From 25ca34923f0374e38a0b0ebd1eecaccc4452f88d Mon Sep 17 00:00:00 2001
From: Fabio Bozzo <fabio.bozzo@consensys.net>
Date: Wed, 13 Nov 2024 18:03:00 +0100
Subject: [PATCH] feat(container): versioning for the CBOR container

---
 pkg/container/car_test.go    |  7 +++++++
 pkg/container/reader.go      | 33 +++++++++++++++++++++++++++------
 pkg/container/serial_test.go | 27 +++++++++++++++++++++++++++
 pkg/container/writer.go      | 12 ++++++++----
 4 files changed, 69 insertions(+), 10 deletions(-)

diff --git a/pkg/container/car_test.go b/pkg/container/car_test.go
index 80f4b5d..5b20b8e 100644
--- a/pkg/container/car_test.go
+++ b/pkg/container/car_test.go
@@ -40,6 +40,13 @@ func TestCarRoundTrip(t *testing.T) {
 }
 
 func FuzzCarRoundTrip(f *testing.F) {
+	// Note: this fuzzing is somewhat broken.
+	// After some time, the fuzzer discover that a varint can be serialized in different
+	// ways that lead to the same integer value. This means that the CAR format can have
+	// multiple legal binary representation for the exact same data, which is what we are
+	// trying to detect here. Ideally, the format would be stricter, but that's how things
+	// are.
+
 	example, err := os.ReadFile("testdata/sample-v1.car")
 	require.NoError(f, err)
 
diff --git a/pkg/container/reader.go b/pkg/container/reader.go
index db1e145..8c29fb1 100644
--- a/pkg/container/reader.go
+++ b/pkg/container/reader.go
@@ -98,15 +98,36 @@ func FromCbor(r io.Reader) (Reader, error) {
 	if err != nil {
 		return nil, err
 	}
-	if n.Kind() != datamodel.Kind_List {
-		return nil, fmt.Errorf("not a list")
+	if n.Kind() != datamodel.Kind_Map {
+		return nil, fmt.Errorf("invalid container format: expected map")
+	}
+	if n.Length() != 1 {
+		return nil, fmt.Errorf("invalid container format: expected single version key")
+	}
+
+	// get the first (and only) key-value pair
+	it := n.MapIterator()
+	key, tokensNode, err := it.Next()
+	if err != nil {
+		return nil, err
 	}
 
-	ctn := make(Reader, n.Length())
+	version, err := key.AsString()
+	if err != nil {
+		return nil, fmt.Errorf("invalid container format: version must be string")
+	}
+	if version != currentContainerVersion {
+		return nil, fmt.Errorf("unsupported container version: %s", version)
+	}
+
+	if tokensNode.Kind() != datamodel.Kind_List {
+		return nil, fmt.Errorf("invalid container format: tokens must be a list")
+	}
 
-	it := n.ListIterator()
-	for !it.Done() {
-		_, val, err := it.Next()
+	ctn := make(Reader, tokensNode.Length())
+	it2 := tokensNode.ListIterator()
+	for !it2.Done() {
+		_, val, err := it2.Next()
 		if err != nil {
 			return nil, err
 		}
diff --git a/pkg/container/serial_test.go b/pkg/container/serial_test.go
index fea2efe..36560b0 100644
--- a/pkg/container/serial_test.go
+++ b/pkg/container/serial_test.go
@@ -176,3 +176,30 @@ func randToken() (*delegation.Token, cid.Cid, []byte) {
 	}
 	return t, c, b
 }
+
+func FuzzContainerRead(f *testing.F) {
+	// Generate a corpus
+	for tokenCount := 0; tokenCount < 10; tokenCount++ {
+		writer := NewWriter()
+		for i := 0; i < tokenCount; i++ {
+			_, c, data := randToken()
+			writer.AddSealed(c, data)
+		}
+		buf := bytes.NewBuffer(nil)
+		err := writer.ToCbor(buf)
+		require.NoError(f, err)
+
+		f.Add(buf.Bytes())
+	}
+
+	f.Fuzz(func(t *testing.T, data []byte) {
+		start := time.Now()
+
+		// search for panics
+		_, _ = FromCbor(bytes.NewReader(data))
+
+		if time.Since(start) > 100*time.Millisecond {
+			panic("too long")
+		}
+	})
+}
diff --git a/pkg/container/writer.go b/pkg/container/writer.go
index 310fa8a..e8d8c32 100644
--- a/pkg/container/writer.go
+++ b/pkg/container/writer.go
@@ -14,6 +14,8 @@ import (
 
 // TODO: should we have a multibase to wrap the cbor? but there is no reader/write in go-multibase :-(
 
+const currentContainerVersion = "ctn-v1"
+
 // Writer is a token container writer. It provides a convenient way to aggregate and serialize tokens together.
 type Writer map[cid.Cid][]byte
 
@@ -43,10 +45,12 @@ func (ctn Writer) ToCarBase64(w io.Writer) error {
 }
 
 func (ctn Writer) ToCbor(w io.Writer) error {
-	node, err := qp.BuildList(basicnode.Prototype.Any, int64(len(ctn)), func(la datamodel.ListAssembler) {
-		for _, bytes := range ctn {
-			qp.ListEntry(la, qp.Bytes(bytes))
-		}
+	node, err := qp.BuildMap(basicnode.Prototype.Any, 1, func(ma datamodel.MapAssembler) {
+		qp.MapEntry(ma, currentContainerVersion, qp.List(int64(len(ctn)), func(la datamodel.ListAssembler) {
+			for _, bytes := range ctn {
+				qp.ListEntry(la, qp.Bytes(bytes))
+			}
+		}))
 	})
 	if err != nil {
 		return err