Merged main:5099dc341f7f into amd-gfx:ab27814b3805

Local branch amd-gfx ab27814 Merged main:7539bcf994ed into amd-gfx:0e8798f5c1bd
Remote branch main 5099dc3 [Clang][CodeGen] Fix use of CXXThisValue with StrictVTablePointers (llvm#68169)

clang/include/clang/Driver/Options.td

@@ -4287,7 +4287,7 @@ def mcmodel_EQ : Joined<["-"], "mcmodel=">, Group<m_Group>,
   MarshallingInfoString<TargetOpts<"CodeModel">, [{"default"}]>;
 def mlarge_data_threshold_EQ : Joined<["-"], "mlarge-data-threshold=">, Group<m_Group>,
   Visibility<[ClangOption, CC1Option]>,
-  MarshallingInfoInt<TargetOpts<"LargeDataThreshold">>;
+  MarshallingInfoInt<TargetOpts<"LargeDataThreshold">, "65535">;
 def mtls_size_EQ : Joined<["-"], "mtls-size=">, Group<m_Group>,
   Flags<[NoXarchOption]>, Visibility<[ClangOption, CC1Option]>,
   HelpText<"Specify bit size of immediate TLS offsets (AArch64 ELF only): "

clang/lib/CodeGen/CGClass.cpp

@@ -28,6 +28,7 @@
 #include "clang/CodeGen/CGFunctionInfo.h"
 #include "llvm/IR/Intrinsics.h"
 #include "llvm/IR/Metadata.h"
+#include "llvm/Support/SaveAndRestore.h"
 #include "llvm/Transforms/Utils/SanitizerStats.h"
 #include <optional>
 
@@ -1291,10 +1292,10 @@ void CodeGenFunction::EmitCtorPrologue(const CXXConstructorDecl *CD,
     assert(BaseCtorContinueBB);
   }
 
-  llvm::Value *const OldThis = CXXThisValue;
   for (; B != E && (*B)->isBaseInitializer() && (*B)->isBaseVirtual(); B++) {
     if (!ConstructVBases)
       continue;
+    SaveAndRestore ThisRAII(CXXThisValue);
     if (CGM.getCodeGenOpts().StrictVTablePointers &&
         CGM.getCodeGenOpts().OptimizationLevel > 0 &&
         isInitializerOfDynamicClass(*B))
@@ -1311,16 +1312,14 @@ void CodeGenFunction::EmitCtorPrologue(const CXXConstructorDecl *CD,
   // Then, non-virtual base initializers.
   for (; B != E && (*B)->isBaseInitializer(); B++) {
     assert(!(*B)->isBaseVirtual());
-
+    SaveAndRestore ThisRAII(CXXThisValue);
     if (CGM.getCodeGenOpts().StrictVTablePointers &&
         CGM.getCodeGenOpts().OptimizationLevel > 0 &&
         isInitializerOfDynamicClass(*B))
       CXXThisValue = Builder.CreateLaunderInvariantGroup(LoadCXXThis());
     EmitBaseInitializer(*this, ClassDecl, *B);
   }
 
-  CXXThisValue = OldThis;
-
   InitializeVTablePointers(ClassDecl);
 
   // And finally, initialize class members.

clang/lib/Format/WhitespaceManager.cpp

@@ -354,6 +354,9 @@ AlignTokenSequence(const FormatStyle &Style, unsigned Start, unsigned End,
       }
     }
 
+    if (Shift == 0)
+      continue;
+
     // This is for function parameters that are split across multiple lines,
     // as mentioned in the ScopeStack comment.
     if (InsideNestedScope && CurrentChange.NewlinesBefore > 0) {
@@ -447,7 +450,7 @@ AlignTokenSequence(const FormatStyle &Style, unsigned Start, unsigned End,
       CurrentChange.Spaces += Shift;
 
     // We should not remove required spaces unless we break the line before.
-    assert(Shift >= 0 || Changes[i].NewlinesBefore > 0 ||
+    assert(Shift > 0 || Changes[i].NewlinesBefore > 0 ||
            CurrentChange.Spaces >=
                static_cast<int>(Changes[i].Tok->SpacesRequiredBefore) ||
            CurrentChange.Tok->is(tok::eof));

clang/test/CodeGen/large-data-threshold.c

@@ -5,7 +5,7 @@
 // RUN: %clang_cc1 -triple x86_64-unknown-unknown -S %s -o - -mcmodel=medium -mlarge-data-threshold=200 | FileCheck %s --check-prefix=ASM-SMALL
 // RUN: %clang_cc1 -triple x86_64-unknown-unknown -S %s -o - -mcmodel=medium -mlarge-data-threshold=2 | FileCheck %s --check-prefix=ASM-LARGE
 
-// IR-DEFAULT: !{i32 1, !"Large Data Threshold", i64 0}
+// IR-DEFAULT: !{i32 1, !"Large Data Threshold", i64 65535}
 // IR-CUSTOM: !{i32 1, !"Large Data Threshold", i64 200}
 
 // ASM-SMALL-NOT: movabsq

clang/test/CodeGenCXX/strict-vtable-pointers-GH67937.cpp

@@ -0,0 +1,22 @@
+// RUN: %clang_cc1 %s -I%S -triple=x86_64-pc-windows-msvc -fstrict-vtable-pointers -disable-llvm-passes -O1 -emit-llvm -o %t.ll
+// RUN: FileCheck %s < %t.ll
+
+struct A {
+  virtual ~A();
+};
+struct B : virtual A {};
+class C : B {};
+C foo;
+
+// CHECK-LABEL: define {{.*}} @"??0C@@QEAA@XZ"(ptr {{.*}} %this, i32 {{.*}} %is_most_derived)
+// CHECK: ctor.init_vbases:
+// CHECK-NEXT: %0 = getelementptr inbounds i8, ptr %this1, i64 0
+// CHECK-NEXT: store ptr @"??_8C@@7B@", ptr %0
+// CHECK-NEXT: %1 = call ptr @llvm.launder.invariant.group.p0(ptr %this1)
+// CHECK-NEXT: %2 = getelementptr inbounds i8, ptr %1, i64 8
+// CHECK-NEXT: %call = call noundef ptr @"??0A@@QEAA@XZ"(ptr {{.*}} %2) #2
+// CHECK-NEXT: br label %ctor.skip_vbases
+// CHECK-EMPTY:
+// CHECK-NEXT: ctor.skip_vbases:
+// CHECK-NEXT: %3 = call ptr @llvm.launder.invariant.group.p0(ptr %this1)
+// CHECK-NEXT: %call3 = call noundef ptr @"??0B@@QEAA@XZ"(ptr {{.*}} %3, i32 noundef 0) #2

libc/src/__support/float_to_string.h

@@ -708,7 +708,7 @@ FloatToString<long double>::get_negative_block(int block_index) {
     const int32_t SHIFT_CONST = TABLE_SHIFT_CONST;
 
     // if the requested block is zero
-    if (block_index <= MIN_BLOCK_2[idx]) {
+    if (block_index < MIN_BLOCK_2[idx]) {
       return 0;
     }
     const uint32_t p = POW10_OFFSET_2[idx] + block_index - MIN_BLOCK_2[idx];

libc/startup/linux/x86_64/CMakeLists.txt

@@ -10,6 +10,7 @@ add_startup_object(
     libc.src.__support.threads.thread
     libc.src.__support.OSUtil.osutil
     libc.src.stdlib.exit
+    libc.src.stdlib.abort
     libc.src.stdlib.atexit
     libc.src.string.memory_utils.inline_memcpy
     libc.src.unistd.environ

libc/startup/linux/x86_64/start.cpp

@@ -7,8 +7,10 @@
 //===----------------------------------------------------------------------===//
 
 #include "config/linux/app.h"
+#include "src/__support/OSUtil/io.h"
 #include "src/__support/OSUtil/syscall.h"
 #include "src/__support/threads/thread.h"
+#include "src/stdlib/abort.h"
 #include "src/stdlib/atexit.h"
 #include "src/stdlib/exit.h"
 #include "src/string/memory_utils/inline_memcpy.h"
@@ -23,6 +25,11 @@
 
 extern "C" int main(int, char **, char **);
 
+extern "C" void __stack_chk_fail() {
+  LIBC_NAMESPACE::write_to_stderr("stack smashing detected");
+  LIBC_NAMESPACE::abort();
+}
+
 namespace LIBC_NAMESPACE {
 
 #ifdef SYS_mmap2
@@ -54,7 +61,9 @@ void init_tls(TLSDescriptor &tls_descriptor) {
   // Per the x86_64 TLS ABI, the entry pointed to by the thread pointer is the
   // address of the TLS block. So, we add more size to accomodate this address
   // entry.
-  uintptr_t tlsSizeWithAddr = tlsSize + sizeof(uintptr_t);
+  // We also need to include space for the stack canary. The canary is at
+  // offset 0x28 (40) and is of size uintptr_t.
+  uintptr_t tlsSizeWithAddr = tlsSize + sizeof(uintptr_t) + 40;
 
   // We cannot call the mmap function here as the functions set errno on
   // failure. Since errno is implemented via a thread local variable, we cannot
@@ -76,6 +85,16 @@ void init_tls(TLSDescriptor &tls_descriptor) {
   LIBC_NAMESPACE::inline_memcpy(reinterpret_cast<char *>(tlsAddr),
                                 reinterpret_cast<const char *>(app.tls.address),
                                 app.tls.init_size);
+  uintptr_t *stackGuardAddr = reinterpret_cast<uintptr_t *>(endPtr + 40);
+  // Setting the stack guard to a random value.
+  // We cannot call the get_random function here as the function sets errno on
+  // failure. Since errno is implemented via a thread local variable, we cannot
+  // use errno before TLS is setup.
+  ssize_t stackGuardRetVal = LIBC_NAMESPACE::syscall_impl<ssize_t>(
+      SYS_getrandom, reinterpret_cast<long>(stackGuardAddr), sizeof(uint64_t),
+      0);
+  if (stackGuardRetVal < 0)
+    LIBC_NAMESPACE::syscall_impl(SYS_exit, 1);
 
   tls_descriptor = {tlsSizeWithAddr, uintptr_t(tlsAddr), endPtr};
   return;

libc/test/integration/src/unistd/CMakeLists.txt

@@ -33,6 +33,29 @@ add_integration_test(
     libc.src.unistd.fork
 )
 
+if((${LIBC_TARGET_OS} STREQUAL "linux") AND (${LIBC_TARGET_ARCHITECTURE_IS_X86}))
+  add_integration_test(
+    stack_smashing_test
+    SUITE
+      unistd-integration-tests
+    SRCS
+      stack_smashing_test.cpp
+    DEPENDS
+      libc.include.errno
+      libc.include.signal
+      libc.include.sys_wait
+      libc.include.unistd
+      libc.src.pthread.pthread_atfork
+      libc.src.signal.raise
+      libc.src.sys.wait.wait
+      libc.src.sys.wait.wait4
+      libc.src.sys.wait.waitpid
+      libc.src.unistd.fork
+    COMPILE_OPTIONS
+      -fstack-protector-all
+  )
+endif()
+
 add_executable(
   libc_execv_test_normal_exit
   EXCLUDE_FROM_ALL

libc/test/integration/src/unistd/stack_smashing_test.cpp

@@ -0,0 +1,68 @@
+//===--- Stack smashing test to check stack canary set up  ----------------===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/__support/CPP/string.h"
+#include "src/__support/OSUtil/io.h"
+#include "src/pthread/pthread_atfork.h"
+#include "src/signal/raise.h"
+#include "src/sys/wait/wait.h"
+#include "src/sys/wait/wait4.h"
+#include "src/sys/wait/waitpid.h"
+#include "src/unistd/fork.h"
+
+#include "test/IntegrationTest/test.h"
+
+#include <errno.h>
+#include <signal.h>
+#include <sys/wait.h>
+#include <unistd.h>
+
+void no_stack_smashing_normal_exit() {
+  pid_t pid = LIBC_NAMESPACE::fork();
+  if (pid == 0) {
+    // Child process
+    char foo[30];
+    for (int i = 0; i < 30; i++)
+      foo[i] = (foo[i] != 42) ? 42 : 24;
+    return;
+  }
+  ASSERT_TRUE(pid > 0);
+  int status;
+  pid_t cpid = LIBC_NAMESPACE::wait(&status);
+  ASSERT_TRUE(cpid > 0);
+  ASSERT_EQ(cpid, pid);
+  ASSERT_TRUE(WIFEXITED(status));
+}
+
+void stack_smashing_abort() {
+  pid_t pid = LIBC_NAMESPACE::fork();
+  if (pid == 0) {
+    // Child process
+    char foo[30];
+    char *frame_ptr = static_cast<char *>(__builtin_frame_address(0));
+    char *cur_ptr = &foo[0];
+    // Corrupt the stack
+    while (cur_ptr != frame_ptr) {
+      *cur_ptr = (*cur_ptr != 42) ? 42 : 24;
+      cur_ptr++;
+    }
+    return;
+  }
+  ASSERT_TRUE(pid > 0);
+  int status;
+  pid_t cpid = LIBC_NAMESPACE::wait(&status);
+  ASSERT_TRUE(cpid > 0);
+  ASSERT_EQ(cpid, pid);
+  ASSERT_TRUE(WTERMSIG(status) == SIGABRT);
+}
+
+TEST_MAIN(int argc, char **argv, char **envp) {
+  no_stack_smashing_normal_exit();
+  stack_smashing_abort();
+  return 0;
+}

libc/test/src/stdio/sprintf_test.cpp

@@ -2433,6 +2433,9 @@ TEST_F(LlvmLibcSPrintfTest, FloatAutoConv) {
   written = LIBC_NAMESPACE::sprintf(buff, "%g", 9999999000000.00);
   ASSERT_STREQ_LEN(written, buff, "1e+13");
 
+  written = LIBC_NAMESPACE::sprintf(buff, "%g", 0xa.aaaaaaaaaaaaaabp-7);
+  ASSERT_STREQ_LEN(written, buff, "0.0833333");
+
   // Simple Subnormal Tests.
 
   written = LIBC_NAMESPACE::sprintf(buff, "%g", 0x1.0p-1027);
@@ -2457,9 +2460,16 @@ TEST_F(LlvmLibcSPrintfTest, FloatAutoConv) {
 
   // Length Modifier Tests.
 
+#if defined(SPECIAL_X86_LONG_DOUBLE)
+
   written = LIBC_NAMESPACE::sprintf(buff, "%Lg", 0xf.fffffffffffffffp+16380L);
   ASSERT_STREQ_LEN(written, buff, "1.18973e+4932");
 
+  written = LIBC_NAMESPACE::sprintf(buff, "%Lg", 0xa.aaaaaaaaaaaaaabp-7L);
+  ASSERT_STREQ_LEN(written, buff, "0.0833333");
+
+#endif // SPECIAL_X86_LONG_DOUBLE
+
   // TODO: Uncomment the below tests after long double support is added
   /*
   written = LIBC_NAMESPACE::sprintf(buff, "%Lf", 1e100L);
@@ -2757,6 +2767,15 @@ TEST_F(LlvmLibcSPrintfTest, FloatAutoConv) {
   written = LIBC_NAMESPACE::sprintf(buff, "%.10g", 0x1.0p-1074);
   ASSERT_STREQ_LEN(written, buff, "4.940656458e-324");
 
+#if defined(SPECIAL_X86_LONG_DOUBLE)
+
+  written = LIBC_NAMESPACE::sprintf(buff, "%.60Lg", 0xa.aaaaaaaaaaaaaabp-7L);
+  ASSERT_STREQ_LEN(
+      written, buff,
+      "0.0833333333333333333355920878593448009041821933351457118988037");
+
+#endif // SPECIAL_X86_LONG_DOUBLE
+
   // Long double precision tests.
   // These are currently commented out because they require long double support
   // that isn't ready yet.

libcxx/include/__algorithm/pstl_backend.h

@@ -49,7 +49,11 @@ A PSTL parallel backend is a tag type to which the following functions are assoc
                                    _Comp __comp);
 
   template <class _ExecutionPolicy, class _InIterator, class _OutIterator, class _UnaryOperation>
-  _OutIterator __pstl_transform(_Backend, _InIterator __first, _InIterator __last, _OutIterator __result, _UnaryOperation __op);
+  _OutIterator __pstl_transform(_Backend,
+                                _InIterator __first,
+                                _InIterator __last,
+                                _OutIterator __result,
+                                _UnaryOperation __op);
 
   template <class _ExecutionPolicy, class _InIterator1, class _InIterator2, class _OutIterator, class _BinaryOperation>
   _OutIterator __pstl_transform(_Backend,

lld/ELF/Arch/X86_64.cpp

@@ -7,12 +7,14 @@
 //===----------------------------------------------------------------------===//
 
 #include "OutputSections.h"
+#include "Relocations.h"
 #include "Symbols.h"
 #include "SyntheticSections.h"
 #include "Target.h"
 #include "lld/Common/ErrorHandler.h"
 #include "llvm/BinaryFormat/ELF.h"
 #include "llvm/Support/Endian.h"
+#include "llvm/Support/MathExtras.h"
 
 using namespace llvm;
 using namespace llvm::object;
@@ -47,6 +49,7 @@ class X86_64 : public TargetInfo {
                                         uint8_t stOther) const override;
   bool deleteFallThruJmpInsn(InputSection &is, InputFile *file,
                              InputSection *nextIS) const override;
+  bool relaxOnce(int pass) const override;
 };
 } // namespace
 
@@ -305,6 +308,43 @@ bool X86_64::deleteFallThruJmpInsn(InputSection &is, InputFile *file,
   return true;
 }
 
+bool X86_64::relaxOnce(int pass) const {
+  uint64_t minVA = UINT64_MAX, maxVA = 0;
+  for (OutputSection *osec : outputSections) {
+    minVA = std::min(minVA, osec->addr);
+    maxVA = std::max(maxVA, osec->addr + osec->size);
+  }
+  // If the max VA difference is under 2^31, GOT-generating relocations with a 32-bit range cannot overflow.
+  if (isUInt<31>(maxVA - minVA))
+    return false;
+
+  SmallVector<InputSection *, 0> storage;
+  bool changed = false;
+  for (OutputSection *osec : outputSections) {
+    if (!(osec->flags & SHF_EXECINSTR))
+      continue;
+    for (InputSection *sec : getInputSections(*osec, storage)) {
+      for (Relocation &rel : sec->relocs()) {
+        if (rel.expr != R_RELAX_GOT_PC)
+          continue;
+
+        uint64_t v = sec->getRelocTargetVA(
+            sec->file, rel.type, rel.addend,
+            sec->getOutputSection()->addr + rel.offset, *rel.sym, rel.expr);
+        if (isInt<32>(v))
+          continue;
+        if (rel.sym->auxIdx == 0) {
+          rel.sym->allocateAux();
+          addGotEntry(*rel.sym);
+          changed = true;
+        }
+        rel.expr = R_GOT_PC;
+      }
+    }
+  }
+  return changed;
+}
+
 RelExpr X86_64::getRelExpr(RelType type, const Symbol &s,
                            const uint8_t *loc) const {
   switch (type) {
@@ -912,7 +952,8 @@ static void relaxGotNoPic(uint8_t *loc, uint64_t val, uint8_t op,
 }
 
 static void relaxGot(uint8_t *loc, const Relocation &rel, uint64_t val) {
-  checkInt(loc, val, 32, rel);
+  assert(isInt<32>(val) &&
+         "GOTPCRELX should not have been relaxed if it overflows");
   const uint8_t op = loc[-2];
   const uint8_t modRm = loc[-1];