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/SmallSet.h"
147 #include "llvm/ADT/SmallString.h"
148 #include "llvm/ADT/SmallVector.h"
149 #include "llvm/ADT/StringExtras.h"
150 #include "llvm/ADT/StringRef.h"
151 #include "llvm/ADT/Triple.h"
153 #include "llvm/IR/Argument.h"
154 #include "llvm/IR/Attributes.h"
155 #include "llvm/IR/BasicBlock.h"
156 #include "llvm/IR/CallSite.h"
157 #include "llvm/IR/CallingConv.h"
158 #include "llvm/IR/Constant.h"
159 #include "llvm/IR/Constants.h"
160 #include "llvm/IR/DataLayout.h"
161 #include "llvm/IR/DerivedTypes.h"
162 #include "llvm/IR/Function.h"
163 #include "llvm/IR/GlobalValue.h"
164 #include "llvm/IR/GlobalVariable.h"
165 #include "llvm/IR/IRBuilder.h"
166 #include "llvm/IR/InlineAsm.h"
167 #include "llvm/IR/InstVisitor.h"
168 #include "llvm/IR/InstrTypes.h"
169 #include "llvm/IR/Instruction.h"
170 #include "llvm/IR/Instructions.h"
171 #include "llvm/IR/IntrinsicInst.h"
172 #include "llvm/IR/Intrinsics.h"
173 #include "llvm/IR/LLVMContext.h"
174 #include "llvm/IR/MDBuilder.h"
175 #include "llvm/IR/Module.h"
176 #include "llvm/IR/Type.h"
177 #include "llvm/IR/Value.h"
178 #include "llvm/IR/ValueMap.h"
179 #include "llvm/Pass.h"
181 #include "llvm/Support/Casting.h"
183 #include "llvm/Support/Compiler.h"
184 #include "llvm/Support/Debug.h"
186 #include "llvm/Support/MathExtras.h"
192 #include <algorithm>
193 #include <cassert>
194 #include <cstddef>
195 #include <cstdint>
196 #include <memory>
197 #include <string>
198 #include <tuple>
199 
200 using namespace llvm;
201 
202 #define DEBUG_TYPE "msan"
203 
204 static const unsigned kOriginSize = 4;
205 static const unsigned kMinOriginAlignment = 4;
206 static const unsigned kShadowTLSAlignment = 8;
207 
208 // These constants must be kept in sync with the ones in msan.h.
209 static const unsigned kParamTLSSize = 800;
210 static const unsigned kRetvalTLSSize = 800;
211 
212 // Accesses sizes are powers of two: 1, 2, 4, 8.
213 static const size_t kNumberOfAccessSizes = 4;
214 
215 /// Track origins of uninitialized values.
216 ///
217 /// Adds a section to MemorySanitizer report that points to the allocation
218 /// (stack or heap) the uninitialized bits came from originally.
219 static cl::opt<int> ClTrackOrigins("msan-track-origins",
220  cl::desc("Track origins (allocation sites) of poisoned memory"),
221  cl::Hidden, cl::init(0));
222 
223 static cl::opt<bool> ClKeepGoing("msan-keep-going",
224  cl::desc("keep going after reporting a UMR"),
225  cl::Hidden, cl::init(false));
226 
227 static cl::opt<bool> ClPoisonStack("msan-poison-stack",
228  cl::desc("poison uninitialized stack variables"),
229  cl::Hidden, cl::init(true));
230 
231 static cl::opt<bool> ClPoisonStackWithCall("msan-poison-stack-with-call",
232  cl::desc("poison uninitialized stack variables with a call"),
233  cl::Hidden, cl::init(false));
234 
235 static cl::opt<int> ClPoisonStackPattern("msan-poison-stack-pattern",
236  cl::desc("poison uninitialized stack variables with the given pattern"),
237  cl::Hidden, cl::init(0xff));
238 
239 static cl::opt<bool> ClPoisonUndef("msan-poison-undef",
240  cl::desc("poison undef temps"),
241  cl::Hidden, cl::init(true));
242 
243 static cl::opt<bool> ClHandleICmp("msan-handle-icmp",
244  cl::desc("propagate shadow through ICmpEQ and ICmpNE"),
245  cl::Hidden, cl::init(true));
246 
247 static cl::opt<bool> ClHandleICmpExact("msan-handle-icmp-exact",
248  cl::desc("exact handling of relational integer ICmp"),
249  cl::Hidden, cl::init(false));
250 
252  "msan-handle-lifetime-intrinsics",
253  cl::desc(
254  "when possible, poison scoped variables at the beginning of the scope "
255  "(slower, but more precise)"),
256  cl::Hidden, cl::init(true));
257 
258 // When compiling the Linux kernel, we sometimes see false positives related to
259 // MSan being unable to understand that inline assembly calls may initialize
260 // local variables.
261 // This flag makes the compiler conservatively unpoison every memory location
262 // passed into an assembly call. Note that this may cause false positives.
263 // Because it's impossible to figure out the array sizes, we can only unpoison
264 // the first sizeof(type) bytes for each type* pointer.
265 // The instrumentation is only enabled in KMSAN builds, and only if
266 // -msan-handle-asm-conservative is on. This is done because we may want to
267 // quickly disable assembly instrumentation when it breaks.
269  "msan-handle-asm-conservative",
270  cl::desc("conservative handling of inline assembly"), cl::Hidden,
271  cl::init(true));
272 
273 // This flag controls whether we check the shadow of the address
274 // operand of load or store. Such bugs are very rare, since load from
275 // a garbage address typically results in SEGV, but still happen
276 // (e.g. only lower bits of address are garbage, or the access happens
277 // early at program startup where malloc-ed memory is more likely to
278 // be zeroed. As of 2012-08-28 this flag adds 20% slowdown.
279 static cl::opt<bool> ClCheckAccessAddress("msan-check-access-address",
280  cl::desc("report accesses through a pointer which has poisoned shadow"),
281  cl::Hidden, cl::init(true));
282 
283 static cl::opt<bool> ClDumpStrictInstructions("msan-dump-strict-instructions",
284  cl::desc("print out instructions with default strict semantics"),
285  cl::Hidden, cl::init(false));
286 
288  "msan-instrumentation-with-call-threshold",
289  cl::desc(
290  "If the function being instrumented requires more than "
291  "this number of checks and origin stores, use callbacks instead of "
292  "inline checks (-1 means never use callbacks)."),
293  cl::Hidden, cl::init(3500));
294 
295 static cl::opt<bool>
296  ClEnableKmsan("msan-kernel",
297  cl::desc("Enable KernelMemorySanitizer instrumentation"),
298  cl::Hidden, cl::init(false));
299 
300 // This is an experiment to enable handling of cases where shadow is a non-zero
301 // compile-time constant. For some unexplainable reason they were silently
302 // ignored in the instrumentation.
303 static cl::opt<bool> ClCheckConstantShadow("msan-check-constant-shadow",
304  cl::desc("Insert checks for constant shadow values"),
305  cl::Hidden, cl::init(false));
306 
307 // This is off by default because of a bug in gold:
308 // https://sourceware.org/bugzilla/show_bug.cgi?id=19002
309 static cl::opt<bool> ClWithComdat("msan-with-comdat",
310  cl::desc("Place MSan constructors in comdat sections"),
311  cl::Hidden, cl::init(false));
312 
313 // These options allow to specify custom memory map parameters
314 // See MemoryMapParams for details.
315 static cl::opt<uint64_t> ClAndMask("msan-and-mask",
316  cl::desc("Define custom MSan AndMask"),
317  cl::Hidden, cl::init(0));
318 
319 static cl::opt<uint64_t> ClXorMask("msan-xor-mask",
320  cl::desc("Define custom MSan XorMask"),
321  cl::Hidden, cl::init(0));
322 
323 static cl::opt<uint64_t> ClShadowBase("msan-shadow-base",
324  cl::desc("Define custom MSan ShadowBase"),
325  cl::Hidden, cl::init(0));
326 
327 static cl::opt<uint64_t> ClOriginBase("msan-origin-base",
328  cl::desc("Define custom MSan OriginBase"),
329  cl::Hidden, cl::init(0));
330 
331 static const char *const kMsanModuleCtorName = "msan.module_ctor";
332 static const char *const kMsanInitName = "__msan_init";
333 
334 namespace {
335 
336 // Memory map parameters used in application-to-shadow address calculation.
337 // Offset = (Addr & ~AndMask) ^ XorMask
338 // Shadow = ShadowBase + Offset
339 // Origin = OriginBase + Offset
340 struct MemoryMapParams {
341  uint64_t AndMask;
342  uint64_t XorMask;
343  uint64_t ShadowBase;
344  uint64_t OriginBase;
345 };
346 
347 struct PlatformMemoryMapParams {
348  const MemoryMapParams *bits32;
349  const MemoryMapParams *bits64;
350 };
351 
352 } // end anonymous namespace
353 
354 // i386 Linux
355 static const MemoryMapParams Linux_I386_MemoryMapParams = {
356  0x000080000000, // AndMask
357  0, // XorMask (not used)
358  0, // ShadowBase (not used)
359  0x000040000000, // OriginBase
360 };
361 
362 // x86_64 Linux
363 static const MemoryMapParams Linux_X86_64_MemoryMapParams = {
364 #ifdef MSAN_LINUX_X86_64_OLD_MAPPING
365  0x400000000000, // AndMask
366  0, // XorMask (not used)
367  0, // ShadowBase (not used)
368  0x200000000000, // OriginBase
369 #else
370  0, // AndMask (not used)
371  0x500000000000, // XorMask
372  0, // ShadowBase (not used)
373  0x100000000000, // OriginBase
374 #endif
375 };
376 
377 // mips64 Linux
378 static const MemoryMapParams Linux_MIPS64_MemoryMapParams = {
379  0, // AndMask (not used)
380  0x008000000000, // XorMask
381  0, // ShadowBase (not used)
382  0x002000000000, // OriginBase
383 };
384 
385 // ppc64 Linux
386 static const MemoryMapParams Linux_PowerPC64_MemoryMapParams = {
387  0xE00000000000, // AndMask
388  0x100000000000, // XorMask
389  0x080000000000, // ShadowBase
390  0x1C0000000000, // OriginBase
391 };
392 
393 // aarch64 Linux
394 static const MemoryMapParams Linux_AArch64_MemoryMapParams = {
395  0, // AndMask (not used)
396  0x06000000000, // XorMask
397  0, // ShadowBase (not used)
398  0x01000000000, // OriginBase
399 };
400 
401 // i386 FreeBSD
402 static const MemoryMapParams FreeBSD_I386_MemoryMapParams = {
403  0x000180000000, // AndMask
404  0x000040000000, // XorMask
405  0x000020000000, // ShadowBase
406  0x000700000000, // OriginBase
407 };
408 
409 // x86_64 FreeBSD
410 static const MemoryMapParams FreeBSD_X86_64_MemoryMapParams = {
411  0xc00000000000, // AndMask
412  0x200000000000, // XorMask
413  0x100000000000, // ShadowBase
414  0x380000000000, // OriginBase
415 };
416 
417 // x86_64 NetBSD
418 static const MemoryMapParams NetBSD_X86_64_MemoryMapParams = {
419  0, // AndMask
420  0x500000000000, // XorMask
421  0, // ShadowBase
422  0x100000000000, // OriginBase
423 };
424 
425 static const PlatformMemoryMapParams Linux_X86_MemoryMapParams = {
428 };
429 
430 static const PlatformMemoryMapParams Linux_MIPS_MemoryMapParams = {
431  nullptr,
433 };
434 
435 static const PlatformMemoryMapParams Linux_PowerPC_MemoryMapParams = {
436  nullptr,
438 };
439 
440 static const PlatformMemoryMapParams Linux_ARM_MemoryMapParams = {
441  nullptr,
443 };
444 
445 static const PlatformMemoryMapParams FreeBSD_X86_MemoryMapParams = {
448 };
449 
450 static const PlatformMemoryMapParams NetBSD_X86_MemoryMapParams = {
451  nullptr,
453 };
454 
455 namespace {
456 
457 /// Instrument functions of a module to detect uninitialized reads.
458 ///
459 /// Instantiating MemorySanitizer inserts the msan runtime library API function
460 /// declarations into the module if they don't exist already. Instantiating
461 /// ensures the __msan_init function is in the list of global constructors for
462 /// the module.
463 class MemorySanitizer {
464 public:
465  MemorySanitizer(Module &M, MemorySanitizerOptions Options) {
466  this->CompileKernel =
467  ClEnableKmsan.getNumOccurrences() > 0 ? ClEnableKmsan : Options.Kernel;
468  if (ClTrackOrigins.getNumOccurrences() > 0)
469  this->TrackOrigins = ClTrackOrigins;
470  else
471  this->TrackOrigins = this->CompileKernel ? 2 : Options.TrackOrigins;
472  this->Recover = ClKeepGoing.getNumOccurrences() > 0
473  ? ClKeepGoing
474  : (this->CompileKernel | Options.Recover);
475  initializeModule(M);
476  }
477 
478  // MSan cannot be moved or copied because of MapParams.
479  MemorySanitizer(MemorySanitizer &&) = delete;
480  MemorySanitizer &operator=(MemorySanitizer &&) = delete;
481  MemorySanitizer(const MemorySanitizer &) = delete;
482  MemorySanitizer &operator=(const MemorySanitizer &) = delete;
483 
484  bool sanitizeFunction(Function &F, TargetLibraryInfo &TLI);
485 
486 private:
487  friend struct MemorySanitizerVisitor;
488  friend struct VarArgAMD64Helper;
489  friend struct VarArgMIPS64Helper;
490  friend struct VarArgAArch64Helper;
491  friend struct VarArgPowerPC64Helper;
492 
493  void initializeModule(Module &M);
494  void initializeCallbacks(Module &M);
495  void createKernelApi(Module &M);
496  void createUserspaceApi(Module &M);
497 
498  /// True if we're compiling the Linux kernel.
499  bool CompileKernel;
500  /// Track origins (allocation points) of uninitialized values.
501  int TrackOrigins;
502  bool Recover;
503 
504  LLVMContext *C;
505  Type *IntptrTy;
506  Type *OriginTy;
507 
508  // XxxTLS variables represent the per-thread state in MSan and per-task state
509  // in KMSAN.
510  // For the userspace these point to thread-local globals. In the kernel land
511  // they point to the members of a per-task struct obtained via a call to
512  // __msan_get_context_state().
513 
514  /// Thread-local shadow storage for function parameters.
515  Value *ParamTLS;
516 
517  /// Thread-local origin storage for function parameters.
518  Value *ParamOriginTLS;
519 
520  /// Thread-local shadow storage for function return value.
521  Value *RetvalTLS;
522 
523  /// Thread-local origin storage for function return value.
524  Value *RetvalOriginTLS;
525 
526  /// Thread-local shadow storage for in-register va_arg function
527  /// parameters (x86_64-specific).
528  Value *VAArgTLS;
529 
530  /// Thread-local shadow storage for in-register va_arg function
531  /// parameters (x86_64-specific).
532  Value *VAArgOriginTLS;
533 
534  /// Thread-local shadow storage for va_arg overflow area
535  /// (x86_64-specific).
536  Value *VAArgOverflowSizeTLS;
537 
538  /// Thread-local space used to pass origin value to the UMR reporting
539  /// function.
540  Value *OriginTLS;
541 
542  /// Are the instrumentation callbacks set up?
543  bool CallbacksInitialized = false;
544 
545  /// The run-time callback to print a warning.
546  FunctionCallee WarningFn;
547 
548  // These arrays are indexed by log2(AccessSize).
549  FunctionCallee MaybeWarningFn[kNumberOfAccessSizes];
550  FunctionCallee MaybeStoreOriginFn[kNumberOfAccessSizes];
551 
552  /// Run-time helper that generates a new origin value for a stack
553  /// allocation.
554  FunctionCallee MsanSetAllocaOrigin4Fn;
555 
556  /// Run-time helper that poisons stack on function entry.
557  FunctionCallee MsanPoisonStackFn;
558 
559  /// Run-time helper that records a store (or any event) of an
560  /// uninitialized value and returns an updated origin id encoding this info.
561  FunctionCallee MsanChainOriginFn;
562 
563  /// MSan runtime replacements for memmove, memcpy and memset.
564  FunctionCallee MemmoveFn, MemcpyFn, MemsetFn;
565 
566  /// KMSAN callback for task-local function argument shadow.
567  StructType *MsanContextStateTy;
568  FunctionCallee MsanGetContextStateFn;
569 
570  /// Functions for poisoning/unpoisoning local variables
571  FunctionCallee MsanPoisonAllocaFn, MsanUnpoisonAllocaFn;
572 
573  /// Each of the MsanMetadataPtrXxx functions returns a pair of shadow/origin
574  /// pointers.
575  FunctionCallee MsanMetadataPtrForLoadN, MsanMetadataPtrForStoreN;
576  FunctionCallee MsanMetadataPtrForLoad_1_8[4];
577  FunctionCallee MsanMetadataPtrForStore_1_8[4];
578  FunctionCallee MsanInstrumentAsmStoreFn;
579 
580  /// Helper to choose between different MsanMetadataPtrXxx().
581  FunctionCallee getKmsanShadowOriginAccessFn(bool isStore, int size);
582 
583  /// Memory map parameters used in application-to-shadow calculation.
584  const MemoryMapParams *MapParams;
585 
586  /// Custom memory map parameters used when -msan-shadow-base or
587  // -msan-origin-base is provided.
588  MemoryMapParams CustomMapParams;
589 
590  MDNode *ColdCallWeights;
591 
592  /// Branch weights for origin store.
593  MDNode *OriginStoreWeights;
594 
595  /// An empty volatile inline asm that prevents callback merge.
596  InlineAsm *EmptyAsm;
597 
598  Function *MsanCtorFunction;
599 };
600 
601 /// A legacy function pass for msan instrumentation.
602 ///
603 /// Instruments functions to detect unitialized reads.
604 struct MemorySanitizerLegacyPass : public FunctionPass {
605  // Pass identification, replacement for typeid.
606  static char ID;
607 
608  MemorySanitizerLegacyPass(MemorySanitizerOptions Options = {})
609  : FunctionPass(ID), Options(Options) {}
610  StringRef getPassName() const override { return "MemorySanitizerLegacyPass"; }
611 
612  void getAnalysisUsage(AnalysisUsage &AU) const override {
614  }
615 
616  bool runOnFunction(Function &F) override {
617  return MSan->sanitizeFunction(
618  F, getAnalysis<TargetLibraryInfoWrapperPass>().getTLI());
619  }
620  bool doInitialization(Module &M) override;
621 
623  MemorySanitizerOptions Options;
624 };
625 
626 } // end anonymous namespace
627 
630  MemorySanitizer Msan(*F.getParent(), Options);
631  if (Msan.sanitizeFunction(F, FAM.getResult<TargetLibraryAnalysis>(F)))
632  return PreservedAnalyses::none();
633  return PreservedAnalyses::all();
634 }
635 
637 
638 INITIALIZE_PASS_BEGIN(MemorySanitizerLegacyPass, "msan",
639  "MemorySanitizer: detects uninitialized reads.", false,
640  false)
642 INITIALIZE_PASS_END(MemorySanitizerLegacyPass, "msan",
643  "MemorySanitizer: detects uninitialized reads.", false,
644  false)
645 
646 FunctionPass *
648  return new MemorySanitizerLegacyPass(Options);
649 }
650 
651 /// Create a non-const global initialized with the given string.
652 ///
653 /// Creates a writable global for Str so that we can pass it to the
654 /// run-time lib. Runtime uses first 4 bytes of the string to store the
655 /// frame ID, so the string needs to be mutable.
657  StringRef Str) {
658  Constant *StrConst = ConstantDataArray::getString(M.getContext(), Str);
659  return new GlobalVariable(M, StrConst->getType(), /*isConstant=*/false,
660  GlobalValue::PrivateLinkage, StrConst, "");
661 }
662 
663 /// Create KMSAN API callbacks.
664 void MemorySanitizer::createKernelApi(Module &M) {
665  IRBuilder<> IRB(*C);
666 
667  // These will be initialized in insertKmsanPrologue().
668  RetvalTLS = nullptr;
669  RetvalOriginTLS = nullptr;
670  ParamTLS = nullptr;
671  ParamOriginTLS = nullptr;
672  VAArgTLS = nullptr;
673  VAArgOriginTLS = nullptr;
674  VAArgOverflowSizeTLS = nullptr;
675  // OriginTLS is unused in the kernel.
676  OriginTLS = nullptr;
677 
678  // __msan_warning() in the kernel takes an origin.
679  WarningFn = M.getOrInsertFunction("__msan_warning", IRB.getVoidTy(),
680  IRB.getInt32Ty());
681  // Requests the per-task context state (kmsan_context_state*) from the
682  // runtime library.
683  MsanContextStateTy = StructType::get(
684  ArrayType::get(IRB.getInt64Ty(), kParamTLSSize / 8),
685  ArrayType::get(IRB.getInt64Ty(), kRetvalTLSSize / 8),
686  ArrayType::get(IRB.getInt64Ty(), kParamTLSSize / 8),
687  ArrayType::get(IRB.getInt64Ty(), kParamTLSSize / 8), /* va_arg_origin */
688  IRB.getInt64Ty(), ArrayType::get(OriginTy, kParamTLSSize / 4), OriginTy,
689  OriginTy);
690  MsanGetContextStateFn = M.getOrInsertFunction(
691  "__msan_get_context_state", PointerType::get(MsanContextStateTy, 0));
692 
693  Type *RetTy = StructType::get(PointerType::get(IRB.getInt8Ty(), 0),
694  PointerType::get(IRB.getInt32Ty(), 0));
695 
696  for (int ind = 0, size = 1; ind < 4; ind++, size <<= 1) {
697  std::string name_load =
698  "__msan_metadata_ptr_for_load_" + std::to_string(size);
699  std::string name_store =
700  "__msan_metadata_ptr_for_store_" + std::to_string(size);
701  MsanMetadataPtrForLoad_1_8[ind] = M.getOrInsertFunction(
702  name_load, RetTy, PointerType::get(IRB.getInt8Ty(), 0));
703  MsanMetadataPtrForStore_1_8[ind] = M.getOrInsertFunction(
704  name_store, RetTy, PointerType::get(IRB.getInt8Ty(), 0));
705  }
706 
707  MsanMetadataPtrForLoadN = M.getOrInsertFunction(
708  "__msan_metadata_ptr_for_load_n", RetTy,
709  PointerType::get(IRB.getInt8Ty(), 0), IRB.getInt64Ty());
710  MsanMetadataPtrForStoreN = M.getOrInsertFunction(
711  "__msan_metadata_ptr_for_store_n", RetTy,
712  PointerType::get(IRB.getInt8Ty(), 0), IRB.getInt64Ty());
713 
714  // Functions for poisoning and unpoisoning memory.
715  MsanPoisonAllocaFn =
716  M.getOrInsertFunction("__msan_poison_alloca", IRB.getVoidTy(),
717  IRB.getInt8PtrTy(), IntptrTy, IRB.getInt8PtrTy());
718  MsanUnpoisonAllocaFn = M.getOrInsertFunction(
719  "__msan_unpoison_alloca", IRB.getVoidTy(), IRB.getInt8PtrTy(), IntptrTy);
720 }
721 
723  return M.getOrInsertGlobal(Name, Ty, [&] {
724  return new GlobalVariable(M, Ty, false, GlobalVariable::ExternalLinkage,
725  nullptr, Name, nullptr,
727  });
728 }
729 
730 /// Insert declarations for userspace-specific functions and globals.
731 void MemorySanitizer::createUserspaceApi(Module &M) {
732  IRBuilder<> IRB(*C);
733  // Create the callback.
734  // FIXME: this function should have "Cold" calling conv,
735  // which is not yet implemented.
736  StringRef WarningFnName = Recover ? "__msan_warning"
737  : "__msan_warning_noreturn";
738  WarningFn = M.getOrInsertFunction(WarningFnName, IRB.getVoidTy());
739 
740  // Create the global TLS variables.
741  RetvalTLS =
742  getOrInsertGlobal(M, "__msan_retval_tls",
743  ArrayType::get(IRB.getInt64Ty(), kRetvalTLSSize / 8));
744 
745  RetvalOriginTLS = getOrInsertGlobal(M, "__msan_retval_origin_tls", OriginTy);
746 
747  ParamTLS =
748  getOrInsertGlobal(M, "__msan_param_tls",
749  ArrayType::get(IRB.getInt64Ty(), kParamTLSSize / 8));
750 
751  ParamOriginTLS =
752  getOrInsertGlobal(M, "__msan_param_origin_tls",
753  ArrayType::get(OriginTy, kParamTLSSize / 4));
754 
755  VAArgTLS =
756  getOrInsertGlobal(M, "__msan_va_arg_tls",
757  ArrayType::get(IRB.getInt64Ty(), kParamTLSSize / 8));
758 
759  VAArgOriginTLS =
760  getOrInsertGlobal(M, "__msan_va_arg_origin_tls",
761  ArrayType::get(OriginTy, kParamTLSSize / 4));
762 
763  VAArgOverflowSizeTLS =
764  getOrInsertGlobal(M, "__msan_va_arg_overflow_size_tls", IRB.getInt64Ty());
765  OriginTLS = getOrInsertGlobal(M, "__msan_origin_tls", IRB.getInt32Ty());
766 
767  for (size_t AccessSizeIndex = 0; AccessSizeIndex < kNumberOfAccessSizes;
768  AccessSizeIndex++) {
769  unsigned AccessSize = 1 << AccessSizeIndex;
770  std::string FunctionName = "__msan_maybe_warning_" + itostr(AccessSize);
771  MaybeWarningFn[AccessSizeIndex] = M.getOrInsertFunction(
772  FunctionName, IRB.getVoidTy(), IRB.getIntNTy(AccessSize * 8),
773  IRB.getInt32Ty());
774 
775  FunctionName = "__msan_maybe_store_origin_" + itostr(AccessSize);
776  MaybeStoreOriginFn[AccessSizeIndex] = M.getOrInsertFunction(
777  FunctionName, IRB.getVoidTy(), IRB.getIntNTy(AccessSize * 8),
778  IRB.getInt8PtrTy(), IRB.getInt32Ty());
779  }
780 
781  MsanSetAllocaOrigin4Fn = M.getOrInsertFunction(
782  "__msan_set_alloca_origin4", IRB.getVoidTy(), IRB.getInt8PtrTy(), IntptrTy,
783  IRB.getInt8PtrTy(), IntptrTy);
784  MsanPoisonStackFn =
785  M.getOrInsertFunction("__msan_poison_stack", IRB.getVoidTy(),
786  IRB.getInt8PtrTy(), IntptrTy);
787 }
788 
789 /// Insert extern declaration of runtime-provided functions and globals.
790 void MemorySanitizer::initializeCallbacks(Module &M) {
791  // Only do this once.
792  if (CallbacksInitialized)
793  return;
794 
795  IRBuilder<> IRB(*C);
796  // Initialize callbacks that are common for kernel and userspace
797  // instrumentation.
798  MsanChainOriginFn = M.getOrInsertFunction(
799  "__msan_chain_origin", IRB.getInt32Ty(), IRB.getInt32Ty());
800  MemmoveFn = M.getOrInsertFunction(
801  "__msan_memmove", IRB.getInt8PtrTy(), IRB.getInt8PtrTy(),
802  IRB.getInt8PtrTy(), IntptrTy);
803  MemcpyFn = M.getOrInsertFunction(
804  "__msan_memcpy", IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IRB.getInt8PtrTy(),
805  IntptrTy);
806  MemsetFn = M.getOrInsertFunction(
807  "__msan_memset", IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IRB.getInt32Ty(),
808  IntptrTy);
809  // We insert an empty inline asm after __msan_report* to avoid callback merge.
810  EmptyAsm = InlineAsm::get(FunctionType::get(IRB.getVoidTy(), false),
811  StringRef(""), StringRef(""),
812  /*hasSideEffects=*/true);
813 
814  MsanInstrumentAsmStoreFn =
815  M.getOrInsertFunction("__msan_instrument_asm_store", IRB.getVoidTy(),
816  PointerType::get(IRB.getInt8Ty(), 0), IntptrTy);
817 
818  if (CompileKernel) {
819  createKernelApi(M);
820  } else {
821  createUserspaceApi(M);
822  }
823  CallbacksInitialized = true;
824 }
825 
826 FunctionCallee MemorySanitizer::getKmsanShadowOriginAccessFn(bool isStore,
827  int size) {
828  FunctionCallee *Fns =
829  isStore ? MsanMetadataPtrForStore_1_8 : MsanMetadataPtrForLoad_1_8;
830  switch (size) {
831  case 1:
832  return Fns[0];
833  case 2:
834  return Fns[1];
835  case 4:
836  return Fns[2];
837  case 8:
838  return Fns[3];
839  default:
840  return nullptr;
841  }
842 }
843 
844 /// Module-level initialization.
845 ///
846 /// inserts a call to __msan_init to the module's constructor list.
847 void MemorySanitizer::initializeModule(Module &M) {
848  auto &DL = M.getDataLayout();
849 
850  bool ShadowPassed = ClShadowBase.getNumOccurrences() > 0;
851  bool OriginPassed = ClOriginBase.getNumOccurrences() > 0;
852  // Check the overrides first
853  if (ShadowPassed || OriginPassed) {
854  CustomMapParams.AndMask = ClAndMask;
855  CustomMapParams.XorMask = ClXorMask;
856  CustomMapParams.ShadowBase = ClShadowBase;
857  CustomMapParams.OriginBase = ClOriginBase;
858  MapParams = &CustomMapParams;
859  } else {
860  Triple TargetTriple(M.getTargetTriple());
861  switch (TargetTriple.getOS()) {
862  case Triple::FreeBSD:
863  switch (TargetTriple.getArch()) {
864  case Triple::x86_64:
865  MapParams = FreeBSD_X86_MemoryMapParams.bits64;
866  break;
867  case Triple::x86:
868  MapParams = FreeBSD_X86_MemoryMapParams.bits32;
869  break;
870  default:
871  report_fatal_error("unsupported architecture");
872  }
873  break;
874  case Triple::NetBSD:
875  switch (TargetTriple.getArch()) {
876  case Triple::x86_64:
877  MapParams = NetBSD_X86_MemoryMapParams.bits64;
878  break;
879  default:
880  report_fatal_error("unsupported architecture");
881  }
882  break;
883  case Triple::Linux:
884  switch (TargetTriple.getArch()) {
885  case Triple::x86_64:
886  MapParams = Linux_X86_MemoryMapParams.bits64;
887  break;
888  case Triple::x86:
889  MapParams = Linux_X86_MemoryMapParams.bits32;
890  break;
891  case Triple::mips64:
892  case Triple::mips64el:
893  MapParams = Linux_MIPS_MemoryMapParams.bits64;
894  break;
895  case Triple::ppc64:
896  case Triple::ppc64le:
897  MapParams = Linux_PowerPC_MemoryMapParams.bits64;
898  break;
899  case Triple::aarch64:
900  case Triple::aarch64_be:
901  MapParams = Linux_ARM_MemoryMapParams.bits64;
902  break;
903  default:
904  report_fatal_error("unsupported architecture");
905  }
906  break;
907  default:
908  report_fatal_error("unsupported operating system");
909  }
910  }
911 
912  C = &(M.getContext());
913  IRBuilder<> IRB(*C);
914  IntptrTy = IRB.getIntPtrTy(DL);
915  OriginTy = IRB.getInt32Ty();
916 
917  ColdCallWeights = MDBuilder(*C).createBranchWeights(1, 1000);
918  OriginStoreWeights = MDBuilder(*C).createBranchWeights(1, 1000);
919 
920  if (!CompileKernel) {
921  std::tie(MsanCtorFunction, std::ignore) =
924  /*InitArgTypes=*/{},
925  /*InitArgs=*/{},
926  // This callback is invoked when the functions are created the first
927  // time. Hook them into the global ctors list in that case:
928  [&](Function *Ctor, FunctionCallee) {
929  if (!ClWithComdat) {
930  appendToGlobalCtors(M, Ctor, 0);
931  return;
932  }
933  Comdat *MsanCtorComdat = M.getOrInsertComdat(kMsanModuleCtorName);
934  Ctor->setComdat(MsanCtorComdat);
935  appendToGlobalCtors(M, Ctor, 0, Ctor);
936  });
937 
938  if (TrackOrigins)
939  M.getOrInsertGlobal("__msan_track_origins", IRB.getInt32Ty(), [&] {
940  return new GlobalVariable(
941  M, IRB.getInt32Ty(), true, GlobalValue::WeakODRLinkage,
942  IRB.getInt32(TrackOrigins), "__msan_track_origins");
943  });
944 
945  if (Recover)
946  M.getOrInsertGlobal("__msan_keep_going", IRB.getInt32Ty(), [&] {
947  return new GlobalVariable(M, IRB.getInt32Ty(), true,
949  IRB.getInt32(Recover), "__msan_keep_going");
950  });
951 }
952 }
953 
954 bool MemorySanitizerLegacyPass::doInitialization(Module &M) {
955  MSan.emplace(M, Options);
956  return true;
957 }
958 
959 namespace {
960 
961 /// A helper class that handles instrumentation of VarArg
962 /// functions on a particular platform.
963 ///
964 /// Implementations are expected to insert the instrumentation
965 /// necessary to propagate argument shadow through VarArg function
966 /// calls. Visit* methods are called during an InstVisitor pass over
967 /// the function, and should avoid creating new basic blocks. A new
968 /// instance of this class is created for each instrumented function.
969 struct VarArgHelper {
970  virtual ~VarArgHelper() = default;
971 
972  /// Visit a CallSite.
973  virtual void visitCallSite(CallSite &CS, IRBuilder<> &IRB) = 0;
974 
975  /// Visit a va_start call.
976  virtual void visitVAStartInst(VAStartInst &I) = 0;
977 
978  /// Visit a va_copy call.
979  virtual void visitVACopyInst(VACopyInst &I) = 0;
980 
981  /// Finalize function instrumentation.
982  ///
983  /// This method is called after visiting all interesting (see above)
984  /// instructions in a function.
985  virtual void finalizeInstrumentation() = 0;
986 };
987 
988 struct MemorySanitizerVisitor;
989 
990 } // end anonymous namespace
991 
992 static VarArgHelper *CreateVarArgHelper(Function &Func, MemorySanitizer &Msan,
993  MemorySanitizerVisitor &Visitor);
994 
995 static unsigned TypeSizeToSizeIndex(unsigned TypeSize) {
996  if (TypeSize <= 8) return 0;
997  return Log2_32_Ceil((TypeSize + 7) / 8);
998 }
999 
1000 namespace {
1001 
1002 /// This class does all the work for a given function. Store and Load
1003 /// instructions store and load corresponding shadow and origin
1004 /// values. Most instructions propagate shadow from arguments to their
1005 /// return values. Certain instructions (most importantly, BranchInst)
1006 /// test their argument shadow and print reports (with a runtime call) if it's
1007 /// non-zero.
1008 struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
1009  Function &F;
1010  MemorySanitizer &MS;
1011  SmallVector<PHINode *, 16> ShadowPHINodes, OriginPHINodes;
1012  ValueMap<Value*, Value*> ShadowMap, OriginMap;
1013  std::unique_ptr<VarArgHelper> VAHelper;
1014  const TargetLibraryInfo *TLI;
1015  BasicBlock *ActualFnStart;
1016 
1017  // The following flags disable parts of MSan instrumentation based on
1018  // blacklist contents and command-line options.
1019  bool InsertChecks;
1020  bool PropagateShadow;
1021  bool PoisonStack;
1022  bool PoisonUndef;
1023  bool CheckReturnValue;
1024 
1025  struct ShadowOriginAndInsertPoint {
1026  Value *Shadow;
1027  Value *Origin;
1028  Instruction *OrigIns;
1029 
1030  ShadowOriginAndInsertPoint(Value *S, Value *O, Instruction *I)
1031  : Shadow(S), Origin(O), OrigIns(I) {}
1032  };
1033  SmallVector<ShadowOriginAndInsertPoint, 16> InstrumentationList;
1034  bool InstrumentLifetimeStart = ClHandleLifetimeIntrinsics;
1035  SmallSet<AllocaInst *, 16> AllocaSet;
1037  SmallVector<StoreInst *, 16> StoreList;
1038 
1039  MemorySanitizerVisitor(Function &F, MemorySanitizer &MS,
1040  const TargetLibraryInfo &TLI)
1041  : F(F), MS(MS), VAHelper(CreateVarArgHelper(F, MS, *this)), TLI(&TLI) {
1042  bool SanitizeFunction = F.hasFnAttribute(Attribute::SanitizeMemory);
1043  InsertChecks = SanitizeFunction;
1044  PropagateShadow = SanitizeFunction;
1045  PoisonStack = SanitizeFunction && ClPoisonStack;
1046  PoisonUndef = SanitizeFunction && ClPoisonUndef;
1047  // FIXME: Consider using SpecialCaseList to specify a list of functions that
1048  // must always return fully initialized values. For now, we hardcode "main".
1049  CheckReturnValue = SanitizeFunction && (F.getName() == "main");
1050 
1051  MS.initializeCallbacks(*F.getParent());
1052  if (MS.CompileKernel)
1053  ActualFnStart = insertKmsanPrologue(F);
1054  else
1055  ActualFnStart = &F.getEntryBlock();
1056 
1057  LLVM_DEBUG(if (!InsertChecks) dbgs()
1058  << "MemorySanitizer is not inserting checks into '"
1059  << F.getName() << "'\n");
1060  }
1061 
1062  Value *updateOrigin(Value *V, IRBuilder<> &IRB) {
1063  if (MS.TrackOrigins <= 1) return V;
1064  return IRB.CreateCall(MS.MsanChainOriginFn, V);
1065  }
1066 
1067  Value *originToIntptr(IRBuilder<> &IRB, Value *Origin) {
1068  const DataLayout &DL = F.getParent()->getDataLayout();
1069  unsigned IntptrSize = DL.getTypeStoreSize(MS.IntptrTy);
1070  if (IntptrSize == kOriginSize) return Origin;
1071  assert(IntptrSize == kOriginSize * 2);
1072  Origin = IRB.CreateIntCast(Origin, MS.IntptrTy, /* isSigned */ false);
1073  return IRB.CreateOr(Origin, IRB.CreateShl(Origin, kOriginSize * 8));
1074  }
1075 
1076  /// Fill memory range with the given origin value.
1077  void paintOrigin(IRBuilder<> &IRB, Value *Origin, Value *OriginPtr,
1078  unsigned Size, unsigned Alignment) {
1079  const DataLayout &DL = F.getParent()->getDataLayout();
1080  unsigned IntptrAlignment = DL.getABITypeAlignment(MS.IntptrTy);
1081  unsigned IntptrSize = DL.getTypeStoreSize(MS.IntptrTy);
1082  assert(IntptrAlignment >= kMinOriginAlignment);
1083  assert(IntptrSize >= kOriginSize);
1084 
1085  unsigned Ofs = 0;
1086  unsigned CurrentAlignment = Alignment;
1087  if (Alignment >= IntptrAlignment && IntptrSize > kOriginSize) {
1088  Value *IntptrOrigin = originToIntptr(IRB, Origin);
1089  Value *IntptrOriginPtr =
1090  IRB.CreatePointerCast(OriginPtr, PointerType::get(MS.IntptrTy, 0));
1091  for (unsigned i = 0; i < Size / IntptrSize; ++i) {
1092  Value *Ptr = i ? IRB.CreateConstGEP1_32(MS.IntptrTy, IntptrOriginPtr, i)
1093  : IntptrOriginPtr;
1094  IRB.CreateAlignedStore(IntptrOrigin, Ptr, CurrentAlignment);
1095  Ofs += IntptrSize / kOriginSize;
1096  CurrentAlignment = IntptrAlignment;
1097  }
1098  }
1099 
1100  for (unsigned i = Ofs; i < (Size + kOriginSize - 1) / kOriginSize; ++i) {
1101  Value *GEP =
1102  i ? IRB.CreateConstGEP1_32(MS.OriginTy, OriginPtr, i) : OriginPtr;
1103  IRB.CreateAlignedStore(Origin, GEP, CurrentAlignment);
1104  CurrentAlignment = kMinOriginAlignment;
1105  }
1106  }
1107 
1108  void storeOrigin(IRBuilder<> &IRB, Value *Addr, Value *Shadow, Value *Origin,
1109  Value *OriginPtr, unsigned Alignment, bool AsCall) {
1110  const DataLayout &DL = F.getParent()->getDataLayout();
1111  unsigned OriginAlignment = std::max(kMinOriginAlignment, Alignment);
1112  unsigned StoreSize = DL.getTypeStoreSize(Shadow->getType());
1113  if (Shadow->getType()->isAggregateType()) {
1114  paintOrigin(IRB, updateOrigin(Origin, IRB), OriginPtr, StoreSize,
1115  OriginAlignment);
1116  } else {
1117  Value *ConvertedShadow = convertToShadowTyNoVec(Shadow, IRB);
1118  Constant *ConstantShadow = dyn_cast_or_null<Constant>(ConvertedShadow);
1119  if (ConstantShadow) {
1120  if (ClCheckConstantShadow && !ConstantShadow->isZeroValue())
1121  paintOrigin(IRB, updateOrigin(Origin, IRB), OriginPtr, StoreSize,
1122  OriginAlignment);
1123  return;
1124  }
1125 
1126  unsigned TypeSizeInBits =
1127  DL.getTypeSizeInBits(ConvertedShadow->getType());
1128  unsigned SizeIndex = TypeSizeToSizeIndex(TypeSizeInBits);
1129  if (AsCall && SizeIndex < kNumberOfAccessSizes && !MS.CompileKernel) {
1130  FunctionCallee Fn = MS.MaybeStoreOriginFn[SizeIndex];
1131  Value *ConvertedShadow2 = IRB.CreateZExt(
1132  ConvertedShadow, IRB.getIntNTy(8 * (1 << SizeIndex)));
1133  IRB.CreateCall(Fn, {ConvertedShadow2,
1134  IRB.CreatePointerCast(Addr, IRB.getInt8PtrTy()),
1135  Origin});
1136  } else {
1137  Value *Cmp = IRB.CreateICmpNE(
1138  ConvertedShadow, getCleanShadow(ConvertedShadow), "_mscmp");
1140  Cmp, &*IRB.GetInsertPoint(), false, MS.OriginStoreWeights);
1141  IRBuilder<> IRBNew(CheckTerm);
1142  paintOrigin(IRBNew, updateOrigin(Origin, IRBNew), OriginPtr, StoreSize,
1143  OriginAlignment);
1144  }
1145  }
1146  }
1147 
1148  void materializeStores(bool InstrumentWithCalls) {
1149  for (StoreInst *SI : StoreList) {
1150  IRBuilder<> IRB(SI);
1151  Value *Val = SI->getValueOperand();
1152  Value *Addr = SI->getPointerOperand();
1153  Value *Shadow = SI->isAtomic() ? getCleanShadow(Val) : getShadow(Val);
1154  Value *ShadowPtr, *OriginPtr;
1155  Type *ShadowTy = Shadow->getType();
1156  unsigned Alignment = SI->getAlignment();
1157  unsigned OriginAlignment = std::max(kMinOriginAlignment, Alignment);
1158  std::tie(ShadowPtr, OriginPtr) =
1159  getShadowOriginPtr(Addr, IRB, ShadowTy, Alignment, /*isStore*/ true);
1160 
1161  StoreInst *NewSI = IRB.CreateAlignedStore(Shadow, ShadowPtr, Alignment);
1162  LLVM_DEBUG(dbgs() << " STORE: " << *NewSI << "\n");
1163  (void)NewSI;
1164 
1165  if (SI->isAtomic())
1166  SI->setOrdering(addReleaseOrdering(SI->getOrdering()));
1167 
1168  if (MS.TrackOrigins && !SI->isAtomic())
1169  storeOrigin(IRB, Addr, Shadow, getOrigin(Val), OriginPtr,
1170  OriginAlignment, InstrumentWithCalls);
1171  }
1172  }
1173 
1174  /// Helper function to insert a warning at IRB's current insert point.
1175  void insertWarningFn(IRBuilder<> &IRB, Value *Origin) {
1176  if (!Origin)
1177  Origin = (Value *)IRB.getInt32(0);
1178  if (MS.CompileKernel) {
1179  IRB.CreateCall(MS.WarningFn, Origin);
1180  } else {
1181  if (MS.TrackOrigins) {
1182  IRB.CreateStore(Origin, MS.OriginTLS);
1183  }
1184  IRB.CreateCall(MS.WarningFn, {});
1185  }
1186  IRB.CreateCall(MS.EmptyAsm, {});
1187  // FIXME: Insert UnreachableInst if !MS.Recover?
1188  // This may invalidate some of the following checks and needs to be done
1189  // at the very end.
1190  }
1191 
1192  void materializeOneCheck(Instruction *OrigIns, Value *Shadow, Value *Origin,
1193  bool AsCall) {
1194  IRBuilder<> IRB(OrigIns);
1195  LLVM_DEBUG(dbgs() << " SHAD0 : " << *Shadow << "\n");
1196  Value *ConvertedShadow = convertToShadowTyNoVec(Shadow, IRB);
1197  LLVM_DEBUG(dbgs() << " SHAD1 : " << *ConvertedShadow << "\n");
1198 
1199  Constant *ConstantShadow = dyn_cast_or_null<Constant>(ConvertedShadow);
1200  if (ConstantShadow) {
1201  if (ClCheckConstantShadow && !ConstantShadow->isZeroValue()) {
1202  insertWarningFn(IRB, Origin);
1203  }
1204  return;
1205  }
1206 
1207  const DataLayout &DL = OrigIns->getModule()->getDataLayout();
1208 
1209  unsigned TypeSizeInBits = DL.getTypeSizeInBits(ConvertedShadow->getType());
1210  unsigned SizeIndex = TypeSizeToSizeIndex(TypeSizeInBits);
1211  if (AsCall && SizeIndex < kNumberOfAccessSizes && !MS.CompileKernel) {
1212  FunctionCallee Fn = MS.MaybeWarningFn[SizeIndex];
1213  Value *ConvertedShadow2 =
1214  IRB.CreateZExt(ConvertedShadow, IRB.getIntNTy(8 * (1 << SizeIndex)));
1215  IRB.CreateCall(Fn, {ConvertedShadow2, MS.TrackOrigins && Origin
1216  ? Origin
1217  : (Value *)IRB.getInt32(0)});
1218  } else {
1219  Value *Cmp = IRB.CreateICmpNE(ConvertedShadow,
1220  getCleanShadow(ConvertedShadow), "_mscmp");
1222  Cmp, OrigIns,
1223  /* Unreachable */ !MS.Recover, MS.ColdCallWeights);
1224 
1225  IRB.SetInsertPoint(CheckTerm);
1226  insertWarningFn(IRB, Origin);
1227  LLVM_DEBUG(dbgs() << " CHECK: " << *Cmp << "\n");
1228  }
1229  }
1230 
1231  void materializeChecks(bool InstrumentWithCalls) {
1232  for (const auto &ShadowData : InstrumentationList) {
1233  Instruction *OrigIns = ShadowData.OrigIns;
1234  Value *Shadow = ShadowData.Shadow;
1235  Value *Origin = ShadowData.Origin;
1236  materializeOneCheck(OrigIns, Shadow, Origin, InstrumentWithCalls);
1237  }
1238  LLVM_DEBUG(dbgs() << "DONE:\n" << F);
1239  }
1240 
1241  BasicBlock *insertKmsanPrologue(Function &F) {
1242  BasicBlock *ret =
1245  Value *ContextState = IRB.CreateCall(MS.MsanGetContextStateFn, {});
1246  Constant *Zero = IRB.getInt32(0);
1247  MS.ParamTLS = IRB.CreateGEP(MS.MsanContextStateTy, ContextState,
1248  {Zero, IRB.getInt32(0)}, "param_shadow");
1249  MS.RetvalTLS = IRB.CreateGEP(MS.MsanContextStateTy, ContextState,
1250  {Zero, IRB.getInt32(1)}, "retval_shadow");
1251  MS.VAArgTLS = IRB.CreateGEP(MS.MsanContextStateTy, ContextState,
1252  {Zero, IRB.getInt32(2)}, "va_arg_shadow");
1253  MS.VAArgOriginTLS = IRB.CreateGEP(MS.MsanContextStateTy, ContextState,
1254  {Zero, IRB.getInt32(3)}, "va_arg_origin");
1255  MS.VAArgOverflowSizeTLS =
1256  IRB.CreateGEP(MS.MsanContextStateTy, ContextState,
1257  {Zero, IRB.getInt32(4)}, "va_arg_overflow_size");
1258  MS.ParamOriginTLS = IRB.CreateGEP(MS.MsanContextStateTy, ContextState,
1259  {Zero, IRB.getInt32(5)}, "param_origin");
1260  MS.RetvalOriginTLS =
1261  IRB.CreateGEP(MS.MsanContextStateTy, ContextState,
1262  {Zero, IRB.getInt32(6)}, "retval_origin");
1263  return ret;
1264  }
1265 
1266  /// Add MemorySanitizer instrumentation to a function.
1267  bool runOnFunction() {
1268  // In the presence of unreachable blocks, we may see Phi nodes with
1269  // incoming nodes from such blocks. Since InstVisitor skips unreachable
1270  // blocks, such nodes will not have any shadow value associated with them.
1271  // It's easier to remove unreachable blocks than deal with missing shadow.
1273 
1274  // Iterate all BBs in depth-first order and create shadow instructions
1275  // for all instructions (where applicable).
1276  // For PHI nodes we create dummy shadow PHIs which will be finalized later.
1277  for (BasicBlock *BB : depth_first(ActualFnStart))
1278  visit(*BB);
1279 
1280  // Finalize PHI nodes.
1281  for (PHINode *PN : ShadowPHINodes) {
1282  PHINode *PNS = cast<PHINode>(getShadow(PN));
1283  PHINode *PNO = MS.TrackOrigins ? cast<PHINode>(getOrigin(PN)) : nullptr;
1284  size_t NumValues = PN->getNumIncomingValues();
1285  for (size_t v = 0; v < NumValues; v++) {
1286  PNS->addIncoming(getShadow(PN, v), PN->getIncomingBlock(v));
1287  if (PNO) PNO->addIncoming(getOrigin(PN, v), PN->getIncomingBlock(v));
1288  }
1289  }
1290 
1291  VAHelper->finalizeInstrumentation();
1292 
1293  // Poison llvm.lifetime.start intrinsics, if we haven't fallen back to
1294  // instrumenting only allocas.
1295  if (InstrumentLifetimeStart) {
1296  for (auto Item : LifetimeStartList) {
1297  instrumentAlloca(*Item.second, Item.first);
1298  AllocaSet.erase(Item.second);
1299  }
1300  }
1301  // Poison the allocas for which we didn't instrument the corresponding
1302  // lifetime intrinsics.
1303  for (AllocaInst *AI : AllocaSet)
1304  instrumentAlloca(*AI);
1305 
1306  bool InstrumentWithCalls = ClInstrumentationWithCallThreshold >= 0 &&
1307  InstrumentationList.size() + StoreList.size() >
1309 
1310  // Insert shadow value checks.
1311  materializeChecks(InstrumentWithCalls);
1312 
1313  // Delayed instrumentation of StoreInst.
1314  // This may not add new address checks.
1315  materializeStores(InstrumentWithCalls);
1316 
1317  return true;
1318  }
1319 
1320  /// Compute the shadow type that corresponds to a given Value.
1321  Type *getShadowTy(Value *V) {
1322  return getShadowTy(V->getType());
1323  }
1324 
1325  /// Compute the shadow type that corresponds to a given Type.
1326  Type *getShadowTy(Type *OrigTy) {
1327  if (!OrigTy->isSized()) {
1328  return nullptr;
1329  }
1330  // For integer type, shadow is the same as the original type.
1331  // This may return weird-sized types like i1.
1332  if (IntegerType *IT = dyn_cast<IntegerType>(OrigTy))
1333  return IT;
1334  const DataLayout &DL = F.getParent()->getDataLayout();
1335  if (VectorType *VT = dyn_cast<VectorType>(OrigTy)) {
1336  uint32_t EltSize = DL.getTypeSizeInBits(VT->getElementType());
1337  return VectorType::get(IntegerType::get(*MS.C, EltSize),
1338  VT->getNumElements());
1339  }
1340  if (ArrayType *AT = dyn_cast<ArrayType>(OrigTy)) {
1341  return ArrayType::get(getShadowTy(AT->getElementType()),
1342  AT->getNumElements());
1343  }
1344  if (StructType *ST = dyn_cast<StructType>(OrigTy)) {
1345  SmallVector<Type*, 4> Elements;
1346  for (unsigned i = 0, n = ST->getNumElements(); i < n; i++)
1347  Elements.push_back(getShadowTy(ST->getElementType(i)));
1348  StructType *Res = StructType::get(*MS.C, Elements, ST->isPacked());
1349  LLVM_DEBUG(dbgs() << "getShadowTy: " << *ST << " ===> " << *Res << "\n");
1350  return Res;
1351  }
1352  uint32_t TypeSize = DL.getTypeSizeInBits(OrigTy);
1353  return IntegerType::get(*MS.C, TypeSize);
1354  }
1355 
1356  /// Flatten a vector type.
1357  Type *getShadowTyNoVec(Type *ty) {
1358  if (VectorType *vt = dyn_cast<VectorType>(ty))
1359  return IntegerType::get(*MS.C, vt->getBitWidth());
1360  return ty;
1361  }
1362 
1363  /// Convert a shadow value to it's flattened variant.
1364  Value *convertToShadowTyNoVec(Value *V, IRBuilder<> &IRB) {
1365  Type *Ty = V->getType();
1366  Type *NoVecTy = getShadowTyNoVec(Ty);
1367  if (Ty == NoVecTy) return V;
1368  return IRB.CreateBitCast(V, NoVecTy);
1369  }
1370 
1371  /// Compute the integer shadow offset that corresponds to a given
1372  /// application address.
1373  ///
1374  /// Offset = (Addr & ~AndMask) ^ XorMask
1375  Value *getShadowPtrOffset(Value *Addr, IRBuilder<> &IRB) {
1376  Value *OffsetLong = IRB.CreatePointerCast(Addr, MS.IntptrTy);
1377 
1378  uint64_t AndMask = MS.MapParams->AndMask;
1379  if (AndMask)
1380  OffsetLong =
1381  IRB.CreateAnd(OffsetLong, ConstantInt::get(MS.IntptrTy, ~AndMask));
1382 
1383  uint64_t XorMask = MS.MapParams->XorMask;
1384  if (XorMask)
1385  OffsetLong =
1386  IRB.CreateXor(OffsetLong, ConstantInt::get(MS.IntptrTy, XorMask));
1387  return OffsetLong;
1388  }
1389 
1390  /// Compute the shadow and origin addresses corresponding to a given
1391  /// application address.
1392  ///
1393  /// Shadow = ShadowBase + Offset
1394  /// Origin = (OriginBase + Offset) & ~3ULL
1395  std::pair<Value *, Value *> getShadowOriginPtrUserspace(Value *Addr,
1396  IRBuilder<> &IRB,
1397  Type *ShadowTy,
1398  unsigned Alignment) {
1399  Value *ShadowOffset = getShadowPtrOffset(Addr, IRB);
1400  Value *ShadowLong = ShadowOffset;
1401  uint64_t ShadowBase = MS.MapParams->ShadowBase;
1402  if (ShadowBase != 0) {
1403  ShadowLong =
1404  IRB.CreateAdd(ShadowLong,
1405  ConstantInt::get(MS.IntptrTy, ShadowBase));
1406  }
1407  Value *ShadowPtr =
1408  IRB.CreateIntToPtr(ShadowLong, PointerType::get(ShadowTy, 0));
1409  Value *OriginPtr = nullptr;
1410  if (MS.TrackOrigins) {
1411  Value *OriginLong = ShadowOffset;
1412  uint64_t OriginBase = MS.MapParams->OriginBase;
1413  if (OriginBase != 0)
1414  OriginLong = IRB.CreateAdd(OriginLong,
1415  ConstantInt::get(MS.IntptrTy, OriginBase));
1416  if (Alignment < kMinOriginAlignment) {
1417  uint64_t Mask = kMinOriginAlignment - 1;
1418  OriginLong =
1419  IRB.CreateAnd(OriginLong, ConstantInt::get(MS.IntptrTy, ~Mask));
1420  }
1421  OriginPtr =
1422  IRB.CreateIntToPtr(OriginLong, PointerType::get(MS.OriginTy, 0));
1423  }
1424  return std::make_pair(ShadowPtr, OriginPtr);
1425  }
1426 
1427  std::pair<Value *, Value *>
1428  getShadowOriginPtrKernel(Value *Addr, IRBuilder<> &IRB, Type *ShadowTy,
1429  unsigned Alignment, bool isStore) {
1430  Value *ShadowOriginPtrs;
1431  const DataLayout &DL = F.getParent()->getDataLayout();
1432  int Size = DL.getTypeStoreSize(ShadowTy);
1433 
1434  FunctionCallee Getter = MS.getKmsanShadowOriginAccessFn(isStore, Size);
1435  Value *AddrCast =
1436  IRB.CreatePointerCast(Addr, PointerType::get(IRB.getInt8Ty(), 0));
1437  if (Getter) {
1438  ShadowOriginPtrs = IRB.CreateCall(Getter, AddrCast);
1439  } else {
1440  Value *SizeVal = ConstantInt::get(MS.IntptrTy, Size);
1441  ShadowOriginPtrs = IRB.CreateCall(isStore ? MS.MsanMetadataPtrForStoreN
1442  : MS.MsanMetadataPtrForLoadN,
1443  {AddrCast, SizeVal});
1444  }
1445  Value *ShadowPtr = IRB.CreateExtractValue(ShadowOriginPtrs, 0);
1446  ShadowPtr = IRB.CreatePointerCast(ShadowPtr, PointerType::get(ShadowTy, 0));
1447  Value *OriginPtr = IRB.CreateExtractValue(ShadowOriginPtrs, 1);
1448 
1449  return std::make_pair(ShadowPtr, OriginPtr);
1450  }
1451 
1452  std::pair<Value *, Value *> getShadowOriginPtr(Value *Addr, IRBuilder<> &IRB,
1453  Type *ShadowTy,
1454  unsigned Alignment,
1455  bool isStore) {
1456  std::pair<Value *, Value *> ret;
1457  if (MS.CompileKernel)
1458  ret = getShadowOriginPtrKernel(Addr, IRB, ShadowTy, Alignment, isStore);
1459  else
1460  ret = getShadowOriginPtrUserspace(Addr, IRB, ShadowTy, Alignment);
1461  return ret;
1462  }
1463 
1464  /// Compute the shadow address for a given function argument.
1465  ///
1466  /// Shadow = ParamTLS+ArgOffset.
1467  Value *getShadowPtrForArgument(Value *A, IRBuilder<> &IRB,
1468  int ArgOffset) {
1469  Value *Base = IRB.CreatePointerCast(MS.ParamTLS, MS.IntptrTy);
1470  if (ArgOffset)
1471  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
1472  return IRB.CreateIntToPtr(Base, PointerType::get(getShadowTy(A), 0),
1473  "_msarg");
1474  }
1475 
1476  /// Compute the origin address for a given function argument.
1477  Value *getOriginPtrForArgument(Value *A, IRBuilder<> &IRB,
1478  int ArgOffset) {
1479  if (!MS.TrackOrigins)
1480  return nullptr;
1481  Value *Base = IRB.CreatePointerCast(MS.ParamOriginTLS, MS.IntptrTy);
1482  if (ArgOffset)
1483  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
1484  return IRB.CreateIntToPtr(Base, PointerType::get(MS.OriginTy, 0),
1485  "_msarg_o");
1486  }
1487 
1488  /// Compute the shadow address for a retval.
1489  Value *getShadowPtrForRetval(Value *A, IRBuilder<> &IRB) {
1490  return IRB.CreatePointerCast(MS.RetvalTLS,
1491  PointerType::get(getShadowTy(A), 0),
1492  "_msret");
1493  }
1494 
1495  /// Compute the origin address for a retval.
1496  Value *getOriginPtrForRetval(IRBuilder<> &IRB) {
1497  // We keep a single origin for the entire retval. Might be too optimistic.
1498  return MS.RetvalOriginTLS;
1499  }
1500 
1501  /// Set SV to be the shadow value for V.
1502  void setShadow(Value *V, Value *SV) {
1503  assert(!ShadowMap.count(V) && "Values may only have one shadow");
1504  ShadowMap[V] = PropagateShadow ? SV : getCleanShadow(V);
1505  }
1506 
1507  /// Set Origin to be the origin value for V.
1508  void setOrigin(Value *V, Value *Origin) {
1509  if (!MS.TrackOrigins) return;
1510  assert(!OriginMap.count(V) && "Values may only have one origin");
1511  LLVM_DEBUG(dbgs() << "ORIGIN: " << *V << " ==> " << *Origin << "\n");
1512  OriginMap[V] = Origin;
1513  }
1514 
1515  Constant *getCleanShadow(Type *OrigTy) {
1516  Type *ShadowTy = getShadowTy(OrigTy);
1517  if (!ShadowTy)
1518  return nullptr;
1519  return Constant::getNullValue(ShadowTy);
1520  }
1521 
1522  /// Create a clean shadow value for a given value.
1523  ///
1524  /// Clean shadow (all zeroes) means all bits of the value are defined
1525  /// (initialized).
1526  Constant *getCleanShadow(Value *V) {
1527  return getCleanShadow(V->getType());
1528  }
1529 
1530  /// Create a dirty shadow of a given shadow type.
1531  Constant *getPoisonedShadow(Type *ShadowTy) {
1532  assert(ShadowTy);
1533  if (isa<IntegerType>(ShadowTy) || isa<VectorType>(ShadowTy))
1534  return Constant::getAllOnesValue(ShadowTy);
1535  if (ArrayType *AT = dyn_cast<ArrayType>(ShadowTy)) {
1536  SmallVector<Constant *, 4> Vals(AT->getNumElements(),
1537  getPoisonedShadow(AT->getElementType()));
1538  return ConstantArray::get(AT, Vals);
1539  }
1540  if (StructType *ST = dyn_cast<StructType>(ShadowTy)) {
1542  for (unsigned i = 0, n = ST->getNumElements(); i < n; i++)
1543  Vals.push_back(getPoisonedShadow(ST->getElementType(i)));
1544  return ConstantStruct::get(ST, Vals);
1545  }
1546  llvm_unreachable("Unexpected shadow type");
1547  }
1548 
1549  /// Create a dirty shadow for a given value.
1550  Constant *getPoisonedShadow(Value *V) {
1551  Type *ShadowTy = getShadowTy(V);
1552  if (!ShadowTy)
1553  return nullptr;
1554  return getPoisonedShadow(ShadowTy);
1555  }
1556 
1557  /// Create a clean (zero) origin.
1558  Value *getCleanOrigin() {
1559  return Constant::getNullValue(MS.OriginTy);
1560  }
1561 
1562  /// Get the shadow value for a given Value.
1563  ///
1564  /// This function either returns the value set earlier with setShadow,
1565  /// or extracts if from ParamTLS (for function arguments).
1566  Value *getShadow(Value *V) {
1567  if (!PropagateShadow) return getCleanShadow(V);
1568  if (Instruction *I = dyn_cast<Instruction>(V)) {
1569  if (I->getMetadata("nosanitize"))
1570  return getCleanShadow(V);
1571  // For instructions the shadow is already stored in the map.
1572  Value *Shadow = ShadowMap[V];
1573  if (!Shadow) {
1574  LLVM_DEBUG(dbgs() << "No shadow: " << *V << "\n" << *(I->getParent()));
1575  (void)I;
1576  assert(Shadow && "No shadow for a value");
1577  }
1578  return Shadow;
1579  }
1580  if (UndefValue *U = dyn_cast<UndefValue>(V)) {
1581  Value *AllOnes = PoisonUndef ? getPoisonedShadow(V) : getCleanShadow(V);
1582  LLVM_DEBUG(dbgs() << "Undef: " << *U << " ==> " << *AllOnes << "\n");
1583  (void)U;
1584  return AllOnes;
1585  }
1586  if (Argument *A = dyn_cast<Argument>(V)) {
1587  // For arguments we compute the shadow on demand and store it in the map.
1588  Value **ShadowPtr = &ShadowMap[V];
1589  if (*ShadowPtr)
1590  return *ShadowPtr;
1591  Function *F = A->getParent();
1592  IRBuilder<> EntryIRB(ActualFnStart->getFirstNonPHI());
1593  unsigned ArgOffset = 0;
1594  const DataLayout &DL = F->getParent()->getDataLayout();
1595  for (auto &FArg : F->args()) {
1596  if (!FArg.getType()->isSized()) {
1597  LLVM_DEBUG(dbgs() << "Arg is not sized\n");
1598  continue;
1599  }
1600  unsigned Size =
1601  FArg.hasByValAttr()
1602  ? DL.getTypeAllocSize(FArg.getType()->getPointerElementType())
1603  : DL.getTypeAllocSize(FArg.getType());
1604  if (A == &FArg) {
1605  bool Overflow = ArgOffset + Size > kParamTLSSize;
1606  Value *Base = getShadowPtrForArgument(&FArg, EntryIRB, ArgOffset);
1607  if (FArg.hasByValAttr()) {
1608  // ByVal pointer itself has clean shadow. We copy the actual
1609  // argument shadow to the underlying memory.
1610  // Figure out maximal valid memcpy alignment.
1611  unsigned ArgAlign = FArg.getParamAlignment();
1612  if (ArgAlign == 0) {
1613  Type *EltType = A->getType()->getPointerElementType();
1614  ArgAlign = DL.getABITypeAlignment(EltType);
1615  }
1616  Value *CpShadowPtr =
1617  getShadowOriginPtr(V, EntryIRB, EntryIRB.getInt8Ty(), ArgAlign,
1618  /*isStore*/ true)
1619  .first;
1620  // TODO(glider): need to copy origins.
1621  if (Overflow) {
1622  // ParamTLS overflow.
1623  EntryIRB.CreateMemSet(
1624  CpShadowPtr, Constant::getNullValue(EntryIRB.getInt8Ty()),
1625  Size, ArgAlign);
1626  } else {
1627  unsigned CopyAlign = std::min(ArgAlign, kShadowTLSAlignment);
1628  Value *Cpy = EntryIRB.CreateMemCpy(CpShadowPtr, CopyAlign, Base,
1629  CopyAlign, Size);
1630  LLVM_DEBUG(dbgs() << " ByValCpy: " << *Cpy << "\n");
1631  (void)Cpy;
1632  }
1633  *ShadowPtr = getCleanShadow(V);
1634  } else {
1635  if (Overflow) {
1636  // ParamTLS overflow.
1637  *ShadowPtr = getCleanShadow(V);
1638  } else {
1639  *ShadowPtr = EntryIRB.CreateAlignedLoad(getShadowTy(&FArg), Base,
1640  kShadowTLSAlignment);
1641  }
1642  }
1643  LLVM_DEBUG(dbgs()
1644  << " ARG: " << FArg << " ==> " << **ShadowPtr << "\n");
1645  if (MS.TrackOrigins && !Overflow) {
1646  Value *OriginPtr =
1647  getOriginPtrForArgument(&FArg, EntryIRB, ArgOffset);
1648  setOrigin(A, EntryIRB.CreateLoad(MS.OriginTy, OriginPtr));
1649  } else {
1650  setOrigin(A, getCleanOrigin());
1651  }
1652  }
1653  ArgOffset += alignTo(Size, kShadowTLSAlignment);
1654  }
1655  assert(*ShadowPtr && "Could not find shadow for an argument");
1656  return *ShadowPtr;
1657  }
1658  // For everything else the shadow is zero.
1659  return getCleanShadow(V);
1660  }
1661 
1662  /// Get the shadow for i-th argument of the instruction I.
1663  Value *getShadow(Instruction *I, int i) {
1664  return getShadow(I->getOperand(i));
1665  }
1666 
1667  /// Get the origin for a value.
1668  Value *getOrigin(Value *V) {
1669  if (!MS.TrackOrigins) return nullptr;
1670  if (!PropagateShadow) return getCleanOrigin();
1671  if (isa<Constant>(V)) return getCleanOrigin();
1672  assert((isa<Instruction>(V) || isa<Argument>(V)) &&
1673  "Unexpected value type in getOrigin()");
1674  if (Instruction *I = dyn_cast<Instruction>(V)) {
1675  if (I->getMetadata("nosanitize"))
1676  return getCleanOrigin();
1677  }
1678  Value *Origin = OriginMap[V];
1679  assert(Origin && "Missing origin");
1680  return Origin;
1681  }
1682 
1683  /// Get the origin for i-th argument of the instruction I.
1684  Value *getOrigin(Instruction *I, int i) {
1685  return getOrigin(I->getOperand(i));
1686  }
1687 
1688  /// Remember the place where a shadow check should be inserted.
1689  ///
1690  /// This location will be later instrumented with a check that will print a
1691  /// UMR warning in runtime if the shadow value is not 0.
1692  void insertShadowCheck(Value *Shadow, Value *Origin, Instruction *OrigIns) {
1693  assert(Shadow);
1694  if (!InsertChecks) return;
1695 #ifndef NDEBUG
1696  Type *ShadowTy = Shadow->getType();
1697  assert((isa<IntegerType>(ShadowTy) || isa<VectorType>(ShadowTy)) &&
1698  "Can only insert checks for integer and vector shadow types");
1699 #endif
1700  InstrumentationList.push_back(
1701  ShadowOriginAndInsertPoint(Shadow, Origin, OrigIns));
1702  }
1703 
1704  /// Remember the place where a shadow check should be inserted.
1705  ///
1706  /// This location will be later instrumented with a check that will print a
1707  /// UMR warning in runtime if the value is not fully defined.
1708  void insertShadowCheck(Value *Val, Instruction *OrigIns) {
1709  assert(Val);
1710  Value *Shadow, *Origin;
1711  if (ClCheckConstantShadow) {
1712  Shadow = getShadow(Val);
1713  if (!Shadow) return;
1714  Origin = getOrigin(Val);
1715  } else {
1716  Shadow = dyn_cast_or_null<Instruction>(getShadow(Val));
1717  if (!Shadow) return;
1718  Origin = dyn_cast_or_null<Instruction>(getOrigin(Val));
1719  }
1720  insertShadowCheck(Shadow, Origin, OrigIns);
1721  }
1722 
1723  AtomicOrdering addReleaseOrdering(AtomicOrdering a) {
1724  switch (a) {
1730  return AtomicOrdering::Release;
1736  }
1737  llvm_unreachable("Unknown ordering");
1738  }
1739 
1740  AtomicOrdering addAcquireOrdering(AtomicOrdering a) {
1741  switch (a) {
1747  return AtomicOrdering::Acquire;
1753  }
1754  llvm_unreachable("Unknown ordering");
1755  }
1756 
1757  // ------------------- Visitors.
1759  void visit(Instruction &I) {
1760  if (!I.getMetadata("nosanitize"))
1762  }
1763 
1764  /// Instrument LoadInst
1765  ///
1766  /// Loads the corresponding shadow and (optionally) origin.
1767  /// Optionally, checks that the load address is fully defined.
1768  void visitLoadInst(LoadInst &I) {
1769  assert(I.getType()->isSized() && "Load type must have size");
1770  assert(!I.getMetadata("nosanitize"));
1771  IRBuilder<> IRB(I.getNextNode());
1772  Type *ShadowTy = getShadowTy(&I);
1773  Value *Addr = I.getPointerOperand();
1774  Value *ShadowPtr, *OriginPtr;
1775  unsigned Alignment = I.getAlignment();
1776  if (PropagateShadow) {
1777  std::tie(ShadowPtr, OriginPtr) =
1778  getShadowOriginPtr(Addr, IRB, ShadowTy, Alignment, /*isStore*/ false);
1779  setShadow(&I,
1780  IRB.CreateAlignedLoad(ShadowTy, ShadowPtr, Alignment, "_msld"));
1781  } else {
1782  setShadow(&I, getCleanShadow(&I));
1783  }
1784 
1786  insertShadowCheck(I.getPointerOperand(), &I);
1787 
1788  if (I.isAtomic())
1789  I.setOrdering(addAcquireOrdering(I.getOrdering()));
1790 
1791  if (MS.TrackOrigins) {
1792  if (PropagateShadow) {
1793  unsigned OriginAlignment = std::max(kMinOriginAlignment, Alignment);
1794  setOrigin(
1795  &I, IRB.CreateAlignedLoad(MS.OriginTy, OriginPtr, OriginAlignment));
1796  } else {
1797  setOrigin(&I, getCleanOrigin());
1798  }
1799  }
1800  }
1801 
1802  /// Instrument StoreInst
1803  ///
1804  /// Stores the corresponding shadow and (optionally) origin.
1805  /// Optionally, checks that the store address is fully defined.
1806  void visitStoreInst(StoreInst &I) {
1807  StoreList.push_back(&I);
1809  insertShadowCheck(I.getPointerOperand(), &I);
1810  }
1811 
1812  void handleCASOrRMW(Instruction &I) {
1813  assert(isa<AtomicRMWInst>(I) || isa<AtomicCmpXchgInst>(I));
1814 
1815  IRBuilder<> IRB(&I);
1816  Value *Addr = I.getOperand(0);
1817  Value *ShadowPtr = getShadowOriginPtr(Addr, IRB, I.getType(),
1818  /*Alignment*/ 1, /*isStore*/ true)
1819  .first;
1820 
1822  insertShadowCheck(Addr, &I);
1823 
1824  // Only test the conditional argument of cmpxchg instruction.
1825  // The other argument can potentially be uninitialized, but we can not
1826  // detect this situation reliably without possible false positives.
1827  if (isa<AtomicCmpXchgInst>(I))
1828  insertShadowCheck(I.getOperand(1), &I);
1829 
1830  IRB.CreateStore(getCleanShadow(&I), ShadowPtr);
1831 
1832  setShadow(&I, getCleanShadow(&I));
1833  setOrigin(&I, getCleanOrigin());
1834  }
1835 
1836  void visitAtomicRMWInst(AtomicRMWInst &I) {
1837  handleCASOrRMW(I);
1838  I.setOrdering(addReleaseOrdering(I.getOrdering()));
1839  }
1840 
1841  void visitAtomicCmpXchgInst(AtomicCmpXchgInst &I) {
1842  handleCASOrRMW(I);
1843  I.setSuccessOrdering(addReleaseOrdering(I.getSuccessOrdering()));
1844  }
1845 
1846  // Vector manipulation.
1847  void visitExtractElementInst(ExtractElementInst &I) {
1848  insertShadowCheck(I.getOperand(1), &I);
1849  IRBuilder<> IRB(&I);
1850  setShadow(&I, IRB.CreateExtractElement(getShadow(&I, 0), I.getOperand(1),
1851  "_msprop"));
1852  setOrigin(&I, getOrigin(&I, 0));
1853  }
1854 
1855  void visitInsertElementInst(InsertElementInst &I) {
1856  insertShadowCheck(I.getOperand(2), &I);
1857  IRBuilder<> IRB(&I);
1858  setShadow(&I, IRB.CreateInsertElement(getShadow(&I, 0), getShadow(&I, 1),
1859  I.getOperand(2), "_msprop"));
1860  setOriginForNaryOp(I);
1861  }
1862 
1863  void visitShuffleVectorInst(ShuffleVectorInst &I) {
1864  insertShadowCheck(I.getOperand(2), &I);
1865  IRBuilder<> IRB(&I);
1866  setShadow(&I, IRB.CreateShuffleVector(getShadow(&I, 0), getShadow(&I, 1),
1867  I.getOperand(2), "_msprop"));
1868  setOriginForNaryOp(I);
1869  }
1870 
1871  // Casts.
1872  void visitSExtInst(SExtInst &I) {
1873  IRBuilder<> IRB(&I);
1874  setShadow(&I, IRB.CreateSExt(getShadow(&I, 0), I.getType(), "_msprop"));
1875  setOrigin(&I, getOrigin(&I, 0));
1876  }
1877 
1878  void visitZExtInst(ZExtInst &I) {
1879  IRBuilder<> IRB(&I);
1880  setShadow(&I, IRB.CreateZExt(getShadow(&I, 0), I.getType(), "_msprop"));
1881  setOrigin(&I, getOrigin(&I, 0));
1882  }
1883 
1884  void visitTruncInst(TruncInst &I) {
1885  IRBuilder<> IRB(&I);
1886  setShadow(&I, IRB.CreateTrunc(getShadow(&I, 0), I.getType(), "_msprop"));
1887  setOrigin(&I, getOrigin(&I, 0));
1888  }
1889 
1890  void visitBitCastInst(BitCastInst &I) {
1891  // Special case: if this is the bitcast (there is exactly 1 allowed) between
1892  // a musttail call and a ret, don't instrument. New instructions are not
1893  // allowed after a musttail call.
1894  if (auto *CI = dyn_cast<CallInst>(I.getOperand(0)))
1895  if (CI->isMustTailCall())
1896  return;
1897  IRBuilder<> IRB(&I);
1898  setShadow(&I, IRB.CreateBitCast(getShadow(&I, 0), getShadowTy(&I)));
1899  setOrigin(&I, getOrigin(&I, 0));
1900  }
1901 
1902  void visitPtrToIntInst(PtrToIntInst &I) {
1903  IRBuilder<> IRB(&I);
1904  setShadow(&I, IRB.CreateIntCast(getShadow(&I, 0), getShadowTy(&I), false,
1905  "_msprop_ptrtoint"));
1906  setOrigin(&I, getOrigin(&I, 0));
1907  }
1908 
1909  void visitIntToPtrInst(IntToPtrInst &I) {
1910  IRBuilder<> IRB(&I);
1911  setShadow(&I, IRB.CreateIntCast(getShadow(&I, 0), getShadowTy(&I), false,
1912  "_msprop_inttoptr"));
1913  setOrigin(&I, getOrigin(&I, 0));
1914  }
1915 
1916  void visitFPToSIInst(CastInst& I) { handleShadowOr(I); }
1917  void visitFPToUIInst(CastInst& I) { handleShadowOr(I); }
1918  void visitSIToFPInst(CastInst& I) { handleShadowOr(I); }
1919  void visitUIToFPInst(CastInst& I) { handleShadowOr(I); }
1920  void visitFPExtInst(CastInst& I) { handleShadowOr(I); }
1921  void visitFPTruncInst(CastInst& I) { handleShadowOr(I); }
1922 
1923  /// Propagate shadow for bitwise AND.
1924  ///
1925  /// This code is exact, i.e. if, for example, a bit in the left argument
1926  /// is defined and 0, then neither the value not definedness of the
1927  /// corresponding bit in B don't affect the resulting shadow.
1928  void visitAnd(BinaryOperator &I) {
1929  IRBuilder<> IRB(&I);
1930  // "And" of 0 and a poisoned value results in unpoisoned value.
1931  // 1&1 => 1; 0&1 => 0; p&1 => p;
1932  // 1&0 => 0; 0&0 => 0; p&0 => 0;
1933  // 1&p => p; 0&p => 0; p&p => p;
1934  // S = (S1 & S2) | (V1 & S2) | (S1 & V2)
1935  Value *S1 = getShadow(&I, 0);
1936  Value *S2 = getShadow(&I, 1);
1937  Value *V1 = I.getOperand(0);
1938  Value *V2 = I.getOperand(1);
1939  if (V1->getType() != S1->getType()) {
1940  V1 = IRB.CreateIntCast(V1, S1->getType(), false);
1941  V2 = IRB.CreateIntCast(V2, S2->getType(), false);
1942  }
1943  Value *S1S2 = IRB.CreateAnd(S1, S2);
1944  Value *V1S2 = IRB.CreateAnd(V1, S2);
1945  Value *S1V2 = IRB.CreateAnd(S1, V2);
1946  setShadow(&I, IRB.CreateOr(S1S2, IRB.CreateOr(V1S2, S1V2)));
1947  setOriginForNaryOp(I);
1948  }
1949 
1950  void visitOr(BinaryOperator &I) {
1951  IRBuilder<> IRB(&I);
1952  // "Or" of 1 and a poisoned value results in unpoisoned value.
1953  // 1|1 => 1; 0|1 => 1; p|1 => 1;
1954  // 1|0 => 1; 0|0 => 0; p|0 => p;
1955  // 1|p => 1; 0|p => p; p|p => p;
1956  // S = (S1 & S2) | (~V1 & S2) | (S1 & ~V2)
1957  Value *S1 = getShadow(&I, 0);
1958  Value *S2 = getShadow(&I, 1);
1959  Value *V1 = IRB.CreateNot(I.getOperand(0));
1960  Value *V2 = IRB.CreateNot(I.getOperand(1));
1961  if (V1->getType() != S1->getType()) {
1962  V1 = IRB.CreateIntCast(V1, S1->getType(), false);
1963  V2 = IRB.CreateIntCast(V2, S2->getType(), false);
1964  }
1965  Value *S1S2 = IRB.CreateAnd(S1, S2);
1966  Value *V1S2 = IRB.CreateAnd(V1, S2);
1967  Value *S1V2 = IRB.CreateAnd(S1, V2);
1968  setShadow(&I, IRB.CreateOr(S1S2, IRB.CreateOr(V1S2, S1V2)));
1969  setOriginForNaryOp(I);
1970  }
1971 
1972  /// Default propagation of shadow and/or origin.
1973  ///
1974  /// This class implements the general case of shadow propagation, used in all
1975  /// cases where we don't know and/or don't care about what the operation
1976  /// actually does. It converts all input shadow values to a common type
1977  /// (extending or truncating as necessary), and bitwise OR's them.
1978  ///
1979  /// This is much cheaper than inserting checks (i.e. requiring inputs to be
1980  /// fully initialized), and less prone to false positives.
1981  ///
1982  /// This class also implements the general case of origin propagation. For a
1983  /// Nary operation, result origin is set to the origin of an argument that is
1984  /// not entirely initialized. If there is more than one such arguments, the
1985  /// rightmost of them is picked. It does not matter which one is picked if all
1986  /// arguments are initialized.
1987  template <bool CombineShadow>
1988  class Combiner {
1989  Value *Shadow = nullptr;
1990  Value *Origin = nullptr;
1991  IRBuilder<> &IRB;
1992  MemorySanitizerVisitor *MSV;
1993 
1994  public:
1995  Combiner(MemorySanitizerVisitor *MSV, IRBuilder<> &IRB)
1996  : IRB(IRB), MSV(MSV) {}
1997 
1998  /// Add a pair of shadow and origin values to the mix.
1999  Combiner &Add(Value *OpShadow, Value *OpOrigin) {
2000  if (CombineShadow) {
2001  assert(OpShadow);
2002  if (!Shadow)
2003  Shadow = OpShadow;
2004  else {
2005  OpShadow = MSV->CreateShadowCast(IRB, OpShadow, Shadow->getType());
2006  Shadow = IRB.CreateOr(Shadow, OpShadow, "_msprop");
2007  }
2008  }
2009 
2010  if (MSV->MS.TrackOrigins) {
2011  assert(OpOrigin);
2012  if (!Origin) {
2013  Origin = OpOrigin;
2014  } else {
2015  Constant *ConstOrigin = dyn_cast<Constant>(OpOrigin);
2016  // No point in adding something that might result in 0 origin value.
2017  if (!ConstOrigin || !ConstOrigin->isNullValue()) {
2018  Value *FlatShadow = MSV->convertToShadowTyNoVec(OpShadow, IRB);
2019  Value *Cond =
2020  IRB.CreateICmpNE(FlatShadow, MSV->getCleanShadow(FlatShadow));
2021  Origin = IRB.CreateSelect(Cond, OpOrigin, Origin);
2022  }
2023  }
2024  }
2025  return *this;
2026  }
2027 
2028  /// Add an application value to the mix.
2029  Combiner &Add(Value *V) {
2030  Value *OpShadow = MSV->getShadow(V);
2031  Value *OpOrigin = MSV->MS.TrackOrigins ? MSV->getOrigin(V) : nullptr;
2032  return Add(OpShadow, OpOrigin);
2033  }
2034 
2035  /// Set the current combined values as the given instruction's shadow
2036  /// and origin.
2037  void Done(Instruction *I) {
2038  if (CombineShadow) {
2039  assert(Shadow);
2040  Shadow = MSV->CreateShadowCast(IRB, Shadow, MSV->getShadowTy(I));
2041  MSV->setShadow(I, Shadow);
2042  }
2043  if (MSV->MS.TrackOrigins) {
2044  assert(Origin);
2045  MSV->setOrigin(I, Origin);
2046  }
2047  }
2048  };
2049 
2050  using ShadowAndOriginCombiner = Combiner<true>;
2051  using OriginCombiner = Combiner<false>;
2052 
2053  /// Propagate origin for arbitrary operation.
2054  void setOriginForNaryOp(Instruction &I) {
2055  if (!MS.TrackOrigins) return;
2056  IRBuilder<> IRB(&I);
2057  OriginCombiner OC(this, IRB);
2058  for (Instruction::op_iterator OI = I.op_begin(); OI != I.op_end(); ++OI)
2059  OC.Add(OI->get());
2060  OC.Done(&I);
2061  }
2062 
2063  size_t VectorOrPrimitiveTypeSizeInBits(Type *Ty) {
2064  assert(!(Ty->isVectorTy() && Ty->getScalarType()->isPointerTy()) &&
2065  "Vector of pointers is not a valid shadow type");
2066  return Ty->isVectorTy() ?
2068  Ty->getPrimitiveSizeInBits();
2069  }
2070 
2071  /// Cast between two shadow types, extending or truncating as
2072  /// necessary.
2073  Value *CreateShadowCast(IRBuilder<> &IRB, Value *V, Type *dstTy,
2074  bool Signed = false) {
2075  Type *srcTy = V->getType();
2076  size_t srcSizeInBits = VectorOrPrimitiveTypeSizeInBits(srcTy);
2077  size_t dstSizeInBits = VectorOrPrimitiveTypeSizeInBits(dstTy);
2078  if (srcSizeInBits > 1 && dstSizeInBits == 1)
2079  return IRB.CreateICmpNE(V, getCleanShadow(V));
2080 
2081  if (dstTy->isIntegerTy() && srcTy->isIntegerTy())
2082  return IRB.CreateIntCast(V, dstTy, Signed);
2083  if (dstTy->isVectorTy() && srcTy->isVectorTy() &&
2084  dstTy->getVectorNumElements() == srcTy->getVectorNumElements())
2085  return IRB.CreateIntCast(V, dstTy, Signed);
2086  Value *V1 = IRB.CreateBitCast(V, Type::getIntNTy(*MS.C, srcSizeInBits));
2087  Value *V2 =
2088  IRB.CreateIntCast(V1, Type::getIntNTy(*MS.C, dstSizeInBits), Signed);
2089  return IRB.CreateBitCast(V2, dstTy);
2090  // TODO: handle struct types.
2091  }
2092 
2093  /// Cast an application value to the type of its own shadow.
2094  Value *CreateAppToShadowCast(IRBuilder<> &IRB, Value *V) {
2095  Type *ShadowTy = getShadowTy(V);
2096  if (V->getType() == ShadowTy)
2097  return V;
2098  if (V->getType()->isPtrOrPtrVectorTy())
2099  return IRB.CreatePtrToInt(V, ShadowTy);
2100  else
2101  return IRB.CreateBitCast(V, ShadowTy);
2102  }
2103 
2104  /// Propagate shadow for arbitrary operation.
2105  void handleShadowOr(Instruction &I) {
2106  IRBuilder<> IRB(&I);
2107  ShadowAndOriginCombiner SC(this, IRB);
2108  for (Instruction::op_iterator OI = I.op_begin(); OI != I.op_end(); ++OI)
2109  SC.Add(OI->get());
2110  SC.Done(&I);
2111  }
2112 
2113  // Handle multiplication by constant.
2114  //
2115  // Handle a special case of multiplication by constant that may have one or
2116  // more zeros in the lower bits. This makes corresponding number of lower bits
2117  // of the result zero as well. We model it by shifting the other operand
2118  // shadow left by the required number of bits. Effectively, we transform
2119  // (X * (A * 2**B)) to ((X << B) * A) and instrument (X << B) as (Sx << B).
2120  // We use multiplication by 2**N instead of shift to cover the case of
2121  // multiplication by 0, which may occur in some elements of a vector operand.
2122  void handleMulByConstant(BinaryOperator &I, Constant *ConstArg,
2123  Value *OtherArg) {
2124  Constant *ShadowMul;
2125  Type *Ty = ConstArg->getType();
2126  if (Ty->isVectorTy()) {
2127  unsigned NumElements = Ty->getVectorNumElements();
2128  Type *EltTy = Ty->getSequentialElementType();
2129  SmallVector<Constant *, 16> Elements;
2130  for (unsigned Idx = 0; Idx < NumElements; ++Idx) {
2131  if (ConstantInt *Elt =
2132  dyn_cast<ConstantInt>(ConstArg->getAggregateElement(Idx))) {
2133  const APInt &V = Elt->getValue();
2134  APInt V2 = APInt(V.getBitWidth(), 1) << V.countTrailingZeros();
2135  Elements.push_back(ConstantInt::get(EltTy, V2));
2136  } else {
2137  Elements.push_back(ConstantInt::get(EltTy, 1));
2138  }
2139  }
2140  ShadowMul = ConstantVector::get(Elements);
2141  } else {
2142  if (ConstantInt *Elt = dyn_cast<ConstantInt>(ConstArg)) {
2143  const APInt &V = Elt->getValue();
2144  APInt V2 = APInt(V.getBitWidth(), 1) << V.countTrailingZeros();
2145  ShadowMul = ConstantInt::get(Ty, V2);
2146  } else {
2147  ShadowMul = ConstantInt::get(Ty, 1);
2148  }
2149  }
2150 
2151  IRBuilder<> IRB(&I);
2152  setShadow(&I,
2153  IRB.CreateMul(getShadow(OtherArg), ShadowMul, "msprop_mul_cst"));
2154  setOrigin(&I, getOrigin(OtherArg));
2155  }
2156 
2157  void visitMul(BinaryOperator &I) {
2158  Constant *constOp0 = dyn_cast<Constant>(I.getOperand(0));
2159  Constant *constOp1 = dyn_cast<Constant>(I.getOperand(1));
2160  if (constOp0 && !constOp1)
2161  handleMulByConstant(I, constOp0, I.getOperand(1));
2162  else if (constOp1 && !constOp0)
2163  handleMulByConstant(I, constOp1, I.getOperand(0));
2164  else
2165  handleShadowOr(I);
2166  }
2167 
2168  void visitFAdd(BinaryOperator &I) { handleShadowOr(I); }
2169  void visitFSub(BinaryOperator &I) { handleShadowOr(I); }
2170  void visitFMul(BinaryOperator &I) { handleShadowOr(I); }
2171  void visitAdd(BinaryOperator &I) { handleShadowOr(I); }
2172  void visitSub(BinaryOperator &I) { handleShadowOr(I); }
2173  void visitXor(BinaryOperator &I) { handleShadowOr(I); }
2174 
2175  void handleIntegerDiv(Instruction &I) {
2176  IRBuilder<> IRB(&I);
2177  // Strict on the second argument.
2178  insertShadowCheck(I.getOperand(1), &I);
2179  setShadow(&I, getShadow(&I, 0));
2180  setOrigin(&I, getOrigin(&I, 0));
2181  }
2182 
2183  void visitUDiv(BinaryOperator &I) { handleIntegerDiv(I); }
2184  void visitSDiv(BinaryOperator &I) { handleIntegerDiv(I); }
2185  void visitURem(BinaryOperator &I) { handleIntegerDiv(I); }
2186  void visitSRem(BinaryOperator &I) { handleIntegerDiv(I); }
2187 
2188  // Floating point division is side-effect free. We can not require that the
2189  // divisor is fully initialized and must propagate shadow. See PR37523.
2190  void visitFDiv(BinaryOperator &I) { handleShadowOr(I); }
2191  void visitFRem(BinaryOperator &I) { handleShadowOr(I); }
2192 
2193  /// Instrument == and != comparisons.
2194  ///
2195  /// Sometimes the comparison result is known even if some of the bits of the
2196  /// arguments are not.
2197  void handleEqualityComparison(ICmpInst &I) {
2198  IRBuilder<> IRB(&I);
2199  Value *A = I.getOperand(0);
2200  Value *B = I.getOperand(1);
2201  Value *Sa = getShadow(A);
2202  Value *Sb = getShadow(B);
2203 
2204  // Get rid of pointers and vectors of pointers.
2205  // For ints (and vectors of ints), types of A and Sa match,
2206  // and this is a no-op.
2207  A = IRB.CreatePointerCast(A, Sa->getType());
2208  B = IRB.CreatePointerCast(B, Sb->getType());
2209 
2210  // A == B <==> (C = A^B) == 0
2211  // A != B <==> (C = A^B) != 0
2212  // Sc = Sa | Sb
2213  Value *C = IRB.CreateXor(A, B);
2214  Value *Sc = IRB.CreateOr(Sa, Sb);
2215  // Now dealing with i = (C == 0) comparison (or C != 0, does not matter now)
2216  // Result is defined if one of the following is true
2217  // * there is a defined 1 bit in C
2218  // * C is fully defined
2219  // Si = !(C & ~Sc) && Sc
2220  Value *Zero = Constant::getNullValue(Sc->getType());
2221  Value *MinusOne = Constant::getAllOnesValue(Sc->getType());
2222  Value *Si =
2223  IRB.CreateAnd(IRB.CreateICmpNE(Sc, Zero),
2224  IRB.CreateICmpEQ(
2225  IRB.CreateAnd(IRB.CreateXor(Sc, MinusOne), C), Zero));
2226  Si->setName("_msprop_icmp");
2227  setShadow(&I, Si);
2228  setOriginForNaryOp(I);
2229  }
2230 
2231  /// Build the lowest possible value of V, taking into account V's
2232  /// uninitialized bits.
2233  Value *getLowestPossibleValue(IRBuilder<> &IRB, Value *A, Value *Sa,
2234  bool isSigned) {
2235  if (isSigned) {
2236  // Split shadow into sign bit and other bits.
2237  Value *SaOtherBits = IRB.CreateLShr(IRB.CreateShl(Sa, 1), 1);
2238  Value *SaSignBit = IRB.CreateXor(Sa, SaOtherBits);
2239  // Maximise the undefined shadow bit, minimize other undefined bits.
2240  return
2241  IRB.CreateOr(IRB.CreateAnd(A, IRB.CreateNot(SaOtherBits)), SaSignBit);
2242  } else {
2243  // Minimize undefined bits.
2244  return IRB.CreateAnd(A, IRB.CreateNot(Sa));
2245  }
2246  }
2247 
2248  /// Build the highest possible value of V, taking into account V's
2249  /// uninitialized bits.
2250  Value *getHighestPossibleValue(IRBuilder<> &IRB, Value *A, Value *Sa,
2251  bool isSigned) {
2252  if (isSigned) {
2253  // Split shadow into sign bit and other bits.
2254  Value *SaOtherBits = IRB.CreateLShr(IRB.CreateShl(Sa, 1), 1);
2255  Value *SaSignBit = IRB.CreateXor(Sa, SaOtherBits);
2256  // Minimise the undefined shadow bit, maximise other undefined bits.
2257  return
2258  IRB.CreateOr(IRB.CreateAnd(A, IRB.CreateNot(SaSignBit)), SaOtherBits);
2259  } else {
2260  // Maximize undefined bits.
2261  return IRB.CreateOr(A, Sa);
2262  }
2263  }
2264 
2265  /// Instrument relational comparisons.
2266  ///
2267  /// This function does exact shadow propagation for all relational
2268  /// comparisons of integers, pointers and vectors of those.
2269  /// FIXME: output seems suboptimal when one of the operands is a constant
2270  void handleRelationalComparisonExact(ICmpInst &I) {
2271  IRBuilder<> IRB(&I);
2272  Value *A = I.getOperand(0);
2273  Value *B = I.getOperand(1);
2274  Value *Sa = getShadow(A);
2275  Value *Sb = getShadow(B);
2276 
2277  // Get rid of pointers and vectors of pointers.
2278  // For ints (and vectors of ints), types of A and Sa match,
2279  // and this is a no-op.
2280  A = IRB.CreatePointerCast(A, Sa->getType());
2281  B = IRB.CreatePointerCast(B, Sb->getType());
2282 
2283  // Let [a0, a1] be the interval of possible values of A, taking into account
2284  // its undefined bits. Let [b0, b1] be the interval of possible values of B.
2285  // Then (A cmp B) is defined iff (a0 cmp b1) == (a1 cmp b0).
2286  bool IsSigned = I.isSigned();
2287  Value *S1 = IRB.CreateICmp(I.getPredicate(),
2288  getLowestPossibleValue(IRB, A, Sa, IsSigned),
2289  getHighestPossibleValue(IRB, B, Sb, IsSigned));
2290  Value *S2 = IRB.CreateICmp(I.getPredicate(),
2291  getHighestPossibleValue(IRB, A, Sa, IsSigned),
2292  getLowestPossibleValue(IRB, B, Sb, IsSigned));
2293  Value *Si = IRB.CreateXor(S1, S2);
2294  setShadow(&I, Si);
2295  setOriginForNaryOp(I);
2296  }
2297 
2298  /// Instrument signed relational comparisons.
2299  ///
2300  /// Handle sign bit tests: x<0, x>=0, x<=-1, x>-1 by propagating the highest
2301  /// bit of the shadow. Everything else is delegated to handleShadowOr().
2302  void handleSignedRelationalComparison(ICmpInst &I) {
2303  Constant *constOp;
2304  Value *op = nullptr;
2305  CmpInst::Predicate pre;
2306  if ((constOp = dyn_cast<Constant>(I.getOperand(1)))) {
2307  op = I.getOperand(0);
2308  pre = I.getPredicate();
2309  } else if ((constOp = dyn_cast<Constant>(I.getOperand(0)))) {
2310  op = I.getOperand(1);
2311  pre = I.getSwappedPredicate();
2312  } else {
2313  handleShadowOr(I);
2314  return;
2315  }
2316 
2317  if ((constOp->isNullValue() &&
2318  (pre == CmpInst::ICMP_SLT || pre == CmpInst::ICMP_SGE)) ||
2319  (constOp->isAllOnesValue() &&
2320  (pre == CmpInst::ICMP_SGT || pre == CmpInst::ICMP_SLE))) {
2321  IRBuilder<> IRB(&I);
2322  Value *Shadow = IRB.CreateICmpSLT(getShadow(op), getCleanShadow(op),
2323  "_msprop_icmp_s");
2324  setShadow(&I, Shadow);
2325  setOrigin(&I, getOrigin(op));
2326  } else {
2327  handleShadowOr(I);
2328  }
2329  }
2330 
2331  void visitICmpInst(ICmpInst &I) {
2332  if (!ClHandleICmp) {
2333  handleShadowOr(I);
2334  return;
2335  }
2336  if (I.isEquality()) {
2337  handleEqualityComparison(I);
2338  return;
2339  }
2340 
2341  assert(I.isRelational());
2342  if (ClHandleICmpExact) {
2343  handleRelationalComparisonExact(I);
2344  return;
2345  }
2346  if (I.isSigned()) {
2347  handleSignedRelationalComparison(I);
2348  return;
2349  }
2350 
2351  assert(I.isUnsigned());
2352  if ((isa<Constant>(I.getOperand(0)) || isa<Constant>(I.getOperand(1)))) {
2353  handleRelationalComparisonExact(I);
2354  return;
2355  }
2356 
2357  handleShadowOr(I);
2358  }
2359 
2360  void visitFCmpInst(FCmpInst &I) {
2361  handleShadowOr(I);
2362  }
2363 
2364  void handleShift(BinaryOperator &I) {
2365  IRBuilder<> IRB(&I);
2366  // If any of the S2 bits are poisoned, the whole thing is poisoned.
2367  // Otherwise perform the same shift on S1.
2368  Value *S1 = getShadow(&I, 0);
2369  Value *S2 = getShadow(&I, 1);
2370  Value *S2Conv = IRB.CreateSExt(IRB.CreateICmpNE(S2, getCleanShadow(S2)),
2371  S2->getType());
2372  Value *V2 = I.getOperand(1);
2373  Value *Shift = IRB.CreateBinOp(I.getOpcode(), S1, V2);
2374  setShadow(&I, IRB.CreateOr(Shift, S2Conv));
2375  setOriginForNaryOp(I);
2376  }
2377 
2378  void visitShl(BinaryOperator &I) { handleShift(I); }
2379  void visitAShr(BinaryOperator &I) { handleShift(I); }
2380  void visitLShr(BinaryOperator &I) { handleShift(I); }
2381 
2382  /// Instrument llvm.memmove
2383  ///
2384  /// At this point we don't know if llvm.memmove will be inlined or not.
2385  /// If we don't instrument it and it gets inlined,
2386  /// our interceptor will not kick in and we will lose the memmove.
2387  /// If we instrument the call here, but it does not get inlined,
2388  /// we will memove the shadow twice: which is bad in case
2389  /// of overlapping regions. So, we simply lower the intrinsic to a call.
2390  ///
2391  /// Similar situation exists for memcpy and memset.
2392  void visitMemMoveInst(MemMoveInst &I) {
2393  IRBuilder<> IRB(&I);
2394  IRB.CreateCall(
2395  MS.MemmoveFn,
2396  {IRB.CreatePointerCast(I.getArgOperand(0), IRB.getInt8PtrTy()),
2397  IRB.CreatePointerCast(I.getArgOperand(1), IRB.getInt8PtrTy()),
2398  IRB.CreateIntCast(I.getArgOperand(2), MS.IntptrTy, false)});
2399  I.eraseFromParent();
2400  }
2401 
2402  // Similar to memmove: avoid copying shadow twice.
2403  // This is somewhat unfortunate as it may slowdown small constant memcpys.
2404  // FIXME: consider doing manual inline for small constant sizes and proper
2405  // alignment.
2406  void visitMemCpyInst(MemCpyInst &I) {
2407  IRBuilder<> IRB(&I);
2408  IRB.CreateCall(
2409  MS.MemcpyFn,
2410  {IRB.CreatePointerCast(I.getArgOperand(0), IRB.getInt8PtrTy()),
2411  IRB.CreatePointerCast(I.getArgOperand(1), IRB.getInt8PtrTy()),
2412  IRB.CreateIntCast(I.getArgOperand(2), MS.IntptrTy, false)});
2413  I.eraseFromParent();
2414  }
2415 
2416  // Same as memcpy.
2417  void visitMemSetInst(MemSetInst &I) {
2418  IRBuilder<> IRB(&I);
2419  IRB.CreateCall(
2420  MS.MemsetFn,
2421  {IRB.CreatePointerCast(I.getArgOperand(0), IRB.getInt8PtrTy()),
2422  IRB.CreateIntCast(I.getArgOperand(1), IRB.getInt32Ty(), false),
2423  IRB.CreateIntCast(I.getArgOperand(2), MS.IntptrTy, false)});
2424  I.eraseFromParent();
2425  }
2426 
2427  void visitVAStartInst(VAStartInst &I) {
2428  VAHelper->visitVAStartInst(I);
2429  }
2430 
2431  void visitVACopyInst(VACopyInst &I) {
2432  VAHelper->visitVACopyInst(I);
2433  }
2434 
2435  /// Handle vector store-like intrinsics.
2436  ///
2437  /// Instrument intrinsics that look like a simple SIMD store: writes memory,
2438  /// has 1 pointer argument and 1 vector argument, returns void.
2439  bool handleVectorStoreIntrinsic(IntrinsicInst &I) {
2440  IRBuilder<> IRB(&I);
2441  Value* Addr = I.getArgOperand(0);
2442  Value *Shadow = getShadow(&I, 1);
2443  Value *ShadowPtr, *OriginPtr;
2444 
2445  // We don't know the pointer alignment (could be unaligned SSE store!).
2446  // Have to assume to worst case.
2447  std::tie(ShadowPtr, OriginPtr) = getShadowOriginPtr(
2448  Addr, IRB, Shadow->getType(), /*Alignment*/ 1, /*isStore*/ true);
2449  IRB.CreateAlignedStore(Shadow, ShadowPtr, 1);
2450 
2452  insertShadowCheck(Addr, &I);
2453 
2454  // FIXME: factor out common code from materializeStores
2455  if (MS.TrackOrigins) IRB.CreateStore(getOrigin(&I, 1), OriginPtr);
2456  return true;
2457  }
2458 
2459  /// Handle vector load-like intrinsics.
2460  ///
2461  /// Instrument intrinsics that look like a simple SIMD load: reads memory,
2462  /// has 1 pointer argument, returns a vector.
2463  bool handleVectorLoadIntrinsic(IntrinsicInst &I) {
2464  IRBuilder<> IRB(&I);
2465  Value *Addr = I.getArgOperand(0);
2466 
2467  Type *ShadowTy = getShadowTy(&I);
2468  Value *ShadowPtr, *OriginPtr;
2469  if (PropagateShadow) {
2470  // We don't know the pointer alignment (could be unaligned SSE load!).
2471  // Have to assume to worst case.
2472  unsigned Alignment = 1;
2473  std::tie(ShadowPtr, OriginPtr) =
2474  getShadowOriginPtr(Addr, IRB, ShadowTy, Alignment, /*isStore*/ false);
2475  setShadow(&I,
2476  IRB.CreateAlignedLoad(ShadowTy, ShadowPtr, Alignment, "_msld"));
2477  } else {
2478  setShadow(&I, getCleanShadow(&I));
2479  }
2480 
2482  insertShadowCheck(Addr, &I);
2483 
2484  if (MS.TrackOrigins) {
2485  if (PropagateShadow)
2486  setOrigin(&I, IRB.CreateLoad(MS.OriginTy, OriginPtr));
2487  else
2488  setOrigin(&I, getCleanOrigin());
2489  }
2490  return true;
2491  }
2492 
2493  /// Handle (SIMD arithmetic)-like intrinsics.
2494  ///
2495  /// Instrument intrinsics with any number of arguments of the same type,
2496  /// equal to the return type. The type should be simple (no aggregates or
2497  /// pointers; vectors are fine).
2498  /// Caller guarantees that this intrinsic does not access memory.
2499  bool maybeHandleSimpleNomemIntrinsic(IntrinsicInst &I) {
2500  Type *RetTy = I.getType();
2501  if (!(RetTy->isIntOrIntVectorTy() ||
2502  RetTy->isFPOrFPVectorTy() ||
2503  RetTy->isX86_MMXTy()))
2504  return false;
2505 
2506  unsigned NumArgOperands = I.getNumArgOperands();
2507 
2508  for (unsigned i = 0; i < NumArgOperands; ++i) {
2509  Type *Ty = I.getArgOperand(i)->getType();
2510  if (Ty != RetTy)
2511  return false;
2512  }
2513 
2514  IRBuilder<> IRB(&I);
2515  ShadowAndOriginCombiner SC(this, IRB);
2516  for (unsigned i = 0; i < NumArgOperands; ++i)
2517  SC.Add(I.getArgOperand(i));
2518  SC.Done(&I);
2519 
2520  return true;
2521  }
2522 
2523  /// Heuristically instrument unknown intrinsics.
2524  ///
2525  /// The main purpose of this code is to do something reasonable with all
2526  /// random intrinsics we might encounter, most importantly - SIMD intrinsics.
2527  /// We recognize several classes of intrinsics by their argument types and
2528  /// ModRefBehaviour and apply special intrumentation when we are reasonably
2529  /// sure that we know what the intrinsic does.
2530  ///
2531  /// We special-case intrinsics where this approach fails. See llvm.bswap
2532  /// handling as an example of that.
2533  bool handleUnknownIntrinsic(IntrinsicInst &I) {
2534  unsigned NumArgOperands = I.getNumArgOperands();
2535  if (NumArgOperands == 0)
2536  return false;
2537 
2538  if (NumArgOperands == 2 &&
2539  I.getArgOperand(0)->getType()->isPointerTy() &&
2540  I.getArgOperand(1)->getType()->isVectorTy() &&
2541  I.getType()->isVoidTy() &&
2542  !I.onlyReadsMemory()) {
2543  // This looks like a vector store.
2544  return handleVectorStoreIntrinsic(I);
2545  }
2546 
2547  if (NumArgOperands == 1 &&
2548  I.getArgOperand(0)->getType()->isPointerTy() &&
2549  I.getType()->isVectorTy() &&
2550  I.onlyReadsMemory()) {
2551  // This looks like a vector load.
2552  return handleVectorLoadIntrinsic(I);
2553  }
2554 
2555  if (I.doesNotAccessMemory())
2556  if (maybeHandleSimpleNomemIntrinsic(I))
2557  return true;
2558 
2559  // FIXME: detect and handle SSE maskstore/maskload
2560  return false;
2561  }
2562 
2563  void handleLifetimeStart(IntrinsicInst &I) {
2564  if (!PoisonStack)
2565  return;
2566  DenseMap<Value *, AllocaInst *> AllocaForValue;
2567  AllocaInst *AI =
2568  llvm::findAllocaForValue(I.getArgOperand(1), AllocaForValue);
2569  if (!AI)
2570  InstrumentLifetimeStart = false;
2571  LifetimeStartList.push_back(std::make_pair(&I, AI));
2572  }
2573 
2574  void handleBswap(IntrinsicInst &I) {
2575  IRBuilder<> IRB(&I);
2576  Value *Op = I.getArgOperand(0);
2577  Type *OpType = Op->getType();
2578  Function *BswapFunc = Intrinsic::getDeclaration(
2579  F.getParent(), Intrinsic::bswap, makeArrayRef(&OpType, 1));
2580  setShadow(&I, IRB.CreateCall(BswapFunc, getShadow(Op)));
2581  setOrigin(&I, getOrigin(Op));
2582  }
2583 
2584  // Instrument vector convert instrinsic.
2585  //
2586  // This function instruments intrinsics like cvtsi2ss:
2587  // %Out = int_xxx_cvtyyy(%ConvertOp)
2588  // or
2589  // %Out = int_xxx_cvtyyy(%CopyOp, %ConvertOp)
2590  // Intrinsic converts \p NumUsedElements elements of \p ConvertOp to the same
2591  // number \p Out elements, and (if has 2 arguments) copies the rest of the
2592  // elements from \p CopyOp.
2593  // In most cases conversion involves floating-point value which may trigger a
2594  // hardware exception when not fully initialized. For this reason we require
2595  // \p ConvertOp[0:NumUsedElements] to be fully initialized and trap otherwise.
2596  // We copy the shadow of \p CopyOp[NumUsedElements:] to \p
2597  // Out[NumUsedElements:]. This means that intrinsics without \p CopyOp always
2598  // return a fully initialized value.
2599  void handleVectorConvertIntrinsic(IntrinsicInst &I, int NumUsedElements) {
2600  IRBuilder<> IRB(&I);
2601  Value *CopyOp, *ConvertOp;
2602 
2603  switch (I.getNumArgOperands()) {
2604  case 3:
2605  assert(isa<ConstantInt>(I.getArgOperand(2)) && "Invalid rounding mode");
2607  case 2:
2608  CopyOp = I.getArgOperand(0);
2609  ConvertOp = I.getArgOperand(1);
2610  break;
2611  case 1:
2612  ConvertOp = I.getArgOperand(0);
2613  CopyOp = nullptr;
2614  break;
2615  default:
2616  llvm_unreachable("Cvt intrinsic with unsupported number of arguments.");
2617  }
2618 
2619  // The first *NumUsedElements* elements of ConvertOp are converted to the
2620  // same number of output elements. The rest of the output is copied from
2621  // CopyOp, or (if not available) filled with zeroes.
2622  // Combine shadow for elements of ConvertOp that are used in this operation,
2623  // and insert a check.
2624  // FIXME: consider propagating shadow of ConvertOp, at least in the case of
2625  // int->any conversion.
2626  Value *ConvertShadow = getShadow(ConvertOp);
2627  Value *AggShadow = nullptr;
2628  if (ConvertOp->getType()->isVectorTy()) {
2629  AggShadow = IRB.CreateExtractElement(
2630  ConvertShadow, ConstantInt::get(IRB.getInt32Ty(), 0));
2631  for (int i = 1; i < NumUsedElements; ++i) {
2632  Value *MoreShadow = IRB.CreateExtractElement(
2633  ConvertShadow, ConstantInt::get(IRB.getInt32Ty(), i));
2634  AggShadow = IRB.CreateOr(AggShadow, MoreShadow);
2635  }
2636  } else {
2637  AggShadow = ConvertShadow;
2638  }
2639  assert(AggShadow->getType()->isIntegerTy());
2640  insertShadowCheck(AggShadow, getOrigin(ConvertOp), &I);
2641 
2642  // Build result shadow by zero-filling parts of CopyOp shadow that come from
2643  // ConvertOp.
2644  if (CopyOp) {
2645  assert(CopyOp->getType() == I.getType());
2646  assert(CopyOp->getType()->isVectorTy());
2647  Value *ResultShadow = getShadow(CopyOp);
2648  Type *EltTy = ResultShadow->getType()->getVectorElementType();
2649  for (int i = 0; i < NumUsedElements; ++i) {
2650  ResultShadow = IRB.CreateInsertElement(
2651  ResultShadow, ConstantInt::getNullValue(EltTy),
2652  ConstantInt::get(IRB.getInt32Ty(), i));
2653  }
2654  setShadow(&I, ResultShadow);
2655  setOrigin(&I, getOrigin(CopyOp));
2656  } else {
2657  setShadow(&I, getCleanShadow(&I));
2658  setOrigin(&I, getCleanOrigin());
2659  }
2660  }
2661 
2662  // Given a scalar or vector, extract lower 64 bits (or less), and return all
2663  // zeroes if it is zero, and all ones otherwise.
2664  Value *Lower64ShadowExtend(IRBuilder<> &IRB, Value *S, Type *T) {
2665  if (S->getType()->isVectorTy())
2666  S = CreateShadowCast(IRB, S, IRB.getInt64Ty(), /* Signed */ true);
2667  assert(S->getType()->getPrimitiveSizeInBits() <= 64);
2668  Value *S2 = IRB.CreateICmpNE(S, getCleanShadow(S));
2669  return CreateShadowCast(IRB, S2, T, /* Signed */ true);
2670  }
2671 
2672  // Given a vector, extract its first element, and return all
2673  // zeroes if it is zero, and all ones otherwise.
2674  Value *LowerElementShadowExtend(IRBuilder<> &IRB, Value *S, Type *T) {
2675  Value *S1 = IRB.CreateExtractElement(S, (uint64_t)0);
2676  Value *S2 = IRB.CreateICmpNE(S1, getCleanShadow(S1));
2677  return CreateShadowCast(IRB, S2, T, /* Signed */ true);
2678  }
2679 
2680  Value *VariableShadowExtend(IRBuilder<> &IRB, Value *S) {
2681  Type *T = S->getType();
2682  assert(T->isVectorTy());
2683  Value *S2 = IRB.CreateICmpNE(S, getCleanShadow(S));
2684  return IRB.CreateSExt(S2, T);
2685  }
2686 
2687  // Instrument vector shift instrinsic.
2688  //
2689  // This function instruments intrinsics like int_x86_avx2_psll_w.
2690  // Intrinsic shifts %In by %ShiftSize bits.
2691  // %ShiftSize may be a vector. In that case the lower 64 bits determine shift
2692  // size, and the rest is ignored. Behavior is defined even if shift size is
2693  // greater than register (or field) width.
2694  void handleVectorShiftIntrinsic(IntrinsicInst &I, bool Variable) {
2695  assert(I.getNumArgOperands() == 2);
2696  IRBuilder<> IRB(&I);
2697  // If any of the S2 bits are poisoned, the whole thing is poisoned.
2698  // Otherwise perform the same shift on S1.
2699  Value *S1 = getShadow(&I, 0);
2700  Value *S2 = getShadow(&I, 1);
2701  Value *S2Conv = Variable ? VariableShadowExtend(IRB, S2)
2702  : Lower64ShadowExtend(IRB, S2, getShadowTy(&I));
2703  Value *V1 = I.getOperand(0);
2704  Value *V2 = I.getOperand(1);
2705  Value *Shift = IRB.CreateCall(I.getFunctionType(), I.getCalledValue(),
2706  {IRB.CreateBitCast(S1, V1->getType()), V2});
2707  Shift = IRB.CreateBitCast(Shift, getShadowTy(&I));
2708  setShadow(&I, IRB.CreateOr(Shift, S2Conv));
2709  setOriginForNaryOp(I);
2710  }
2711 
2712  // Get an X86_MMX-sized vector type.
2713  Type *getMMXVectorTy(unsigned EltSizeInBits) {
2714  const unsigned X86_MMXSizeInBits = 64;
2715  assert(EltSizeInBits != 0 && (X86_MMXSizeInBits % EltSizeInBits) == 0 &&
2716  "Illegal MMX vector element size");
2717  return VectorType::get(IntegerType::get(*MS.C, EltSizeInBits),
2718  X86_MMXSizeInBits / EltSizeInBits);
2719  }
2720 
2721  // Returns a signed counterpart for an (un)signed-saturate-and-pack
2722  // intrinsic.
2723  Intrinsic::ID getSignedPackIntrinsic(Intrinsic::ID id) {
2724  switch (id) {
2725  case Intrinsic::x86_sse2_packsswb_128:
2726  case Intrinsic::x86_sse2_packuswb_128:
2727  return Intrinsic::x86_sse2_packsswb_128;
2728 
2729  case Intrinsic::x86_sse2_packssdw_128:
2730  case Intrinsic::x86_sse41_packusdw:
2731  return Intrinsic::x86_sse2_packssdw_128;
2732 
2733  case Intrinsic::x86_avx2_packsswb:
2734  case Intrinsic::x86_avx2_packuswb:
2735  return Intrinsic::x86_avx2_packsswb;
2736 
2737  case Intrinsic::x86_avx2_packssdw:
2738  case Intrinsic::x86_avx2_packusdw:
2739  return Intrinsic::x86_avx2_packssdw;
2740 
2741  case Intrinsic::x86_mmx_packsswb:
2742  case Intrinsic::x86_mmx_packuswb:
2743  return Intrinsic::x86_mmx_packsswb;
2744 
2745  case Intrinsic::x86_mmx_packssdw:
2746  return Intrinsic::x86_mmx_packssdw;
2747  default:
2748  llvm_unreachable("unexpected intrinsic id");
2749  }
2750  }
2751 
2752  // Instrument vector pack instrinsic.
2753  //
2754  // This function instruments intrinsics like x86_mmx_packsswb, that
2755  // packs elements of 2 input vectors into half as many bits with saturation.
2756  // Shadow is propagated with the signed variant of the same intrinsic applied
2757  // to sext(Sa != zeroinitializer), sext(Sb != zeroinitializer).
2758  // EltSizeInBits is used only for x86mmx arguments.
2759  void handleVectorPackIntrinsic(IntrinsicInst &I, unsigned EltSizeInBits = 0) {
2760  assert(I.getNumArgOperands() == 2);
2761  bool isX86_MMX = I.getOperand(0)->getType()->isX86_MMXTy();
2762  IRBuilder<> IRB(&I);
2763  Value *S1 = getShadow(&I, 0);
2764  Value *S2 = getShadow(&I, 1);
2765  assert(isX86_MMX || S1->getType()->isVectorTy());
2766 
2767  // SExt and ICmpNE below must apply to individual elements of input vectors.
2768  // In case of x86mmx arguments, cast them to appropriate vector types and
2769  // back.
2770  Type *T = isX86_MMX ? getMMXVectorTy(EltSizeInBits) : S1->getType();
2771  if (isX86_MMX) {
2772  S1 = IRB.CreateBitCast(S1, T);
2773  S2 = IRB.CreateBitCast(S2, T);
2774  }
2775  Value *S1_ext = IRB.CreateSExt(
2776  IRB.CreateICmpNE(S1, Constant::getNullValue(T)), T);
2777  Value *S2_ext = IRB.CreateSExt(
2778  IRB.CreateICmpNE(S2, Constant::getNullValue(T)), T);
2779  if (isX86_MMX) {
2780  Type *X86_MMXTy = Type::getX86_MMXTy(*MS.C);
2781  S1_ext = IRB.CreateBitCast(S1_ext, X86_MMXTy);
2782  S2_ext = IRB.CreateBitCast(S2_ext, X86_MMXTy);
2783  }
2784 
2785  Function *ShadowFn = Intrinsic::getDeclaration(
2786  F.getParent(), getSignedPackIntrinsic(I.getIntrinsicID()));
2787 
2788  Value *S =
2789  IRB.CreateCall(ShadowFn, {S1_ext, S2_ext}, "_msprop_vector_pack");
2790  if (isX86_MMX) S = IRB.CreateBitCast(S, getShadowTy(&I));
2791  setShadow(&I, S);
2792  setOriginForNaryOp(I);
2793  }
2794 
2795  // Instrument sum-of-absolute-differencies intrinsic.
2796  void handleVectorSadIntrinsic(IntrinsicInst &I) {
2797  const unsigned SignificantBitsPerResultElement = 16;
2798  bool isX86_MMX = I.getOperand(0)->getType()->isX86_MMXTy();
2799  Type *ResTy = isX86_MMX ? IntegerType::get(*MS.C, 64) : I.getType();
2800  unsigned ZeroBitsPerResultElement =
2801  ResTy->getScalarSizeInBits() - SignificantBitsPerResultElement;
2802 
2803  IRBuilder<> IRB(&I);
2804  Value *S = IRB.CreateOr(getShadow(&I, 0), getShadow(&I, 1));
2805  S = IRB.CreateBitCast(S, ResTy);
2806  S = IRB.CreateSExt(IRB.CreateICmpNE(S, Constant::getNullValue(ResTy)),
2807  ResTy);
2808  S = IRB.CreateLShr(S, ZeroBitsPerResultElement);
2809  S = IRB.CreateBitCast(S, getShadowTy(&I));
2810  setShadow(&I, S);
2811  setOriginForNaryOp(I);
2812  }
2813 
2814  // Instrument multiply-add intrinsic.
2815  void handleVectorPmaddIntrinsic(IntrinsicInst &I,
2816  unsigned EltSizeInBits = 0) {
2817  bool isX86_MMX = I.getOperand(0)->getType()->isX86_MMXTy();
2818  Type *ResTy = isX86_MMX ? getMMXVectorTy(EltSizeInBits * 2) : I.getType();
2819  IRBuilder<> IRB(&I);
2820  Value *S = IRB.CreateOr(getShadow(&I, 0), getShadow(&I, 1));
2821  S = IRB.CreateBitCast(S, ResTy);
2822  S = IRB.CreateSExt(IRB.CreateICmpNE(S, Constant::getNullValue(ResTy)),
2823  ResTy);
2824  S = IRB.CreateBitCast(S, getShadowTy(&I));
2825  setShadow(&I, S);
2826  setOriginForNaryOp(I);
2827  }
2828 
2829  // Instrument compare-packed intrinsic.
2830  // Basically, an or followed by sext(icmp ne 0) to end up with all-zeros or
2831  // all-ones shadow.
2832  void handleVectorComparePackedIntrinsic(IntrinsicInst &I) {
2833  IRBuilder<> IRB(&I);
2834  Type *ResTy = getShadowTy(&I);
2835  Value *S0 = IRB.CreateOr(getShadow(&I, 0), getShadow(&I, 1));
2836  Value *S = IRB.CreateSExt(
2837  IRB.CreateICmpNE(S0, Constant::getNullValue(ResTy)), ResTy);
2838  setShadow(&I, S);
2839  setOriginForNaryOp(I);
2840  }
2841 
2842  // Instrument compare-scalar intrinsic.
2843  // This handles both cmp* intrinsics which return the result in the first
2844  // element of a vector, and comi* which return the result as i32.
2845  void handleVectorCompareScalarIntrinsic(IntrinsicInst &I) {
2846  IRBuilder<> IRB(&I);
2847  Value *S0 = IRB.CreateOr(getShadow(&I, 0), getShadow(&I, 1));
2848  Value *S = LowerElementShadowExtend(IRB, S0, getShadowTy(&I));
2849  setShadow(&I, S);
2850  setOriginForNaryOp(I);
2851  }
2852 
2853  void handleStmxcsr(IntrinsicInst &I) {
2854  IRBuilder<> IRB(&I);
2855  Value* Addr = I.getArgOperand(0);
2856  Type *Ty = IRB.getInt32Ty();
2857  Value *ShadowPtr =
2858  getShadowOriginPtr(Addr, IRB, Ty, /*Alignment*/ 1, /*isStore*/ true)
2859  .first;
2860 
2861  IRB.CreateStore(getCleanShadow(Ty),
2862  IRB.CreatePointerCast(ShadowPtr, Ty->getPointerTo()));
2863 
2865  insertShadowCheck(Addr, &I);
2866  }
2867 
2868  void handleLdmxcsr(IntrinsicInst &I) {
2869  if (!InsertChecks) return;
2870 
2871  IRBuilder<> IRB(&I);
2872  Value *Addr = I.getArgOperand(0);
2873  Type *Ty = IRB.getInt32Ty();
2874  unsigned Alignment = 1;
2875  Value *ShadowPtr, *OriginPtr;
2876  std::tie(ShadowPtr, OriginPtr) =
2877  getShadowOriginPtr(Addr, IRB, Ty, Alignment, /*isStore*/ false);
2878 
2880  insertShadowCheck(Addr, &I);
2881 
2882  Value *Shadow = IRB.CreateAlignedLoad(Ty, ShadowPtr, Alignment, "_ldmxcsr");
2883  Value *Origin = MS.TrackOrigins ? IRB.CreateLoad(MS.OriginTy, OriginPtr)
2884  : getCleanOrigin();
2885  insertShadowCheck(Shadow, Origin, &I);
2886  }
2887 
2888  void handleMaskedStore(IntrinsicInst &I) {
2889  IRBuilder<> IRB(&I);
2890  Value *V = I.getArgOperand(0);
2891  Value *Addr = I.getArgOperand(1);
2892  unsigned Align = cast<ConstantInt>(I.getArgOperand(2))->getZExtValue();
2893  Value *Mask = I.getArgOperand(3);
2894  Value *Shadow = getShadow(V);
2895 
2896  Value *ShadowPtr;
2897  Value *OriginPtr;
2898  std::tie(ShadowPtr, OriginPtr) = getShadowOriginPtr(
2899  Addr, IRB, Shadow->getType(), Align, /*isStore*/ true);
2900 
2901  if (ClCheckAccessAddress) {
2902  insertShadowCheck(Addr, &I);
2903  // Uninitialized mask is kind of like uninitialized address, but not as
2904  // scary.
2905  insertShadowCheck(Mask, &I);
2906  }
2907 
2908  IRB.CreateMaskedStore(Shadow, ShadowPtr, Align, Mask);
2909 
2910  if (MS.TrackOrigins) {
2911  auto &DL = F.getParent()->getDataLayout();
2912  paintOrigin(IRB, getOrigin(V), OriginPtr,
2913  DL.getTypeStoreSize(Shadow->getType()),
2914  std::max(Align, kMinOriginAlignment));
2915  }
2916  }
2917 
2918  bool handleMaskedLoad(IntrinsicInst &I) {
2919  IRBuilder<> IRB(&I);
2920  Value *Addr = I.getArgOperand(0);
2921  unsigned Align = cast<ConstantInt>(I.getArgOperand(1))->getZExtValue();
2922  Value *Mask = I.getArgOperand(2);
2923  Value *PassThru = I.getArgOperand(3);
2924 
2925  Type *ShadowTy = getShadowTy(&I);
2926  Value *ShadowPtr, *OriginPtr;
2927  if (PropagateShadow) {
2928  std::tie(ShadowPtr, OriginPtr) =
2929  getShadowOriginPtr(Addr, IRB, ShadowTy, Align, /*isStore*/ false);
2930  setShadow(&I, IRB.CreateMaskedLoad(ShadowPtr, Align, Mask,
2931  getShadow(PassThru), "_msmaskedld"));
2932  } else {
2933  setShadow(&I, getCleanShadow(&I));
2934  }
2935 
2936  if (ClCheckAccessAddress) {
2937  insertShadowCheck(Addr, &I);
2938  insertShadowCheck(Mask, &I);
2939  }
2940 
2941  if (MS.TrackOrigins) {
2942  if (PropagateShadow) {
2943  // Choose between PassThru's and the loaded value's origins.
2944  Value *MaskedPassThruShadow = IRB.CreateAnd(
2945  getShadow(PassThru), IRB.CreateSExt(IRB.CreateNeg(Mask), ShadowTy));
2946 
2947  Value *Acc = IRB.CreateExtractElement(
2948  MaskedPassThruShadow, ConstantInt::get(IRB.getInt32Ty(), 0));
2949  for (int i = 1, N = PassThru->getType()->getVectorNumElements(); i < N;
2950  ++i) {
2951  Value *More = IRB.CreateExtractElement(
2952  MaskedPassThruShadow, ConstantInt::get(IRB.getInt32Ty(), i));
2953  Acc = IRB.CreateOr(Acc, More);
2954  }
2955 
2956  Value *Origin = IRB.CreateSelect(
2957  IRB.CreateICmpNE(Acc, Constant::getNullValue(Acc->getType())),
2958  getOrigin(PassThru), IRB.CreateLoad(MS.OriginTy, OriginPtr));
2959 
2960  setOrigin(&I, Origin);
2961  } else {
2962  setOrigin(&I, getCleanOrigin());
2963  }
2964  }
2965  return true;
2966  }
2967 
2968  // Instrument BMI / BMI2 intrinsics.
2969  // All of these intrinsics are Z = I(X, Y)
2970  // where the types of all operands and the result match, and are either i32 or i64.
2971  // The following instrumentation happens to work for all of them:
2972  // Sz = I(Sx, Y) | (sext (Sy != 0))
2973  void handleBmiIntrinsic(IntrinsicInst &I) {
2974  IRBuilder<> IRB(&I);
2975  Type *ShadowTy = getShadowTy(&I);
2976 
2977  // If any bit of the mask operand is poisoned, then the whole thing is.
2978  Value *SMask = getShadow(&I, 1);
2979  SMask = IRB.CreateSExt(IRB.CreateICmpNE(SMask, getCleanShadow(ShadowTy)),
2980  ShadowTy);
2981  // Apply the same intrinsic to the shadow of the first operand.
2982  Value *S = IRB.CreateCall(I.getCalledFunction(),
2983  {getShadow(&I, 0), I.getOperand(1)});
2984  S = IRB.CreateOr(SMask, S);
2985  setShadow(&I, S);
2986  setOriginForNaryOp(I);
2987  }
2988 
2989  void visitIntrinsicInst(IntrinsicInst &I) {
2990  switch (I.getIntrinsicID()) {
2991  case Intrinsic::lifetime_start:
2992  handleLifetimeStart(I);
2993  break;
2994  case Intrinsic::bswap:
2995  handleBswap(I);
2996  break;
2997  case Intrinsic::masked_store:
2998  handleMaskedStore(I);
2999  break;
3000  case Intrinsic::masked_load:
3001  handleMaskedLoad(I);
3002  break;
3003  case Intrinsic::x86_sse_stmxcsr:
3004  handleStmxcsr(I);
3005  break;
3006  case Intrinsic::x86_sse_ldmxcsr:
3007  handleLdmxcsr(I);
3008  break;
3009  case Intrinsic::x86_avx512_vcvtsd2usi64:
3010  case Intrinsic::x86_avx512_vcvtsd2usi32:
3011  case Intrinsic::x86_avx512_vcvtss2usi64:
3012  case Intrinsic::x86_avx512_vcvtss2usi32:
3013  case Intrinsic::x86_avx512_cvttss2usi64:
3014  case Intrinsic::x86_avx512_cvttss2usi:
3015  case Intrinsic::x86_avx512_cvttsd2usi64:
3016  case Intrinsic::x86_avx512_cvttsd2usi:
3017  case Intrinsic::x86_avx512_cvtusi2ss:
3018  case Intrinsic::x86_avx512_cvtusi642sd:
3019  case Intrinsic::x86_avx512_cvtusi642ss:
3020  case Intrinsic::x86_sse2_cvtsd2si64:
3021  case Intrinsic::x86_sse2_cvtsd2si:
3022  case Intrinsic::x86_sse2_cvtsd2ss:
3023  case Intrinsic::x86_sse2_cvttsd2si64:
3024  case Intrinsic::x86_sse2_cvttsd2si:
3025  case Intrinsic::x86_sse_cvtss2si64:
3026  case Intrinsic::x86_sse_cvtss2si:
3027  case Intrinsic::x86_sse_cvttss2si64:
3028  case Intrinsic::x86_sse_cvttss2si:
3029  handleVectorConvertIntrinsic(I, 1);
3030  break;
3031  case Intrinsic::x86_sse_cvtps2pi:
3032  case Intrinsic::x86_sse_cvttps2pi:
3033  handleVectorConvertIntrinsic(I, 2);
3034  break;
3035 
3036  case Intrinsic::x86_avx512_psll_w_512:
3037  case Intrinsic::x86_avx512_psll_d_512:
3038  case Intrinsic::x86_avx512_psll_q_512:
3039  case Intrinsic::x86_avx512_pslli_w_512:
3040  case Intrinsic::x86_avx512_pslli_d_512:
3041  case Intrinsic::x86_avx512_pslli_q_512:
3042  case Intrinsic::x86_avx512_psrl_w_512:
3043  case Intrinsic::x86_avx512_psrl_d_512:
3044  case Intrinsic::x86_avx512_psrl_q_512:
3045  case Intrinsic::x86_avx512_psra_w_512:
3046  case Intrinsic::x86_avx512_psra_d_512:
3047  case Intrinsic::x86_avx512_psra_q_512:
3048  case Intrinsic::x86_avx512_psrli_w_512:
3049  case Intrinsic::x86_avx512_psrli_d_512:
3050  case Intrinsic::x86_avx512_psrli_q_512:
3051  case Intrinsic::x86_avx512_psrai_w_512:
3052  case Intrinsic::x86_avx512_psrai_d_512:
3053  case Intrinsic::x86_avx512_psrai_q_512:
3054  case Intrinsic::x86_avx512_psra_q_256:
3055  case Intrinsic::x86_avx512_psra_q_128:
3056  case Intrinsic::x86_avx512_psrai_q_256:
3057  case Intrinsic::x86_avx512_psrai_q_128:
3058  case Intrinsic::x86_avx2_psll_w:
3059  case Intrinsic::x86_avx2_psll_d:
3060  case Intrinsic::x86_avx2_psll_q:
3061  case Intrinsic::x86_avx2_pslli_w:
3062  case Intrinsic::x86_avx2_pslli_d:
3063  case Intrinsic::x86_avx2_pslli_q:
3064  case Intrinsic::x86_avx2_psrl_w:
3065  case Intrinsic::x86_avx2_psrl_d:
3066  case Intrinsic::x86_avx2_psrl_q:
3067  case Intrinsic::x86_avx2_psra_w:
3068  case Intrinsic::x86_avx2_psra_d:
3069  case Intrinsic::x86_avx2_psrli_w:
3070  case Intrinsic::x86_avx2_psrli_d:
3071  case Intrinsic::x86_avx2_psrli_q:
3072  case Intrinsic::x86_avx2_psrai_w:
3073  case Intrinsic::x86_avx2_psrai_d:
3074  case Intrinsic::x86_sse2_psll_w:
3075  case Intrinsic::x86_sse2_psll_d:
3076  case Intrinsic::x86_sse2_psll_q:
3077  case Intrinsic::x86_sse2_pslli_w:
3078  case Intrinsic::x86_sse2_pslli_d:
3079  case Intrinsic::x86_sse2_pslli_q:
3080  case Intrinsic::x86_sse2_psrl_w:
3081  case Intrinsic::x86_sse2_psrl_d:
3082  case Intrinsic::x86_sse2_psrl_q:
3083  case Intrinsic::x86_sse2_psra_w:
3084  case Intrinsic::x86_sse2_psra_d:
3085  case Intrinsic::x86_sse2_psrli_w:
3086  case Intrinsic::x86_sse2_psrli_d:
3087  case Intrinsic::x86_sse2_psrli_q:
3088  case Intrinsic::x86_sse2_psrai_w:
3089  case Intrinsic::x86_sse2_psrai_d:
3090  case Intrinsic::x86_mmx_psll_w:
3091  case Intrinsic::x86_mmx_psll_d:
3092  case Intrinsic::x86_mmx_psll_q:
3093  case Intrinsic::x86_mmx_pslli_w:
3094  case Intrinsic::x86_mmx_pslli_d:
3095  case Intrinsic::x86_mmx_pslli_q:
3096  case Intrinsic::x86_mmx_psrl_w:
3097  case Intrinsic::x86_mmx_psrl_d:
3098  case Intrinsic::x86_mmx_psrl_q:
3099  case Intrinsic::x86_mmx_psra_w:
3100  case Intrinsic::x86_mmx_psra_d:
3101  case Intrinsic::x86_mmx_psrli_w:
3102  case Intrinsic::x86_mmx_psrli_d:
3103  case Intrinsic::x86_mmx_psrli_q:
3104  case Intrinsic::x86_mmx_psrai_w:
3105  case Intrinsic::x86_mmx_psrai_d:
3106  handleVectorShiftIntrinsic(I, /* Variable */ false);
3107  break;
3108  case Intrinsic::x86_avx2_psllv_d:
3109  case Intrinsic::x86_avx2_psllv_d_256:
3110  case Intrinsic::x86_avx512_psllv_d_512:
3111  case Intrinsic::x86_avx2_psllv_q:
3112  case Intrinsic::x86_avx2_psllv_q_256:
3113  case Intrinsic::x86_avx512_psllv_q_512:
3114  case Intrinsic::x86_avx2_psrlv_d:
3115  case Intrinsic::x86_avx2_psrlv_d_256:
3116  case Intrinsic::x86_avx512_psrlv_d_512:
3117  case Intrinsic::x86_avx2_psrlv_q:
3118  case Intrinsic::x86_avx2_psrlv_q_256:
3119  case Intrinsic::x86_avx512_psrlv_q_512:
3120  case Intrinsic::x86_avx2_psrav_d:
3121  case Intrinsic::x86_avx2_psrav_d_256:
3122  case Intrinsic::x86_avx512_psrav_d_512:
3123  case Intrinsic::x86_avx512_psrav_q_128:
3124  case Intrinsic::x86_avx512_psrav_q_256:
3125  case Intrinsic::x86_avx512_psrav_q_512:
3126  handleVectorShiftIntrinsic(I, /* Variable */ true);
3127  break;
3128 
3129  case Intrinsic::x86_sse2_packsswb_128:
3130  case Intrinsic::x86_sse2_packssdw_128:
3131  case Intrinsic::x86_sse2_packuswb_128:
3132  case Intrinsic::x86_sse41_packusdw:
3133  case Intrinsic::x86_avx2_packsswb:
3134  case Intrinsic::x86_avx2_packssdw:
3135  case Intrinsic::x86_avx2_packuswb:
3136  case Intrinsic::x86_avx2_packusdw:
3137  handleVectorPackIntrinsic(I);
3138  break;
3139 
3140  case Intrinsic::x86_mmx_packsswb:
3141  case Intrinsic::x86_mmx_packuswb:
3142  handleVectorPackIntrinsic(I, 16);
3143  break;
3144 
3145  case Intrinsic::x86_mmx_packssdw:
3146  handleVectorPackIntrinsic(I, 32);
3147  break;
3148 
3149  case Intrinsic::x86_mmx_psad_bw:
3150  case Intrinsic::x86_sse2_psad_bw:
3151  case Intrinsic::x86_avx2_psad_bw:
3152  handleVectorSadIntrinsic(I);
3153  break;
3154 
3155  case Intrinsic::x86_sse2_pmadd_wd:
3156  case Intrinsic::x86_avx2_pmadd_wd:
3157  case Intrinsic::x86_ssse3_pmadd_ub_sw_128:
3158  case Intrinsic::x86_avx2_pmadd_ub_sw:
3159  handleVectorPmaddIntrinsic(I);
3160  break;
3161 
3162  case Intrinsic::x86_ssse3_pmadd_ub_sw:
3163  handleVectorPmaddIntrinsic(I, 8);
3164  break;
3165 
3166  case Intrinsic::x86_mmx_pmadd_wd:
3167  handleVectorPmaddIntrinsic(I, 16);
3168  break;
3169 
3170  case Intrinsic::x86_sse_cmp_ss:
3171  case Intrinsic::x86_sse2_cmp_sd:
3172  case Intrinsic::x86_sse_comieq_ss:
3173  case Intrinsic::x86_sse_comilt_ss:
3174  case Intrinsic::x86_sse_comile_ss:
3175  case Intrinsic::x86_sse_comigt_ss:
3176  case Intrinsic::x86_sse_comige_ss:
3177  case Intrinsic::x86_sse_comineq_ss:
3178  case Intrinsic::x86_sse_ucomieq_ss:
3179  case Intrinsic::x86_sse_ucomilt_ss:
3180  case Intrinsic::x86_sse_ucomile_ss:
3181  case Intrinsic::x86_sse_ucomigt_ss:
3182  case Intrinsic::x86_sse_ucomige_ss:
3183  case Intrinsic::x86_sse_ucomineq_ss:
3184  case Intrinsic::x86_sse2_comieq_sd:
3185  case Intrinsic::x86_sse2_comilt_sd:
3186  case Intrinsic::x86_sse2_comile_sd:
3187  case Intrinsic::x86_sse2_comigt_sd:
3188  case Intrinsic::x86_sse2_comige_sd:
3189  case Intrinsic::x86_sse2_comineq_sd:
3190  case Intrinsic::x86_sse2_ucomieq_sd:
3191  case Intrinsic::x86_sse2_ucomilt_sd:
3192  case Intrinsic::x86_sse2_ucomile_sd:
3193  case Intrinsic::x86_sse2_ucomigt_sd:
3194  case Intrinsic::x86_sse2_ucomige_sd:
3195  case Intrinsic::x86_sse2_ucomineq_sd:
3196  handleVectorCompareScalarIntrinsic(I);
3197  break;
3198 
3199  case Intrinsic::x86_sse_cmp_ps:
3200  case Intrinsic::x86_sse2_cmp_pd:
3201  // FIXME: For x86_avx_cmp_pd_256 and x86_avx_cmp_ps_256 this function
3202  // generates reasonably looking IR that fails in the backend with "Do not
3203  // know how to split the result of this operator!".
3204  handleVectorComparePackedIntrinsic(I);
3205  break;
3206 
3207  case Intrinsic::x86_bmi_bextr_32:
3208  case Intrinsic::x86_bmi_bextr_64:
3209  case Intrinsic::x86_bmi_bzhi_32:
3210  case Intrinsic::x86_bmi_bzhi_64:
3211  case Intrinsic::x86_bmi_pdep_32:
3212  case Intrinsic::x86_bmi_pdep_64:
3213  case Intrinsic::x86_bmi_pext_32:
3214  case Intrinsic::x86_bmi_pext_64:
3215  handleBmiIntrinsic(I);
3216  break;
3217 
3218  case Intrinsic::is_constant:
3219  // The result of llvm.is.constant() is always defined.
3220  setShadow(&I, getCleanShadow(&I));
3221  setOrigin(&I, getCleanOrigin());
3222  break;
3223 
3224  default:
3225  if (!handleUnknownIntrinsic(I))
3226  visitInstruction(I);
3227  break;
3228  }
3229  }
3230 
3231  void visitCallSite(CallSite CS) {
3232  Instruction &I = *CS.getInstruction();
3233  assert(!I.getMetadata("nosanitize"));
3234  assert((CS.isCall() || CS.isInvoke()) && "Unknown type of CallSite");
3235  if (CS.isCall()) {
3236  CallInst *Call = cast<CallInst>(&I);
3237 
3238  // For inline asm, do the usual thing: check argument shadow and mark all
3239  // outputs as clean. Note that any side effects of the inline asm that are
3240  // not immediately visible in its constraints are not handled.
3241  if (Call->isInlineAsm()) {
3242  if (ClHandleAsmConservative && MS.CompileKernel)
3243  visitAsmInstruction(I);
3244  else
3245  visitInstruction(I);
3246  return;
3247  }
3248 
3249  assert(!isa<IntrinsicInst>(&I) && "intrinsics are handled elsewhere");
3250 
3251  // We are going to insert code that relies on the fact that the callee
3252  // will become a non-readonly function after it is instrumented by us. To
3253  // prevent this code from being optimized out, mark that function
3254  // non-readonly in advance.
3255  if (Function *Func = Call->getCalledFunction()) {
3256  // Clear out readonly/readnone attributes.
3257  AttrBuilder B;
3258  B.addAttribute(Attribute::ReadOnly)
3259  .addAttribute(Attribute::ReadNone);
3261  }
3262 
3264  }
3265  IRBuilder<> IRB(&I);
3266 
3267  unsigned ArgOffset = 0;
3268  LLVM_DEBUG(dbgs() << " CallSite: " << I << "\n");
3269  for (CallSite::arg_iterator ArgIt = CS.arg_begin(), End = CS.arg_end();
3270  ArgIt != End; ++ArgIt) {
3271  Value *A = *ArgIt;
3272  unsigned i = ArgIt - CS.arg_begin();
3273  if (!A->getType()->isSized()) {
3274  LLVM_DEBUG(dbgs() << "Arg " << i << " is not sized: " << I << "\n");
3275  continue;
3276  }
3277  unsigned Size = 0;
3278  Value *Store = nullptr;
3279  // Compute the Shadow for arg even if it is ByVal, because
3280  // in that case getShadow() will copy the actual arg shadow to
3281  // __msan_param_tls.
3282  Value *ArgShadow = getShadow(A);
3283  Value *ArgShadowBase = getShadowPtrForArgument(A, IRB, ArgOffset);
3284  LLVM_DEBUG(dbgs() << " Arg#" << i << ": " << *A
3285  << " Shadow: " << *ArgShadow << "\n");
3286  bool ArgIsInitialized = false;
3287  const DataLayout &DL = F.getParent()->getDataLayout();
3288  if (CS.paramHasAttr(i, Attribute::ByVal)) {
3289  assert(A->getType()->isPointerTy() &&
3290  "ByVal argument is not a pointer!");
3291  Size = DL.getTypeAllocSize(A->getType()->getPointerElementType());
3292  if (ArgOffset + Size > kParamTLSSize) break;
3293  unsigned ParamAlignment = CS.getParamAlignment(i);
3294  unsigned Alignment = std::min(ParamAlignment, kShadowTLSAlignment);
3295  Value *AShadowPtr =
3296  getShadowOriginPtr(A, IRB, IRB.getInt8Ty(), Alignment,
3297  /*isStore*/ false)
3298  .first;
3299 
3300  Store = IRB.CreateMemCpy(ArgShadowBase, Alignment, AShadowPtr,
3301  Alignment, Size);
3302  // TODO(glider): need to copy origins.
3303  } else {
3304  Size = DL.getTypeAllocSize(A->getType());
3305  if (ArgOffset + Size > kParamTLSSize) break;
3306  Store = IRB.CreateAlignedStore(ArgShadow, ArgShadowBase,
3307  kShadowTLSAlignment);
3308  Constant *Cst = dyn_cast<Constant>(ArgShadow);
3309  if (Cst && Cst->isNullValue()) ArgIsInitialized = true;
3310  }
3311  if (MS.TrackOrigins && !ArgIsInitialized)
3312  IRB.CreateStore(getOrigin(A),
3313  getOriginPtrForArgument(A, IRB, ArgOffset));
3314  (void)Store;
3315  assert(Size != 0 && Store != nullptr);
3316  LLVM_DEBUG(dbgs() << " Param:" << *Store << "\n");
3317  ArgOffset += alignTo(Size, 8);
3318  }
3319  LLVM_DEBUG(dbgs() << " done with call args\n");
3320 
3321  FunctionType *FT = CS.getFunctionType();
3322  if (FT->isVarArg()) {
3323  VAHelper->visitCallSite(CS, IRB);
3324  }
3325 
3326  // Now, get the shadow for the RetVal.
3327  if (!I.getType()->isSized()) return;
3328  // Don't emit the epilogue for musttail call returns.
3329  if (CS.isCall() && cast<CallInst>(&I)->isMustTailCall()) return;
3330  IRBuilder<> IRBBefore(&I);
3331  // Until we have full dynamic coverage, make sure the retval shadow is 0.
3332  Value *Base = getShadowPtrForRetval(&I, IRBBefore);
3333  IRBBefore.CreateAlignedStore(getCleanShadow(&I), Base, kShadowTLSAlignment);
3334  BasicBlock::iterator NextInsn;
3335  if (CS.isCall()) {
3336  NextInsn = ++I.getIterator();
3337  assert(NextInsn != I.getParent()->end());
3338  } else {
3339  BasicBlock *NormalDest = cast<InvokeInst>(&I)->getNormalDest();
3340  if (!NormalDest->getSinglePredecessor()) {
3341  // FIXME: this case is tricky, so we are just conservative here.
3342  // Perhaps we need to split the edge between this BB and NormalDest,
3343  // but a naive attempt to use SplitEdge leads to a crash.
3344  setShadow(&I, getCleanShadow(&I));
3345  setOrigin(&I, getCleanOrigin());
3346  return;
3347  }
3348  // FIXME: NextInsn is likely in a basic block that has not been visited yet.
3349  // Anything inserted there will be instrumented by MSan later!
3350  NextInsn = NormalDest->getFirstInsertionPt();
3351  assert(NextInsn != NormalDest->end() &&
3352  "Could not find insertion point for retval shadow load");
3353  }
3354  IRBuilder<> IRBAfter(&*NextInsn);
3355  Value *RetvalShadow = IRBAfter.CreateAlignedLoad(
3356  getShadowTy(&I), getShadowPtrForRetval(&I, IRBAfter),
3357  kShadowTLSAlignment, "_msret");
3358  setShadow(&I, RetvalShadow);
3359  if (MS.TrackOrigins)
3360  setOrigin(&I, IRBAfter.CreateLoad(MS.OriginTy,
3361  getOriginPtrForRetval(IRBAfter)));
3362  }
3363 
3364  bool isAMustTailRetVal(Value *RetVal) {
3365  if (auto *I = dyn_cast<BitCastInst>(RetVal)) {
3366  RetVal = I->getOperand(0);
3367  }
3368  if (auto *I = dyn_cast<CallInst>(RetVal)) {
3369  return I->isMustTailCall();
3370  }
3371  return false;
3372  }
3373 
3374  void visitReturnInst(ReturnInst &I) {
3375  IRBuilder<> IRB(&I);
3376  Value *RetVal = I.getReturnValue();
3377  if (!RetVal) return;
3378  // Don't emit the epilogue for musttail call returns.
3379  if (isAMustTailRetVal(RetVal)) return;
3380  Value *ShadowPtr = getShadowPtrForRetval(RetVal, IRB);
3381  if (CheckReturnValue) {
3382  insertShadowCheck(RetVal, &I);
3383  Value *Shadow = getCleanShadow(RetVal);
3384  IRB.CreateAlignedStore(Shadow, ShadowPtr, kShadowTLSAlignment);
3385  } else {
3386  Value *Shadow = getShadow(RetVal);
3387  IRB.CreateAlignedStore(Shadow, ShadowPtr, kShadowTLSAlignment);
3388  if (MS.TrackOrigins)
3389  IRB.CreateStore(getOrigin(RetVal), getOriginPtrForRetval(IRB));
3390  }
3391  }
3392 
3393  void visitPHINode(PHINode &I) {
3394  IRBuilder<> IRB(&I);
3395  if (!PropagateShadow) {
3396  setShadow(&I, getCleanShadow(&I));
3397  setOrigin(&I, getCleanOrigin());
3398  return;
3399  }
3400 
3401  ShadowPHINodes.push_back(&I);
3402  setShadow(&I, IRB.CreatePHI(getShadowTy(&I), I.getNumIncomingValues(),
3403  "_msphi_s"));
3404  if (MS.TrackOrigins)
3405  setOrigin(&I, IRB.CreatePHI(MS.OriginTy, I.getNumIncomingValues(),
3406  "_msphi_o"));
3407  }
3408 
3409  Value *getLocalVarDescription(AllocaInst &I) {
3410  SmallString<2048> StackDescriptionStorage;
3411  raw_svector_ostream StackDescription(StackDescriptionStorage);
3412  // We create a string with a description of the stack allocation and
3413  // pass it into __msan_set_alloca_origin.
3414  // It will be printed by the run-time if stack-originated UMR is found.
3415  // The first 4 bytes of the string are set to '----' and will be replaced
3416  // by __msan_va_arg_overflow_size_tls at the first call.
3417  StackDescription << "----" << I.getName() << "@" << F.getName();
3419  StackDescription.str());
3420  }
3421 
3422  void poisonAllocaUserspace(AllocaInst &I, IRBuilder<> &IRB, Value *Len) {
3423  if (PoisonStack && ClPoisonStackWithCall) {
3424  IRB.CreateCall(MS.MsanPoisonStackFn,
3425  {IRB.CreatePointerCast(&I, IRB.getInt8PtrTy()), Len});
3426  } else {
3427  Value *ShadowBase, *OriginBase;
3428  std::tie(ShadowBase, OriginBase) =
3429  getShadowOriginPtr(&I, IRB, IRB.getInt8Ty(), 1, /*isStore*/ true);
3430 
3431  Value *PoisonValue = IRB.getInt8(PoisonStack ? ClPoisonStackPattern : 0);
3432  IRB.CreateMemSet(ShadowBase, PoisonValue, Len, I.getAlignment());
3433  }
3434 
3435  if (PoisonStack && MS.TrackOrigins) {
3436  Value *Descr = getLocalVarDescription(I);
3437  IRB.CreateCall(MS.MsanSetAllocaOrigin4Fn,
3438  {IRB.CreatePointerCast(&I, IRB.getInt8PtrTy()), Len,
3439  IRB.CreatePointerCast(Descr, IRB.getInt8PtrTy()),
3440  IRB.CreatePointerCast(&F, MS.IntptrTy)});
3441  }
3442  }
3443 
3444  void poisonAllocaKmsan(AllocaInst &I, IRBuilder<> &IRB, Value *Len) {
3445  Value *Descr = getLocalVarDescription(I);
3446  if (PoisonStack) {
3447  IRB.CreateCall(MS.MsanPoisonAllocaFn,
3448  {IRB.CreatePointerCast(&I, IRB.getInt8PtrTy()), Len,
3449  IRB.CreatePointerCast(Descr, IRB.getInt8PtrTy())});
3450  } else {
3451  IRB.CreateCall(MS.MsanUnpoisonAllocaFn,
3452  {IRB.CreatePointerCast(&I, IRB.getInt8PtrTy()), Len});
3453  }
3454  }
3455 
3456  void instrumentAlloca(AllocaInst &I, Instruction *InsPoint = nullptr) {
3457  if (!InsPoint)
3458  InsPoint = &I;
3459  IRBuilder<> IRB(InsPoint->getNextNode());
3460  const DataLayout &DL = F.getParent()->getDataLayout();
3461  uint64_t TypeSize = DL.getTypeAllocSize(I.getAllocatedType());
3462  Value *Len = ConstantInt::get(MS.IntptrTy, TypeSize);
3463  if (I.isArrayAllocation())
3464  Len = IRB.CreateMul(Len, I.getArraySize());
3465 
3466  if (MS.CompileKernel)
3467  poisonAllocaKmsan(I, IRB, Len);
3468  else
3469  poisonAllocaUserspace(I, IRB, Len);
3470  }
3471 
3472  void visitAllocaInst(AllocaInst &I) {
3473  setShadow(&I, getCleanShadow(&I));
3474  setOrigin(&I, getCleanOrigin());
3475  // We'll get to this alloca later unless it's poisoned at the corresponding
3476  // llvm.lifetime.start.
3477  AllocaSet.insert(&I);
3478  }
3479 
3480  void visitSelectInst(SelectInst& I) {
3481  IRBuilder<> IRB(&I);
3482  // a = select b, c, d
3483  Value *B = I.getCondition();
3484  Value *C = I.getTrueValue();
3485  Value *D = I.getFalseValue();
3486  Value *Sb = getShadow(B);
3487  Value *Sc = getShadow(C);
3488  Value *Sd = getShadow(D);
3489 
3490  // Result shadow if condition shadow is 0.
3491  Value *Sa0 = IRB.CreateSelect(B, Sc, Sd);
3492  Value *Sa1;
3493  if (I.getType()->isAggregateType()) {
3494  // To avoid "sign extending" i1 to an arbitrary aggregate type, we just do
3495  // an extra "select". This results in much more compact IR.
3496  // Sa = select Sb, poisoned, (select b, Sc, Sd)
3497  Sa1 = getPoisonedShadow(getShadowTy(I.getType()));
3498  } else {
3499  // Sa = select Sb, [ (c^d) | Sc | Sd ], [ b ? Sc : Sd ]
3500  // If Sb (condition is poisoned), look for bits in c and d that are equal
3501  // and both unpoisoned.
3502  // If !Sb (condition is unpoisoned), simply pick one of Sc and Sd.
3503 
3504  // Cast arguments to shadow-compatible type.
3505  C = CreateAppToShadowCast(IRB, C);
3506  D = CreateAppToShadowCast(IRB, D);
3507 
3508  // Result shadow if condition shadow is 1.
3509  Sa1 = IRB.CreateOr(IRB.CreateXor(C, D), IRB.CreateOr(Sc, Sd));
3510  }
3511  Value *Sa = IRB.CreateSelect(Sb, Sa1, Sa0, "_msprop_select");
3512  setShadow(&I, Sa);
3513  if (MS.TrackOrigins) {
3514  // Origins are always i32, so any vector conditions must be flattened.
3515  // FIXME: consider tracking vector origins for app vectors?
3516  if (B->getType()->isVectorTy()) {
3517  Type *FlatTy = getShadowTyNoVec(B->getType());
3518  B = IRB.CreateICmpNE(IRB.CreateBitCast(B, FlatTy),
3519  ConstantInt::getNullValue(FlatTy));
3520  Sb = IRB.CreateICmpNE(IRB.CreateBitCast(Sb, FlatTy),
3521  ConstantInt::getNullValue(FlatTy));
3522  }
3523  // a = select b, c, d
3524  // Oa = Sb ? Ob : (b ? Oc : Od)
3525  setOrigin(
3526  &I, IRB.CreateSelect(Sb, getOrigin(I.getCondition()),
3527  IRB.CreateSelect(B, getOrigin(I.getTrueValue()),
3528  getOrigin(I.getFalseValue()))));
3529  }
3530  }
3531 
3532  void visitLandingPadInst(LandingPadInst &I) {
3533  // Do nothing.
3534  // See https://github.com/google/sanitizers/issues/504
3535  setShadow(&I, getCleanShadow(&I));
3536  setOrigin(&I, getCleanOrigin());
3537  }
3538 
3539  void visitCatchSwitchInst(CatchSwitchInst &I) {
3540  setShadow(&I, getCleanShadow(&I));
3541  setOrigin(&I, getCleanOrigin());
3542  }
3543 
3544  void visitFuncletPadInst(FuncletPadInst &I) {
3545  setShadow(&I, getCleanShadow(&I));
3546  setOrigin(&I, getCleanOrigin());
3547  }
3548 
3549  void visitGetElementPtrInst(GetElementPtrInst &I) {
3550  handleShadowOr(I);
3551  }
3552 
3553  void visitExtractValueInst(ExtractValueInst &I) {
3554  IRBuilder<> IRB(&I);
3555  Value *Agg = I.getAggregateOperand();
3556  LLVM_DEBUG(dbgs() << "ExtractValue: " << I << "\n");
3557  Value *AggShadow = getShadow(Agg);
3558  LLVM_DEBUG(dbgs() << " AggShadow: " << *AggShadow << "\n");
3559  Value *ResShadow = IRB.CreateExtractValue(AggShadow, I.getIndices());
3560  LLVM_DEBUG(dbgs() << " ResShadow: " << *ResShadow << "\n");
3561  setShadow(&I, ResShadow);
3562  setOriginForNaryOp(I);
3563  }
3564 
3565  void visitInsertValueInst(InsertValueInst &I) {
3566  IRBuilder<> IRB(&I);
3567  LLVM_DEBUG(dbgs() << "InsertValue: " << I << "\n");
3568  Value *AggShadow = getShadow(I.getAggregateOperand());
3569  Value *InsShadow = getShadow(I.getInsertedValueOperand());
3570  LLVM_DEBUG(dbgs() << " AggShadow: " << *AggShadow << "\n");
3571  LLVM_DEBUG(dbgs() << " InsShadow: " << *InsShadow << "\n");
3572  Value *Res = IRB.CreateInsertValue(AggShadow, InsShadow, I.getIndices());
3573  LLVM_DEBUG(dbgs() << " Res: " << *Res << "\n");
3574  setShadow(&I, Res);
3575  setOriginForNaryOp(I);
3576  }
3577 
3578  void dumpInst(Instruction &I) {
3579  if (CallInst *CI = dyn_cast<CallInst>(&I)) {
3580  errs() << "ZZZ call " << CI->getCalledFunction()->getName() << "\n";
3581  } else {
3582  errs() << "ZZZ " << I.getOpcodeName() << "\n";
3583  }
3584  errs() << "QQQ " << I << "\n";
3585  }
3586 
3587  void visitResumeInst(ResumeInst &I) {
3588  LLVM_DEBUG(dbgs() << "Resume: " << I << "\n");
3589  // Nothing to do here.
3590  }
3591 
3592  void visitCleanupReturnInst(CleanupReturnInst &CRI) {
3593  LLVM_DEBUG(dbgs() << "CleanupReturn: " << CRI << "\n");
3594  // Nothing to do here.
3595  }
3596 
3597  void visitCatchReturnInst(CatchReturnInst &CRI) {
3598  LLVM_DEBUG(dbgs() << "CatchReturn: " << CRI << "\n");
3599  // Nothing to do here.
3600  }
3601 
3602  void instrumentAsmArgument(Value *Operand, Instruction &I, IRBuilder<> &IRB,
3603  const DataLayout &DL, bool isOutput) {
3604  // For each assembly argument, we check its value for being initialized.
3605  // If the argument is a pointer, we assume it points to a single element
3606  // of the corresponding type (or to a 8-byte word, if the type is unsized).
3607  // Each such pointer is instrumented with a call to the runtime library.
3608  Type *OpType = Operand->getType();
3609  // Check the operand value itself.
3610  insertShadowCheck(Operand, &I);
3611  if (!OpType->isPointerTy() || !isOutput) {
3612  assert(!isOutput);
3613  return;
3614  }
3615  Type *ElType = OpType->getPointerElementType();
3616  if (!ElType->isSized())
3617  return;
3618  int Size = DL.getTypeStoreSize(ElType);
3619  Value *Ptr = IRB.CreatePointerCast(Operand, IRB.getInt8PtrTy());
3620  Value *SizeVal = ConstantInt::get(MS.IntptrTy, Size);
3621  IRB.CreateCall(MS.MsanInstrumentAsmStoreFn, {Ptr, SizeVal});
3622  }
3623 
3624  /// Get the number of output arguments returned by pointers.
3625  int getNumOutputArgs(InlineAsm *IA, CallInst *CI) {
3626  int NumRetOutputs = 0;
3627  int NumOutputs = 0;
3628  Type *RetTy = dyn_cast<Value>(CI)->getType();
3629  if (!RetTy->isVoidTy()) {
3630  // Register outputs are returned via the CallInst return value.
3631  StructType *ST = dyn_cast_or_null<StructType>(RetTy);
3632  if (ST)
3633  NumRetOutputs = ST->getNumElements();
3634  else
3635  NumRetOutputs = 1;
3636  }
3638  for (size_t i = 0, n = Constraints.size(); i < n; i++) {
3639  InlineAsm::ConstraintInfo Info = Constraints[i];
3640  switch (Info.Type) {
3641  case InlineAsm::isOutput:
3642  NumOutputs++;
3643  break;
3644  default:
3645  break;
3646  }
3647  }
3648  return NumOutputs - NumRetOutputs;
3649  }
3650 
3651  void visitAsmInstruction(Instruction &I) {
3652  // Conservative inline assembly handling: check for poisoned shadow of
3653  // asm() arguments, then unpoison the result and all the memory locations
3654  // pointed to by those arguments.
3655  // An inline asm() statement in C++ contains lists of input and output
3656  // arguments used by the assembly code. These are mapped to operands of the
3657  // CallInst as follows:
3658  // - nR register outputs ("=r) are returned by value in a single structure
3659  // (SSA value of the CallInst);
3660  // - nO other outputs ("=m" and others) are returned by pointer as first
3661  // nO operands of the CallInst;
3662  // - nI inputs ("r", "m" and others) are passed to CallInst as the
3663  // remaining nI operands.
3664  // The total number of asm() arguments in the source is nR+nO+nI, and the
3665  // corresponding CallInst has nO+nI+1 operands (the last operand is the
3666  // function to be called).
3667  const DataLayout &DL = F.getParent()->getDataLayout();
3668  CallInst *CI = dyn_cast<CallInst>(&I);
3669  IRBuilder<> IRB(&I);
3670  InlineAsm *IA = cast<InlineAsm>(CI->getCalledValue());
3671  int OutputArgs = getNumOutputArgs(IA, CI);
3672  // The last operand of a CallInst is the function itself.
3673  int NumOperands = CI->getNumOperands() - 1;
3674 
3675  // Check input arguments. Doing so before unpoisoning output arguments, so
3676  // that we won't overwrite uninit values before checking them.
3677  for (int i = OutputArgs; i < NumOperands; i++) {
3678  Value *Operand = CI->getOperand(i);
3679  instrumentAsmArgument(Operand, I, IRB, DL, /*isOutput*/ false);
3680  }
3681  // Unpoison output arguments. This must happen before the actual InlineAsm
3682  // call, so that the shadow for memory published in the asm() statement
3683  // remains valid.
3684  for (int i = 0; i < OutputArgs; i++) {
3685  Value *Operand = CI->getOperand(i);
3686  instrumentAsmArgument(Operand, I, IRB, DL, /*isOutput*/ true);
3687  }
3688 
3689  setShadow(&I, getCleanShadow(&I));
3690  setOrigin(&I, getCleanOrigin());
3691  }
3692 
3693  void visitInstruction(Instruction &I) {
3694  // Everything else: stop propagating and check for poisoned shadow.
3696  dumpInst(I);
3697  LLVM_DEBUG(dbgs() << "DEFAULT: " << I << "\n");
3698  for (size_t i = 0, n = I.getNumOperands(); i < n; i++) {
3699  Value *Operand = I.getOperand(i);
3700  if (Operand->getType()->isSized())
3701  insertShadowCheck(Operand, &I);
3702  }
3703  setShadow(&I, getCleanShadow(&I));
3704  setOrigin(&I, getCleanOrigin());
3705  }
3706 };
3707 
3708 /// AMD64-specific implementation of VarArgHelper.
3709 struct VarArgAMD64Helper : public VarArgHelper {
3710  // An unfortunate workaround for asymmetric lowering of va_arg stuff.
3711  // See a comment in visitCallSite for more details.
3712  static const unsigned AMD64GpEndOffset = 48; // AMD64 ABI Draft 0.99.6 p3.5.7
3713  static const unsigned AMD64FpEndOffsetSSE = 176;
3714  // If SSE is disabled, fp_offset in va_list is zero.
3715  static const unsigned AMD64FpEndOffsetNoSSE = AMD64GpEndOffset;
3716 
3717  unsigned AMD64FpEndOffset;
3718  Function &F;
3719  MemorySanitizer &MS;
3720  MemorySanitizerVisitor &MSV;
3721  Value *VAArgTLSCopy = nullptr;
3722  Value *VAArgTLSOriginCopy = nullptr;
3723  Value *VAArgOverflowSize = nullptr;
3724 
3725  SmallVector<CallInst*, 16> VAStartInstrumentationList;
3726 
3727  enum ArgKind { AK_GeneralPurpose, AK_FloatingPoint, AK_Memory };
3728 
3729  VarArgAMD64Helper(Function &F, MemorySanitizer &MS,
3730  MemorySanitizerVisitor &MSV)
3731  : F(F), MS(MS), MSV(MSV) {
3732  AMD64FpEndOffset = AMD64FpEndOffsetSSE;
3733  for (const auto &Attr : F.getAttributes().getFnAttributes()) {
3734  if (Attr.isStringAttribute() &&
3735  (Attr.getKindAsString() == "target-features")) {
3736  if (Attr.getValueAsString().contains("-sse"))
3737  AMD64FpEndOffset = AMD64FpEndOffsetNoSSE;
3738  break;
3739  }
3740  }
3741  }
3742 
3743  ArgKind classifyArgument(Value* arg) {
3744  // A very rough approximation of X86_64 argument classification rules.
3745  Type *T = arg->getType();
3746  if (T->isFPOrFPVectorTy() || T->isX86_MMXTy())
3747  return AK_FloatingPoint;
3748  if (T->isIntegerTy() && T->getPrimitiveSizeInBits() <= 64)
3749  return AK_GeneralPurpose;
3750  if (T->isPointerTy())
3751  return AK_GeneralPurpose;
3752  return AK_Memory;
3753  }
3754 
3755  // For VarArg functions, store the argument shadow in an ABI-specific format
3756  // that corresponds to va_list layout.
3757  // We do this because Clang lowers va_arg in the frontend, and this pass
3758  // only sees the low level code that deals with va_list internals.
3759  // A much easier alternative (provided that Clang emits va_arg instructions)
3760  // would have been to associate each live instance of va_list with a copy of
3761  // MSanParamTLS, and extract shadow on va_arg() call in the argument list
3762  // order.
3763  void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {
3764  unsigned GpOffset = 0;
3765  unsigned FpOffset = AMD64GpEndOffset;
3766  unsigned OverflowOffset = AMD64FpEndOffset;
3767  const DataLayout &DL = F.getParent()->getDataLayout();
3768  for (CallSite::arg_iterator ArgIt = CS.arg_begin(), End = CS.arg_end();
3769  ArgIt != End; ++ArgIt) {
3770  Value *A = *ArgIt;
3771  unsigned ArgNo = CS.getArgumentNo(ArgIt);
3772  bool IsFixed = ArgNo < CS.getFunctionType()->getNumParams();
3773  bool IsByVal = CS.paramHasAttr(ArgNo, Attribute::ByVal);
3774  if (IsByVal) {
3775  // ByVal arguments always go to the overflow area.
3776  // Fixed arguments passed through the overflow area will be stepped
3777  // over by va_start, so don't count them towards the offset.
3778  if (IsFixed)
3779  continue;
3780  assert(A->getType()->isPointerTy());
3781  Type *RealTy = A->getType()->getPointerElementType();
3782  uint64_t ArgSize = DL.getTypeAllocSize(RealTy);
3783  Value *ShadowBase = getShadowPtrForVAArgument(
3784  RealTy, IRB, OverflowOffset, alignTo(ArgSize, 8));
3785  Value *OriginBase = nullptr;
3786  if (MS.TrackOrigins)
3787  OriginBase = getOriginPtrForVAArgument(RealTy, IRB, OverflowOffset);
3788  OverflowOffset += alignTo(ArgSize, 8);
3789  if (!ShadowBase)
3790  continue;
3791  Value *ShadowPtr, *OriginPtr;
3792  std::tie(ShadowPtr, OriginPtr) =
3793  MSV.getShadowOriginPtr(A, IRB, IRB.getInt8Ty(), kShadowTLSAlignment,
3794  /*isStore*/ false);
3795 
3796  IRB.CreateMemCpy(ShadowBase, kShadowTLSAlignment, ShadowPtr,
3797  kShadowTLSAlignment, ArgSize);
3798  if (MS.TrackOrigins)
3799  IRB.CreateMemCpy(OriginBase, kShadowTLSAlignment, OriginPtr,
3800  kShadowTLSAlignment, ArgSize);
3801  } else {
3802  ArgKind AK = classifyArgument(A);
3803  if (AK == AK_GeneralPurpose && GpOffset >= AMD64GpEndOffset)
3804  AK = AK_Memory;
3805  if (AK == AK_FloatingPoint && FpOffset >= AMD64FpEndOffset)
3806  AK = AK_Memory;
3807  Value *ShadowBase, *OriginBase = nullptr;
3808  switch (AK) {
3809  case AK_GeneralPurpose:
3810  ShadowBase =
3811  getShadowPtrForVAArgument(A->getType(), IRB, GpOffset, 8);
3812  if (MS.TrackOrigins)
3813  OriginBase =
3814  getOriginPtrForVAArgument(A->getType(), IRB, GpOffset);
3815  GpOffset += 8;
3816  break;
3817  case AK_FloatingPoint:
3818  ShadowBase =
3819  getShadowPtrForVAArgument(A->getType(), IRB, FpOffset, 16);
3820  if (MS.TrackOrigins)
3821  OriginBase =
3822  getOriginPtrForVAArgument(A->getType(), IRB, FpOffset);
3823  FpOffset += 16;
3824  break;
3825  case AK_Memory:
3826  if (IsFixed)
3827  continue;
3828  uint64_t ArgSize = DL.getTypeAllocSize(A->getType());
3829  ShadowBase =
3830  getShadowPtrForVAArgument(A->getType(), IRB, OverflowOffset, 8);
3831  if (MS.TrackOrigins)
3832  OriginBase =
3833  getOriginPtrForVAArgument(A->getType(), IRB, OverflowOffset);
3834  OverflowOffset += alignTo(ArgSize, 8);
3835  }
3836  // Take fixed arguments into account for GpOffset and FpOffset,
3837  // but don't actually store shadows for them.
3838  // TODO(glider): don't call get*PtrForVAArgument() for them.
3839  if (IsFixed)
3840  continue;
3841  if (!ShadowBase)
3842  continue;
3843  Value *Shadow = MSV.getShadow(A);
3844  IRB.CreateAlignedStore(Shadow, ShadowBase, kShadowTLSAlignment);
3845  if (MS.TrackOrigins) {
3846  Value *Origin = MSV.getOrigin(A);
3847  unsigned StoreSize = DL.getTypeStoreSize(Shadow->getType());
3848  MSV.paintOrigin(IRB, Origin, OriginBase, StoreSize,
3849  std::max(kShadowTLSAlignment, kMinOriginAlignment));
3850  }
3851  }
3852  }
3853  Constant *OverflowSize =
3854  ConstantInt::get(IRB.getInt64Ty(), OverflowOffset - AMD64FpEndOffset);
3855  IRB.CreateStore(OverflowSize, MS.VAArgOverflowSizeTLS);
3856  }
3857 
3858  /// Compute the shadow address for a given va_arg.
3859  Value *getShadowPtrForVAArgument(Type *Ty, IRBuilder<> &IRB,
3860  unsigned ArgOffset, unsigned ArgSize) {
3861  // Make sure we don't overflow __msan_va_arg_tls.
3862  if (ArgOffset + ArgSize > kParamTLSSize)
3863  return nullptr;
3864  Value *Base = IRB.CreatePointerCast(MS.VAArgTLS, MS.IntptrTy);
3865  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
3866  return IRB.CreateIntToPtr(Base, PointerType::get(MSV.getShadowTy(Ty), 0),
3867  "_msarg_va_s");
3868  }
3869 
3870  /// Compute the origin address for a given va_arg.
3871  Value *getOriginPtrForVAArgument(Type *Ty, IRBuilder<> &IRB, int ArgOffset) {
3872  Value *Base = IRB.CreatePointerCast(MS.VAArgOriginTLS, MS.IntptrTy);
3873  // getOriginPtrForVAArgument() is always called after
3874  // getShadowPtrForVAArgument(), so __msan_va_arg_origin_tls can never
3875  // overflow.
3876  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
3877  return IRB.CreateIntToPtr(Base, PointerType::get(MS.OriginTy, 0),
3878  "_msarg_va_o");
3879  }
3880 
3881  void unpoisonVAListTagForInst(IntrinsicInst &I) {
3882  IRBuilder<> IRB(&I);
3883  Value *VAListTag = I.getArgOperand(0);
3884  Value *ShadowPtr, *OriginPtr;
3885  unsigned Alignment = 8;
3886  std::tie(ShadowPtr, OriginPtr) =
3887  MSV.getShadowOriginPtr(VAListTag, IRB, IRB.getInt8Ty(), Alignment,
3888  /*isStore*/ true);
3889 
3890  // Unpoison the whole __va_list_tag.
3891  // FIXME: magic ABI constants.
3892  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
3893  /* size */ 24, Alignment, false);
3894  // We shouldn't need to zero out the origins, as they're only checked for
3895  // nonzero shadow.
3896  }
3897 
3898  void visitVAStartInst(VAStartInst &I) override {
3900  return;
3901  VAStartInstrumentationList.push_back(&I);
3902  unpoisonVAListTagForInst(I);
3903  }
3904 
3905  void visitVACopyInst(VACopyInst &I) override {
3906  if (F.getCallingConv() == CallingConv::Win64) return;
3907  unpoisonVAListTagForInst(I);
3908  }
3909 
3910  void finalizeInstrumentation() override {
3911  assert(!VAArgOverflowSize && !VAArgTLSCopy &&
3912  "finalizeInstrumentation called twice");
3913  if (!VAStartInstrumentationList.empty()) {
3914  // If there is a va_start in this function, make a backup copy of
3915  // va_arg_tls somewhere in the function entry block.
3916  IRBuilder<> IRB(MSV.ActualFnStart->getFirstNonPHI());
3917  VAArgOverflowSize =
3918  IRB.CreateLoad(IRB.getInt64Ty(), MS.VAArgOverflowSizeTLS);
3919  Value *CopySize =
3920  IRB.CreateAdd(ConstantInt::get(MS.IntptrTy, AMD64FpEndOffset),
3921  VAArgOverflowSize);
3922  VAArgTLSCopy = IRB.CreateAlloca(Type::getInt8Ty(*MS.C), CopySize);
3923  IRB.CreateMemCpy(VAArgTLSCopy, 8, MS.VAArgTLS, 8, CopySize);
3924  if (MS.TrackOrigins) {
3925  VAArgTLSOriginCopy = IRB.CreateAlloca(Type::getInt8Ty(*MS.C), CopySize);
3926  IRB.CreateMemCpy(VAArgTLSOriginCopy, 8, MS.VAArgOriginTLS, 8, CopySize);
3927  }
3928  }
3929 
3930  // Instrument va_start.
3931  // Copy va_list shadow from the backup copy of the TLS contents.
3932  for (size_t i = 0, n = VAStartInstrumentationList.size(); i < n; i++) {
3933  CallInst *OrigInst = VAStartInstrumentationList[i];
3934  IRBuilder<> IRB(OrigInst->getNextNode());
3935  Value *VAListTag = OrigInst->getArgOperand(0);
3936 
3937  Type *RegSaveAreaPtrTy = Type::getInt64PtrTy(*MS.C);
3938  Value *RegSaveAreaPtrPtr = IRB.CreateIntToPtr(
3939  IRB.CreateAdd(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
3940  ConstantInt::get(MS.IntptrTy, 16)),
3941  PointerType::get(RegSaveAreaPtrTy, 0));
3942  Value *RegSaveAreaPtr =
3943  IRB.CreateLoad(RegSaveAreaPtrTy, RegSaveAreaPtrPtr);
3944  Value *RegSaveAreaShadowPtr, *RegSaveAreaOriginPtr;
3945  unsigned Alignment = 16;
3946  std::tie(RegSaveAreaShadowPtr, RegSaveAreaOriginPtr) =
3947  MSV.getShadowOriginPtr(RegSaveAreaPtr, IRB, IRB.getInt8Ty(),
3948  Alignment, /*isStore*/ true);
3949  IRB.CreateMemCpy(RegSaveAreaShadowPtr, Alignment, VAArgTLSCopy, Alignment,
3950  AMD64FpEndOffset);
3951  if (MS.TrackOrigins)
3952  IRB.CreateMemCpy(RegSaveAreaOriginPtr, Alignment, VAArgTLSOriginCopy,
3953  Alignment, AMD64FpEndOffset);
3954  Type *OverflowArgAreaPtrTy = Type::getInt64PtrTy(*MS.C);
3955  Value *OverflowArgAreaPtrPtr = IRB.CreateIntToPtr(
3956  IRB.CreateAdd(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
3957  ConstantInt::get(MS.IntptrTy, 8)),
3958  PointerType::get(OverflowArgAreaPtrTy, 0));
3959  Value *OverflowArgAreaPtr =
3960  IRB.CreateLoad(OverflowArgAreaPtrTy, OverflowArgAreaPtrPtr);
3961  Value *OverflowArgAreaShadowPtr, *OverflowArgAreaOriginPtr;
3962  std::tie(OverflowArgAreaShadowPtr, OverflowArgAreaOriginPtr) =
3963  MSV.getShadowOriginPtr(OverflowArgAreaPtr, IRB, IRB.getInt8Ty(),
3964  Alignment, /*isStore*/ true);
3965  Value *SrcPtr = IRB.CreateConstGEP1_32(IRB.getInt8Ty(), VAArgTLSCopy,
3966  AMD64FpEndOffset);
3967  IRB.CreateMemCpy(OverflowArgAreaShadowPtr, Alignment, SrcPtr, Alignment,
3968  VAArgOverflowSize);
3969  if (MS.TrackOrigins) {
3970  SrcPtr = IRB.CreateConstGEP1_32(IRB.getInt8Ty(), VAArgTLSOriginCopy,
3971  AMD64FpEndOffset);
3972  IRB.CreateMemCpy(OverflowArgAreaOriginPtr, Alignment, SrcPtr, Alignment,
3973  VAArgOverflowSize);
3974  }
3975  }
3976  }
3977 };
3978 
3979 /// MIPS64-specific implementation of VarArgHelper.
3980 struct VarArgMIPS64Helper : public VarArgHelper {
3981  Function &F;
3982  MemorySanitizer &MS;
3983  MemorySanitizerVisitor &MSV;
3984  Value *VAArgTLSCopy = nullptr;
3985  Value *VAArgSize = nullptr;
3986 
3987  SmallVector<CallInst*, 16> VAStartInstrumentationList;
3988 
3989  VarArgMIPS64Helper(Function &F, MemorySanitizer &MS,
3990  MemorySanitizerVisitor &MSV) : F(F), MS(MS), MSV(MSV) {}
3991 
3992  void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {
3993  unsigned VAArgOffset = 0;
3994  const DataLayout &DL = F.getParent()->getDataLayout();
3995  for (CallSite::arg_iterator ArgIt = CS.arg_begin() +
3996  CS.getFunctionType()->getNumParams(), End = CS.arg_end();
3997  ArgIt != End; ++ArgIt) {
3998  Triple TargetTriple(F.getParent()->getTargetTriple());
3999  Value *A = *ArgIt;
4000  Value *Base;
4001  uint64_t ArgSize = DL.getTypeAllocSize(A->getType());
4002  if (TargetTriple.getArch() == Triple::mips64) {
4003  // Adjusting the shadow for argument with size < 8 to match the placement
4004  // of bits in big endian system
4005  if (ArgSize < 8)
4006  VAArgOffset += (8 - ArgSize);
4007  }
4008  Base = getShadowPtrForVAArgument(A->getType(), IRB, VAArgOffset, ArgSize);
4009  VAArgOffset += ArgSize;
4010  VAArgOffset = alignTo(VAArgOffset, 8);
4011  if (!Base)
4012  continue;
4013  IRB.CreateAlignedStore(MSV.getShadow(A), Base, kShadowTLSAlignment);
4014  }
4015 
4016  Constant *TotalVAArgSize = ConstantInt::get(IRB.getInt64Ty(), VAArgOffset);
4017  // Here using VAArgOverflowSizeTLS as VAArgSizeTLS to avoid creation of
4018  // a new class member i.e. it is the total size of all VarArgs.
4019  IRB.CreateStore(TotalVAArgSize, MS.VAArgOverflowSizeTLS);
4020  }
4021 
4022  /// Compute the shadow address for a given va_arg.
4023  Value *getShadowPtrForVAArgument(Type *Ty, IRBuilder<> &IRB,
4024  unsigned ArgOffset, unsigned ArgSize) {
4025  // Make sure we don't overflow __msan_va_arg_tls.
4026  if (ArgOffset + ArgSize > kParamTLSSize)
4027  return nullptr;
4028  Value *Base = IRB.CreatePointerCast(MS.VAArgTLS, MS.IntptrTy);
4029  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
4030  return IRB.CreateIntToPtr(Base, PointerType::get(MSV.getShadowTy(Ty), 0),
4031  "_msarg");
4032  }
4033 
4034  void visitVAStartInst(VAStartInst &I) override {
4035  IRBuilder<> IRB(&I);
4036  VAStartInstrumentationList.push_back(&I);
4037  Value *VAListTag = I.getArgOperand(0);
4038  Value *ShadowPtr, *OriginPtr;
4039  unsigned Alignment = 8;
4040  std::tie(ShadowPtr, OriginPtr) = MSV.getShadowOriginPtr(
4041  VAListTag, IRB, IRB.getInt8Ty(), Alignment, /*isStore*/ true);
4042  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
4043  /* size */ 8, Alignment, false);
4044  }
4045 
4046  void visitVACopyInst(VACopyInst &I) override {
4047  IRBuilder<> IRB(&I);
4048  VAStartInstrumentationList.push_back(&I);
4049  Value *VAListTag = I.getArgOperand(0);
4050  Value *ShadowPtr, *OriginPtr;
4051  unsigned Alignment = 8;
4052  std::tie(ShadowPtr, OriginPtr) = MSV.getShadowOriginPtr(
4053  VAListTag, IRB, IRB.getInt8Ty(), Alignment, /*isStore*/ true);
4054  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
4055  /* size */ 8, Alignment, false);
4056  }
4057 
4058  void finalizeInstrumentation() override {
4059  assert(!VAArgSize && !VAArgTLSCopy &&
4060  "finalizeInstrumentation called twice");
4061  IRBuilder<> IRB(MSV.ActualFnStart->getFirstNonPHI());
4062  VAArgSize = IRB.CreateLoad(IRB.getInt64Ty(), MS.VAArgOverflowSizeTLS);
4063  Value *CopySize = IRB.CreateAdd(ConstantInt::get(MS.IntptrTy, 0),
4064  VAArgSize);
4065 
4066  if (!VAStartInstrumentationList.empty()) {
4067  // If there is a va_start in this function, make a backup copy of
4068  // va_arg_tls somewhere in the function entry block.
4069  VAArgTLSCopy = IRB.CreateAlloca(Type::getInt8Ty(*MS.C), CopySize);
4070  IRB.CreateMemCpy(VAArgTLSCopy, 8, MS.VAArgTLS, 8, CopySize);
4071  }
4072 
4073  // Instrument va_start.
4074  // Copy va_list shadow from the backup copy of the TLS contents.
4075  for (size_t i = 0, n = VAStartInstrumentationList.size(); i < n; i++) {
4076  CallInst *OrigInst = VAStartInstrumentationList[i];
4077  IRBuilder<> IRB(OrigInst->getNextNode());
4078  Value *VAListTag = OrigInst->getArgOperand(0);
4079  Type *RegSaveAreaPtrTy = Type::getInt64PtrTy(*MS.C);
4080  Value *RegSaveAreaPtrPtr =
4081  IRB.CreateIntToPtr(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
4082  PointerType::get(RegSaveAreaPtrTy, 0));
4083  Value *RegSaveAreaPtr =
4084  IRB.CreateLoad(RegSaveAreaPtrTy, RegSaveAreaPtrPtr);
4085  Value *RegSaveAreaShadowPtr, *RegSaveAreaOriginPtr;
4086  unsigned Alignment = 8;
4087  std::tie(RegSaveAreaShadowPtr, RegSaveAreaOriginPtr) =
4088  MSV.getShadowOriginPtr(RegSaveAreaPtr, IRB, IRB.getInt8Ty(),
4089  Alignment, /*isStore*/ true);
4090  IRB.CreateMemCpy(RegSaveAreaShadowPtr, Alignment, VAArgTLSCopy, Alignment,
4091  CopySize);
4092  }
4093  }
4094 };
4095 
4096 /// AArch64-specific implementation of VarArgHelper.
4097 struct VarArgAArch64Helper : public VarArgHelper {
4098  static const unsigned kAArch64GrArgSize = 64;
4099  static const unsigned kAArch64VrArgSize = 128;
4100 
4101  static const unsigned AArch64GrBegOffset = 0;
4102  static const unsigned AArch64GrEndOffset = kAArch64GrArgSize;
4103  // Make VR space aligned to 16 bytes.
4104  static const unsigned AArch64VrBegOffset = AArch64GrEndOffset;
4105  static const unsigned AArch64VrEndOffset = AArch64VrBegOffset
4106  + kAArch64VrArgSize;
4107  static const unsigned AArch64VAEndOffset = AArch64VrEndOffset;
4108 
4109  Function &F;
4110  MemorySanitizer &MS;
4111  MemorySanitizerVisitor &MSV;
4112  Value *VAArgTLSCopy = nullptr;
4113  Value *VAArgOverflowSize = nullptr;
4114 
4115  SmallVector<CallInst*, 16> VAStartInstrumentationList;
4116 
4117  enum ArgKind { AK_GeneralPurpose, AK_FloatingPoint, AK_Memory };
4118 
4119  VarArgAArch64Helper(Function &F, MemorySanitizer &MS,
4120  MemorySanitizerVisitor &MSV) : F(F), MS(MS), MSV(MSV) {}
4121 
4122  ArgKind classifyArgument(Value* arg) {
4123  Type *T = arg->getType();
4124  if (T->isFPOrFPVectorTy())
4125  return AK_FloatingPoint;
4126  if ((T->isIntegerTy() && T->getPrimitiveSizeInBits() <= 64)
4127  || (T->isPointerTy()))
4128  return AK_GeneralPurpose;
4129  return AK_Memory;
4130  }
4131 
4132  // The instrumentation stores the argument shadow in a non ABI-specific
4133  // format because it does not know which argument is named (since Clang,
4134  // like x86_64 case, lowers the va_args in the frontend and this pass only
4135  // sees the low level code that deals with va_list internals).
4136  // The first seven GR registers are saved in the first 56 bytes of the
4137  // va_arg tls arra, followers by the first 8 FP/SIMD registers, and then
4138  // the remaining arguments.
4139  // Using constant offset within the va_arg TLS array allows fast copy
4140  // in the finalize instrumentation.
4141  void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {
4142  unsigned GrOffset = AArch64GrBegOffset;
4143  unsigned VrOffset = AArch64VrBegOffset;
4144  unsigned OverflowOffset = AArch64VAEndOffset;
4145 
4146  const DataLayout &DL = F.getParent()->getDataLayout();
4147  for (CallSite::arg_iterator ArgIt = CS.arg_begin(), End = CS.arg_end();
4148  ArgIt != End; ++ArgIt) {
4149  Value *A = *ArgIt;
4150  unsigned ArgNo = CS.getArgumentNo(ArgIt);
4151  bool IsFixed = ArgNo < CS.getFunctionType()->getNumParams();
4152  ArgKind AK = classifyArgument(A);
4153  if (AK == AK_GeneralPurpose && GrOffset >= AArch64GrEndOffset)
4154  AK = AK_Memory;
4155  if (AK == AK_FloatingPoint && VrOffset >= AArch64VrEndOffset)
4156  AK = AK_Memory;
4157  Value *Base;
4158  switch (AK) {
4159  case AK_GeneralPurpose:
4160  Base = getShadowPtrForVAArgument(A->getType(), IRB, GrOffset, 8);
4161  GrOffset += 8;
4162  break;
4163  case AK_FloatingPoint:
4164  Base = getShadowPtrForVAArgument(A->getType(), IRB, VrOffset, 8);
4165  VrOffset += 16;
4166  break;
4167  case AK_Memory:
4168  // Don't count fixed arguments in the overflow area - va_start will
4169  // skip right over them.
4170  if (IsFixed)
4171  continue;
4172  uint64_t ArgSize = DL.getTypeAllocSize(A->getType());
4173  Base = getShadowPtrForVAArgument(A->getType(), IRB, OverflowOffset,
4174  alignTo(ArgSize, 8));
4175  OverflowOffset += alignTo(ArgSize, 8);
4176  break;
4177  }
4178  // Count Gp/Vr fixed arguments to their respective offsets, but don't
4179  // bother to actually store a shadow.
4180  if (IsFixed)
4181  continue;
4182  if (!Base)
4183  continue;
4184  IRB.CreateAlignedStore(MSV.getShadow(A), Base, kShadowTLSAlignment);
4185  }
4186  Constant *OverflowSize =
4187  ConstantInt::get(IRB.getInt64Ty(), OverflowOffset - AArch64VAEndOffset);
4188  IRB.CreateStore(OverflowSize, MS.VAArgOverflowSizeTLS);
4189  }
4190 
4191  /// Compute the shadow address for a given va_arg.
4192  Value *getShadowPtrForVAArgument(Type *Ty, IRBuilder<> &IRB,
4193  unsigned ArgOffset, unsigned ArgSize) {
4194  // Make sure we don't overflow __msan_va_arg_tls.
4195  if (ArgOffset + ArgSize > kParamTLSSize)
4196  return nullptr;
4197  Value *Base = IRB.CreatePointerCast(MS.VAArgTLS, MS.IntptrTy);
4198  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
4199  return IRB.CreateIntToPtr(Base, PointerType::get(MSV.getShadowTy(Ty), 0),
4200  "_msarg");
4201  }
4202 
4203  void visitVAStartInst(VAStartInst &I) override {
4204  IRBuilder<> IRB(&I);
4205  VAStartInstrumentationList.push_back(&I);
4206  Value *VAListTag = I.getArgOperand(0);
4207  Value *ShadowPtr, *OriginPtr;
4208  unsigned Alignment = 8;
4209  std::tie(ShadowPtr, OriginPtr) = MSV.getShadowOriginPtr(
4210  VAListTag, IRB, IRB.getInt8Ty(), Alignment, /*isStore*/ true);
4211  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
4212  /* size */ 32, Alignment, false);
4213  }
4214 
4215  void visitVACopyInst(VACopyInst &I) override {
4216  IRBuilder<> IRB(&I);
4217  VAStartInstrumentationList.push_back(&I);
4218  Value *VAListTag = I.getArgOperand(0);
4219  Value *ShadowPtr, *OriginPtr;
4220  unsigned Alignment = 8;
4221  std::tie(ShadowPtr, OriginPtr) = MSV.getShadowOriginPtr(
4222  VAListTag, IRB, IRB.getInt8Ty(), Alignment, /*isStore*/ true);
4223  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
4224  /* size */ 32, Alignment, false);
4225  }
4226 
4227  // Retrieve a va_list field of 'void*' size.
4228  Value* getVAField64(IRBuilder<> &IRB, Value *VAListTag, int offset) {
4229  Value *SaveAreaPtrPtr =
4230  IRB.CreateIntToPtr(
4231  IRB.CreateAdd(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
4232  ConstantInt::get(MS.IntptrTy, offset)),
4233  Type::getInt64PtrTy(*MS.C));
4234  return IRB.CreateLoad(Type::getInt64Ty(*MS.C), SaveAreaPtrPtr);
4235  }
4236 
4237  // Retrieve a va_list field of 'int' size.
4238  Value* getVAField32(IRBuilder<> &IRB, Value *VAListTag, int offset) {
4239  Value *SaveAreaPtr =
4240  IRB.CreateIntToPtr(
4241  IRB.CreateAdd(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
4242  ConstantInt::get(MS.IntptrTy, offset)),
4243  Type::getInt32PtrTy(*MS.C));
4244  Value *SaveArea32 = IRB.CreateLoad(IRB.getInt32Ty(), SaveAreaPtr);
4245  return IRB.CreateSExt(SaveArea32, MS.IntptrTy);
4246  }
4247 
4248  void finalizeInstrumentation() override {
4249  assert(!VAArgOverflowSize && !VAArgTLSCopy &&
4250  "finalizeInstrumentation called twice");
4251  if (!VAStartInstrumentationList.empty()) {
4252  // If there is a va_start in this function, make a backup copy of
4253  // va_arg_tls somewhere in the function entry block.
4254  IRBuilder<> IRB(MSV.ActualFnStart->getFirstNonPHI());
4255  VAArgOverflowSize =
4256  IRB.CreateLoad(IRB.getInt64Ty(), MS.VAArgOverflowSizeTLS);
4257  Value *CopySize =
4258  IRB.CreateAdd(ConstantInt::get(MS.IntptrTy, AArch64VAEndOffset),
4259  VAArgOverflowSize);
4260  VAArgTLSCopy = IRB.CreateAlloca(Type::getInt8Ty(*MS.C), CopySize);
4261  IRB.CreateMemCpy(VAArgTLSCopy, 8, MS.VAArgTLS, 8, CopySize);
4262  }
4263 
4264  Value *GrArgSize = ConstantInt::get(MS.IntptrTy, kAArch64GrArgSize);
4265  Value *VrArgSize = ConstantInt::get(MS.IntptrTy, kAArch64VrArgSize);
4266 
4267  // Instrument va_start, copy va_list shadow from the backup copy of
4268  // the TLS contents.
4269  for (size_t i = 0, n = VAStartInstrumentationList.size(); i < n; i++) {
4270  CallInst *OrigInst = VAStartInstrumentationList[i];
4271  IRBuilder<> IRB(OrigInst->getNextNode());
4272 
4273  Value *VAListTag = OrigInst->getArgOperand(0);
4274 
4275  // The variadic ABI for AArch64 creates two areas to save the incoming
4276  // argument registers (one for 64-bit general register xn-x7 and another
4277  // for 128-bit FP/SIMD vn-v7).
4278  // We need then to propagate the shadow arguments on both regions
4279  // 'va::__gr_top + va::__gr_offs' and 'va::__vr_top + va::__vr_offs'.
4280  // The remaning arguments are saved on shadow for 'va::stack'.
4281  // One caveat is it requires only to propagate the non-named arguments,
4282  // however on the call site instrumentation 'all' the arguments are
4283  // saved. So to copy the shadow values from the va_arg TLS array
4284  // we need to adjust the offset for both GR and VR fields based on
4285  // the __{gr,vr}_offs value (since they are stores based on incoming
4286  // named arguments).
4287 
4288  // Read the stack pointer from the va_list.
4289  Value *StackSaveAreaPtr = getVAField64(IRB, VAListTag, 0);
4290 
4291  // Read both the __gr_top and __gr_off and add them up.
4292  Value *GrTopSaveAreaPtr = getVAField64(IRB, VAListTag, 8);
4293  Value *GrOffSaveArea = getVAField32(IRB, VAListTag, 24);
4294 
4295  Value *GrRegSaveAreaPtr = IRB.CreateAdd(GrTopSaveAreaPtr, GrOffSaveArea);
4296 
4297  // Read both the __vr_top and __vr_off and add them up.
4298  Value *VrTopSaveAreaPtr = getVAField64(IRB, VAListTag, 16);
4299  Value *VrOffSaveArea = getVAField32(IRB, VAListTag, 28);
4300 
4301  Value *VrRegSaveAreaPtr = IRB.CreateAdd(VrTopSaveAreaPtr, VrOffSaveArea);
4302 
4303  // It does not know how many named arguments is being used and, on the
4304  // callsite all the arguments were saved. Since __gr_off is defined as
4305  // '0 - ((8 - named_gr) * 8)', the idea is to just propagate the variadic
4306  // argument by ignoring the bytes of shadow from named arguments.
4307  Value *GrRegSaveAreaShadowPtrOff =
4308  IRB.CreateAdd(GrArgSize, GrOffSaveArea);
4309 
4310  Value *GrRegSaveAreaShadowPtr =
4311  MSV.getShadowOriginPtr(GrRegSaveAreaPtr, IRB, IRB.getInt8Ty(),
4312  /*Alignment*/ 8, /*isStore*/ true)
4313  .first;
4314 
4315  Value *GrSrcPtr = IRB.CreateInBoundsGEP(IRB.getInt8Ty(), VAArgTLSCopy,
4316  GrRegSaveAreaShadowPtrOff);
4317  Value *GrCopySize = IRB.CreateSub(GrArgSize, GrRegSaveAreaShadowPtrOff);
4318 
4319  IRB.CreateMemCpy(GrRegSaveAreaShadowPtr, 8, GrSrcPtr, 8, GrCopySize);
4320 
4321  // Again, but for FP/SIMD values.
4322  Value *VrRegSaveAreaShadowPtrOff =
4323  IRB.CreateAdd(VrArgSize, VrOffSaveArea);
4324 
4325  Value *VrRegSaveAreaShadowPtr =
4326  MSV.getShadowOriginPtr(VrRegSaveAreaPtr, IRB, IRB.getInt8Ty(),
4327  /*Alignment*/ 8, /*isStore*/ true)
4328  .first;
4329 
4330  Value *VrSrcPtr = IRB.CreateInBoundsGEP(
4331  IRB.getInt8Ty(),
4332  IRB.CreateInBoundsGEP(IRB.getInt8Ty(), VAArgTLSCopy,
4333  IRB.getInt32(AArch64VrBegOffset)),
4334  VrRegSaveAreaShadowPtrOff);
4335  Value *VrCopySize = IRB.CreateSub(VrArgSize, VrRegSaveAreaShadowPtrOff);
4336 
4337  IRB.CreateMemCpy(VrRegSaveAreaShadowPtr, 8, VrSrcPtr, 8, VrCopySize);
4338 
4339  // And finally for remaining arguments.
4340  Value *StackSaveAreaShadowPtr =
4341  MSV.getShadowOriginPtr(StackSaveAreaPtr, IRB, IRB.getInt8Ty(),
4342  /*Alignment*/ 16, /*isStore*/ true)
4343  .first;
4344 
4345  Value *StackSrcPtr =
4346  IRB.CreateInBoundsGEP(IRB.getInt8Ty(), VAArgTLSCopy,
4347  IRB.getInt32(AArch64VAEndOffset));
4348 
4349  IRB.CreateMemCpy(StackSaveAreaShadowPtr, 16, StackSrcPtr, 16,
4350  VAArgOverflowSize);
4351  }
4352  }
4353 };
4354 
4355 /// PowerPC64-specific implementation of VarArgHelper.
4356 struct VarArgPowerPC64Helper : public VarArgHelper {
4357  Function &F;
4358  MemorySanitizer &MS;
4359  MemorySanitizerVisitor &MSV;
4360  Value *VAArgTLSCopy = nullptr;
4361  Value *VAArgSize = nullptr;
4362 
4363  SmallVector<CallInst*, 16> VAStartInstrumentationList;
4364 
4365  VarArgPowerPC64Helper(Function &F, MemorySanitizer &MS,
4366  MemorySanitizerVisitor &MSV) : F(F), MS(MS), MSV(MSV) {}
4367 
4368  void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {
4369  // For PowerPC, we need to deal with alignment of stack arguments -
4370  // they are mostly aligned to 8 bytes, but vectors and i128 arrays
4371  // are aligned to 16 bytes, byvals can be aligned to 8 or 16 bytes,
4372  // and QPX vectors are aligned to 32 bytes. For that reason, we
4373  // compute current offset from stack pointer (which is always properly
4374  // aligned), and offset for the first vararg, then subtract them.
4375  unsigned VAArgBase;
4376  Triple TargetTriple(F.getParent()->getTargetTriple());
4377  // Parameter save area starts at 48 bytes from frame pointer for ABIv1,
4378  // and 32 bytes for ABIv2. This is usually determined by target
4379  // endianness, but in theory could be overriden by function attribute.
4380  // For simplicity, we ignore it here (it'd only matter for QPX vectors).
4381  if (TargetTriple.getArch() == Triple::ppc64)
4382  VAArgBase = 48;
4383  else
4384  VAArgBase = 32;
4385  unsigned VAArgOffset = VAArgBase;
4386  const DataLayout &DL = F.getParent()->getDataLayout();
4387  for (CallSite::arg_iterator ArgIt = CS.arg_begin(), End = CS.arg_end();
4388  ArgIt != End; ++ArgIt) {
4389  Value *A = *ArgIt;
4390  unsigned ArgNo = CS.getArgumentNo(ArgIt);
4391  bool IsFixed = ArgNo < CS.getFunctionType()->getNumParams();
4392  bool IsByVal = CS.paramHasAttr(ArgNo, Attribute::ByVal);
4393  if (IsByVal) {
4394  assert(A->getType()->isPointerTy());
4395  Type *RealTy = A->getType()->getPointerElementType();
4396  uint64_t ArgSize = DL.getTypeAllocSize(RealTy);
4397  uint64_t ArgAlign = CS.getParamAlignment(ArgNo);
4398  if (ArgAlign < 8)
4399  ArgAlign = 8;
4400  VAArgOffset = alignTo(VAArgOffset, ArgAlign);
4401  if (!IsFixed) {
4402  Value *Base = getShadowPtrForVAArgument(
4403  RealTy, IRB, VAArgOffset - VAArgBase, ArgSize);
4404  if (Base) {
4405  Value *AShadowPtr, *AOriginPtr;
4406  std::tie(AShadowPtr, AOriginPtr) =
4407  MSV.getShadowOriginPtr(A, IRB, IRB.getInt8Ty(),
4408  kShadowTLSAlignment, /*isStore*/ false);
4409 
4410  IRB.CreateMemCpy(Base, kShadowTLSAlignment, AShadowPtr,
4411  kShadowTLSAlignment, ArgSize);
4412  }
4413  }
4414  VAArgOffset += alignTo(ArgSize, 8);
4415  } else {
4416  Value *Base;
4417  uint64_t ArgSize = DL.getTypeAllocSize(A->getType());
4418  uint64_t ArgAlign = 8;
4419  if (A->getType()->isArrayTy()) {
4420  // Arrays are aligned to element size, except for long double
4421  // arrays, which are aligned to 8 bytes.
4422  Type *ElementTy = A->getType()->getArrayElementType();
4423  if (!ElementTy->isPPC_FP128Ty())
4424  ArgAlign = DL.getTypeAllocSize(ElementTy);
4425  } else if (A->getType()->isVectorTy()) {
4426  // Vectors are naturally aligned.
4427  ArgAlign = DL.getTypeAllocSize(A->getType());
4428  }
4429  if (ArgAlign < 8)
4430  ArgAlign = 8;
4431  VAArgOffset = alignTo(VAArgOffset, ArgAlign);
4432  if (DL.isBigEndian()) {
4433  // Adjusting the shadow for argument with size < 8 to match the placement
4434  // of bits in big endian system
4435  if (ArgSize < 8)
4436  VAArgOffset += (8 - ArgSize);
4437  }
4438  if (!IsFixed) {
4439  Base = getShadowPtrForVAArgument(A->getType(), IRB,
4440  VAArgOffset - VAArgBase, ArgSize);
4441  if (Base)
4442  IRB.CreateAlignedStore(MSV.getShadow(A), Base, kShadowTLSAlignment);
4443  }
4444  VAArgOffset += ArgSize;
4445  VAArgOffset = alignTo(VAArgOffset, 8);
4446  }
4447  if (IsFixed)
4448  VAArgBase = VAArgOffset;
4449  }
4450 
4451  Constant *TotalVAArgSize = ConstantInt::get(IRB.getInt64Ty(),
4452  VAArgOffset - VAArgBase);
4453  // Here using VAArgOverflowSizeTLS as VAArgSizeTLS to avoid creation of
4454  // a new class member i.e. it is the total size of all VarArgs.
4455  IRB.CreateStore(TotalVAArgSize, MS.VAArgOverflowSizeTLS);
4456  }
4457 
4458  /// Compute the shadow address for a given va_arg.
4459  Value *getShadowPtrForVAArgument(Type *Ty, IRBuilder<> &IRB,
4460  unsigned ArgOffset, unsigned ArgSize) {
4461  // Make sure we don't overflow __msan_va_arg_tls.
4462  if (ArgOffset + ArgSize > kParamTLSSize)
4463  return nullptr;
4464  Value *Base = IRB.CreatePointerCast(MS.VAArgTLS, MS.IntptrTy);
4465  Base = IRB.CreateAdd(Base, ConstantInt::get(MS.IntptrTy, ArgOffset));
4466  return IRB.CreateIntToPtr(Base, PointerType::get(MSV.getShadowTy(Ty), 0),
4467  "_msarg");
4468  }
4469 
4470  void visitVAStartInst(VAStartInst &I) override {
4471  IRBuilder<> IRB(&I);
4472  VAStartInstrumentationList.push_back(&I);
4473  Value *VAListTag = I.getArgOperand(0);
4474  Value *ShadowPtr, *OriginPtr;
4475  unsigned Alignment = 8;
4476  std::tie(ShadowPtr, OriginPtr) = MSV.getShadowOriginPtr(
4477  VAListTag, IRB, IRB.getInt8Ty(), Alignment, /*isStore*/ true);
4478  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
4479  /* size */ 8, Alignment, false);
4480  }
4481 
4482  void visitVACopyInst(VACopyInst &I) override {
4483  IRBuilder<> IRB(&I);
4484  Value *VAListTag = I.getArgOperand(0);
4485  Value *ShadowPtr, *OriginPtr;
4486  unsigned Alignment = 8;
4487  std::tie(ShadowPtr, OriginPtr) = MSV.getShadowOriginPtr(
4488  VAListTag, IRB, IRB.getInt8Ty(), Alignment, /*isStore*/ true);
4489  // Unpoison the whole __va_list_tag.
4490  // FIXME: magic ABI constants.
4491  IRB.CreateMemSet(ShadowPtr, Constant::getNullValue(IRB.getInt8Ty()),
4492  /* size */ 8, Alignment, false);
4493  }
4494 
4495  void finalizeInstrumentation() override {
4496  assert(!VAArgSize && !VAArgTLSCopy &&
4497  "finalizeInstrumentation called twice");
4498  IRBuilder<> IRB(MSV.ActualFnStart->getFirstNonPHI());
4499  VAArgSize = IRB.CreateLoad(IRB.getInt64Ty(), MS.VAArgOverflowSizeTLS);
4500  Value *CopySize = IRB.CreateAdd(ConstantInt::get(MS.IntptrTy, 0),
4501  VAArgSize);
4502 
4503  if (!VAStartInstrumentationList.empty()) {
4504  // If there is a va_start in this function, make a backup copy of
4505  // va_arg_tls somewhere in the function entry block.
4506  VAArgTLSCopy = IRB.CreateAlloca(Type::getInt8Ty(*MS.C), CopySize);
4507  IRB.CreateMemCpy(VAArgTLSCopy, 8, MS.VAArgTLS, 8, CopySize);
4508  }
4509 
4510  // Instrument va_start.
4511  // Copy va_list shadow from the backup copy of the TLS contents.
4512  for (size_t i = 0, n = VAStartInstrumentationList.size(); i < n; i++) {
4513  CallInst *OrigInst = VAStartInstrumentationList[i];
4514  IRBuilder<> IRB(OrigInst->getNextNode());
4515  Value *VAListTag = OrigInst->getArgOperand(0);
4516  Type *RegSaveAreaPtrTy = Type::getInt64PtrTy(*MS.C);
4517  Value *RegSaveAreaPtrPtr =
4518  IRB.CreateIntToPtr(IRB.CreatePtrToInt(VAListTag, MS.IntptrTy),
4519  PointerType::get(RegSaveAreaPtrTy, 0));
4520  Value *RegSaveAreaPtr =
4521  IRB.CreateLoad(RegSaveAreaPtrTy, RegSaveAreaPtrPtr);
4522  Value *RegSaveAreaShadowPtr, *RegSaveAreaOriginPtr;
4523  unsigned Alignment = 8;
4524  std::tie(RegSaveAreaShadowPtr, RegSaveAreaOriginPtr) =
4525  MSV.getShadowOriginPtr(RegSaveAreaPtr, IRB, IRB.getInt8Ty(),
4526  Alignment, /*isStore*/ true);
4527  IRB.CreateMemCpy(RegSaveAreaShadowPtr, Alignment, VAArgTLSCopy, Alignment,
4528  CopySize);
4529  }
4530  }
4531 };
4532 
4533 /// A no-op implementation of VarArgHelper.
4534 struct VarArgNoOpHelper : public VarArgHelper {
4535  VarArgNoOpHelper(Function &F, MemorySanitizer &MS,
4536  MemorySanitizerVisitor &MSV) {}
4537 
4538  void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {}
4539 
4540  void visitVAStartInst(VAStartInst &I) override {}
4541 
4542  void visitVACopyInst(VACopyInst &I) override {}
4543 
4544  void finalizeInstrumentation() override {}
4545 };
4546 
4547 } // end anonymous namespace
4548 
4549 static VarArgHelper *CreateVarArgHelper(Function &Func, MemorySanitizer &Msan,
4550  MemorySanitizerVisitor &Visitor) {
4551  // VarArg handling is only implemented on AMD64. False positives are possible
4552  // on other platforms.
4553  Triple TargetTriple(Func.getParent()->getTargetTriple());
4554  if (TargetTriple.getArch() == Triple::x86_64)
4555  return new VarArgAMD64Helper(Func, Msan, Visitor);
4556  else if (TargetTriple.isMIPS64())
4557  return new VarArgMIPS64Helper(Func, Msan, Visitor);
4558  else if (TargetTriple.getArch() == Triple::aarch64)
4559  return new VarArgAArch64Helper(Func, Msan, Visitor);
4560  else if (TargetTriple.getArch() == Triple::ppc64 ||
4561  TargetTriple.getArch() == Triple::ppc64le)
4562  return new VarArgPowerPC64Helper(Func, Msan, Visitor);
4563  else
4564  return new VarArgNoOpHelper(Func, Msan, Visitor);
4565 }
4566 
4567 bool MemorySanitizer::sanitizeFunction(Function &F, TargetLibraryInfo &TLI) {
4568  if (!CompileKernel && (&F == MsanCtorFunction))
4569  return false;
4570  MemorySanitizerVisitor Visitor(F, *this, TLI);
4571 
4572  // Clear out readonly/readnone attributes.
4573  AttrBuilder B;
4574  B.addAttribute(Attribute::ReadOnly)
4575  .addAttribute(Attribute::ReadNone);
4577 
4578  return Visitor.runOnFunction();
4579 }
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:2557
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:1528
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:776
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:379
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:2228
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:323
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)
bool isMustTailCall() const
static Type * getX86_MMXTy(LLVMContext &C)
Definition: Type.cpp:170
Use * op_iterator
Definition: User.h:224
bool isPPC_FP128Ty() const
Return true if this is powerpc long double.
Definition: Type.h:158
bool erase(const T &V)
Definition: SmallSet.h:207
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:1218
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:879
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:416
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:885
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...
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.
static cl::opt< uint64_t > ClOriginBase("msan-origin-base", cl::desc("Define custom MSan OriginBase"), cl::Hidden, cl::init(0))
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:1022
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:645
an instruction for type-safe pointer arithmetic to access elements of arrays and structs ...
Definition: Instructions.h:873
static bool runOnFunction(Function &F, bool PostInlining)
initializer< Ty > init(const Ty &Val)
Definition: CommandLine.h:432
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:1257
static Constant * getOrInsertGlobal(Module &M, StringRef Name, Type *Ty)
LLVM Basic Block Representation.
Definition: BasicBlock.h:57
static cl::opt< uint64_t > ClShadowBase("msan-shadow-base", cl::desc("Define custom MSan ShadowBase"), cl::Hidden, cl::init(0))
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:1333
Resume the propagation of an exception.
SmallSet - This maintains a set of unique values, optimizing for the case when the set is small (less...
Definition: SmallSet.h:134
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
static cl::opt< uint64_t > ClXorMask("msan-xor-mask", cl::desc("Define custom MSan XorMask"), cl::Hidden, cl::init(0))
FunctionType * getFunctionType() const
Definition: InstrTypes.h:1121
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
AllocaInst * findAllocaForValue(Value *V, DenseMap< Value *, AllocaInst *> &AllocaForValue)
Finds alloca where the value comes from.
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:709
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
std::pair< NoneType, bool > insert(const T &V)
insert - Insert an element into the set if it isn&#39;t already there.
Definition: SmallSet.h:180
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:2079
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:2121
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:736
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:2092
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:738
CallInst * CreateMaskedStore(Value *Val, Value *Ptr, unsigned Align, Value *Mask)
Create a call to Masked Store intrinsic.
Definition: IRBuilder.cpp:489
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:63
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:468
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:739
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
static cl::opt< uint64_t > ClAndMask("msan-and-mask", cl::desc("Define custom MSan AndMask"), cl::Hidden, cl::init(0))
uint64_t getTypeSizeInBits(Type *Ty) const
Size examples:
Definition: DataLayout.h:601
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:469
void removeAttributes(unsigned i, const AttrBuilder &Attrs)
removes the attributes from the list of attributes.
Definition: Function.cpp:419
Predicate getPredicate() const
Return the predicate for this instruction.
Definition: InstrTypes.h:784
unsigned getNumArgOperands() const
Definition: InstrTypes.h:1216
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:1264
bool onlyReadsMemory(unsigned OpNo) const
Definition: InstrTypes.h:1534
#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
static cl::opt< bool > ClHandleLifetimeIntrinsics("msan-handle-lifetime-intrinsics", cl::desc("when possible, poison scoped variables at the beginning of the scope " "(slower, but more precise)"), cl::Hidden, cl::init(true))
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:332
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:2856
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
The C convention as implemented on Windows/x86-64 and AArch64.
Definition: CallingConv.h:153
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:824
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:2129
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:694
signed greater or equal
Definition: InstrTypes.h:737