[LLVM-C] Move and Clean Up The Instrumentation Pass Bindings

[CodaFi]

Summary

Moves the bindings to ASAN, MSAN, TSAN, and DFSAN out of the Go bindings and into the C bindings.

Additionally, add bindings to Code Coverage Sanitizer.

The notable addition to the API is the options wrappers. Memory Sanitizer, and Code Coverage Sanitizer offer structures that encapsulate their configuration options. As these structures are subject to change in future releases of LLVM, the bindings have corresponding opaque wrapper structs that have explicit allocation/destruction routines. This way, we can support future added options by adding bindings that act as setters. We support future removed options by marking their corresponding parameter dead in the documentation/deprecating the setter wrapper.

[whitequark]

llvm/include/llvm-c/Instrumentation.h

46

I think enormous parameter lists are a very error-prone way to do this kind of binding. What about making a C structure reflecting the currently recognized options, and passing the length of this structure to the LLVMCreate*Options functions? This is similar to what WinAPI is doing.

I.e.

struct SanitizerCoverageOptions {
  unsigned CoverageType;
  LLVMBool IndirectCalls;
  // ...
}

LLVMSanitizerCoverageOptions LLVMCreateSanitizerCoverageOptions(
    SanitizerCoverageOptions *Options,
    unsigned OptionsLength
);

So, we could deprecate options by making them dead (and hopefully there's some attribute we can use to cause a warning to be emitted on access), and add new ones by appending them towards the end of the structure. For C/C++ users of the C API as well as any autogenerated bindings, the upgrades are now completely transparent. For non-autogenerated bindings, it's still less work to add new options.

@echristo @deadalnix What do you think?

[CodaFi]

Looks like ESan is dead https://reviews.llvm.org/rL355862. I'll update this patch.

[CodaFi]

Remove support for Efficiency Sanitizer

llvm/bindings/go/llvm/InstrumentationBindings.h

-//===- InstrumentationBindings.h - instrumentation bindings -----*- C++ -*-===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-//
-// This file defines C bindings for the Transforms/Instrumentation component.
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef LLVM_BINDINGS_GO_LLVM_INSTRUMENTATIONBINDINGS_H
-#define LLVM_BINDINGS_GO_LLVM_INSTRUMENTATIONBINDINGS_H
-
-#include "llvm-c/Core.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-// FIXME: These bindings shouldn't be Go-specific and should eventually move to
-// a (somewhat) less stable collection of C APIs for use in creating bindings of
-// LLVM in other languages.
-
-void LLVMAddAddressSanitizerFunctionPass(LLVMPassManagerRef PM);
-void LLVMAddAddressSanitizerModulePass(LLVMPassManagerRef PM);
-void LLVMAddThreadSanitizerPass(LLVMPassManagerRef PM);
-void LLVMAddMemorySanitizerLegacyPassPass(LLVMPassManagerRef PM);
-void LLVMAddDataFlowSanitizerPass(LLVMPassManagerRef PM, int ABIListFilesNum,
-                                  const char **ABIListFiles);
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif

llvm/bindings/go/llvm/InstrumentationBindings.cpp

-//===- InstrumentationBindings.cpp - instrumentation bindings -------------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-//
-// This file defines C bindings for the instrumentation component.
-//
-//===----------------------------------------------------------------------===//
-
-#include "InstrumentationBindings.h"
-#include "llvm-c/Core.h"
-#include "llvm/IR/LegacyPassManager.h"
-#include "llvm/IR/Module.h"
-#include "llvm/Transforms/Instrumentation.h"
-#include "llvm/Transforms/Instrumentation/AddressSanitizer.h"
-#include "llvm/Transforms/Instrumentation/MemorySanitizer.h"
-#include "llvm/Transforms/Instrumentation/ThreadSanitizer.h"
-
-using namespace llvm;
-
-void LLVMAddAddressSanitizerFunctionPass(LLVMPassManagerRef PM) {
-  unwrap(PM)->add(createAddressSanitizerFunctionPass());
-}
-
-void LLVMAddAddressSanitizerModulePass(LLVMPassManagerRef PM) {
-  unwrap(PM)->add(createModuleAddressSanitizerLegacyPassPass());
-}
-
-void LLVMAddThreadSanitizerPass(LLVMPassManagerRef PM) {
-  unwrap(PM)->add(createThreadSanitizerLegacyPassPass());
-}
-
-void LLVMAddMemorySanitizerLegacyPassPass(LLVMPassManagerRef PM) {
-  unwrap(PM)->add(createMemorySanitizerLegacyPassPass());
-}
-
-void LLVMAddDataFlowSanitizerPass(LLVMPassManagerRef PM,
-                                  int ABIListFilesNum,
-                                  const char **ABIListFiles) {
-  std::vector<std::string> ABIListFilesVec;
-  for (int i = 0; i != ABIListFilesNum; ++i) {
-    ABIListFilesVec.push_back(ABIListFiles[i]);
-  }
-  unwrap(PM)->add(createDataFlowSanitizerPass(ABIListFilesVec));
-}

llvm/bindings/go/llvm/transforms_instrumentation.go

 //===- transforms_instrumentation.go - Bindings for instrumentation -------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // This file defines bindings for the instrumentation component.
 //
 //===----------------------------------------------------------------------===//
 
 package llvm
 
 /*
-#include "InstrumentationBindings.h"
+#include "IRBindings.h"
+#include "llvm-c/Instrumentation.h"
 #include <stdlib.h>
 */
 import "C"
 import "unsafe"
 
 func (pm PassManager) AddAddressSanitizerFunctionPass() {
-  C.LLVMAddAddressSanitizerFunctionPass(pm.C)
+  C.LLVMAddAddressSanitizerFunctionPass(pm.C, C.LLVMBool(0), 
+                                          C.LLVMBool(0), C.LLVMBool(0))
 }
 
 func (pm PassManager) AddAddressSanitizerModulePass() {
-  C.LLVMAddAddressSanitizerModulePass(pm.C)
+  C.LLVMAddAddressSanitizerModulePass(pm.C, C.LLVMBool(0), 
+                                        C.LLVMBool(0), C.LLVMBool(0))
 }
 
 func (pm PassManager) AddThreadSanitizerPass() {
   C.LLVMAddThreadSanitizerPass(pm.C)
 }
 
 func (pm PassManager) AddMemorySanitizerLegacyPassPass() {
-  C.LLVMAddMemorySanitizerLegacyPassPass(pm.C)
+  C.LLVMAddMemorySanitizerPass(pm.C, nil)
 }
 
 func (pm PassManager) AddDataFlowSanitizerPass(abilist []string) {
   abiliststrs := make([]*C.char, len(abilist))
   for i, arg := range abilist {
     abiliststrs[i] = C.CString(arg)
     defer C.free(unsafe.Pointer(abiliststrs[i]))
   }
-  C.LLVMAddDataFlowSanitizerPass(pm.C, C.int(len(abilist)), &abiliststrs[0])
+  C.LLVMAddDataFlowSanitizerPass(pm.C, &abiliststrs[0], C.size_t(len(abilist)), nil, nil)
 }

llvm/include/llvm-c/Instrumentation.h

+/*===-- llvm-c/Instrumentation.h ---- Instrumentation Pass Bindings - C++ -*-=*\
+|*                                                                            *|
+|* Part of the LLVM Project, under the Apache License v2.0 with LLVM          *|
+|* Exceptions.                                                                *|
+|* See https://llvm.org/LICENSE.txt for license information.                  *|
+|* SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception                    *|
+|*                                                                            *|
+|*===----------------------------------------------------------------------===*|
+|*                                                                            *|
+|* This header declares the C interface to the instrumentation passes defined
+|* in libLLVMInstrumentation.                                                 *|
+|*                                                                            *|
+\*===----------------------------------------------------------------------===*/
+
+#ifndef LLVM_C_INSTRUMENTATION_H
+#define LLVM_C_INSTRUMENTATION_H
+
+#include "llvm-c/DataTypes.h"
+#include "llvm-c/Types.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * Creates and initializes options suitable for configuring the code coverage
+ * instrumentation passes. The caller is responsible for freeing this by calling
+ * \c LLVMDisposeSanitizerCoverageOptions.
+ *
+ * @see llvm::SanitizerCoverageOptions::SanitizerCoverageOptions()
+ */
+LLVMSanitizerCoverageOptions LLVMCreateSanitizerCoverageOptions(
+    unsigned CoverageType,
+    LLVMBool IndirectCalls,
+    LLVMBool TraceBB,
+    LLVMBool TraceCmp,
+    LLVMBool TraceDiv,
+    LLVMBool TraceGep,
+    LLVMBool Use8bitCounters,
+    LLVMBool TracePC,
+    LLVMBool TracePCGuard,
+    LLVMBool Inline8bitCounters,
+    LLVMBool PCTable,
+    LLVMBool NoPrune,
+    LLVMBool StackDepth);
+
+/**
+ * Destroys code coverage instrumentation options.
+ */
+void LLVMDisposeSanitizerCoverageOptions(LLVMSanitizerCoverageOptions Opts);
+
+/* @see llvm::createSanitizerCoverageModulePass */
+void LLVMAddSanitizerCoverageModulePass(LLVMPassManagerRef PM,
+                                        LLVMSanitizerCoverageOptions Options);
+
+/* @see llvm::createAddressSanitizerFunctionPass */
+void LLVMAddAddressSanitizerFunctionPass(LLVMPassManagerRef PM,
+                                         LLVMBool CompileKernel,
+                                         LLVMBool Recover,
+                                         LLVMBool UseAfterScope);
+
+/* @see llvm::createModuleAddressSanitizerLegacyPassPass */
+void LLVMAddAddressSanitizerModulePass(LLVMPassManagerRef PM,
+                                       LLVMBool CompileKernel,
+                                       LLVMBool Recover,
+                                       LLVMBool UseGlobalsGC);
+
+/* @see llvm::createHWAddressSanitizerPass */
+void LLVMAddHWAddressSanitizerPasses(LLVMPassManagerRef PM,
+                                     LLVMBool CompileKernel,
+                                     LLVMBool Recover);
+
+/**
+ * Creates and initializes options suitable for configuring the memory sanitizer
+ * instrumentation passes. The caller is responsible for freeing this by calling
+ * \c LLVMDisposeMemorySanitizerOptions.
+ *
+ * @see llvm::MemorySanitizerOptions::MemorySanitizerOptions()
+ */
+LLVMMemorySanitizerOptions LLVMCreateMemorySanitizerOptions(int TrackOrigins,
+                                                            LLVMBool Recover,
+                                                            LLVMBool Kernel);
+
+/**
+ * Destroys memory sanitizer instrumentation options.
+ */
+void LLVMDisposeMemorySanitizerOptions(LLVMMemorySanitizerOptions Opts);
+
+/* llvm::createMemorySanitizerLegacyPassPass */
+void LLVMAddMemorySanitizerPass(LLVMPassManagerRef PM,
+                                LLVMMemorySanitizerOptions Options);
+
+
+/* @see llvm::createThreadSanitizerLegacyPassPass */
+void LLVMAddThreadSanitizerPass(LLVMPassManagerRef PM);
+
+
+/* @see llvm::createDataFlowSanitizerPass */
+void LLVMAddDataFlowSanitizerPass(LLVMPassManagerRef PM,
+                                  char **ABIListFiles,
+                                  size_t ABIListFilesCount,
+                                  void *(*getArgTLS)(),
+                                  void *(*getRetValTLS)());
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
+

llvm/include/llvm-c/Types.h

 typedef struct LLVMOpaqueModuleFlagEntry LLVMModuleFlagEntry;
 
 /**
  * @see llvm::JITEventListener
  */
 typedef struct LLVMOpaqueJITEventListener *LLVMJITEventListenerRef;
 
 /**
+ * @see llvm::SanitizerCoverageOptions
+ */
+typedef struct LLVMOpaqueSanitizerCoverageOptions *LLVMSanitizerCoverageOptions;
+
+/**
+ * @see llvm::MemorySanitizerOptions
+ */
+typedef struct LLVMOpaqueMemorySanitizerOptions *LLVMMemorySanitizerOptions;
+
+/**
  * @}
  */
 
 #ifdef __cplusplus
 }
 #endif
 
 #endif

llvm/lib/Transforms/Instrumentation/Instrumentation.cpp

 //===-- Instrumentation.cpp - TransformUtils Infrastructure ---------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // This file defines the common initialization infrastructure for the
 // Instrumentation library.
 //
 //===----------------------------------------------------------------------===//
 
 #include "llvm/Transforms/Instrumentation.h"
 #include "llvm-c/Initialization.h"
+#include "llvm-c/Transforms/Utils.h"
 #include "llvm/ADT/Triple.h"
 #include "llvm/IR/IntrinsicInst.h"
+#include "llvm/IR/LegacyPassManager.h"
 #include "llvm/IR/Module.h"
 #include "llvm/InitializePasses.h"
 #include "llvm/PassRegistry.h"
+#include "llvm/Support/CBindingWrapping.h"
+#include "llvm/Transforms/Instrumentation.h"
+#include "llvm/Transforms/Instrumentation/AddressSanitizer.h"
+#include "llvm/Transforms/Instrumentation/MemorySanitizer.h"
+#include "llvm/Transforms/Instrumentation/ThreadSanitizer.h"
 
 using namespace llvm;
 
 /// Moves I before IP. Returns new insert point.
-static BasicBlock::iterator moveBeforeInsertPoint(BasicBlock::iterator I, BasicBlock::iterator IP) {
+static BasicBlock::iterator moveBeforeInsertPoint(BasicBlock::iterator I,
+                                                  BasicBlock::iterator IP) {
   // If I is IP, move the insert point down.
   if (I == IP)
     return ++IP;
   // Otherwise, move I before IP and return IP.
   I->moveBefore(&*IP);
   return IP;
 }
 
void llvm::initializeInstrumentation(PassRegistry &Registry) {
   initializeDataFlowSanitizerPass(Registry);
 }
 
 /// LLVMInitializeInstrumentation - C binding for
 /// initializeInstrumentation.
 void LLVMInitializeInstrumentation(LLVMPassRegistryRef R) {
   initializeInstrumentation(*unwrap(R));
 }
+
+struct LLVMOpaqueSanitizerCoverageOptions {
+  llvm::SanitizerCoverageOptions Opts;
+};
+
+LLVMSanitizerCoverageOptions LLVMCreateSanitizerCoverageOptions(
+    unsigned CoverageType,
+    LLVMBool IndirectCalls,
+    LLVMBool TraceBB,
+    LLVMBool TraceCmp,
+    LLVMBool TraceDiv,
+    LLVMBool TraceGep,
+    LLVMBool Use8bitCounters,
+    LLVMBool TracePC,
+    LLVMBool TracePCGuard,
+    LLVMBool Inline8bitCounters,
+    LLVMBool PCTable,
+    LLVMBool NoPrune,
+    LLVMBool StackDepth) {
+  SanitizerCoverageOptions Opts;
+  Opts.CoverageType =
+    static_cast<SanitizerCoverageOptions::Type>(CoverageType);
+  Opts.IndirectCalls = IndirectCalls;
+  Opts.TraceBB = TraceBB;
+  Opts.TraceCmp = TraceCmp;
+  Opts.TraceDiv = TraceDiv;
+  Opts.TraceGep = TraceGep;
+  Opts.Use8bitCounters = Use8bitCounters;
+  Opts.TracePC = TracePC;
+  Opts.TracePCGuard = TracePCGuard;
+  Opts.NoPrune = NoPrune;
+  Opts.Inline8bitCounters = Inline8bitCounters;
+  Opts.PCTable = PCTable;
+  Opts.StackDepth = StackDepth;
+  LLVMOpaqueSanitizerCoverageOptions *Result =
+    static_cast<LLVMOpaqueSanitizerCoverageOptions *>(
+      safe_malloc(sizeof(LLVMOpaqueSanitizerCoverageOptions)));
+  Result->Opts = Opts;
+  return Result;
+}
+
+void LLVMDisposeSanitizerCoverageOptions(LLVMSanitizerCoverageOptions Opts) {
+  free(Opts);
+}
+
+void LLVMAddSanitizerCoverageModulePass(LLVMPassManagerRef PM,
+                                        LLVMSanitizerCoverageOptions Options) {
+  SanitizerCoverageOptions Opts;
+  if (Options) {
+    Opts = Options->Opts;
+  }
+  unwrap(PM)->add(createSanitizerCoverageModulePass(Opts));
+}
+
+void LLVMAddAddressSanitizerFunctionPass(LLVMPassManagerRef PM,
+                                         LLVMBool CompileKernel,
+                                         LLVMBool Recover,
+                                         LLVMBool UseAfterScope) {
+  unwrap(PM)->add(createAddressSanitizerFunctionPass(CompileKernel,
+                                                     Recover,
+                                                     UseAfterScope));
+}
+
+void LLVMAddAddressSanitizerModulePass(LLVMPassManagerRef PM,
+                                       LLVMBool CompileKernel,
+                                       LLVMBool Recover,
+                                       LLVMBool UseGlobalsGC) {
+  unwrap(PM)->add(createModuleAddressSanitizerLegacyPassPass(CompileKernel,
+                                                             Recover,
+                                                             UseGlobalsGC));
+}
+
+void LLVMAddHWAddressSanitizerPasses(LLVMPassManagerRef PM,
+                                     LLVMBool CompileKernel,
+                                     LLVMBool Recover) {
+  unwrap(PM)->add(createHWAddressSanitizerPass(CompileKernel, Recover));
+}
+
+struct LLVMOpaqueMemorySanitizerOptions {
+  llvm::MemorySanitizerOptions Opts;
+};
+
+LLVMMemorySanitizerOptions LLVMCreateMemorySanitizerOptions(int TrackOrigins,
+                                                            LLVMBool Recover,
+                                                            LLVMBool Kernel) {
+  MemorySanitizerOptions Opts;
+  Opts.TrackOrigins = TrackOrigins;
+  Opts.Recover = Recover;
+  Opts.Kernel = Kernel;
+  LLVMOpaqueMemorySanitizerOptions *Result =
+    static_cast<LLVMOpaqueMemorySanitizerOptions *>(
+      safe_malloc(sizeof(LLVMOpaqueMemorySanitizerOptions)));
+  Result->Opts = Opts;
+  return Result;
+}
+
+void LLVMDisposeMemorySanitizerOptions(LLVMMemorySanitizerOptions Opts) {
+  free(Opts);
+}
+
+void LLVMAddMemorySanitizerPass(LLVMPassManagerRef PM,
+                                LLVMMemorySanitizerOptions Options) {
+  MemorySanitizerOptions Opts;
+  if (Options) {
+    Opts = Options->Opts;
+  }
+  unwrap(PM)->add(createMemorySanitizerLegacyPassPass(Opts));
+}
+
+void LLVMAddThreadSanitizerPass(LLVMPassManagerRef PM) {
+  unwrap(PM)->add(createThreadSanitizerLegacyPassPass());
+}
+
+void LLVMAddDataFlowSanitizerPass(LLVMPassManagerRef PM,
+                                  char **ABIListFiles,
+                                  size_t ABIListFilesCount,
+                                  void *(*getArgTLS)(),
+                                  void *(*getRetValTLS)()) {
+  std::vector<std::string> ABIListFilesVec;
+  for (size_t i = 0; i != ABIListFilesCount; ++i) {
+    ABIListFilesVec.push_back(ABIListFiles[i]);
+  }
+  unwrap(PM)->add(createDataFlowSanitizerPass(ABIListFilesVec,
+                                              getArgTLS, getRetValTLS));
+}