Add some basic security policies with sensible defaults

This ommmit contains some security hardening measures for the Hugo build runtime.

There are some rarely used features in Hugo that would be good to have disabled by default. One example would be the "external helpers".

For `asciidoctor` and some others we use Go's `os/exec` package to start a new process.

These are a predefined set of binary names, all loaded from `PATH` and with a predefined set of arguments. Still, if you don't use `asciidoctor` in your project, you might as well have it turned off.

You can configure your own in the new `security` configuration section, but the defaults are configured to create a minimal amount of site breakage. And if that do happen, you will get clear instructions in the loa about what to do.

The default configuration is listed below. Note that almost all of these options are regular expression _whitelists_ (a string or a slice); the value `none` will block all.

```toml
[security]
  enableInlineShortcodes = false
  [security.exec]
    allow = ['^dart-sass-embedded$', '^go$', '^npx$', '^postcss$']
    osEnv = ['(?i)^(PATH|PATHEXT|APPDATA|TMP|TEMP|TERM)$']

  [security.funcs]
    getenv = ['^HUGO_']

  [security.http]
    methods = ['(?i)GET|POST']
    urls = ['.*']
```

markup/asciidocext/convert.go

@@ -21,10 +21,9 @@ import (
 	"path/filepath"
 	"strings"
 
+	"github.com/gohugoio/hugo/common/hexec"
 	"github.com/gohugoio/hugo/htesting"
 
-	"github.com/cli/safeexec"
-
 	"github.com/gohugoio/hugo/identity"
 	"github.com/gohugoio/hugo/markup/asciidocext/asciidocext_config"
 	"github.com/gohugoio/hugo/markup/converter"
@@ -67,7 +66,11 @@ type asciidocConverter struct {
 }
 
 func (a *asciidocConverter) Convert(ctx converter.RenderContext) (converter.Result, error) {
-	content, toc, err := a.extractTOC(a.getAsciidocContent(ctx.Src, a.ctx))
+	b, err := a.getAsciidocContent(ctx.Src, a.ctx)
+	if err != nil {
+		return nil, err
+	}
+	content, toc, err := a.extractTOC(b)
 	if err != nil {
 		return nil, err
 	}
@@ -83,20 +86,19 @@ func (a *asciidocConverter) Supports(_ identity.Identity) bool {
 
 // getAsciidocContent calls asciidoctor as an external helper
 // to convert AsciiDoc content to HTML.
-func (a *asciidocConverter) getAsciidocContent(src []byte, ctx converter.DocumentContext) []byte {
-	path := getAsciidoctorExecPath()
-	if path == "" {
+func (a *asciidocConverter) getAsciidocContent(src []byte, ctx converter.DocumentContext) ([]byte, error) {
+	if !hasAsciiDoc() {
 		a.cfg.Logger.Errorln("asciidoctor not found in $PATH: Please install.\n",
 			"                 Leaving AsciiDoc content unrendered.")
-		return src
+		return src, nil
 	}
 
 	args := a.parseArgs(ctx)
 	args = append(args, "-")
 
-	a.cfg.Logger.Infoln("Rendering", ctx.DocumentName, "with", path, "using asciidoctor args", args, "...")
+	a.cfg.Logger.Infoln("Rendering", ctx.DocumentName, " using asciidoctor args", args, "...")
 
-	return internal.ExternallyRenderContent(a.cfg, ctx, src, path, args)
+	return internal.ExternallyRenderContent(a.cfg, ctx, src, asciiDocBinaryName, args)
 }
 
 func (a *asciidocConverter) parseArgs(ctx converter.DocumentContext) []string {
@@ -195,12 +197,10 @@ func (a *asciidocConverter) appendArg(args []string, option, value, defaultValue
 	return args
 }
 
-func getAsciidoctorExecPath() string {
-	path, err := safeexec.LookPath("asciidoctor")
-	if err != nil {
-		return ""
-	}
-	return path
+const asciiDocBinaryName = "asciidoctor"
+
+func hasAsciiDoc() bool {
+	return hexec.InPath(asciiDocBinaryName)
 }
 
 // extractTOC extracts the toc from the given src html.
@@ -311,8 +311,12 @@ func nodeContent(node *html.Node) string {
 
 // Supports returns whether Asciidoctor is installed on this computer.
 func Supports() bool {
+	hasBin := hasAsciiDoc()
 	if htesting.SupportsAll() {
+		if !hasBin {
+			panic("asciidoctor not installed")
+		}
 		return true
 	}
-	return getAsciidoctorExecPath() != ""
+	return hasBin
 }

markup/asciidocext/convert_test.go

@@ -21,8 +21,10 @@ import (
 	"path/filepath"
 	"testing"
 
+	"github.com/gohugoio/hugo/common/hexec"
 	"github.com/gohugoio/hugo/common/loggers"
 	"github.com/gohugoio/hugo/config"
+	"github.com/gohugoio/hugo/config/security"
 	"github.com/gohugoio/hugo/markup/converter"
 	"github.com/gohugoio/hugo/markup/markup_config"
 	"github.com/gohugoio/hugo/markup/tableofcontents"
@@ -280,20 +282,28 @@ func TestAsciidoctorAttributes(t *testing.T) {
 	c.Assert(args[4], qt.Equals, "--no-header-footer")
 }
 
+func getProvider(c *qt.C, mconf markup_config.Config) converter.Provider {
+	sc := security.DefaultConfig
+	sc.Exec.Allow = security.NewWhitelist("asciidoctor")
+
+	p, err := Provider.New(
+		converter.ProviderConfig{
+			MarkupConfig: mconf,
+			Logger:       loggers.NewErrorLogger(),
+			Exec:         hexec.New(sc),
+		},
+	)
+	c.Assert(err, qt.IsNil)
+	return p
+}
+
 func TestConvert(t *testing.T) {
 	if !Supports() {
 		t.Skip("asciidoctor not installed")
 	}
 	c := qt.New(t)
 
-	mconf := markup_config.Default
-	p, err := Provider.New(
-		converter.ProviderConfig{
-			MarkupConfig: mconf,
-			Logger:       loggers.NewErrorLogger(),
-		},
-	)
-	c.Assert(err, qt.IsNil)
+	p := getProvider(c, markup_config.Default)
 
 	conv, err := p.New(converter.DocumentContext{})
 	c.Assert(err, qt.IsNil)
@@ -308,14 +318,8 @@ func TestTableOfContents(t *testing.T) {
 		t.Skip("asciidoctor not installed")
 	}
 	c := qt.New(t)
-	mconf := markup_config.Default
-	p, err := Provider.New(
-		converter.ProviderConfig{
-			MarkupConfig: mconf,
-			Logger:       loggers.NewErrorLogger(),
-		},
-	)
-	c.Assert(err, qt.IsNil)
+	p := getProvider(c, markup_config.Default)
+
 	conv, err := p.New(converter.DocumentContext{})
 	c.Assert(err, qt.IsNil)
 	r, err := conv.Convert(converter.RenderContext{Src: []byte(`:toc: macro
@@ -390,14 +394,7 @@ func TestTableOfContentsWithCode(t *testing.T) {
 		t.Skip("asciidoctor not installed")
 	}
 	c := qt.New(t)
-	mconf := markup_config.Default
-	p, err := Provider.New(
-		converter.ProviderConfig{
-			MarkupConfig: mconf,
-			Logger:       loggers.NewErrorLogger(),
-		},
-	)
-	c.Assert(err, qt.IsNil)
+	p := getProvider(c, markup_config.Default)
 	conv, err := p.New(converter.DocumentContext{})
 	c.Assert(err, qt.IsNil)
 	r, err := conv.Convert(converter.RenderContext{Src: []byte(`:toc: auto
@@ -433,13 +430,8 @@ func TestTableOfContentsPreserveTOC(t *testing.T) {
 	c := qt.New(t)
 	mconf := markup_config.Default
 	mconf.AsciidocExt.PreserveTOC = true
-	p, err := Provider.New(
-		converter.ProviderConfig{
-			MarkupConfig: mconf,
-			Logger:       loggers.NewErrorLogger(),
-		},
-	)
-	c.Assert(err, qt.IsNil)
+	p := getProvider(c, mconf)
+
 	conv, err := p.New(converter.DocumentContext{})
 	c.Assert(err, qt.IsNil)
 	r, err := conv.Convert(converter.RenderContext{Src: []byte(`:toc:

markup/converter/converter.go

@@ -16,6 +16,7 @@ package converter
 import (
 	"bytes"
 
+	"github.com/gohugoio/hugo/common/hexec"
 	"github.com/gohugoio/hugo/common/loggers"
 	"github.com/gohugoio/hugo/config"
 	"github.com/gohugoio/hugo/identity"
@@ -32,6 +33,7 @@ type ProviderConfig struct {
 	Cfg       config.Provider // Site config
 	ContentFs afero.Fs
 	Logger    loggers.Logger
+	Exec      *hexec.Exec
 	Highlight func(code, lang, optsStr string) (string, error)
 }
 

markup/internal/external.go

@@ -2,42 +2,56 @@ package internal
 
 import (
 	"bytes"
+	"fmt"
 	"strings"
 
-	"github.com/cli/safeexec"
+	"github.com/gohugoio/hugo/common/collections"
 	"github.com/gohugoio/hugo/common/hexec"
-
 	"github.com/gohugoio/hugo/markup/converter"
 )
 
 func ExternallyRenderContent(
 	cfg converter.ProviderConfig,
 	ctx converter.DocumentContext,
-	content []byte, path string, args []string) []byte {
+	content []byte, binaryName string, args []string) ([]byte, error) {
 	logger := cfg.Logger
-	cmd, err := hexec.SafeCommand(path, args...)
-	if err != nil {
-		logger.Errorf("%s rendering %s: %v", path, ctx.DocumentName, err)
-		return nil
+
+	if strings.Contains(binaryName, "/") {
+		panic(fmt.Sprintf("should be no slash in %q", binaryName))
 	}
-	cmd.Stdin = bytes.NewReader(content)
+
+	argsv := collections.StringSliceToInterfaceSlice(args)
+
 	var out, cmderr bytes.Buffer
-	cmd.Stdout = &out
-	cmd.Stderr = &cmderr
+	argsv = append(argsv, hexec.WithStdout(&out))
+	argsv = append(argsv, hexec.WithStderr(&cmderr))
+	argsv = append(argsv, hexec.WithStdin(bytes.NewReader(content)))
+
+	cmd, err := cfg.Exec.New(binaryName, argsv...)
+	if err != nil {
+		return nil, err
+	}
+
 	err = cmd.Run()
+
 	// Most external helpers exit w/ non-zero exit code only if severe, i.e.
 	// halting errors occurred. -> log stderr output regardless of state of err
 	for _, item := range strings.Split(cmderr.String(), "\n") {
 		item := strings.TrimSpace(item)
 		if item != "" {
-			logger.Errorf("%s: %s", ctx.DocumentName, item)
+			if err == nil {
+				logger.Warnf("%s: %s", ctx.DocumentName, item)
+			} else {
+				logger.Errorf("%s: %s", ctx.DocumentName, item)
+			}
 		}
 	}
+
 	if err != nil {
-		logger.Errorf("%s rendering %s: %v", path, ctx.DocumentName, err)
+		logger.Errorf("%s rendering %s: %v", binaryName, ctx.DocumentName, err)
 	}
 
-	return normalizeExternalHelperLineFeeds(out.Bytes())
+	return normalizeExternalHelperLineFeeds(out.Bytes()), nil
 }
 
 // Strips carriage returns from third-party / external processes (useful for Windows)
@@ -45,13 +59,13 @@ func normalizeExternalHelperLineFeeds(content []byte) []byte {
 	return bytes.Replace(content, []byte("\r"), []byte(""), -1)
 }
 
-func GetPythonExecPath() string {
-	path, err := safeexec.LookPath("python")
-	if err != nil {
-		path, err = safeexec.LookPath("python.exe")
-		if err != nil {
-			return ""
+var pythonBinaryCandidates = []string{"python", "python.exe"}
+
+func GetPythonBinaryAndExecPath() (string, string) {
+	for _, p := range pythonBinaryCandidates {
+		if pth := hexec.LookPath(p); pth != "" {
+			return p, pth
 		}
 	}
-	return path
+	return "", ""
 }

markup/pandoc/convert.go

@@ -15,7 +15,7 @@
 package pandoc
 
 import (
-	"github.com/cli/safeexec"
+	"github.com/gohugoio/hugo/common/hexec"
 	"github.com/gohugoio/hugo/htesting"
 	"github.com/gohugoio/hugo/identity"
 	"github.com/gohugoio/hugo/markup/internal"
@@ -44,7 +44,11 @@ type pandocConverter struct {
 }
 
 func (c *pandocConverter) Convert(ctx converter.RenderContext) (converter.Result, error) {
-	return converter.Bytes(c.getPandocContent(ctx.Src, c.ctx)), nil
+	b, err := c.getPandocContent(ctx.Src, c.ctx)
+	if err != nil {
+		return nil, err
+	}
+	return converter.Bytes(b), nil
 }
 
 func (c *pandocConverter) Supports(feature identity.Identity) bool {
@@ -52,31 +56,35 @@ func (c *pandocConverter) Supports(feature identity.Identity) bool {
 }
 
 // getPandocContent calls pandoc as an external helper to convert pandoc markdown to HTML.
-func (c *pandocConverter) getPandocContent(src []byte, ctx converter.DocumentContext) []byte {
+func (c *pandocConverter) getPandocContent(src []byte, ctx converter.DocumentContext) ([]byte, error) {
 	logger := c.cfg.Logger
-	path := getPandocExecPath()
-	if path == "" {
+	binaryName := getPandocBinaryName()
+	if binaryName == "" {
 		logger.Println("pandoc not found in $PATH: Please install.\n",
 			"                 Leaving pandoc content unrendered.")
-		return src
+		return src, nil
 	}
 	args := []string{"--mathjax"}
-	return internal.ExternallyRenderContent(c.cfg, ctx, src, path, args)
+	return internal.ExternallyRenderContent(c.cfg, ctx, src, binaryName, args)
 }
 
-func getPandocExecPath() string {
-	path, err := safeexec.LookPath("pandoc")
-	if err != nil {
-		return ""
-	}
+const pandocBinary = "pandoc"
 
-	return path
+func getPandocBinaryName() string {
+	if hexec.InPath(pandocBinary) {
+		return pandocBinary
+	}
+	return ""
 }
 
 // Supports returns whether Pandoc is installed on this computer.
 func Supports() bool {
+	hasBin := getPandocBinaryName() != ""
 	if htesting.SupportsAll() {
+		if !hasBin {
+			panic("pandoc not installed")
+		}
 		return true
 	}
-	return getPandocExecPath() != ""
+	return hasBin
 }

markup/pandoc/convert_test.go

@@ -16,7 +16,9 @@ package pandoc
 import (
 	"testing"
 
+	"github.com/gohugoio/hugo/common/hexec"
 	"github.com/gohugoio/hugo/common/loggers"
+	"github.com/gohugoio/hugo/config/security"
 
 	"github.com/gohugoio/hugo/markup/converter"
 
@@ -28,7 +30,9 @@ func TestConvert(t *testing.T) {
 		t.Skip("pandoc not installed")
 	}
 	c := qt.New(t)
-	p, err := Provider.New(converter.ProviderConfig{Logger: loggers.NewErrorLogger()})
+	sc := security.DefaultConfig
+	sc.Exec.Allow = security.NewWhitelist("pandoc")
+	p, err := Provider.New(converter.ProviderConfig{Exec: hexec.New(sc), Logger: loggers.NewErrorLogger()})
 	c.Assert(err, qt.IsNil)
 	conv, err := p.New(converter.DocumentContext{})
 	c.Assert(err, qt.IsNil)

markup/rst/convert.go

@@ -18,7 +18,7 @@ import (
 	"bytes"
 	"runtime"
 
-	"github.com/cli/safeexec"
+	"github.com/gohugoio/hugo/common/hexec"
 	"github.com/gohugoio/hugo/htesting"
 
 	"github.com/gohugoio/hugo/identity"
@@ -48,7 +48,11 @@ type rstConverter struct {
 }
 
 func (c *rstConverter) Convert(ctx converter.RenderContext) (converter.Result, error) {
-	return converter.Bytes(c.getRstContent(ctx.Src, c.ctx)), nil
+	b, err := c.getRstContent(ctx.Src, c.ctx)
+	if err != nil {
+		return nil, err
+	}
+	return converter.Bytes(b), nil
 }
 
 func (c *rstConverter) Supports(feature identity.Identity) bool {
@@ -57,31 +61,38 @@ func (c *rstConverter) Supports(feature identity.Identity) bool {
 
 // getRstContent calls the Python script rst2html as an external helper
 // to convert reStructuredText content to HTML.
-func (c *rstConverter) getRstContent(src []byte, ctx converter.DocumentContext) []byte {
+func (c *rstConverter) getRstContent(src []byte, ctx converter.DocumentContext) ([]byte, error) {
 	logger := c.cfg.Logger
-	path := getRstExecPath()
+	binaryName, binaryPath := getRstBinaryNameAndPath()
 
-	if path == "" {
+	if binaryName == "" {
 		logger.Println("rst2html / rst2html.py not found in $PATH: Please install.\n",
 			"                 Leaving reStructuredText content unrendered.")
-		return src
+		return src, nil
 	}
 
-	logger.Infoln("Rendering", ctx.DocumentName, "with", path, "...")
+	logger.Infoln("Rendering", ctx.DocumentName, "with", binaryName, "...")
 
 	var result []byte
+	var err error
+
 	// certain *nix based OSs wrap executables in scripted launchers
 	// invoking binaries on these OSs via python interpreter causes SyntaxError
 	// invoke directly so that shebangs work as expected
 	// handle Windows manually because it doesn't do shebangs
 	if runtime.GOOS == "windows" {
-		python := internal.GetPythonExecPath()
-		args := []string{path, "--leave-comments", "--initial-header-level=2"}
-		result = internal.ExternallyRenderContent(c.cfg, ctx, src, python, args)
+		pythonBinary, _ := internal.GetPythonBinaryAndExecPath()
+		args := []string{binaryPath, "--leave-comments", "--initial-header-level=2"}
+		result, err = internal.ExternallyRenderContent(c.cfg, ctx, src, pythonBinary, args)
 	} else {
 		args := []string{"--leave-comments", "--initial-header-level=2"}
-		result = internal.ExternallyRenderContent(c.cfg, ctx, src, path, args)
+		result, err = internal.ExternallyRenderContent(c.cfg, ctx, src, binaryName, args)
 	}
+
+	if err != nil {
+		return nil, err
+	}
+
 	// TODO(bep) check if rst2html has a body only option.
 	bodyStart := bytes.Index(result, []byte("<body>\n"))
 	if bodyStart < 0 {
@@ -96,24 +107,29 @@ func (c *rstConverter) getRstContent(src []byte, ctx converter.DocumentContext)
 		}
 	}
 
-	return result[bodyStart+7 : bodyEnd]
+	return result[bodyStart+7 : bodyEnd], err
 }
 
-func getRstExecPath() string {
-	path, err := safeexec.LookPath("rst2html")
-	if err != nil {
-		path, err = safeexec.LookPath("rst2html.py")
-		if err != nil {
-			return ""
+var rst2Binaries = []string{"rst2html", "rst2html.py"}
+
+func getRstBinaryNameAndPath() (string, string) {
+	for _, candidate := range rst2Binaries {
+		if pth := hexec.LookPath(candidate); pth != "" {
+			return candidate, pth
 		}
 	}
-	return path
+	return "", ""
 }
 
-// Supports returns whether rst is installed on this computer.
+// Supports returns whether rst is (or should be) installed on this computer.
 func Supports() bool {
+	name, _ := getRstBinaryNameAndPath()
+	hasBin := name != ""
 	if htesting.SupportsAll() {
+		if !hasBin {
+			panic("rst not installed")
+		}
 		return true
 	}
-	return getRstExecPath() != ""
+	return hasBin
 }

markup/rst/convert_test.go

@@ -16,7 +16,9 @@ package rst
 import (
 	"testing"
 
+	"github.com/gohugoio/hugo/common/hexec"
 	"github.com/gohugoio/hugo/common/loggers"
+	"github.com/gohugoio/hugo/config/security"
 
 	"github.com/gohugoio/hugo/markup/converter"
 
@@ -28,7 +30,14 @@ func TestConvert(t *testing.T) {
 		t.Skip("rst not installed")
 	}
 	c := qt.New(t)
-	p, err := Provider.New(converter.ProviderConfig{Logger: loggers.NewErrorLogger()})
+	sc := security.DefaultConfig
+	sc.Exec.Allow = security.NewWhitelist("rst", "python")
+
+	p, err := Provider.New(
+		converter.ProviderConfig{
+			Logger: loggers.NewErrorLogger(),
+			Exec:   hexec.New(sc),
+		})
 	c.Assert(err, qt.IsNil)
 	conv, err := p.New(converter.DocumentContext{})
 	c.Assert(err, qt.IsNil)