS, const Twine &Name="")
Definition: IRBuilder.h:1878
Atomic ordering constants.
NodeTy * getNextNode()
Get the next node, or nullptr for the list tail.
Definition: ilist_node.h:288
PassT::Result & getResult(IRUnitT &IR, ExtraArgTs... ExtraArgs)
Get the result of an analysis pass for a given IR unit.
Definition: PassManager.h:769
IterTy arg_begin() const
Definition: CallSite.h:579
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:139
This class represents lattice values for constants.
Definition: AllocatorList.h:23
BinaryOps getOpcode() const
Definition: InstrTypes.h:316
bool isAtomic() const
Return true if this instruction has an AtomicOrdering of unordered or higher.
LoadInst * CreateLoad(Type *Ty, Value *Ptr, const char *Name)
Provided to resolve &#39;CreateLoad(Ty, Ptr, "...")&#39; correctly, instead of converting the string to &#39;bool...
Definition: IRBuilder.h:1392
Value * CreateXor(Value *LHS, Value *RHS, const Twine &Name="")
Definition: IRBuilder.h:1235
A Module instance is used to store all the information related to an LLVM module. ...
Definition: Module.h:65
bool isSized(SmallPtrSetImpl< Type *> *Visited=nullptr) const
Return true if it makes sense to take the size of this type.
Definition: Type.h:264
LoadInst * CreateAlignedLoad(Type *Ty, Value *Ptr, unsigned Align, const char *Name)
Provided to resolve &#39;CreateAlignedLoad(Ptr, Align, "...")&#39; correctly, instead of converting the strin...
Definition: IRBuilder.h:1428
void setOrdering(AtomicOrdering Ordering)
Sets the ordering constraint of this rmw instruction.
Definition: Instructions.h:794
static const MemoryMapParams Linux_I386_MemoryMapParams
Same, but only replaced by something equivalent.
Definition: GlobalValue.h:53
amdgpu Simplify well known AMD library false FunctionCallee Value const Twine & Name
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:528
static const MemoryMapParams NetBSD_X86_64_MemoryMapParams
A handy container for a FunctionType+Callee-pointer pair, which can be passed around as a single enti...
Definition: DerivedTypes.h:164
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
unsigned getNumElements() const
Random access to the elements.
Definition: DerivedTypes.h:344
bool removeUnreachableBlocks(Function &F, LazyValueInfo *LVI=nullptr, DomTreeUpdater *DTU=nullptr, MemorySSAUpdater *MSSAU=nullptr)
Remove all blocks that can not be reached from the function&#39;s entry.
Definition: Local.cpp:2215
Value * CreateICmpSLT(Value *LHS, Value *RHS, const Twine &Name="")
Definition: IRBuilder.h:1906
This class represents a function call, abstracting a target machine&#39;s calling convention.
static cl::opt< bool > ClHandleAsmConservative("msan-handle-asm-conservative", cl::desc("conservative handling of inline assembly"), cl::Hidden, cl::init(true))
static PointerType * getInt32PtrTy(LLVMContext &C, unsigned AS=0)
Definition: Type.cpp:227
void setOrdering(AtomicOrdering Ordering)
Sets the ordering constraint of this load instruction.
Definition: Instructions.h:253
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:629
const Value * getTrueValue() const
AtomicOrdering getOrdering() const
Returns the ordering constraint of this load instruction.
Definition: Instructions.h:247
Like Internal, but omit from symbol table.
Definition: GlobalValue.h:56
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:48
bool hasFnAttribute(Attribute::AttrKind Kind) const
Return true if the function has the attribute.
Definition: Function.h:320
A raw_ostream that writes to an SmallVector or SmallString.
Definition: raw_ostream.h:509
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:1698
Metadata node.
Definition: Metadata.h:863
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:684
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:403
An instruction for reading from memory.
Definition: Instructions.h:167
static IntegerType * getInt64Ty(LLVMContext &C)
Definition: Type.cpp:176
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:691
Hexagon Common GEP
bool isVectorTy() const
True if this is an instance of VectorType.
Definition: Type.h:229
#define op(i)
The C convention as implemented on Windows/x86-64 and AArch64.
Definition: CallingConv.h:153
bool isMustTailCall() const
static Type * getX86_MMXTy(LLVMContext &C)
Definition: Type.cpp:170
static cl::opt< unsigned long long > ClXorMask("msan-xor-mask", cl::desc("Define custom MSan XorMask"), cl::Hidden, cl::init(0))
Use * op_iterator
Definition: User.h:224
bool isPPC_FP128Ty() const
Return true if this is powerpc long double.
Definition: Type.h:158
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:229
static PointerType * getInt64PtrTy(LLVMContext &C, unsigned AS=0)
Definition: Type.cpp:231
static Constant * get(ArrayType *T, ArrayRef< Constant *> V)
Definition: Constants.cpp:992
FunctionType * getFunctionType() const
Definition: CallSite.h:328
unsigned getBitWidth() const
Return the number of bits in the APInt.
Definition: APInt.h:1508
static Constant * getNullValue(Type *Ty)
Constructor to create a &#39;0&#39; constant of arbitrary type.
Definition: Constants.cpp:274
Value * CreateNot(Value *V, const Twine &Name="")
Definition: IRBuilder.h:1369
StoreInst * CreateAlignedStore(Value *Val, Value *Ptr, unsigned Align, bool isVolatile=false)
Definition: IRBuilder.h:1465
IntegerType * getInt32Ty()
Fetch the type representing a 32-bit integer.
Definition: IRBuilder.h:346
Value * getArgOperand(unsigned i) const
Definition: InstrTypes.h:1155
unsigned countTrailingZeros() const
Count the number of trailing zero bits.
Definition: APInt.h:1631
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:50
bool isSigned() const
Definition: InstrTypes.h:816
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:375
const DataLayout & getDataLayout() const
Get the data layout for the module&#39;s target platform.
Definition: Module.cpp:369
unsigned getAlignment() const
Return the alignment of the memory that is being allocated by the instruction.
Definition: Instructions.h:112
IntegerType * getInt64Ty()
Fetch the type representing a 64-bit integer.
Definition: IRBuilder.h:351
This is the base class for all instructions that perform data casts.
Definition: InstrTypes.h:353
ArrayRef< T > makeArrayRef(const T &OneElt)
Construct an ArrayRef from a single element.
Definition: ArrayRef.h:450
&#39;undef&#39; values are things that do not have specified contents.
Definition: Constants.h:1285
This class wraps the llvm.memmove intrinsic.
Class to represent struct types.
Definition: DerivedTypes.h:232
LLVMContext & getContext() const
Get the global data context.
Definition: Module.h:244
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:651
bool isUnsigned() const
Definition: InstrTypes.h:822
static cl::opt< unsigned long long > ClAndMask("msan-and-mask", cl::desc("Define custom MSan AndMask"), cl::Hidden, cl::init(0))
bool isIntegerTy() const
True if this is an instance of IntegerType.
Definition: Type.h:196
This provides a uniform API for creating instructions and inserting them into a basic block: either a...
Definition: IRBuilder.h:742
IntegerType * getIntPtrTy(const DataLayout &DL, unsigned AddrSpace=0)
Fetch the type representing a pointer to an integer value.
Definition: IRBuilder.h:389
unsigned getArgumentNo(Value::const_user_iterator I) const
Given a value use iterator, returns the argument that corresponds to it.
Definition: CallSite.h:206
This file contains the simple types necessary to represent the attributes associated with functions a...
static cl::opt< unsigned long long > ClOriginBase("msan-origin-base", cl::desc("Define custom MSan OriginBase"), cl::Hidden, cl::init(0))
Value * CreateAdd(Value *LHS, Value *RHS, const Twine &Name="", bool HasNUW=false, bool HasNSW=false)
Definition: IRBuilder.h:1049
void setName(const Twine &Name)
Change the name of the value.
Definition: Value.cpp:285
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:341
This file implements a class to represent arbitrary precision integral constant values and operations...
Type * getVoidTy()
Fetch the type representing void.
Definition: IRBuilder.h:379
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:1421
InstrTy * getInstruction() const
Definition: CallSite.h:96
Value * CreateIntToPtr(Value *V, Type *DestTy, const Twine &Name="")
Definition: IRBuilder.h:1762
bool isNullValue() const
Return true if this is the value that would be returned by getNullValue.
Definition: Constants.cpp:84
Class to represent function types.
Definition: DerivedTypes.h:102
Value * CreateBitCast(Value *V, Type *DestTy, const Twine &Name="")
Definition: IRBuilder.h:1767
Type * getType() const
All values are typed, get the type of this value.
Definition: Value.h:244
This represents the llvm.va_start intrinsic.
static const char *const kMsanInitName
std::string itostr(int64_t X)
Definition: StringExtras.h:238
AtomicOrdering getSuccessOrdering() const
Returns the success ordering constraint of this cmpxchg instruction.
Definition: Instructions.h:582
#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:400
std::pair< Function *, FunctionCallee > getOrCreateSanitizerCtorAndInitFunctions(Module &M, StringRef CtorName, StringRef InitName, ArrayRef< Type *> InitArgTypes, ArrayRef< Value *> InitArgs, function_ref< void(Function *, FunctionCallee)> FunctionsCreatedCallback, StringRef VersionCheckName=StringRef())
Creates sanitizer constructor function lazily.
This instruction compares its operands according to the predicate given to the constructor.
static bool isStore(int Opcode)
void setComdat(Comdat *C)
Definition: GlobalObject.h:102
bool isVarArg() const
Definition: DerivedTypes.h:122
This class represents a no-op cast from one type to another.
MDNode * getMetadata(unsigned KindID) const
Get the metadata of given kind attached to this Instruction.
Definition: Instruction.h:234
Value * getInsertedValueOperand()
SmallString - A SmallString is just a SmallVector with methods and accessors that make it work better...
Definition: SmallString.h:25
bool paramHasAttr(unsigned ArgNo, Attribute::AttrKind Kind) const
Return true if the call or the callee has the given attribute.
Definition: CallSite.h:385
Value * CreateSub(Value *LHS, Value *RHS, const Twine &Name="", bool HasNUW=false, bool HasNSW=false)
Definition: IRBuilder.h:1066
AttributeList getAttributes() const
Return the attribute list for this Function.
Definition: Function.h:223
An instruction for storing to memory.
Definition: Instructions.h:320
bool isIntOrIntVectorTy() const
Return true if this is an integer type or a vector of integer types.
Definition: Type.h:202
static const unsigned kParamTLSSize
Value * CreateZExt(Value *V, Type *DestTy, const Twine &Name="")
Definition: IRBuilder.h:1694
ConstraintPrefix Type
Type - The basic type of the constraint: input/output/clobber.
Definition: InlineAsm.h:120
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:1018
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:126
Value * getOperand(unsigned i) const
Definition: User.h:169
Analysis containing CSE Info
Definition: CSEInfo.cpp:20
static PreservedAnalyses none()
Convenience factory function for the empty preserved set.
Definition: PassManager.h:156
static const PlatformMemoryMapParams Linux_MIPS_MemoryMapParams
bool isRelational() const
Return true if the predicate is relational (not EQ or NE).
Value * CreateOr(Value *LHS, Value *RHS, const Twine &Name="")
Definition: IRBuilder.h:1217
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:344
Type * getScalarType() const
If this is a vector type, return the element type, otherwise return &#39;this&#39;.
Definition: Type.h:303
bool isZeroValue() const
Return true if the value is negative zero or null value.
Definition: Constants.cpp:64
bool isVoidTy() const
Return true if this is &#39;void&#39;.
Definition: Type.h:140
const BasicBlock & getEntryBlock() const
Definition: Function.h:639
an instruction for type-safe pointer arithmetic to access elements of arrays and structs ...
Definition: Instructions.h:873
static cl::opt< unsigned long long > ClShadowBase("msan-shadow-base", cl::desc("Define custom MSan ShadowBase"), cl::Hidden, cl::init(0))
static bool runOnFunction(Function &F, bool PostInlining)
initializer< Ty > init(const Ty &Val)
Definition: CommandLine.h:423
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:189
static GCRegistry::Add< OcamlGC > B("ocaml", "ocaml 3.10-compatible GC")
A set of analyses that are preserved following a run of a transformation pass.
Definition: PassManager.h:153
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:216
const BasicBlock * getSinglePredecessor() const
Return the predecessor of this block if it has a single predecessor block.
Definition: BasicBlock.cpp:233
Value * getCalledValue() const
Definition: InstrTypes.h:1194
static Constant * getOrInsertGlobal(Module &M, StringRef Name, Type *Ty)
LLVM Basic Block Representation.
Definition: BasicBlock.h:57
The instances of the Type class are immutable: once they are created, they are never changed...
Definition: Type.h:45
This is an important class for using LLVM in a threaded context.
Definition: LLVMContext.h:64
const char * getOpcodeName() const
Definition: Instruction.h:127
This is an important base class in LLVM.
Definition: Constant.h:41
bool isInlineAsm() const
Check if this call is an inline asm statement.
Definition: InstrTypes.h:1270
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:2057
bool isPointerTy() const
True if this is an instance of PointerType.
Definition: Type.h:223
FunctionType * getFunctionType() const
Definition: InstrTypes.h:1058
unsigned getNumParams() const
Return the number of fixed parameters this function type requires.
Definition: DerivedTypes.h:138
Represent the analysis usage information of a pass.
op_iterator op_end()
Definition: User.h:231
Value * CreateNeg(Value *V, const Twine &Name="", bool HasNUW=false, bool HasNSW=false)
Definition: IRBuilder.h:1343
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:646
FunctionPass class - This class is used to implement most global optimizations.
Definition: Pass.h:284
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:296
static Constant * get(StructType *T, ArrayRef< Constant *> V)
Definition: Constants.cpp:1053
Value * getPointerOperand()
Definition: Instructions.h:284
bool isX86_MMXTy() const
Return true if this is X86 MMX.
Definition: Type.h:181
Value * CreateICmpEQ(Value *LHS, Value *RHS, const Twine &Name="")
Definition: IRBuilder.h:1874
self_iterator getIterator()
Definition: ilist_node.h:81
Class to represent integer types.
Definition: DerivedTypes.h:39
IntegerType * getIntNTy(unsigned N)
Fetch the type representing an N-bit integer.
Definition: IRBuilder.h:359
Value * CreateExtractElement(Value *Vec, Value *Idx, const Twine &Name="")
Definition: IRBuilder.h:2077
This class represents a cast from an integer to a pointer.
const Value * getCondition() const
static Constant * getAllOnesValue(Type *Ty)
Definition: Constants.cpp:328
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:482
INITIALIZE_PASS_BEGIN(MemorySanitizerLegacyPass, "msan", "MemorySanitizer: detects uninitialized reads.", false, false) INITIALIZE_PASS_END(MemorySanitizerLegacyPass
static PreservedAnalyses all()
Construct a special preserved set that preserves all passes.
Definition: PassManager.h:159
const Value * getArraySize() const
Get the number of elements allocated.
Definition: Instructions.h:92
Value * CreateExtractValue(Value *Agg, ArrayRef< unsigned > Idxs, const Twine &Name="")
Definition: IRBuilder.h:2119
size_t size() const
Definition: SmallVector.h:52
static wasm::ValType getType(const TargetRegisterClass *RC)
PointerType * getInt8PtrTy(unsigned AddrSpace=0)
Fetch the type representing a pointer to an 8-bit integer value.
Definition: IRBuilder.h:384
MDNode * createBranchWeights(uint32_t TrueWeight, uint32_t FalseWeight)
Return metadata containing two branch weights.
Definition: MDBuilder.cpp:37
AtomicOrdering getOrdering() const
Returns the ordering constraint of this rmw instruction.
Definition: Instructions.h:789
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:1083
Value * CreateTrunc(Value *V, Type *DestTy, const Twine &Name="")
Definition: IRBuilder.h:1690
Type * getAllocatedType() const
Return the type that is being allocated by the instruction.
Definition: Instructions.h:105
Triple - Helper class for working with autoconf configuration names.
Definition: Triple.h:43
signed greater than
Definition: InstrTypes.h:673
std::vector< ConstraintInfo > ConstraintInfoVector
Definition: InlineAsm.h:115
unsigned first
Intrinsic::ID getIntrinsicID() const
Return the intrinsic ID of this intrinsic.
Definition: IntrinsicInst.h:50
PHINode * CreatePHI(Type *Ty, unsigned NumReservedValues, const Twine &Name="")
Definition: IRBuilder.h:2004
Value * CreateGEP(Value *Ptr, ArrayRef< Value *> IdxList, const Twine &Name="")
Definition: IRBuilder.h:1493
bool isPtrOrPtrVectorTy() const
Return true if this is a pointer type or a vector of pointer types.
Definition: Type.h:226
static IntegerType * get(LLVMContext &C, unsigned NumBits)
This static method is the primary way of constructing an IntegerType.
Definition: Type.cpp:239
Type * getSequentialElementType() const
Definition: Type.h:357
Iterator for intrusive lists based on ilist_node.
unsigned getNumOperands() const
Definition: User.h:191
See the file comment.
Definition: ValueMap.h:85
This is the shared class of boolean and integer constants.
Definition: Constants.h:83
auto size(R &&Range, typename std::enable_if< std::is_same< typename std::iterator_traits< decltype(Range.begin())>::iterator_category, std::random_access_iterator_tag >::value, void >::type *=nullptr) -> decltype(std::distance(Range.begin(), Range.end()))
Get the size of a range.
Definition: STLExtras.h:1166
iterator end()
Definition: BasicBlock.h:270
unsigned getScalarSizeInBits() const LLVM_READONLY
If this is a vector type, return the getPrimitiveSizeInBits value for the element type...
Definition: Type.cpp:129
CallingConv::ID getCallingConv() const
getCallingConv()/setCallingConv(CC) - These method get and set the calling convention of this functio...
Definition: Function.h:212
Value * CreateIntCast(Value *V, Type *DestTy, bool isSigned, const Twine &Name="")
Definition: IRBuilder.h:1836
This is a &#39;vector&#39; (really, a variable-sized array), optimized for the case when the array is small...
Definition: SmallVector.h:841
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:2090
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:749
bool isAggregateType() const
Return true if the type is an aggregate type.
Definition: Type.h:257
signed less than
Definition: InstrTypes.h:675
CallInst * CreateMaskedStore(Value *Val, Value *Ptr, unsigned Align, Value *Mask)
Create a call to Masked Store intrinsic.
Definition: IRBuilder.cpp:491
CHAIN = SC CHAIN, Imm128 - System call.
ConstantInt * getInt32(uint32_t C)
Get a constant 32-bit value.
Definition: IRBuilder.h:306
static IntegerType * getIntNTy(LLVMContext &C, unsigned N)
Definition: Type.cpp:179
StringRef str()
Return a StringRef for the vector contents.
Definition: raw_ostream.h:534
static GCRegistry::Add< StatepointGC > D("statepoint-example", "an example strategy for statepoint")
CallInst * CreateMemCpy(Value *Dst, unsigned DstAlign, Value *Src, unsigned SrcAlign, uint64_t Size, 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:445
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:631
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:83
FunctionCallee getOrInsertFunction(StringRef Name, FunctionType *T, AttributeList AttributeList)
Look up the specified function in the module symbol table.
Definition: Module.cpp:143
CallInst * CreateMaskedLoad(Value *Ptr, unsigned Align, Value *Mask, Value *PassThru=nullptr, const Twine &Name="")
Create a call to Masked Load intrinsic.
Definition: IRBuilder.cpp:470
User::op_iterator arg_iterator
The type of iterator to use when looping over actual arguments at this call site. ...
Definition: CallSite.h:220
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:493
signed less or equal
Definition: InstrTypes.h:676
Class to represent vector types.
Definition: DerivedTypes.h:424
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:336
Value * CreateShl(Value *LHS, Value *RHS, const Twine &Name="", bool HasNUW=false, bool HasNSW=false)
Definition: IRBuilder.h:1138
const Value * getFalseValue() const
Value * CreatePointerCast(Value *V, Type *DestTy, const Twine &Name="")
Definition: IRBuilder.h:1813
uint64_t getTypeSizeInBits(Type *Ty) const
Size examples:
Definition: DataLayout.h:593
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:461
void removeAttributes(unsigned i, const AttrBuilder &Attrs)
removes the attributes from the list of attributes.
Definition: Function.cpp:415
Predicate getPredicate() const
Return the predicate for this instruction.
Definition: InstrTypes.h:721
unsigned getNumArgOperands() const
Definition: InstrTypes.h:1153
static cl::opt< bool > ClEnableKmsan("msan-kernel", cl::desc("Enable KernelMemorySanitizer instrumentation"), cl::Hidden, cl::init(false))
Constant * getOrInsertGlobal(StringRef Name, Type *Ty, function_ref< GlobalVariable *()> CreateGlobalCallback)
Look up the specified global in the module symbol table.
Definition: Module.cpp:204
static cl::opt< bool > ClHandleICmpExact("msan-handle-icmp-exact", cl::desc("exact handling of relational integer ICmp"), cl::Hidden, cl::init(false))
unsigned getAlignment() const
Return the alignment of the access that is being performed.
Definition: Instructions.h:240
LLVM_NODISCARD bool empty() const
Definition: SmallVector.h:55
StringRef getName() const
Return a constant reference to the value&#39;s name.
Definition: Value.cpp:214
Function * getCalledFunction() const
Returns the function called, or null if this is an indirect function invocation.
Definition: InstrTypes.h:1201
bool onlyReadsMemory(unsigned OpNo) const
Definition: InstrTypes.h:1471
#define I(x, y, z)
Definition: MD5.cpp:58
#define N
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:580
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:322
This instruction extracts a single (scalar) element from a VectorType value.
uint32_t Size
Definition: Profile.cpp:46
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:2837
CallInst * CreateCall(FunctionType *FTy, Value *Callee, ArrayRef< Value *> Args=None, const Twine &Name="", MDNode *FPMathTag=nullptr)
Definition: IRBuilder.h:2009
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:42
PreservedAnalyses run(Function &F, FunctionAnalysisManager &FAM)
static cl::opt< bool > ClCheckConstantShadow("msan-check-constant-shadow", cl::desc("Insert checks for constant shadow values"), cl::Hidden, cl::init(false))
bool isInvoke() const
Return true if a InvokeInst is enclosed.
Definition: CallSite.h:91
static const MemoryMapParams Linux_MIPS64_MemoryMapParams
Value * CreateAnd(Value *LHS, Value *RHS, const Twine &Name="")
Definition: IRBuilder.h:1199
Value * CreatePtrToInt(Value *V, Type *DestTy, const Twine &Name="")
Definition: IRBuilder.h:1757
static const unsigned kOriginSize
const std::string to_string(const T &Value)
Definition: ScopedPrinter.h:61
static const MemoryMapParams FreeBSD_I386_MemoryMapParams
Analysis pass providing the TargetLibraryInfo.
bool isFPOrFPVectorTy() const
Return true if this is a FP type or a vector of FP.
Definition: Type.h:184
iterator_range< df_iterator< T > > depth_first(const T &G)
bool isCall() const
Return true if a CallInst is enclosed.
Definition: CallSite.h:87
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:157
assert(ImpDefSCC.getReg()==AMDGPU::SCC &&ImpDefSCC.isDef())
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:114
void setSuccessOrdering(AtomicOrdering Ordering)
Sets the success ordering constraint of this cmpxchg instruction.
Definition: Instructions.h:587
ArrayRef< unsigned > getIndices() const
Module * getParent()
Get the module that this global value is contained inside of...
Definition: GlobalValue.h:565
LLVM Value Representation.
Definition: Value.h:72
uint64_t getTypeStoreSize(Type *Ty) const
Returns the maximum number of bytes that may be overwritten by storing the specified type...
Definition: DataLayout.h:444
static VectorType * get(Type *ElementType, unsigned NumElements)
This static method is the primary way to construct an VectorType.
Definition: Type.cpp:605
unsigned getParamAlignment(unsigned ArgNo) const
Extract the alignment for a call or parameter (0=unknown).
Definition: CallSite.h:414
#define LLVM_FALLTHROUGH
LLVM_FALLTHROUGH - Mark fallthrough cases in switch statements.
Definition: Compiler.h:250
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:80
AttributeSet getFnAttributes() const
The function attributes are returned.
BasicBlock::iterator GetInsertPoint() const
Definition: IRBuilder.h:121
BasicBlock * SplitBlock(BasicBlock *Old, Instruction *SplitPt, DominatorTree *DT=nullptr, LoopInfo *LI=nullptr, MemorySSAUpdater *MSSAU=nullptr)
Split the specified block at the specified instruction - everything before SplitPt stays in Old and e...
Value * CreateLShr(Value *LHS, Value *RHS, const Twine &Name="", bool isExact=false)
Definition: IRBuilder.h:1159
static cl::opt< bool > ClPoisonUndef("msan-poison-undef", cl::desc("poison undef temps"), cl::Hidden, cl::init(true))
ConstantInt * getInt8(uint8_t C)
Get a constant 8-bit value.
Definition: IRBuilder.h:296
StringRef - Represent a constant reference to a string, i.e.
Definition: StringRef.h:48
Predicate getSwappedPredicate() const
For example, EQ->EQ, SLE->SGE, ULT->UGT, OEQ->OEQ, ULE->UGE, OLT->OGT, etc.
Definition: InstrTypes.h:761
FunctionPass * createMemorySanitizerLegacyPassPass(MemorySanitizerOptions Options={})
static const PlatformMemoryMapParams FreeBSD_X86_MemoryMapParams
A container for analyses that lazily runs them and caches their results.
Type * getArrayElementType() const
Definition: Type.h:364
Value * CreateInsertValue(Value *Agg, Value *Val, ArrayRef< unsigned > Idxs, const Twine &Name="")
Definition: IRBuilder.h:2127
bool isBigEndian() const
Definition: DataLayout.h:233
#define LLVM_DEBUG(X)
Definition: Debug.h:122
Value * getPointerOperand()
Definition: Instructions.h:412
static IntegerType * getInt8Ty(LLVMContext &C)
Definition: Type.cpp:173
static const char *const kMsanModuleCtorName
static Constant * get(ArrayRef< Constant *> V)
Definition: Constants.cpp:1088
Instruction * SplitBlockAndInsertIfThen(Value *Cond, Instruction *SplitBefore, bool Unreachable, MDNode *BranchWeights=nullptr, DominatorTree *DT=nullptr, LoopInfo *LI=nullptr, BasicBlock *ThenBlock=nullptr)
Split the containing block at the specified instruction - everything before SplitBefore stays in the ...
iterator_range< arg_iterator > args()
Definition: Function.h:688
signed greater or equal
Definition: InstrTypes.h:674
static ConstraintInfoVector ParseConstraints(StringRef ConstraintString)
ParseConstraints - Split up the constraint string into the specific constraints and their prefixes...
Definition: InlineAsm.cpp:207
bool isArrayTy() const
True if this is an instance of ArrayType.
Definition: Type.h:220
A wrapper class for inspecting calls to intrinsic functions.
Definition: IntrinsicInst.h:43
const BasicBlock * getParent() const
Definition: Instruction.h:66
an instruction to allocate memory on the stack
Definition: Instructions.h:59
This instruction inserts a struct field of array element value into an aggregate value.