[IR] Introduce captures attribute

This introduces the `captures` attribute as described in:
https://discourse.llvm.org/t/rfc-improvements-to-capture-tracking/81420

This initial patch only introduces the IR/bitcode support for
the attribute and its in-memory representation as `CaptureInfo`.
This will be followed by a patch to remove (and upgrade) the
`nocapture` attribute, and then by actual inference/analysis
support.

Based on the RFC feedback, I've used a syntax similar to the
`memory` attribute, though the only "location" that can be
specified right now is `ret`.

I've added some pretty extensive documentation to LangRef on the
semantics. One non-obvious bit here is that using ptrtoint will
not result in a "return-only" capture, even if the ptrtoint
result is only used in the return value. Without this requirement
we wouldn't be able to continue ordinary capture analysis on the
return value.

Author: nikic

llvm/docs/LangRef.rst

@@ -1397,6 +1397,36 @@ Currently, only the following parameter attributes are defined:
     function, returning a pointer to allocated storage disjoint from the
     storage for any other object accessible to the caller.
 
+``captures(...)``
+    This attributes restrict the ways in which the callee may capture the
+    pointer. This is not a valid attribute for return values. This attribute
+    applies only to the particular copy of the pointer passed in this argument.
+
+    The arguments of ``captures`` is a list of captured pointer components,
+    which may be ``none``, or a combination of:
+
+    - ``address``: The integral address of the pointer.
+    - ``provenance``: The ability to access the pointer for both read and write
+      after the function returns.
+    - ``read_provenance``: The ability to access the pointer only for reads
+      after the function returns.
+
+    Additionally, it is possible to specify that the pointer is captured via
+    the return value only, by using ``caputres(ret: ...)``.
+
+    The `pointer capture section <pointercapture>` discusses these semantics
+    in more detail.
+
+    Some examples of how to use the attribute:
+
+    - ``captures(none)``: Pointer not captured.
+    - ``captures(address, provenance)``: Equivalent to omitting the attribute.
+    - ``captures(address)``: Address may be captured, but not provenance.
+    - ``captures(address, read_provenance)``: Both address and provenance
+      captured, but only for read-only access.
+    - ``captures(ret: address, provenance)``: Pointer captured through return
+      value only.
+
 .. _nocapture:
 
 ``nocapture``
@@ -3339,10 +3369,91 @@ Pointer Capture
 ---------------
 
 Given a function call and a pointer that is passed as an argument or stored in
-the memory before the call, a pointer is *captured* by the call if it makes a
-copy of any part of the pointer that outlives the call.
-To be precise, a pointer is captured if one or more of the following conditions
-hold:
+memory before the call, the call may capture two components of the pointer:
+
+  * The address of the pointer, which is its integral value. This also includes
+    parts of the address or any information about the address, including the
+    fact that it does not equal one specific value.
+  * The provenance of the pointer, which is the ability to perform memory
+    accesses through the pointer, in the sense of the :ref:`pointer aliasing
+    rules <pointeraliasing>`. We further distinguish whether only read acceses
+    are allowed, or both reads and writes.
+
+For example, the following function captures the address of ``%a``, because
+it is compared to a pointer, leaking information about the identitiy of the
+pointer:
+
+.. code-block:: llvm
+
+    @glb = global i8 0
+
+    define i1 @f(ptr %a) {
+      %c = icmp eq ptr %a, @glb
+      ret i1 %c
+    }
+
+The function does not capture the provenance of the pointer, because the
+``icmp`` instruction only operates on the pointer address. The following
+function captures both the address and provenance of the pointer, as both
+may be read from ``@glb`` after the function returns:
+
+.. code-block:: llvm
+
+    @glb = global ptr null
+
+    define void @f(ptr %a) {
+      store ptr %a, ptr @glb
+      ret void
+    }
+
+The following function captures *neither* the address nor the provenance of
+the pointer:
+
+.. code-block:: llvm
+
+    define i32 @f(ptr %a) {
+      %v = load i32, ptr %a
+      ret i32
+    }
+
+While address capture includes uses of the address within the body of the
+function, provenance capture refers exclusively to the ability to perform
+accesses *after* the function returns. Memory accesses within the function
+itself are not considered pointer captures.
+
+We can further say that the capture only occurs through a specific location.
+In the following example, the pointer (both address and provenance) is captured
+through the return value only:
+
+.. code-block:: llvm
+
+    define ptr @f(ptr %a) {
+      %gep = getelementptr i8, ptr %a, i64 4
+      ret ptr %gep
+    }
+
+However, we always consider direct inspection of the pointer address
+(e.g. using ``ptrtoint``) to be location-independent. The following example
+is *not* considered a return-only capture, even though the ``ptrtoint``
+ultimately only contribues to the return value:
+
+.. code-block:: llvm
+
+    @lookup = constant [4 x i8] [i8 0, i8 1, i8 2, i8 3]
+
+    define ptr @f(ptr %a) {
+      %a.addr = ptrtoint ptr %a to i64
+      %mask = and i64 %a.addr, 3
+      %gep = getelementptr i8, ptr @lookup, i64 %mask
+      ret ptr %gep
+    }
+
+This definition is chosen to allow capture analysis to continue with the return
+value in the usual fashion.
+
+The following describes possible ways to capture a pointer in more detail,
+where unqualified uses of the word "capture" refer to capturing both address
+and provenance.
 
 1. The call stores any bit of the pointer carrying information into a place,
    and the stored bits can be read from the place by the caller after this call
@@ -3381,30 +3492,30 @@ hold:
     @lock = global i1 true
 
     define void @f(ptr %a) {
-      store ptr %a, ptr* @glb
+      store ptr %a, ptr @glb
       store atomic i1 false, ptr @lock release ; %a is captured because another thread can safely read @glb
       store ptr null, ptr @glb
       ret void
     }
 
-3. The call's behavior depends on any bit of the pointer carrying information.
+3. The call's behavior depends on any bit of the pointer carrying information
+   (address capture only).
 
 .. code-block:: llvm
 
     @glb = global i8 0
 
     define void @f(ptr %a) {
       %c = icmp eq ptr %a, @glb
-      br i1 %c, label %BB_EXIT, label %BB_CONTINUE ; escapes %a
+      br i1 %c, label %BB_EXIT, label %BB_CONTINUE ; captures address of %a only
     BB_EXIT:
       call void @exit()
       unreachable
     BB_CONTINUE:
       ret void
     }
 
-4. The pointer is used in a volatile access as its address.
-
+4. The pointer is used as the pointer operand of a volatile access.
 
 .. _volatile:
 

llvm/include/llvm/AsmParser/LLParser.h

@@ -379,6 +379,7 @@ namespace llvm {
                                     bool inAttrGrp, LocTy &BuiltinLoc);
     bool parseRangeAttr(AttrBuilder &B);
     bool parseInitializesAttr(AttrBuilder &B);
+    bool parseCapturesAttr(AttrBuilder &B);
     bool parseRequiredTypeAttr(AttrBuilder &B, lltok::Kind AttrToken,
                                Attribute::AttrKind AttrKind);
 

llvm/include/llvm/AsmParser/LLToken.h

@@ -207,6 +207,11 @@ enum Kind {
   kw_inaccessiblememonly,
   kw_inaccessiblemem_or_argmemonly,
 
+  // Captures attribute:
+  kw_address,
+  kw_provenance,
+  kw_read_provenance,
+
   // nofpclass attribute:
   kw_all,
   kw_nan,

llvm/include/llvm/Bitcode/LLVMBitCodes.h

@@ -788,6 +788,7 @@ enum AttributeKindCodes {
   ATTR_KIND_NO_EXT = 99,
   ATTR_KIND_NO_DIVERGENCE_SOURCE = 100,
   ATTR_KIND_SANITIZE_TYPE = 101,
+  ATTR_KIND_CAPTURES = 102,
 };
 
 enum ComdatSelectionKindCodes {

llvm/include/llvm/IR/Attributes.h

@@ -284,6 +284,9 @@ class Attribute {
   /// Returns memory effects.
   MemoryEffects getMemoryEffects() const;
 
+  /// Returns information from captures attribute.
+  CaptureInfo getCaptureInfo() const;
+
   /// Return the FPClassTest for nofpclass
   FPClassTest getNoFPClass() const;
 
@@ -436,6 +439,7 @@ class AttributeSet {
   UWTableKind getUWTableKind() const;
   AllocFnKind getAllocKind() const;
   MemoryEffects getMemoryEffects() const;
+  CaptureInfo getCaptureInfo() const;
   FPClassTest getNoFPClass() const;
   std::string getAsString(bool InAttrGrp = false) const;
 
@@ -1260,6 +1264,9 @@ class AttrBuilder {
   /// Add memory effect attribute.
   AttrBuilder &addMemoryAttr(MemoryEffects ME);
 
+  /// Add captures attribute.
+  AttrBuilder &addCapturesAttr(CaptureInfo CI);
+
   // Add nofpclass attribute
   AttrBuilder &addNoFPClassAttr(FPClassTest NoFPClassMask);
 

llvm/include/llvm/IR/Attributes.td

@@ -183,6 +183,9 @@ def NoCallback : EnumAttr<"nocallback", IntersectAnd, [FnAttr]>;
 /// Function creates no aliases of pointer.
 def NoCapture : EnumAttr<"nocapture", IntersectAnd, [ParamAttr]>;
 
+/// Specify how the pointer may be captured.
+def Captures : IntAttr<"captures", IntersectCustom, [ParamAttr]>;
+
 /// Function is not a source of divergence.
 def NoDivergenceSource : EnumAttr<"nodivergencesource", IntersectAnd, [FnAttr]>;
 

llvm/include/llvm/Support/ModRef.h

@@ -273,6 +273,93 @@ raw_ostream &operator<<(raw_ostream &OS, MemoryEffects RMRB);
 // Legacy alias.
 using FunctionModRefBehavior = MemoryEffects;
 
+/// Components of the pointer that may be captured.
+enum class CaptureComponents : uint8_t {
+  None = 0,
+  Address = (1 << 0),
+  ReadProvenance = (1 << 1),
+  Provenance = (1 << 2) | ReadProvenance,
+  All = Address | Provenance,
+  LLVM_MARK_AS_BITMASK_ENUM(Provenance),
+};
+
+inline bool capturesNothing(CaptureComponents CC) {
+  return CC == CaptureComponents::None;
+}
+
+inline bool capturesAnything(CaptureComponents CC) {
+  return CC != CaptureComponents::None;
+}
+
+inline bool capturesAddress(CaptureComponents CC) {
+  return (CC & CaptureComponents::Address) != CaptureComponents::None;
+}
+
+inline bool capturesReadProvenanceOnly(CaptureComponents CC) {
+  return (CC & CaptureComponents::Provenance) ==
+         CaptureComponents::ReadProvenance;
+}
+
+inline bool capturesFullProvenance(CaptureComponents CC) {
+  return (CC & CaptureComponents::Provenance) == CaptureComponents::Provenance;
+}
+
+raw_ostream &operator<<(raw_ostream &OS, CaptureComponents CC);
+
+/// Represents which components of the pointer may be captured and whether
+/// the capture is via the return value only. This represents the captures(...)
+/// attribute in IR.
+///
+/// For more information on the precise semantics see LangRef.
+class CaptureInfo {
+  CaptureComponents Components;
+  bool ReturnOnly;
+
+public:
+  CaptureInfo(CaptureComponents Components, bool ReturnOnly = false)
+      : Components(Components),
+        ReturnOnly(capturesAnything(Components) && ReturnOnly) {}
+
+  /// Create CaptureInfo that may capture all components of the pointer.
+  static CaptureInfo all() { return CaptureInfo(CaptureComponents::All); }
+
+  /// Get the potentially captured components of the pointer.
+  operator CaptureComponents() const { return Components; }
+
+  /// Whether the pointer is captured through the return value only.
+  bool isReturnOnly() const { return ReturnOnly; }
+
+  bool operator==(CaptureInfo Other) const {
+    return Components == Other.Components && ReturnOnly == Other.ReturnOnly;
+  }
+
+  bool operator!=(CaptureInfo Other) const { return !(*this == Other); }
+
+  /// Compute union of CaptureInfos.
+  CaptureInfo operator|(CaptureInfo Other) const {
+    return CaptureInfo(Components | Other.Components,
+                       ReturnOnly && Other.ReturnOnly);
+  }
+
+  /// Compute intersection of CaptureInfos.
+  CaptureInfo operator&(CaptureInfo Other) const {
+    return CaptureInfo(Components & Other.Components,
+                       ReturnOnly || Other.ReturnOnly);
+  }
+
+  static CaptureInfo createFromIntValue(uint32_t Data) {
+    return CaptureInfo(CaptureComponents(Data >> 1), Data & 1);
+  }
+
+  /// Convert CaptureInfo into an encoded integer value (used by captures
+  /// attribute).
+  uint32_t toIntValue() const {
+    return (uint32_t(Components) << 1) | ReturnOnly;
+  }
+};
+
+raw_ostream &operator<<(raw_ostream &OS, CaptureInfo Info);
+
 } // namespace llvm
 
 #endif

llvm/lib/AsmParser/LLLexer.cpp

@@ -704,6 +704,9 @@ lltok::Kind LLLexer::LexIdentifier() {
   KEYWORD(argmemonly);
   KEYWORD(inaccessiblememonly);
   KEYWORD(inaccessiblemem_or_argmemonly);
+  KEYWORD(address);
+  KEYWORD(provenance);
+  KEYWORD(read_provenance);
 
   // nofpclass attribute
   KEYWORD(all);