kgo / kmsg: process the RecordBatches field in FetchResponses properly

As it turns out, a FetchResponse can be of high enough version to use
actual record batches, but still use message sets. Rather than relying
on the response version, we need to check the magic in each batch.

It may be possible that each batch has a different magic from one batch
to the next, so we cannot just check the first batch and use that for
all decoding.

For message set v1, it is also possible that a client could have used
messages v0, compressed them, and used that as the "inner messages" in a
message v1.

All of these cases are now handled properly, which necessitated the
removal of some functions from kmsg. The functions in kmsg were not
necessarily correct. Further, it was a bit odd to stuff message decoding
and validating into kmsg. This makes kmsg a more dedicated encoding /
decoding package.

pkg/kgo/client.go

@@ -20,6 +20,7 @@ package kgo
 import (
  "context"
  "fmt"
+ "hash/crc32"
  "math/rand"
  "reflect"
  "sort"
@@ -33,6 +34,8 @@ import (
  "github.com/twmb/franz-go/pkg/kmsg"
 )
 
+var crc32c = crc32.MakeTable(crc32.Castagnoli) // record crc's use Castagnoli table; for consuming/producing
+
 // Client issues requests and handles responses to a Kafka cluster.
 type Client struct {
  cfg cfg

pkg/kgo/sink.go

@@ -1581,8 +1581,6 @@ func (r seqRecBatch) appendTo(
  return dst
 }
 
-var crc32c = crc32.MakeTable(crc32.Castagnoli) // record crc's use Castagnoli table
-
 func (pnr promisedNumberedRecord) appendTo(dst []byte, offsetDelta int32) []byte {
  dst = kbin.AppendVarint(dst, pnr.lengthField)
  dst = kbin.AppendInt8(dst, 0) // attributes, currently unused

pkg/kgo/source.go

@@ -2,15 +2,22 @@ package kgo
 
 import (
  "context"
+ "encoding/binary"
  "fmt"
+ "hash/crc32"
  "sync"
  "sync/atomic"
  "time"
 
+ "github.com/twmb/franz-go/pkg/kbin"
  "github.com/twmb/franz-go/pkg/kerr"
  "github.com/twmb/franz-go/pkg/kmsg"
 )
 
+type readerFrom interface {
+ ReadFrom([]byte) error
+}
+
 // A source consumes from an individual broker.
 //
 // As long as there is at least one active cursor, a source aims to have *one*
@@ -764,37 +771,91 @@ func (o *cursorOffsetNext) processRespPartition(version int16, rp *kmsg.FetchRes
  LogStartOffset: rp.LogStartOffset,
  }
 
- switch version {
- case 0, 1:
- msgs, err := kmsg.ReadV0Messages(rp.RecordBatches)
- if err != nil {
- fp.Err = err
+ aborter := buildAborter(rp)
+
+ // A response could contain any of message v0, message v1, or record
+ // batches, and this is solely dictated by the magic byte (not the
+ // fetch response version). The magic byte is located at byte 17.
+ //
+ // 0 thru 8: int64 offset / first offset
+ // 9 thru 12: int32 length
+ // 13 thru 16: crc (magic 0 or 1), or partition leader epoch (magic 2)
+ // 17: magic
+ //
+ // We decode and validate similarly for messages and record batches, so
+ // we "abstract" away the high level stuff into a check function just
+ // below, and then switch based on the magic for how to process.
+ var (
+ in = rp.RecordBatches
+
+ r readerFrom
+ length int32
+ lengthField *int32
+ crcField *int32
+ crcTable *crc32.Table
+ crcAt int
+
+ check = func() bool {
+ if err := r.ReadFrom(in[:length]); err != nil {
+ return false
+ }
+ if length := int32(len(in[12:length])); length != *lengthField {
+ fp.Err = fmt.Errorf("encoded length %d does not match read length %d", *lengthField, length)
+ return false
+ }
+ if crcCalc := int32(crc32.Checksum(in[crcAt:length], crcTable)); crcCalc != *crcField {
+ fp.Err = fmt.Errorf("encoded crc %x does not match calculated crc %x", *crcField, crcCalc)
+ return false
+ }
+ return true
  }
-  o.processV0Messages(&fp, msgs, decompressor)
+ )
 
- case 2, 3:
- msgs, err := kmsg.ReadV1Messages(rp.RecordBatches)
- if err != nil {
- fp.Err = err
+ for len(in) > 17 && fp.Err == nil {
+ length = int32(binary.BigEndian.Uint32(in[8:]))
+ length += 12 // for the int64 offset we skipped and int32 length field itself
+ if len(in) < int(length) {
+ break
  }
- o.processV1Messages(&fp, msgs, decompressor)
 
- default:
- batches, err := kmsg.ReadRecordBatches(rp.RecordBatches)
- if err != nil {
- fp.Err = err
+ switch magic := in[16]; magic {
+ case 0:
+ m := new(kmsg.MessageV0)
+ lengthField = &m.MessageSize
+ crcField = &m.CRC
+ crcTable = crc32.IEEETable
+ crcAt = 16
+ r = m
+ case 1:
+ m := new(kmsg.MessageV1)
+ lengthField = &m.MessageSize
+ crcField = &m.CRC
+ crcTable = crc32.IEEETable
+ crcAt = 16
+ r = m
+ case 2:
+ rb := new(kmsg.RecordBatch)
+ lengthField = &rb.Length
+ crcField = &rb.CRC
+ crcTable = crc32c
+ crcAt = 21
+ r = rb
+
  }
- var numPartitionRecords int
- for i := range batches {
- numPartitionRecords += int(batches[i].NumRecords)
+
+ if !check() {
+ break
  }
- fp.Records = make([]*Record, 0, numPartitionRecords)
- aborter := buildAborter(rp)
- for i := range batches {
- o.processRecordBatch(&fp, &batches[i], aborter, decompressor)
- if fp.Err != nil {
- break
- }
+
+ in = in[length:]
+
+ switch t := r.(type) {
+ case *kmsg.MessageV0:
+ o.processV0OuterMessage(&fp, t, decompressor)
+ case *kmsg.MessageV1:
+ o.processV1OuterMessage(&fp, t, decompressor)
+ case *kmsg.RecordBatch:
+ o.processRecordBatch(&fp, t, aborter, decompressor)
  }
  }
 
@@ -843,6 +904,24 @@ func (a aborter) trackAbortedPID(producerID int64) {
 // processing records to fetch part //
 //////////////////////////////////////
 
+// readRawRecords reads n records from in and returns them, returning
+// kbin.ErrNotEnoughData if in does not contain enough data.
+func readRawRecords(n int, in []byte) ([]kmsg.Record, error) {
+ rs := make([]kmsg.Record, n)
+ for i := 0; i < n; i++ {
+ length, used := kbin.Varint(in)
+ total := used + int(length)
+ if used == 0 || length < 0 || len(in) < total {
+ return nil, kbin.ErrNotEnoughData
+ }
+ if err := (&rs[i]).ReadFrom(in[:total]); err != nil {
+ return nil, err
+ }
+ in = in[total:]
+ }
+ return rs, nil
+}
+
 func (o *cursorOffsetNext) processRecordBatch(
  fp *FetchPartition,
  batch *kmsg.RecordBatch,
@@ -870,7 +949,7 @@ func (o *cursorOffsetNext) processRecordBatch(
  return
  }
 
- krecords, err := kmsg.ReadRecords(int(batch.NumRecords), rawRecords)
+ krecords, err := readRawRecords(int(batch.NumRecords), rawRecords)
  if err != nil {
  fp.Err = fmt.Errorf("invalid record batch: %v", err)
  return
@@ -903,36 +982,88 @@ func (o *cursorOffsetNext) processRecordBatch(
  }
 }
 
-func (o *cursorOffsetNext) processV1Messages(
+// Processes an outer v1 message. There could be no inner message, which makes
+// this easy, but if not, we decompress and process each inner message as
+// either v0 or v1. We only expect the inner message to be v1, but technically
+// a crazy pipeline could have v0 anywhere.
+func (o *cursorOffsetNext) processV1OuterMessage(
  fp *FetchPartition,
- messages []kmsg.MessageV1,
+ message *kmsg.MessageV1,
  decompressor *decompressor,
 ) {
- for i := range messages {
- message := &messages[i]
- compression := byte(message.Attributes & 0x0003)
- if compression == 0 {
- if !o.processV1Message(fp, message) {
- return
- }
- continue
+ compression := byte(message.Attributes & 0x0003)
+ if compression == 0 {
+ o.processV1Message(fp, message)
+ return
+ }
+
+ rawInner, err := decompressor.decompress(message.Value, compression)
+ if err != nil {
+ fp.Err = fmt.Errorf("unable to decompress messages: %v", err)
+ return
+ }
+
+ var innerMessages []readerFrom
+ for len(rawInner) > 17 { // magic at byte 17
+ length := int32(binary.BigEndian.Uint32(rawInner[8:]))
+ length += 12 // skip offset and length fields
+ if len(rawInner) < int(length) {
+ break
  }
 
- rawMessages, err := decompressor.decompress(message.Value, compression)
- if err != nil {
- fp.Err = fmt.Errorf("unable to decompress messages: %v", err)
- return
+ var (
+ magic = rawInner[16]
+
+ msg readerFrom
+ lengthField *int32
+ crcField *int32
+ )
+
+ switch magic {
+ case 0:
+ m := new(kmsg.MessageV0)
+ msg = m
+ lengthField = &m.MessageSize
+ crcField = &m.CRC
+ case 1:
+ m := new(kmsg.MessageV1)
+ msg = m
+ lengthField = &m.MessageSize
+ crcField = &m.CRC
+
+ default:
+ fp.Err = fmt.Errorf("message set v1 has inner message with invalid magic %d", magic)
+ break
  }
- innerMessages, err := kmsg.ReadV1Messages(rawMessages)
- if err != nil {
- fp.Err = err
+
+ if err := msg.ReadFrom(rawInner[:length]); err != nil {
+ break
  }
- if len(innerMessages) == 0 {
- return
+ if length := int32(len(rawInner[12:length])); length != *lengthField {
+ fp.Err = fmt.Errorf("encoded length %d does not match read length %d", *lengthField, length)
+ break
  }
- firstOffset := message.Offset - int64(len(innerMessages)) + 1
- for i := range innerMessages {
- innerMessage := &innerMessages[i]
+ if crcCalc := int32(crc32.ChecksumIEEE(rawInner[16:length])); crcCalc != *crcField {
+ fp.Err = fmt.Errorf("encoded crc %x does not match calculated crc %x", *crcField, crcCalc)
+ break
+ }
+ innerMessages = append(innerMessages, msg)
+ rawInner = rawInner[length:]
+ }
+ if len(innerMessages) == 0 {
+ return
+ }
+
+ firstOffset := message.Offset - int64(len(innerMessages)) + 1
+ for i := range innerMessages {
+ innerMessage := innerMessages[i]
+ switch innerMessage := innerMessage.(type) {
+ case *kmsg.MessageV0:
+ innerMessage.Offset = firstOffset + int64(i)
+ if !o.processV0Message(fp, innerMessage) {
+ return
+ }
+ case *kmsg.MessageV1:
  innerMessage.Offset = firstOffset + int64(i)
  if !o.processV1Message(fp, innerMessage) {
  return
@@ -958,40 +1089,57 @@ func (o *cursorOffsetNext) processV1Message(
  return true
 }
 
-func (o *cursorOffsetNext) processV0Messages(
+// Processes an outer v0 message. We expect inner messages to be entirely v0 as
+// well, so this only tries v0 always.
+func (o *cursorOffsetNext) processV0OuterMessage(
  fp *FetchPartition,
- messages []kmsg.MessageV0,
+ message *kmsg.MessageV0,
  decompressor *decompressor,
 ) {
- for i := range messages {
- message := &messages[i]
- compression := byte(message.Attributes & 0x0003)
- if compression == 0 {
- if !o.processV0Message(fp, message) {
- return
- }
- continue
- }
+ compression := byte(message.Attributes & 0x0003)
+ if compression == 0 {
+ o.processV0Message(fp, message)
+ return
+ }
 
- rawMessages, err := decompressor.decompress(message.Value, compression)
- if err != nil {
- fp.Err = fmt.Errorf("unable to decompress messages: %v", err)
- return
+ rawInner, err := decompressor.decompress(message.Value, compression)
+ if err != nil {
+ fp.Err = fmt.Errorf("unable to decompress messages: %v", err)
+ return
+ }
+
+ var innerMessages []kmsg.MessageV0
+ for len(rawInner) > 17 { // magic at byte 17
+ length := int32(binary.BigEndian.Uint32(rawInner[8:]))
+ length += 12 // skip offset and length fields
+ if len(rawInner) < int(length) {
+ break
  }
- innerMessages, err := kmsg.ReadV0Messages(rawMessages)
- if err != nil {
- fp.Err = err
+ var m kmsg.MessageV0
+ if err := m.ReadFrom(rawInner[:length]); err != nil {
+ break
  }
- if len(innerMessages) == 0 {
- return
+ if length := int32(len(rawInner[12:length])); length != m.MessageSize {
+ fp.Err = fmt.Errorf("encoded length %d does not match read length %d", m.MessageSize, length)
+ break
  }
- firstOffset := message.Offset - int64(len(innerMessages)) + 1
- for i := range innerMessages {
- innerMessage := &innerMessages[i]
- innerMessage.Offset = firstOffset + int64(i)
- if !o.processV0Message(fp, innerMessage) {
- return
- }
+ if crcCalc := int32(crc32.ChecksumIEEE(rawInner[16:length])); crcCalc != m.CRC {
+ fp.Err = fmt.Errorf("encoded crc %x does not match calculated crc %x", m.CRC, crcCalc)
+ break
+ }
+ innerMessages = append(innerMessages, m)
+ rawInner = rawInner[length:]
+ }
+ if len(innerMessages) == 0 {
+ return
+ }
+
+ firstOffset := message.Offset - int64(len(innerMessages)) + 1
+ for i := range innerMessages {
+ innerMessage := &innerMessages[i]
+ innerMessage.Offset = firstOffset + int64(i)
+ if !o.processV0Message(fp, innerMessage) {
+ return
  }
  }
 }