Skip to content

zstd: Data corruption on SpeedBestCompression level #875

@MichaelEischer

Description

@MichaelEischer

Following a bug report at restic/restic#4523, I've traced the data corruption back to the zstd library.

The following two minimized data chunks return corrupt data after compressing it at SpeedBestCompression level and decompressing it afterwards. There is no data corruption at the default compression level. I've tested versions v1.16.7 (via restic using Go 1.20.7) and v1.17.1 (using Go 1.21.3).

correct-3d0e366bad4a5e9b443f407b114756a0f5a8153dbc242e0b2601c707136815eb.bin-minimized.txt
correct-dc82d97be7683ecd41097ab02d7b15de81e8bbcd1c476c50b254b1f458090929.bin-minimized.txt

To produce the minimized examples, I've used the following code snippet, which also serves as a reproducer:

package main

import (
	"bytes"
	"os"
	"sort"

	"github.com/klauspost/compress/zstd"
)

func buildEncoder() *zstd.Encoder {
	level := zstd.SpeedBestCompression

	opts := []zstd.EOption{
		// Set the compression level configured.
		zstd.WithEncoderLevel(level),
		// Disable CRC, we have enough checks in place, makes the
		// compressed data four bytes shorter.
		zstd.WithEncoderCRC(false),
		// Set a window of 512kbyte, so we have good lookbehind for usual
		// blob sizes.
		zstd.WithWindowSize(512 * 1024),
	}

	enc, err := zstd.NewWriter(nil, opts...)
	if err != nil {
		panic(err)
	}
	return enc
}

func buildDecoder() *zstd.Decoder {
	opts := []zstd.DOption{
		// Use all available cores.
		zstd.WithDecoderConcurrency(0),
		// Limit the maximum decompressed memory. Set to a very high,
		// conservative value.
		zstd.WithDecoderMaxMemory(16 * 1024 * 1024 * 1024),
	}

	dec, err := zstd.NewReader(nil, opts...)
	if err != nil {
		panic(err)
	}
	return dec
}

var enc = buildEncoder()
var dec = buildDecoder()

func verifyCompression(data []byte) bool {
	compressed := enc.EncodeAll(data, nil)
	decompressed, err := dec.DecodeAll(compressed, make([]byte, 0, len(data)))
	if err != nil {
		panic(err)
	}

	return bytes.Equal(data, decompressed)
}

func main() {
	data, err := os.ReadFile(os.Args[1])
	if err != nil {
		panic(err)
	}

	idx := sort.Search(len(data), func(i int) bool {
		return !verifyCompression(data[:i])
	})
	if verifyCompression(data[:idx]) {
		panic("missing compression error")
	}

	startIdx := sort.Search(idx, func(i int) bool {
		return verifyCompression(data[i:idx])
	}) - 1
	if verifyCompression(data[startIdx:idx]) {
		panic("missing compression error")
	}

	println("minimal example from", startIdx, "to", idx)

	f, err := os.OpenFile(os.Args[1]+"-minimized", os.O_WRONLY|os.O_CREATE|os.O_EXCL, 0666)
	if err != nil {
		panic(err)
	}
	_, err = f.Write(data[startIdx:idx])
	if err != nil {
		panic(err)
	}
	err = f.Close()
	if err != nil {
		panic(err)
	}
}

Run it using

go mod init test
go mod tidy
go run main.go correct-3d0e366bad4a5e9b443f407b114756a0f5a8153dbc242e0b2601c707136815eb.bin

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions