[XRay][compiler-rt] Allow for building the XRay runtime without PREINIT initialization.

[dberris]

Summary

Define a build-time configuration option for the XRay runtime to
determine whether the archive will add an entry to the .preinit_array
section of the binary. We also allow for initializing the XRay data
structures with an explicit call to __xray_init(). This allows us to
give users the capability to initialize the XRay data structures on
demand.

This can allow us to start porting XRay to platforms where
.preinit_array isn't a supported section. It also allows us to limit
the effects of XRay in the initialization sequence for applications that
are sensitive to this kind of interference (i.e. large binaries) or
those that want to package XRay control in libraries.

Future changes should allow us to build two different library archives
for the XRay runtime, and allow clang users to determine which version
to link.

[kpw]

include/xray/xray_interface.h

109–113 ↗

(On Diff #108894)

Are callers expected to synchronize before calling __xray_init()? Is it even legal to call this function more than once? This should be documented in the header.

lib/xray/xray_init.cc

50–52 ↗

(On Diff #108894)

Is it safe for initializeFlags() to be called twice? Called by different callers at the same time? Depending on what you expect of callers invoking __xray_init, these things can happen.

I suggest that we add an atomic_uint8_t XRayFlagsInitialized above and check that within a XrayFlagInitMutex, then it will be safe for callers to make multiple unsynchronized invocations of __xray_init().

I will also suggest early return if this atomic_load is true. As written, this just wraps the initializeFlags() call. XRayInstrMap is still reloaded even if XRayInitialized is true.

[dberris]

• fixup: synchronize better

include/xray/xray_interface.h

109–113 ↗

(On Diff #108894)

Good point. Updated doc and implementation to allow for multiple calls (and documenting the semantics of that).

lib/xray/xray_init.cc

50–52 ↗

(On Diff #108894)

Good point! Done, PTAL.

compiler-rt/trunk/CMakeLists.txt

 include(CompilerRTUtils)
 
 option(COMPILER_RT_BUILD_BUILTINS "Build builtins" ON)
 mark_as_advanced(COMPILER_RT_BUILD_BUILTINS)
 option(COMPILER_RT_BUILD_SANITIZERS "Build sanitizers" ON)
 mark_as_advanced(COMPILER_RT_BUILD_SANITIZERS)
 option(COMPILER_RT_BUILD_XRAY "Build xray" ON)
 mark_as_advanced(COMPILER_RT_BUILD_XRAY)
+option(COMPILER_RT_BUILD_XRAY_NO_PREINIT "Build xray with no preinit patching" OFF)
+mark_as_advanced(COMPILER_RT_BUILD_XRAY_NO_PREINIT)
 
 set(COMPILER_RT_BAREMETAL_BUILD OFF CACHE BOOLEAN
   "Build for a bare-metal target.")
 
 if (COMPILER_RT_STANDALONE_BUILD)
   load_llvm_config()
 
   # Find Python interpreter.

compiler-rt/trunk/include/xray/xray_interface.h

 /// id. We return 0 if we encounter any error, even if 0 may be a valid function
 /// address.
 extern uintptr_t __xray_function_address(int32_t FuncId);
 
 /// This function returns the maximum valid function id. Returns 0 if we
 /// encounter errors (when there are no instrumented functions, etc.).
 extern size_t __xray_max_function_id();
 
+/// Initialize the required XRay data structures. This is useful in cases where
+/// users want to control precisely when the XRay instrumentation data
+/// structures are initialized, for example when the XRay library is built with
+/// the XRAY_NO_PREINIT preprocessor definition.
+///
+/// Calling __xray_init() more than once is safe across multiple threads.
+extern void __xray_init();
+
 } // end extern "C"
 
 #endif // XRAY_XRAY_INTERFACE_H

compiler-rt/trunk/lib/xray/CMakeLists.txt

 
 include_directories(..)
 include_directories(../../include)
 
 set(XRAY_CFLAGS ${SANITIZER_COMMON_CFLAGS})
 set(XRAY_COMMON_DEFINITIONS XRAY_HAS_EXCEPTIONS=1)
 append_list_if(
   COMPILER_RT_HAS_XRAY_COMPILER_FLAG XRAY_SUPPORTED=1 XRAY_COMMON_DEFINITIONS)
+append_list_if(
+  COMPILER_RT_BUILD_XRAY_NO_PREINIT XRAY_NO_PREINIT XRAY_COMMON_DEFINITIONS)
 
 add_compiler_rt_object_libraries(RTXray
   ARCHS ${XRAY_SUPPORTED_ARCH}
   SOURCES ${XRAY_SOURCES} CFLAGS ${XRAY_CFLAGS}
   DEFS ${XRAY_COMMON_DEFINITIONS})
 
 add_compiler_rt_component(xray)
 

compiler-rt/trunk/lib/xray/xray_init.cc

 // FIXME: Support DSO instrumentation maps too. The current solution only works
 // for statically linked executables.
 __sanitizer::atomic_uint8_t XRayInitialized{0};
 
 // This should always be updated before XRayInitialized is updated.
 __sanitizer::SpinMutex XRayInstrMapMutex;
 XRaySledMap XRayInstrMap;
 
+// Global flag to determine whether the flags have been initialized.
+__sanitizer::atomic_uint8_t XRayFlagsInitialized{0};
+
+// A mutex to allow only one thread to initialize the XRay data structures.
+__sanitizer::SpinMutex XRayInitMutex;
+
 // __xray_init() will do the actual loading of the current process' memory map
 // and then proceed to look for the .xray_instr_map section/segment.
 void __xray_init() XRAY_NEVER_INSTRUMENT {
+  __sanitizer::SpinMutexLock Guard(&XRayInitMutex);
+  // Short-circuit if we've already initialized XRay before.
+  if (__sanitizer::atomic_load(&XRayInitialized,
+                               __sanitizer::memory_order_acquire))
+    return;
+
+  if (!__sanitizer::atomic_load(&XRayFlagsInitialized,
+                                __sanitizer::memory_order_acquire)) {
     initializeFlags();
+    __sanitizer::atomic_store(&XRayFlagsInitialized, true,
+                              __sanitizer::memory_order_release);
+  }
+
   if (__start_xray_instr_map == nullptr) {
     if (Verbosity())
       Report("XRay instrumentation map missing. Not initializing XRay.\n");
     return;
   }
 
   {
     __sanitizer::SpinMutexLock Guard(&XRayInstrMapMutex);
     XRayInstrMap.Sleds = __start_xray_instr_map;
     XRayInstrMap.Entries = __stop_xray_instr_map - __start_xray_instr_map;
     XRayInstrMap.SledsIndex = __start_xray_fn_idx;
     XRayInstrMap.Functions = __stop_xray_fn_idx - __start_xray_fn_idx;
   }
   __sanitizer::atomic_store(&XRayInitialized, true,
                             __sanitizer::memory_order_release);
 
+#ifndef XRAY_NO_PREINIT
   if (flags()->patch_premain)
     __xray_patch();
+#endif
 }
 
+#ifndef XRAY_NO_PREINIT
 __attribute__((section(".preinit_array"),
                used)) void (*__local_xray_preinit)(void) = __xray_init;
+#endif