[clang] [clang][bytecode] Check composite bitcasts for indeterminate bits (PR #118988)
https://github.com/tbaederr closed https://github.com/llvm/llvm-project/pull/118988 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang][bytecode] Check composite bitcasts for indeterminate bits (PR #118988)
https://github.com/tbaederr updated
https://github.com/llvm/llvm-project/pull/118988
>From c45a6b422ff2ad9f60e7e39d335be2e3d3fe4465 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?=
Date: Fri, 6 Dec 2024 15:52:38 +0100
Subject: [PATCH] [clang][bytecode] Check composite bitcasts for indeterminate
bits
---
clang/lib/AST/ByteCode/BitcastBuffer.cpp | 34 ---
clang/lib/AST/ByteCode/BitcastBuffer.h| 4 +++
.../lib/AST/ByteCode/InterpBuiltinBitCast.cpp | 33 ++
.../ByteCode/builtin-bit-cast-bitfields.cpp | 31 -
4 files changed, 82 insertions(+), 20 deletions(-)
diff --git a/clang/lib/AST/ByteCode/BitcastBuffer.cpp
b/clang/lib/AST/ByteCode/BitcastBuffer.cpp
index 7f29c7c2db0147..fbd500fd8f5f4d 100644
--- a/clang/lib/AST/ByteCode/BitcastBuffer.cpp
+++ b/clang/lib/AST/ByteCode/BitcastBuffer.cpp
@@ -62,11 +62,7 @@ BitcastBuffer::copyBits(Bits BitOffset, Bits BitWidth, Bits
FullBitWidth,
}
bool BitcastBuffer::allInitialized() const {
- Bits Sum;
- for (BitRange BR : InitializedBits)
-Sum += BR.size();
-
- return Sum == FinalBitSize;
+ return rangeInitialized(Bits::zero(), FinalBitSize);
}
void BitcastBuffer::markInitialized(Bits Offset, Bits Length) {
@@ -111,6 +107,34 @@ void BitcastBuffer::markInitialized(Bits Offset, Bits
Length) {
#endif
}
+bool BitcastBuffer::rangeInitialized(Bits Offset, Bits Length) const {
+ if (Length.isZero())
+return true;
+
+ BitRange Range(Offset, Offset + Length - Bits(1));
+ Bits Sum;
+ bool FoundStart = false;
+ for (BitRange BR : InitializedBits) {
+if (FoundStart) {
+ if (BR.contains(Range.End)) {
+Sum += (Range.End - BR.Start + Bits(1));
+break;
+ }
+
+ // Else, BR is completely inside Range.
+ Sum += BR.size();
+}
+if (BR.contains(Range.Start)) {
+ Sum += (BR.End - Range.Start + Bits(1));
+ FoundStart = true;
+}
+ }
+
+ // Note that Sum can be larger than Range, e.g. when Range is fully
+ // contained in one range.
+ return Sum >= Range.size();
+}
+
#if 0
template
static std::string hex(T t) {
diff --git a/clang/lib/AST/ByteCode/BitcastBuffer.h
b/clang/lib/AST/ByteCode/BitcastBuffer.h
index 00fbdc9b85421d..2a0d8a0cd9a81f 100644
--- a/clang/lib/AST/ByteCode/BitcastBuffer.h
+++ b/clang/lib/AST/ByteCode/BitcastBuffer.h
@@ -55,6 +55,7 @@ struct Bytes {
Bits toBits() const { return Bits(N * 8); }
};
+/// A bit range. Both Start and End are inclusive.
struct BitRange {
Bits Start;
Bits End;
@@ -62,6 +63,8 @@ struct BitRange {
BitRange(Bits Start, Bits End) : Start(Start), End(End) {}
Bits size() const { return End - Start + Bits(1); }
bool operator<(BitRange Other) const { return Start.N < Other.Start.N; }
+
+ bool contains(Bits B) { return Start <= B && End >= B; }
};
/// Track what bits have been initialized to known values and which ones
@@ -85,6 +88,7 @@ struct BitcastBuffer {
/// Marks the bits in the given range as initialized.
/// FIXME: Can we do this automatically in pushData()?
void markInitialized(Bits Start, Bits Length);
+ bool rangeInitialized(Bits Offset, Bits Length) const;
/// Push \p BitWidth bits at \p BitOffset from \p In into the buffer.
/// \p TargetEndianness is the endianness of the target we're compiling for.
diff --git a/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
b/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
index 4c25a3bb132fcf..e26410c0b35e86 100644
--- a/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
+++ b/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
@@ -248,12 +248,10 @@ static bool readPointerToBuffer(const Context &Ctx, const
Pointer &FromPtr,
if (BitWidth.isZero())
return true;
-if (!P.isInitialized()) {
- assert(false && "Implement uninitialized value tracking");
- return ReturnOnUninit;
-}
+// Bits will be left uninitialized and diagnosed when reading.
+if (!P.isInitialized())
+ return true;
-assert(P.isInitialized());
if (T == PT_Ptr) {
assert(P.getType()->isNullPtrType());
// Clang treats nullptr_t has having NO bits in its value
@@ -262,6 +260,7 @@ static bool readPointerToBuffer(const Context &Ctx, const
Pointer &FromPtr,
return true;
}
+assert(P.isInitialized());
auto Buff =
std::make_unique(ObjectReprChars.getQuantity());
// Work around floating point types that contain unused padding bytes.
@@ -355,10 +354,11 @@ bool clang::interp::DoBitCastPtr(InterpState &S, CodePtr
OpPC,
ToPtr, S.getContext(), Buffer.size(),
[&](const Pointer &P, PrimType T, Bits BitOffset,
bool PackedBools) -> bool {
-CharUnits ObjectReprChars = ASTCtx.getTypeSizeInChars(P.getType());
+QualType PtrType = P.getType();
+CharUnits ObjectReprChars = ASTCtx.getTypeSizeInChars(PtrType);
Bits FullBi
[clang] [clang][bytecode] Check composite bitcasts for indeterminate bits (PR #118988)
https://github.com/tbaederr updated
https://github.com/llvm/llvm-project/pull/118988
>From 7e723fd3ee3844ad83fa61118dd263a84a6145e3 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?=
Date: Fri, 6 Dec 2024 15:52:38 +0100
Subject: [PATCH] [clang][bytecode] Check composite bitcasts for indeterminate
bits
---
clang/lib/AST/ByteCode/BitcastBuffer.cpp | 34 ---
clang/lib/AST/ByteCode/BitcastBuffer.h| 4 +++
clang/lib/AST/ByteCode/Interp.h | 5 +--
.../lib/AST/ByteCode/InterpBuiltinBitCast.cpp | 33 ++
.../ByteCode/builtin-bit-cast-bitfields.cpp | 31 -
5 files changed, 85 insertions(+), 22 deletions(-)
diff --git a/clang/lib/AST/ByteCode/BitcastBuffer.cpp
b/clang/lib/AST/ByteCode/BitcastBuffer.cpp
index 7f29c7c2db0147..fbd500fd8f5f4d 100644
--- a/clang/lib/AST/ByteCode/BitcastBuffer.cpp
+++ b/clang/lib/AST/ByteCode/BitcastBuffer.cpp
@@ -62,11 +62,7 @@ BitcastBuffer::copyBits(Bits BitOffset, Bits BitWidth, Bits
FullBitWidth,
}
bool BitcastBuffer::allInitialized() const {
- Bits Sum;
- for (BitRange BR : InitializedBits)
-Sum += BR.size();
-
- return Sum == FinalBitSize;
+ return rangeInitialized(Bits::zero(), FinalBitSize);
}
void BitcastBuffer::markInitialized(Bits Offset, Bits Length) {
@@ -111,6 +107,34 @@ void BitcastBuffer::markInitialized(Bits Offset, Bits
Length) {
#endif
}
+bool BitcastBuffer::rangeInitialized(Bits Offset, Bits Length) const {
+ if (Length.isZero())
+return true;
+
+ BitRange Range(Offset, Offset + Length - Bits(1));
+ Bits Sum;
+ bool FoundStart = false;
+ for (BitRange BR : InitializedBits) {
+if (FoundStart) {
+ if (BR.contains(Range.End)) {
+Sum += (Range.End - BR.Start + Bits(1));
+break;
+ }
+
+ // Else, BR is completely inside Range.
+ Sum += BR.size();
+}
+if (BR.contains(Range.Start)) {
+ Sum += (BR.End - Range.Start + Bits(1));
+ FoundStart = true;
+}
+ }
+
+ // Note that Sum can be larger than Range, e.g. when Range is fully
+ // contained in one range.
+ return Sum >= Range.size();
+}
+
#if 0
template
static std::string hex(T t) {
diff --git a/clang/lib/AST/ByteCode/BitcastBuffer.h
b/clang/lib/AST/ByteCode/BitcastBuffer.h
index 00fbdc9b85421d..2a0d8a0cd9a81f 100644
--- a/clang/lib/AST/ByteCode/BitcastBuffer.h
+++ b/clang/lib/AST/ByteCode/BitcastBuffer.h
@@ -55,6 +55,7 @@ struct Bytes {
Bits toBits() const { return Bits(N * 8); }
};
+/// A bit range. Both Start and End are inclusive.
struct BitRange {
Bits Start;
Bits End;
@@ -62,6 +63,8 @@ struct BitRange {
BitRange(Bits Start, Bits End) : Start(Start), End(End) {}
Bits size() const { return End - Start + Bits(1); }
bool operator<(BitRange Other) const { return Start.N < Other.Start.N; }
+
+ bool contains(Bits B) { return Start <= B && End >= B; }
};
/// Track what bits have been initialized to known values and which ones
@@ -85,6 +88,7 @@ struct BitcastBuffer {
/// Marks the bits in the given range as initialized.
/// FIXME: Can we do this automatically in pushData()?
void markInitialized(Bits Start, Bits Length);
+ bool rangeInitialized(Bits Offset, Bits Length) const;
/// Push \p BitWidth bits at \p BitOffset from \p In into the buffer.
/// \p TargetEndianness is the endianness of the target we're compiling for.
diff --git a/clang/lib/AST/ByteCode/Interp.h b/clang/lib/AST/ByteCode/Interp.h
index c5c2a5ef19cc4d..01ea32fa3928b8 100644
--- a/clang/lib/AST/ByteCode/Interp.h
+++ b/clang/lib/AST/ByteCode/Interp.h
@@ -3043,8 +3043,9 @@ inline bool BitCast(InterpState &S, CodePtr OpPC, bool
TargetIsUCharOrByte,
uint32_t ResultBitWidth, const llvm::fltSemantics *Sem) {
const Pointer &FromPtr = S.Stk.pop();
- if (!CheckLoad(S, OpPC, FromPtr))
-return false;
+ // Only reject uninitialized global variables.
+ if (FromPtr.isStatic() && !FromPtr.isInitialized())
+return CheckInitialized(S, OpPC, FromPtr, AK_Read);
size_t BuffSize = ResultBitWidth / 8;
llvm::SmallVector Buff(BuffSize);
diff --git a/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
b/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
index 4c25a3bb132fcf..e26410c0b35e86 100644
--- a/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
+++ b/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
@@ -248,12 +248,10 @@ static bool readPointerToBuffer(const Context &Ctx, const
Pointer &FromPtr,
if (BitWidth.isZero())
return true;
-if (!P.isInitialized()) {
- assert(false && "Implement uninitialized value tracking");
- return ReturnOnUninit;
-}
+// Bits will be left uninitialized and diagnosed when reading.
+if (!P.isInitialized())
+ return true;
-assert(P.isInitialized());
if (T == PT_Ptr) {
assert(P.getType()->isNullPtrType());
// Clang treats nullptr_t has having NO bits in its val
[clang] [clang][bytecode] Check composite bitcasts for indeterminate bits (PR #118988)
https://github.com/tbaederr updated
https://github.com/llvm/llvm-project/pull/118988
>From d3e02b47e352447647930cd1ca6142276e39b870 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?=
Date: Fri, 6 Dec 2024 15:52:38 +0100
Subject: [PATCH] [clang][bytecode] Check composite bitcasts for indeterminate
bits
---
clang/lib/AST/ByteCode/BitcastBuffer.cpp | 34 ---
clang/lib/AST/ByteCode/BitcastBuffer.h| 4 +++
.../lib/AST/ByteCode/InterpBuiltinBitCast.cpp | 20 +--
.../ByteCode/builtin-bit-cast-bitfields.cpp | 15
4 files changed, 66 insertions(+), 7 deletions(-)
diff --git a/clang/lib/AST/ByteCode/BitcastBuffer.cpp
b/clang/lib/AST/ByteCode/BitcastBuffer.cpp
index 7f29c7c2db0147..fbd500fd8f5f4d 100644
--- a/clang/lib/AST/ByteCode/BitcastBuffer.cpp
+++ b/clang/lib/AST/ByteCode/BitcastBuffer.cpp
@@ -62,11 +62,7 @@ BitcastBuffer::copyBits(Bits BitOffset, Bits BitWidth, Bits
FullBitWidth,
}
bool BitcastBuffer::allInitialized() const {
- Bits Sum;
- for (BitRange BR : InitializedBits)
-Sum += BR.size();
-
- return Sum == FinalBitSize;
+ return rangeInitialized(Bits::zero(), FinalBitSize);
}
void BitcastBuffer::markInitialized(Bits Offset, Bits Length) {
@@ -111,6 +107,34 @@ void BitcastBuffer::markInitialized(Bits Offset, Bits
Length) {
#endif
}
+bool BitcastBuffer::rangeInitialized(Bits Offset, Bits Length) const {
+ if (Length.isZero())
+return true;
+
+ BitRange Range(Offset, Offset + Length - Bits(1));
+ Bits Sum;
+ bool FoundStart = false;
+ for (BitRange BR : InitializedBits) {
+if (FoundStart) {
+ if (BR.contains(Range.End)) {
+Sum += (Range.End - BR.Start + Bits(1));
+break;
+ }
+
+ // Else, BR is completely inside Range.
+ Sum += BR.size();
+}
+if (BR.contains(Range.Start)) {
+ Sum += (BR.End - Range.Start + Bits(1));
+ FoundStart = true;
+}
+ }
+
+ // Note that Sum can be larger than Range, e.g. when Range is fully
+ // contained in one range.
+ return Sum >= Range.size();
+}
+
#if 0
template
static std::string hex(T t) {
diff --git a/clang/lib/AST/ByteCode/BitcastBuffer.h
b/clang/lib/AST/ByteCode/BitcastBuffer.h
index 00fbdc9b85421d..2a0d8a0cd9a81f 100644
--- a/clang/lib/AST/ByteCode/BitcastBuffer.h
+++ b/clang/lib/AST/ByteCode/BitcastBuffer.h
@@ -55,6 +55,7 @@ struct Bytes {
Bits toBits() const { return Bits(N * 8); }
};
+/// A bit range. Both Start and End are inclusive.
struct BitRange {
Bits Start;
Bits End;
@@ -62,6 +63,8 @@ struct BitRange {
BitRange(Bits Start, Bits End) : Start(Start), End(End) {}
Bits size() const { return End - Start + Bits(1); }
bool operator<(BitRange Other) const { return Start.N < Other.Start.N; }
+
+ bool contains(Bits B) { return Start <= B && End >= B; }
};
/// Track what bits have been initialized to known values and which ones
@@ -85,6 +88,7 @@ struct BitcastBuffer {
/// Marks the bits in the given range as initialized.
/// FIXME: Can we do this automatically in pushData()?
void markInitialized(Bits Start, Bits Length);
+ bool rangeInitialized(Bits Offset, Bits Length) const;
/// Push \p BitWidth bits at \p BitOffset from \p In into the buffer.
/// \p TargetEndianness is the endianness of the target we're compiling for.
diff --git a/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
b/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
index 4c25a3bb132fcf..aa5eab8335d0b7 100644
--- a/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
+++ b/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
@@ -355,10 +355,11 @@ bool clang::interp::DoBitCastPtr(InterpState &S, CodePtr
OpPC,
ToPtr, S.getContext(), Buffer.size(),
[&](const Pointer &P, PrimType T, Bits BitOffset,
bool PackedBools) -> bool {
-CharUnits ObjectReprChars = ASTCtx.getTypeSizeInChars(P.getType());
+QualType PtrType = P.getType();
+CharUnits ObjectReprChars = ASTCtx.getTypeSizeInChars(PtrType);
Bits FullBitWidth = Bits(ASTCtx.toBits(ObjectReprChars));
if (T == PT_Float) {
- const auto &Semantics = ASTCtx.getFloatTypeSemantics(P.getType());
+ const auto &Semantics = ASTCtx.getFloatTypeSemantics(PtrType);
Bits NumBits = Bits(llvm::APFloatBase::getSizeInBits(Semantics));
assert(NumBits.isFullByte());
assert(NumBits.getQuantity() <= FullBitWidth.getQuantity());
@@ -382,6 +383,21 @@ bool clang::interp::DoBitCastPtr(InterpState &S, CodePtr
OpPC,
else
BitWidth = FullBitWidth;
+// If any of the bits are uninitialized, we need to abort unless the
+// target type is std::byte or unsigned char.
+if (!Buffer.rangeInitialized(BitOffset, BitWidth)) {
+ if (!PtrType->isStdByteType() &&
+ !PtrType->isSpecificBuiltinType(BuiltinType::UChar) &&
+ !PtrType->isSpecificBuiltinType(BuiltinType::Char_U)) {
+const Expr *E = S.Cur
[clang] [clang][bytecode] Check composite bitcasts for indeterminate bits (PR #118988)
https://github.com/tbaederr updated
https://github.com/llvm/llvm-project/pull/118988
>From 08ff053907e66539afd497cf790141c285151958 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?=
Date: Fri, 6 Dec 2024 15:52:38 +0100
Subject: [PATCH] [clang][bytecode] Check composite bitcasts for indeterminate
bits
---
clang/lib/AST/ByteCode/BitcastBuffer.cpp | 34 ---
clang/lib/AST/ByteCode/BitcastBuffer.h| 4 +++
.../lib/AST/ByteCode/InterpBuiltinBitCast.cpp | 21 ++--
.../ByteCode/builtin-bit-cast-bitfields.cpp | 14
4 files changed, 66 insertions(+), 7 deletions(-)
diff --git a/clang/lib/AST/ByteCode/BitcastBuffer.cpp
b/clang/lib/AST/ByteCode/BitcastBuffer.cpp
index 7f29c7c2db0147..fbd500fd8f5f4d 100644
--- a/clang/lib/AST/ByteCode/BitcastBuffer.cpp
+++ b/clang/lib/AST/ByteCode/BitcastBuffer.cpp
@@ -62,11 +62,7 @@ BitcastBuffer::copyBits(Bits BitOffset, Bits BitWidth, Bits
FullBitWidth,
}
bool BitcastBuffer::allInitialized() const {
- Bits Sum;
- for (BitRange BR : InitializedBits)
-Sum += BR.size();
-
- return Sum == FinalBitSize;
+ return rangeInitialized(Bits::zero(), FinalBitSize);
}
void BitcastBuffer::markInitialized(Bits Offset, Bits Length) {
@@ -111,6 +107,34 @@ void BitcastBuffer::markInitialized(Bits Offset, Bits
Length) {
#endif
}
+bool BitcastBuffer::rangeInitialized(Bits Offset, Bits Length) const {
+ if (Length.isZero())
+return true;
+
+ BitRange Range(Offset, Offset + Length - Bits(1));
+ Bits Sum;
+ bool FoundStart = false;
+ for (BitRange BR : InitializedBits) {
+if (FoundStart) {
+ if (BR.contains(Range.End)) {
+Sum += (Range.End - BR.Start + Bits(1));
+break;
+ }
+
+ // Else, BR is completely inside Range.
+ Sum += BR.size();
+}
+if (BR.contains(Range.Start)) {
+ Sum += (BR.End - Range.Start + Bits(1));
+ FoundStart = true;
+}
+ }
+
+ // Note that Sum can be larger than Range, e.g. when Range is fully
+ // contained in one range.
+ return Sum >= Range.size();
+}
+
#if 0
template
static std::string hex(T t) {
diff --git a/clang/lib/AST/ByteCode/BitcastBuffer.h
b/clang/lib/AST/ByteCode/BitcastBuffer.h
index 00fbdc9b85421d..2a0d8a0cd9a81f 100644
--- a/clang/lib/AST/ByteCode/BitcastBuffer.h
+++ b/clang/lib/AST/ByteCode/BitcastBuffer.h
@@ -55,6 +55,7 @@ struct Bytes {
Bits toBits() const { return Bits(N * 8); }
};
+/// A bit range. Both Start and End are inclusive.
struct BitRange {
Bits Start;
Bits End;
@@ -62,6 +63,8 @@ struct BitRange {
BitRange(Bits Start, Bits End) : Start(Start), End(End) {}
Bits size() const { return End - Start + Bits(1); }
bool operator<(BitRange Other) const { return Start.N < Other.Start.N; }
+
+ bool contains(Bits B) { return Start <= B && End >= B; }
};
/// Track what bits have been initialized to known values and which ones
@@ -85,6 +88,7 @@ struct BitcastBuffer {
/// Marks the bits in the given range as initialized.
/// FIXME: Can we do this automatically in pushData()?
void markInitialized(Bits Start, Bits Length);
+ bool rangeInitialized(Bits Offset, Bits Length) const;
/// Push \p BitWidth bits at \p BitOffset from \p In into the buffer.
/// \p TargetEndianness is the endianness of the target we're compiling for.
diff --git a/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
b/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
index 4c25a3bb132fcf..dc5f0c6e6ab6ea 100644
--- a/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
+++ b/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
@@ -355,10 +355,11 @@ bool clang::interp::DoBitCastPtr(InterpState &S, CodePtr
OpPC,
ToPtr, S.getContext(), Buffer.size(),
[&](const Pointer &P, PrimType T, Bits BitOffset,
bool PackedBools) -> bool {
-CharUnits ObjectReprChars = ASTCtx.getTypeSizeInChars(P.getType());
+QualType PtrType = P.getType();
+CharUnits ObjectReprChars = ASTCtx.getTypeSizeInChars(PtrType);
Bits FullBitWidth = Bits(ASTCtx.toBits(ObjectReprChars));
if (T == PT_Float) {
- const auto &Semantics = ASTCtx.getFloatTypeSemantics(P.getType());
+ const auto &Semantics = ASTCtx.getFloatTypeSemantics(PtrType);
Bits NumBits = Bits(llvm::APFloatBase::getSizeInBits(Semantics));
assert(NumBits.isFullByte());
assert(NumBits.getQuantity() <= FullBitWidth.getQuantity());
@@ -382,6 +383,22 @@ bool clang::interp::DoBitCastPtr(InterpState &S, CodePtr
OpPC,
else
BitWidth = FullBitWidth;
+// If any of the bits are uninitialized, we need to abort unless the
+// target type is std::byte or unsigned char.
+if (!Buffer.rangeInitialized(BitOffset, BitWidth)) {
+ if (!PtrType->isStdByteType() &&
+ !PtrType->isSpecificBuiltinType(BuiltinType::UChar) &&
+ !PtrType->isSpecificBuiltinType(BuiltinType::Char_U)) {
+const Expr *E = S.Cu
[clang] [clang][bytecode] Check composite bitcasts for indeterminate bits (PR #118988)
llvmbot wrote:
@llvm/pr-subscribers-clang
Author: Timm Baeder (tbaederr)
Changes
---
Full diff: https://github.com/llvm/llvm-project/pull/118988.diff
4 Files Affected:
- (modified) clang/lib/AST/ByteCode/BitcastBuffer.cpp (+29-5)
- (modified) clang/lib/AST/ByteCode/BitcastBuffer.h (+4)
- (modified) clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp (+19-2)
- (modified) clang/test/AST/ByteCode/builtin-bit-cast-bitfields.cpp (+14)
``diff
diff --git a/clang/lib/AST/ByteCode/BitcastBuffer.cpp
b/clang/lib/AST/ByteCode/BitcastBuffer.cpp
index 7f29c7c2db0147..fbd500fd8f5f4d 100644
--- a/clang/lib/AST/ByteCode/BitcastBuffer.cpp
+++ b/clang/lib/AST/ByteCode/BitcastBuffer.cpp
@@ -62,11 +62,7 @@ BitcastBuffer::copyBits(Bits BitOffset, Bits BitWidth, Bits
FullBitWidth,
}
bool BitcastBuffer::allInitialized() const {
- Bits Sum;
- for (BitRange BR : InitializedBits)
-Sum += BR.size();
-
- return Sum == FinalBitSize;
+ return rangeInitialized(Bits::zero(), FinalBitSize);
}
void BitcastBuffer::markInitialized(Bits Offset, Bits Length) {
@@ -111,6 +107,34 @@ void BitcastBuffer::markInitialized(Bits Offset, Bits
Length) {
#endif
}
+bool BitcastBuffer::rangeInitialized(Bits Offset, Bits Length) const {
+ if (Length.isZero())
+return true;
+
+ BitRange Range(Offset, Offset + Length - Bits(1));
+ Bits Sum;
+ bool FoundStart = false;
+ for (BitRange BR : InitializedBits) {
+if (FoundStart) {
+ if (BR.contains(Range.End)) {
+Sum += (Range.End - BR.Start + Bits(1));
+break;
+ }
+
+ // Else, BR is completely inside Range.
+ Sum += BR.size();
+}
+if (BR.contains(Range.Start)) {
+ Sum += (BR.End - Range.Start + Bits(1));
+ FoundStart = true;
+}
+ }
+
+ // Note that Sum can be larger than Range, e.g. when Range is fully
+ // contained in one range.
+ return Sum >= Range.size();
+}
+
#if 0
template
static std::string hex(T t) {
diff --git a/clang/lib/AST/ByteCode/BitcastBuffer.h
b/clang/lib/AST/ByteCode/BitcastBuffer.h
index 00fbdc9b85421d..2a0d8a0cd9a81f 100644
--- a/clang/lib/AST/ByteCode/BitcastBuffer.h
+++ b/clang/lib/AST/ByteCode/BitcastBuffer.h
@@ -55,6 +55,7 @@ struct Bytes {
Bits toBits() const { return Bits(N * 8); }
};
+/// A bit range. Both Start and End are inclusive.
struct BitRange {
Bits Start;
Bits End;
@@ -62,6 +63,8 @@ struct BitRange {
BitRange(Bits Start, Bits End) : Start(Start), End(End) {}
Bits size() const { return End - Start + Bits(1); }
bool operator<(BitRange Other) const { return Start.N < Other.Start.N; }
+
+ bool contains(Bits B) { return Start <= B && End >= B; }
};
/// Track what bits have been initialized to known values and which ones
@@ -85,6 +88,7 @@ struct BitcastBuffer {
/// Marks the bits in the given range as initialized.
/// FIXME: Can we do this automatically in pushData()?
void markInitialized(Bits Start, Bits Length);
+ bool rangeInitialized(Bits Offset, Bits Length) const;
/// Push \p BitWidth bits at \p BitOffset from \p In into the buffer.
/// \p TargetEndianness is the endianness of the target we're compiling for.
diff --git a/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
b/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
index 4c25a3bb132fcf..dc5f0c6e6ab6ea 100644
--- a/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
+++ b/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
@@ -355,10 +355,11 @@ bool clang::interp::DoBitCastPtr(InterpState &S, CodePtr
OpPC,
ToPtr, S.getContext(), Buffer.size(),
[&](const Pointer &P, PrimType T, Bits BitOffset,
bool PackedBools) -> bool {
-CharUnits ObjectReprChars = ASTCtx.getTypeSizeInChars(P.getType());
+QualType PtrType = P.getType();
+CharUnits ObjectReprChars = ASTCtx.getTypeSizeInChars(PtrType);
Bits FullBitWidth = Bits(ASTCtx.toBits(ObjectReprChars));
if (T == PT_Float) {
- const auto &Semantics = ASTCtx.getFloatTypeSemantics(P.getType());
+ const auto &Semantics = ASTCtx.getFloatTypeSemantics(PtrType);
Bits NumBits = Bits(llvm::APFloatBase::getSizeInBits(Semantics));
assert(NumBits.isFullByte());
assert(NumBits.getQuantity() <= FullBitWidth.getQuantity());
@@ -382,6 +383,22 @@ bool clang::interp::DoBitCastPtr(InterpState &S, CodePtr
OpPC,
else
BitWidth = FullBitWidth;
+// If any of the bits are uninitialized, we need to abort unless the
+// target type is std::byte or unsigned char.
+if (!Buffer.rangeInitialized(BitOffset, BitWidth)) {
+ if (!PtrType->isStdByteType() &&
+ !PtrType->isSpecificBuiltinType(BuiltinType::UChar) &&
+ !PtrType->isSpecificBuiltinType(BuiltinType::Char_U)) {
+const Expr *E = S.Current->getExpr(OpPC);
+QualType ExprType = E->getType();
+S.FFDiag(E, diag::note_constexpr_bit_cast_indet_dest)
+<< ExprType
[clang] [clang][bytecode] Check composite bitcasts for indeterminate bits (PR #118988)
https://github.com/tbaederr created
https://github.com/llvm/llvm-project/pull/118988
None
>From 0086686210e22c7e47444f2e46a262683d9c4594 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?=
Date: Fri, 6 Dec 2024 15:52:38 +0100
Subject: [PATCH] [clang][bytecode] Check composite bitcasts for indeterminate
bits
---
clang/lib/AST/ByteCode/BitcastBuffer.cpp | 34 ---
clang/lib/AST/ByteCode/BitcastBuffer.h| 4 +++
.../lib/AST/ByteCode/InterpBuiltinBitCast.cpp | 21 ++--
.../ByteCode/builtin-bit-cast-bitfields.cpp | 14
4 files changed, 66 insertions(+), 7 deletions(-)
diff --git a/clang/lib/AST/ByteCode/BitcastBuffer.cpp
b/clang/lib/AST/ByteCode/BitcastBuffer.cpp
index 7f29c7c2db0147..fbd500fd8f5f4d 100644
--- a/clang/lib/AST/ByteCode/BitcastBuffer.cpp
+++ b/clang/lib/AST/ByteCode/BitcastBuffer.cpp
@@ -62,11 +62,7 @@ BitcastBuffer::copyBits(Bits BitOffset, Bits BitWidth, Bits
FullBitWidth,
}
bool BitcastBuffer::allInitialized() const {
- Bits Sum;
- for (BitRange BR : InitializedBits)
-Sum += BR.size();
-
- return Sum == FinalBitSize;
+ return rangeInitialized(Bits::zero(), FinalBitSize);
}
void BitcastBuffer::markInitialized(Bits Offset, Bits Length) {
@@ -111,6 +107,34 @@ void BitcastBuffer::markInitialized(Bits Offset, Bits
Length) {
#endif
}
+bool BitcastBuffer::rangeInitialized(Bits Offset, Bits Length) const {
+ if (Length.isZero())
+return true;
+
+ BitRange Range(Offset, Offset + Length - Bits(1));
+ Bits Sum;
+ bool FoundStart = false;
+ for (BitRange BR : InitializedBits) {
+if (FoundStart) {
+ if (BR.contains(Range.End)) {
+Sum += (Range.End - BR.Start + Bits(1));
+break;
+ }
+
+ // Else, BR is completely inside Range.
+ Sum += BR.size();
+}
+if (BR.contains(Range.Start)) {
+ Sum += (BR.End - Range.Start + Bits(1));
+ FoundStart = true;
+}
+ }
+
+ // Note that Sum can be larger than Range, e.g. when Range is fully
+ // contained in one range.
+ return Sum >= Range.size();
+}
+
#if 0
template
static std::string hex(T t) {
diff --git a/clang/lib/AST/ByteCode/BitcastBuffer.h
b/clang/lib/AST/ByteCode/BitcastBuffer.h
index 00fbdc9b85421d..2a0d8a0cd9a81f 100644
--- a/clang/lib/AST/ByteCode/BitcastBuffer.h
+++ b/clang/lib/AST/ByteCode/BitcastBuffer.h
@@ -55,6 +55,7 @@ struct Bytes {
Bits toBits() const { return Bits(N * 8); }
};
+/// A bit range. Both Start and End are inclusive.
struct BitRange {
Bits Start;
Bits End;
@@ -62,6 +63,8 @@ struct BitRange {
BitRange(Bits Start, Bits End) : Start(Start), End(End) {}
Bits size() const { return End - Start + Bits(1); }
bool operator<(BitRange Other) const { return Start.N < Other.Start.N; }
+
+ bool contains(Bits B) { return Start <= B && End >= B; }
};
/// Track what bits have been initialized to known values and which ones
@@ -85,6 +88,7 @@ struct BitcastBuffer {
/// Marks the bits in the given range as initialized.
/// FIXME: Can we do this automatically in pushData()?
void markInitialized(Bits Start, Bits Length);
+ bool rangeInitialized(Bits Offset, Bits Length) const;
/// Push \p BitWidth bits at \p BitOffset from \p In into the buffer.
/// \p TargetEndianness is the endianness of the target we're compiling for.
diff --git a/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
b/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
index 4c25a3bb132fcf..dc5f0c6e6ab6ea 100644
--- a/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
+++ b/clang/lib/AST/ByteCode/InterpBuiltinBitCast.cpp
@@ -355,10 +355,11 @@ bool clang::interp::DoBitCastPtr(InterpState &S, CodePtr
OpPC,
ToPtr, S.getContext(), Buffer.size(),
[&](const Pointer &P, PrimType T, Bits BitOffset,
bool PackedBools) -> bool {
-CharUnits ObjectReprChars = ASTCtx.getTypeSizeInChars(P.getType());
+QualType PtrType = P.getType();
+CharUnits ObjectReprChars = ASTCtx.getTypeSizeInChars(PtrType);
Bits FullBitWidth = Bits(ASTCtx.toBits(ObjectReprChars));
if (T == PT_Float) {
- const auto &Semantics = ASTCtx.getFloatTypeSemantics(P.getType());
+ const auto &Semantics = ASTCtx.getFloatTypeSemantics(PtrType);
Bits NumBits = Bits(llvm::APFloatBase::getSizeInBits(Semantics));
assert(NumBits.isFullByte());
assert(NumBits.getQuantity() <= FullBitWidth.getQuantity());
@@ -382,6 +383,22 @@ bool clang::interp::DoBitCastPtr(InterpState &S, CodePtr
OpPC,
else
BitWidth = FullBitWidth;
+// If any of the bits are uninitialized, we need to abort unless the
+// target type is std::byte or unsigned char.
+if (!Buffer.rangeInitialized(BitOffset, BitWidth)) {
+ if (!PtrType->isStdByteType() &&
+ !PtrType->isSpecificBuiltinType(BuiltinType::UChar) &&
+ !PtrType->isSpecificBuiltinType(BuiltinType::Char_U)) {
+const Expr *E
