Security Fix: clean paths correctly to avoid circumvention of BlacklistedPaths

server/context/context.go

@@ -48,11 +48,9 @@ func (c *Context) Redirect(uri string, statusCode int) {
 	http.Redirect(c.RespWriter, c.Req, uri, statusCode)
 }
 
-// Path returns requested path.
+// Path returns the cleaned requested path.
-//
-// The returned bytes are valid until your request handler returns.
 func (c *Context) Path() string {
-	return c.Req.URL.Path
+	return utils.CleanPath(c.Req.URL.Path)
 }
 
 func (c *Context) Host() string {

server/utils/utils.go

@@ -1,6 +1,8 @@
 package utils
 
 import (
+	"net/url"
+	"path"
 	"strings"
 )
 
@@ -11,3 +13,15 @@ func TrimHostPort(host string) string {
 	}
 	return host
 }
+
+func CleanPath(uriPath string) string {
+	unescapedPath, _ := url.PathUnescape(uriPath)
+	cleanedPath := path.Join("/", unescapedPath)
+
+	// If the path refers to a directory, add a trailing slash.
+	if !strings.HasSuffix(cleanedPath, "/") && (strings.HasSuffix(unescapedPath, "/") || strings.HasSuffix(unescapedPath, "/.") || strings.HasSuffix(unescapedPath, "/..")) {
+		cleanedPath += "/"
+	}
+
+	return cleanedPath
+}

server/utils/utils_test.go

@@ -11,3 +11,59 @@ func TestTrimHostPort(t *testing.T) {
 	assert.EqualValues(t, "", TrimHostPort(":"))
 	assert.EqualValues(t, "example.com", TrimHostPort("example.com:80"))
 }
+
+// TestCleanPath is mostly copied from fasthttp, to keep the behaviour we had before migrating away from it.
+// Source (MIT licensed): https://github.com/valyala/fasthttp/blob/v1.48.0/uri_test.go#L154
+// Copyright (c) 2015-present Aliaksandr Valialkin, VertaMedia, Kirill Danshin, Erik Dubbelboer, FastHTTP Authors
+func TestCleanPath(t *testing.T) {
+	// double slash
+	testURIPathNormalize(t, "/aa//bb", "/aa/bb")
+
+	// triple slash
+	testURIPathNormalize(t, "/x///y/", "/x/y/")
+
+	// multi slashes
+	testURIPathNormalize(t, "/abc//de///fg////", "/abc/de/fg/")
+
+	// encoded slashes
+	testURIPathNormalize(t, "/xxxx%2fyyy%2f%2F%2F", "/xxxx/yyy/")
+
+	// dotdot
+	testURIPathNormalize(t, "/aaa/..", "/")
+
+	// dotdot with trailing slash
+	testURIPathNormalize(t, "/xxx/yyy/../", "/xxx/")
+
+	// multi dotdots
+	testURIPathNormalize(t, "/aaa/bbb/ccc/../../ddd", "/aaa/ddd")
+
+	// dotdots separated by other data
+	testURIPathNormalize(t, "/a/b/../c/d/../e/..", "/a/c/")
+
+	// too many dotdots
+	testURIPathNormalize(t, "/aaa/../../../../xxx", "/xxx")
+	testURIPathNormalize(t, "/../../../../../..", "/")
+	testURIPathNormalize(t, "/../../../../../../", "/")
+
+	// encoded dotdots
+	testURIPathNormalize(t, "/aaa%2Fbbb%2F%2E.%2Fxxx", "/aaa/xxx")
+
+	// double slash with dotdots
+	testURIPathNormalize(t, "/aaa////..//b", "/b")
+
+	// fake dotdot
+	testURIPathNormalize(t, "/aaa/..bbb/ccc/..", "/aaa/..bbb/")
+
+	// single dot
+	testURIPathNormalize(t, "/a/./b/././c/./d.html", "/a/b/c/d.html")
+	testURIPathNormalize(t, "./foo/", "/foo/")
+	testURIPathNormalize(t, "./../.././../../aaa/bbb/../../../././../", "/")
+	testURIPathNormalize(t, "./a/./.././../b/./foo.html", "/b/foo.html")
+}
+
+func testURIPathNormalize(t *testing.T, requestURI, expectedPath string) {
+	cleanedPath := CleanPath(requestURI)
+	if cleanedPath != expectedPath {
+		t.Fatalf("Unexpected path %q. Expected %q. requestURI=%q", cleanedPath, expectedPath, requestURI)
+	}
+}