Giant blob of minor changes
[dotfiles/.git] / .config / coc / extensions / coc-go-data / tools / pkg / mod / golang.org / x / tools@v0.0.0-20201028153306-37f0764111ff / go / analysis / passes / unsafeptr / unsafeptr.go
diff --git a/.config/coc/extensions/coc-go-data/tools/pkg/mod/golang.org/x/tools@v0.0.0-20201028153306-37f0764111ff/go/analysis/passes/unsafeptr/unsafeptr.go b/.config/coc/extensions/coc-go-data/tools/pkg/mod/golang.org/x/tools@v0.0.0-20201028153306-37f0764111ff/go/analysis/passes/unsafeptr/unsafeptr.go
new file mode 100644 (file)
index 0000000..ed86e5e
--- /dev/null
@@ -0,0 +1,168 @@
+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Package unsafeptr defines an Analyzer that checks for invalid
+// conversions of uintptr to unsafe.Pointer.
+package unsafeptr
+
+import (
+       "go/ast"
+       "go/token"
+       "go/types"
+
+       "golang.org/x/tools/go/analysis"
+       "golang.org/x/tools/go/analysis/passes/inspect"
+       "golang.org/x/tools/go/analysis/passes/internal/analysisutil"
+       "golang.org/x/tools/go/ast/inspector"
+)
+
+const Doc = `check for invalid conversions of uintptr to unsafe.Pointer
+
+The unsafeptr analyzer reports likely incorrect uses of unsafe.Pointer
+to convert integers to pointers. A conversion from uintptr to
+unsafe.Pointer is invalid if it implies that there is a uintptr-typed
+word in memory that holds a pointer value, because that word will be
+invisible to stack copying and to the garbage collector.`
+
+var Analyzer = &analysis.Analyzer{
+       Name:     "unsafeptr",
+       Doc:      Doc,
+       Requires: []*analysis.Analyzer{inspect.Analyzer},
+       Run:      run,
+}
+
+func run(pass *analysis.Pass) (interface{}, error) {
+       inspect := pass.ResultOf[inspect.Analyzer].(*inspector.Inspector)
+
+       nodeFilter := []ast.Node{
+               (*ast.CallExpr)(nil),
+               (*ast.StarExpr)(nil),
+               (*ast.UnaryExpr)(nil),
+       }
+       inspect.Preorder(nodeFilter, func(n ast.Node) {
+               switch x := n.(type) {
+               case *ast.CallExpr:
+                       if len(x.Args) == 1 &&
+                               hasBasicType(pass.TypesInfo, x.Fun, types.UnsafePointer) &&
+                               hasBasicType(pass.TypesInfo, x.Args[0], types.Uintptr) &&
+                               !isSafeUintptr(pass.TypesInfo, x.Args[0]) {
+                               pass.ReportRangef(x, "possible misuse of unsafe.Pointer")
+                       }
+               case *ast.StarExpr:
+                       if t := pass.TypesInfo.Types[x].Type; isReflectHeader(t) {
+                               pass.ReportRangef(x, "possible misuse of %s", t)
+                       }
+               case *ast.UnaryExpr:
+                       if x.Op != token.AND {
+                               return
+                       }
+                       if t := pass.TypesInfo.Types[x.X].Type; isReflectHeader(t) {
+                               pass.ReportRangef(x, "possible misuse of %s", t)
+                       }
+               }
+       })
+       return nil, nil
+}
+
+// isSafeUintptr reports whether x - already known to be a uintptr -
+// is safe to convert to unsafe.Pointer.
+func isSafeUintptr(info *types.Info, x ast.Expr) bool {
+       // Check unsafe.Pointer safety rules according to
+       // https://golang.org/pkg/unsafe/#Pointer.
+
+       switch x := analysisutil.Unparen(x).(type) {
+       case *ast.SelectorExpr:
+               // "(6) Conversion of a reflect.SliceHeader or
+               // reflect.StringHeader Data field to or from Pointer."
+               if x.Sel.Name != "Data" {
+                       break
+               }
+               // reflect.SliceHeader and reflect.StringHeader are okay,
+               // but only if they are pointing at a real slice or string.
+               // It's not okay to do:
+               //      var x SliceHeader
+               //      x.Data = uintptr(unsafe.Pointer(...))
+               //      ... use x ...
+               //      p := unsafe.Pointer(x.Data)
+               // because in the middle the garbage collector doesn't
+               // see x.Data as a pointer and so x.Data may be dangling
+               // by the time we get to the conversion at the end.
+               // For now approximate by saying that *Header is okay
+               // but Header is not.
+               pt, ok := info.Types[x.X].Type.(*types.Pointer)
+               if ok && isReflectHeader(pt.Elem()) {
+                       return true
+               }
+
+       case *ast.CallExpr:
+               // "(5) Conversion of the result of reflect.Value.Pointer or
+               // reflect.Value.UnsafeAddr from uintptr to Pointer."
+               if len(x.Args) != 0 {
+                       break
+               }
+               sel, ok := x.Fun.(*ast.SelectorExpr)
+               if !ok {
+                       break
+               }
+               switch sel.Sel.Name {
+               case "Pointer", "UnsafeAddr":
+                       t, ok := info.Types[sel.X].Type.(*types.Named)
+                       if ok && t.Obj().Pkg().Path() == "reflect" && t.Obj().Name() == "Value" {
+                               return true
+                       }
+               }
+       }
+
+       // "(3) Conversion of a Pointer to a uintptr and back, with arithmetic."
+       return isSafeArith(info, x)
+}
+
+// isSafeArith reports whether x is a pointer arithmetic expression that is safe
+// to convert to unsafe.Pointer.
+func isSafeArith(info *types.Info, x ast.Expr) bool {
+       switch x := analysisutil.Unparen(x).(type) {
+       case *ast.CallExpr:
+               // Base case: initial conversion from unsafe.Pointer to uintptr.
+               return len(x.Args) == 1 &&
+                       hasBasicType(info, x.Fun, types.Uintptr) &&
+                       hasBasicType(info, x.Args[0], types.UnsafePointer)
+
+       case *ast.BinaryExpr:
+               // "It is valid both to add and to subtract offsets from a
+               // pointer in this way. It is also valid to use &^ to round
+               // pointers, usually for alignment."
+               switch x.Op {
+               case token.ADD, token.SUB, token.AND_NOT:
+                       // TODO(mdempsky): Match compiler
+                       // semantics. ADD allows a pointer on either
+                       // side; SUB and AND_NOT don't care about RHS.
+                       return isSafeArith(info, x.X) && !isSafeArith(info, x.Y)
+               }
+       }
+
+       return false
+}
+
+// hasBasicType reports whether x's type is a types.Basic with the given kind.
+func hasBasicType(info *types.Info, x ast.Expr, kind types.BasicKind) bool {
+       t := info.Types[x].Type
+       if t != nil {
+               t = t.Underlying()
+       }
+       b, ok := t.(*types.Basic)
+       return ok && b.Kind() == kind
+}
+
+// isReflectHeader reports whether t is reflect.SliceHeader or reflect.StringHeader.
+func isReflectHeader(t types.Type) bool {
+       if named, ok := t.(*types.Named); ok {
+               if obj := named.Obj(); obj.Pkg() != nil && obj.Pkg().Path() == "reflect" {
+                       switch obj.Name() {
+                       case "SliceHeader", "StringHeader":
+                               return true
+                       }
+               }
+       }
+       return false
+}