update libcxxabi to llvm 16

lib/libcxxabi/src/aix_state_tab_eh.inc

@@ -14,6 +14,10 @@
 #include <stdio.h>
 #include <sys/debug.h>
 
+#if !__has_cpp_attribute(clang::optnone)
+#error This file requires clang::optnone attribute support
+#endif
+
 /*
   The legacy IBM xlC and xlclang++ compilers use the state table for EH
   instead of the range table. Destructors, or addresses of the possible catch
@@ -183,10 +187,6 @@ enum FSMMagic : uint32_t {
   number3 = 0x1cedbeef  // State table generated by xlclang++ compiler.
 };
 
-constexpr uint32_t REG_EXCP_OBJ = 14; // Register to pass the address of the exception
-                                      // object from the personality to xlclang++
-                                      // compiled code.
-
 constexpr size_t dtorArgument = 0x02; // Flag to destructor indicating to free
                                       // virtual bases, don't delete object.
 
@@ -555,8 +555,16 @@ __xlcxx_personality_v0(int version, _Unwind_Action actions, uint64_t exceptionCl
   if (actions & _UA_CLEANUP_PHASE) {
     // Phase 2 cleanup:
     if (results.reason == _URC_HANDLER_FOUND) {
+      // Store the address of unwind_exception in the stack field
+      // reserved for compilers (SP + 3 * sizeof(uintptr_t)) in the stack of
+      // the caller of the function containing the landing pad (within the link
+      // area for the call to the latter) for __xlc_exception_handle()
+      // to retrieve when it is called by the landing pad.
+      uintptr_t *currentSP = reinterpret_cast<uintptr_t*>(_Unwind_GetGR(context, 1));
+      uintptr_t *callersSP = reinterpret_cast<uintptr_t*>(currentSP[0]);
+      callersSP[3] = reinterpret_cast<uintptr_t>(unwind_exception);
+      _LIBCXXABI_TRACE_STATETAB("Handshake: set unwind_exception=%p in stack=%p\n", reinterpret_cast<void*>(unwind_exception), reinterpret_cast<void*>(callersSP));
       // Jump to the handler.
-      _Unwind_SetGR(context, REG_EXCP_OBJ, reinterpret_cast<uintptr_t>(unwind_exception));
       _Unwind_SetIP(context, results.landingPad);
       return _URC_INSTALL_CONTEXT;
     }
@@ -633,12 +641,38 @@ _LIBCXXABI_FUNC_VIS void __xlc_throw_badexception() {
   __cxa_throw(newexception, const_cast<std::type_info*>(&typeid(std::bad_exception)), 0);
 }
 
+// force_a_stackframe
+// This function is called by __xlc_exception_handle() to ensure a stack frame
+// is created for __xlc_exception_handle().
+__attribute__((noinline, optnone))
+static void force_a_stackframe() {}
+
 // __xlc_exception_handle
 // This function is for xlclang++. It returns the address of the exception
-// object set in gpr14 by the personality routine for xlclang++ compiled code.
+// object stored in the reserved field in the stack of the caller of the
+// function that calls __xlc_exception_handle() (within the link area for the
+// call to the latter). The address is stored by the personality routine for
+// xlclang++ compiled code. The implementation of __xlc_exception_handle()
+// assumes a stack frame is created for it. The following ensures this
+// assumption holds true: 1) a call to force_a_stackframe() is made inside
+// __xlc_exception_handle() to make it non-leaf; and 2) optimizations are
+// disabled for this function with attribute 'optnone'. Note: this function
+// may not work as expected if these are changed.
+__attribute__((optnone))
 _LIBCXXABI_FUNC_VIS uintptr_t __xlc_exception_handle() {
-  uintptr_t exceptionObject;
-  asm("mr %0, 14" : "=r"(exceptionObject));
+  // Make a call to force_a_stackframe() so that the compiler creates a stack
+  // frame for this function.
+  force_a_stackframe();
+
+  // Get the SP of this function, i.e., __xlc_exception_handle().
+  uintptr_t *lastStack;
+  asm("mr %0, 1" : "=r"(lastStack));
+  // Get the SP of the caller of __xlc_exception_handle().
+  uintptr_t *callerStack = reinterpret_cast<uintptr_t*>(lastStack[0]);
+  // Get the SP of the caller of the caller.
+  uintptr_t *callerStack2 = reinterpret_cast<uintptr_t*>(callerStack[0]);
+  uintptr_t exceptionObject = callerStack2[3];
+  _LIBCXXABI_TRACE_STATETAB("Handshake: exceptionObject=%p from stack=%p\n", reinterpret_cast<void*>(exceptionObject), reinterpret_cast<void*>(callerStack2));
   return exceptionObject;
 }
 

lib/libcxxabi/src/cxa_demangle.cpp

@@ -386,15 +386,12 @@ __cxa_demangle(const char *MangledName, char *Buf, size_t *N, int *Status) {
 
   int InternalStatus = demangle_success;
   Demangler Parser(MangledName, MangledName + std::strlen(MangledName));
-  OutputBuffer O;
-
   Node *AST = Parser.parse();
 
   if (AST == nullptr)
     InternalStatus = demangle_invalid_mangled_name;
-  else if (!initializeOutputBuffer(Buf, N, O, 1024))
-    InternalStatus = demangle_memory_alloc_failure;
   else {
+    OutputBuffer O(Buf, N);
     assert(Parser.ForwardTemplateRefs.empty());
     AST->print(O);
     O += '\0';

lib/libcxxabi/src/cxa_guard_impl.h

@@ -5,6 +5,7 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
+
 #ifndef LIBCXXABI_SRC_INCLUDE_CXA_GUARD_IMPL_H
 #define LIBCXXABI_SRC_INCLUDE_CXA_GUARD_IMPL_H
 
@@ -54,9 +55,12 @@
 #  endif
 #endif
 
+#include <__threading_support>
+#include <cstdint>
+#include <cstring>
 #include <limits.h>
 #include <stdlib.h>
-#include <__threading_support>
+
 #ifndef _LIBCXXABI_HAS_NO_THREADS
 #  if defined(__ELF__) && defined(_LIBCXXABI_LINK_PTHREAD_LIB)
 #    pragma comment(lib, "pthread")

lib/libcxxabi/src/demangle/ItaniumDemangle.h

@@ -26,6 +26,7 @@
 #include <cstdlib>
 #include <cstring>
 #include <limits>
+#include <new>
 #include <utility>
 
 DEMANGLE_NAMESPACE_BEGIN
@@ -369,6 +370,10 @@ public:
   VendorExtQualType(const Node *Ty_, StringView Ext_, const Node *TA_)
       : Node(KVendorExtQualType), Ty(Ty_), Ext(Ext_), TA(TA_) {}
 
+  const Node *getTy() const { return Ty; }
+  StringView getExt() const { return Ext; }
+  const Node *getTA() const { return TA; }
+
   template <typename Fn> void match(Fn F) const { F(Ty, Ext, TA); }
 
   void printLeft(OutputBuffer &OB) const override {
@@ -417,6 +422,9 @@ public:
              Child_->ArrayCache, Child_->FunctionCache),
         Quals(Quals_), Child(Child_) {}
 
+  Qualifiers getQuals() const { return Quals; }
+  const Node *getChild() const { return Child; }
+
   template<typename Fn> void match(Fn F) const { F(Child, Quals); }
 
   bool hasRHSComponentSlow(OutputBuffer &OB) const override {
@@ -585,6 +593,8 @@ public:
       : Node(KPointerType, Pointee_->RHSComponentCache),
         Pointee(Pointee_) {}
 
+  const Node *getPointee() const { return Pointee; }
+
   template<typename Fn> void match(Fn F) const { F(Pointee); }
 
   bool hasRHSComponentSlow(OutputBuffer &OB) const override {
@@ -1070,6 +1080,9 @@ public:
   VectorType(const Node *BaseType_, const Node *Dimension_)
       : Node(KVectorType), BaseType(BaseType_), Dimension(Dimension_) {}
 
+  const Node *getBaseType() const { return BaseType; }
+  const Node *getDimension() const { return Dimension; }
+
   template<typename Fn> void match(Fn F) const { F(BaseType, Dimension); }
 
   void printLeft(OutputBuffer &OB) const override {
@@ -3019,14 +3032,21 @@ AbstractManglingParser<Derived, Alloc>::parseOperatorEncoding() {
   if (numLeft() < 2)
     return nullptr;
 
-  auto Op = std::lower_bound(
-      &Ops[0], &Ops[NumOps], First,
-      [](const OperatorInfo &Op_, const char *Enc_) { return Op_ < Enc_; });
-  if (Op == &Ops[NumOps] || *Op != First)
+  // We can't use lower_bound as that can link to symbols in the C++ library,
+  // and this must remain independant of that.
+  size_t lower = 0u, upper = NumOps - 1; // Inclusive bounds.
+  while (upper != lower) {
+    size_t middle = (upper + lower) / 2;
+    if (Ops[middle] < First)
+      lower = middle + 1;
+    else
+      upper = middle;
+  }
+  if (Ops[lower] != First)
     return nullptr;
 
   First += 2;
-  return Op;
+  return &Ops[lower];
 }
 
 //   <operator-name> ::= See parseOperatorEncoding()
@@ -5099,7 +5119,7 @@ template <>
 struct FloatData<long double>
 {
 #if defined(__mips__) && defined(__mips_n64) || defined(__aarch64__) || \
-    defined(__wasm__) || defined(__riscv)
+    defined(__wasm__) || defined(__riscv) || defined(__loongarch__)
     static const size_t mangled_size = 32;
 #elif defined(__arm__) || defined(__mips__) || defined(__hexagon__)
     static const size_t mangled_size = 16;

lib/libcxxabi/src/demangle/Utility.h

@@ -69,7 +69,9 @@ class OutputBuffer {
 
 public:
   OutputBuffer(char *StartBuf, size_t Size)
-      : Buffer(StartBuf), CurrentPosition(0), BufferCapacity(Size) {}
+      : Buffer(StartBuf), BufferCapacity(Size) {}
+  OutputBuffer(char *StartBuf, size_t *SizePtr)
+      : OutputBuffer(StartBuf, StartBuf ? *SizePtr : 0) {}
   OutputBuffer() = default;
   // Non-copyable
   OutputBuffer(const OutputBuffer &) = delete;
@@ -77,12 +79,6 @@ public:
 
   operator StringView() const { return StringView(Buffer, CurrentPosition); }
 
-  void reset(char *Buffer_, size_t BufferCapacity_) {
-    CurrentPosition = 0;
-    Buffer = Buffer_;
-    BufferCapacity = BufferCapacity_;
-  }
-
   /// If a ParameterPackExpansion (or similar type) is encountered, the offset
   /// into the pack that we're currently printing.
   unsigned CurrentPackIndex = std::numeric_limits<unsigned>::max();
@@ -198,21 +194,6 @@ public:
   ScopedOverride &operator=(const ScopedOverride &) = delete;
 };
 
-inline bool initializeOutputBuffer(char *Buf, size_t *N, OutputBuffer &OB,
-                                   size_t InitSize) {
-  size_t BufferSize;
-  if (Buf == nullptr) {
-    Buf = static_cast<char *>(std::malloc(InitSize));
-    if (Buf == nullptr)
-      return false;
-    BufferSize = InitSize;
-  } else
-    BufferSize = *N;
-
-  OB.reset(Buf, BufferSize);
-  return true;
-}
-
 DEMANGLE_NAMESPACE_END
 
 #endif

lib/libcxxabi/src/fallback_malloc.cpp

@@ -15,6 +15,7 @@
 #endif
 #endif
 
+#include <assert.h>
 #include <stdlib.h> // for malloc, calloc, free
 #include <string.h> // for memset
 #include <new> // for std::__libcpp_aligned_{alloc,free}
@@ -63,11 +64,28 @@ char heap[HEAP_SIZE] __attribute__((aligned));
 typedef unsigned short heap_offset;
 typedef unsigned short heap_size;
 
+// On both 64 and 32 bit targets heap_node should have the following properties
+// Size: 4
+// Alignment: 2
 struct heap_node {
   heap_offset next_node; // offset into heap
   heap_size len;         // size in units of "sizeof(heap_node)"
 };
 
+// All pointers returned by fallback_malloc must be at least aligned
+// as RequiredAligned. Note that RequiredAlignment can be greater than
+// alignof(std::max_align_t) on 64 bit systems compiling 32 bit code.
+struct FallbackMaxAlignType {
+} __attribute__((aligned));
+const size_t RequiredAlignment = alignof(FallbackMaxAlignType);
+
+static_assert(alignof(FallbackMaxAlignType) % sizeof(heap_node) == 0,
+              "The required alignment must be evenly divisible by the sizeof(heap_node)");
+
+// The number of heap_node's that can fit in a chunk of memory with the size
+// of the RequiredAlignment. On 64 bit targets NodesPerAlignment should be 4.
+const size_t NodesPerAlignment = alignof(FallbackMaxAlignType) / sizeof(heap_node);
+
 static const heap_node* list_end =
     (heap_node*)(&heap[HEAP_SIZE]); // one past the end of the heap
 static heap_node* freelist = NULL;
@@ -82,10 +100,23 @@ heap_offset offset_from_node(const heap_node* ptr) {
       sizeof(heap_node));
 }
 
+// Return a pointer to the first address, 'A', in `heap` that can actually be
+// used to represent a heap_node. 'A' must be aligned so that
+// '(A + sizeof(heap_node)) % RequiredAlignment == 0'. On 64 bit systems this
+// address should be 12 bytes after the first 16 byte boundary.
+heap_node* getFirstAlignedNodeInHeap() {
+  heap_node* node = (heap_node*)heap;
+  const size_t alignNBytesAfterBoundary = RequiredAlignment - sizeof(heap_node);
+  size_t boundaryOffset = reinterpret_cast<size_t>(node) % RequiredAlignment;
+  size_t requiredOffset = alignNBytesAfterBoundary - boundaryOffset;
+  size_t NElemOffset = requiredOffset / sizeof(heap_node);
+  return node + NElemOffset;
+}
+
 void init_heap() {
-  freelist = (heap_node*)heap;
+  freelist = getFirstAlignedNodeInHeap();
   freelist->next_node = offset_from_node(list_end);
-  freelist->len = HEAP_SIZE / sizeof(heap_node);
+  freelist->len = static_cast<heap_size>(list_end - freelist);
 }
 
 //  How big a chunk we allocate
@@ -109,23 +140,44 @@ void* fallback_malloc(size_t len) {
   for (p = freelist, prev = 0; p && p != list_end;
        prev = p, p = node_from_offset(p->next_node)) {
 
-    if (p->len > nelems) { //  chunk is larger, shorten, and return the tail
-      heap_node* q;
+    // Check the invariant that all heap_nodes pointers 'p' are aligned
+    // so that 'p + 1' has an alignment of at least RequiredAlignment
+    assert(reinterpret_cast<size_t>(p + 1) % RequiredAlignment == 0);
 
-      p->len = static_cast<heap_size>(p->len - nelems);
-      q = p + p->len;
-      q->next_node = 0;
-      q->len = static_cast<heap_size>(nelems);
-      return (void*)(q + 1);
+    // Calculate the number of extra padding elements needed in order
+    // to split 'p' and create a properly aligned heap_node from the tail
+    // of 'p'. We calculate aligned_nelems such that 'p->len - aligned_nelems'
+    // will be a multiple of NodesPerAlignment.
+    size_t aligned_nelems = nelems;
+    if (p->len > nelems) {
+      heap_size remaining_len = static_cast<heap_size>(p->len - nelems);
+      aligned_nelems += remaining_len % NodesPerAlignment;
     }
 
-    if (p->len == nelems) { // exact size match
+    // chunk is larger and we can create a properly aligned heap_node
+    // from the tail. In this case we shorten 'p' and return the tail.
+    if (p->len > aligned_nelems) {
+      heap_node* q;
+      p->len = static_cast<heap_size>(p->len - aligned_nelems);
+      q = p + p->len;
+      q->next_node = 0;
+      q->len = static_cast<heap_size>(aligned_nelems);
+      void* ptr = q + 1;
+      assert(reinterpret_cast<size_t>(ptr) % RequiredAlignment == 0);
+      return ptr;
+    }
+
+    // The chunk is the exact size or the chunk is larger but not large
+    // enough to split due to alignment constraints.
+    if (p->len >= nelems) {
       if (prev == 0)
         freelist = node_from_offset(p->next_node);
       else
         prev->next_node = p->next_node;
       p->next_node = 0;
-      return (void*)(p + 1);
+      void* ptr = p + 1;
+      assert(reinterpret_cast<size_t>(ptr) % RequiredAlignment == 0);
+      return ptr;
     }
   }
   return NULL; // couldn't find a spot big enough