t5de-api/cfl/cfl.go

// Package cfl reads and writes CFL files, IMVU's product container: a product's files in one archive, as the
// client downloads and caches them.
//
// Layout (little-endian): "CFL3", u32 directory offset, u32 directory size (uncompressed). Each entry's data is a
// u32 stored length and the stored bytes. At the directory offset: u32 compression flag, u32 stored length, the
// directory. Each directory entry: i32 size, i32 offset (of the entry's length field), i32 compression flag, i16
// name length, the name (Latin-1). "DFL3" files are the same, with each directory entry followed by an i32 length
// and that many bytes of content hash.
package cfl

import (
	"bytes"
	"encoding/binary"
	"errors"
	"fmt"
	"io"
	"math"
	"unicode/utf8"
)

// Compression flags.
const (
	None = 0
	LZMA = 4 // pylzma's format: 5 property bytes, then the raw stream
)

// Entry is one file in a CFL.
type Entry struct {
	Name string
	Data []byte
}

// ErrLZMA is returned for CFLs with LZMA-compressed entries or directories.
var ErrLZMA = errors.New("cfl: LZMA-compressed CFLs aren't supported")

const headerSize = 12

// Write writes entries as a CFL, in order and uncompressed. The output depends only on the entries.
func Write(w io.Writer, entries []Entry) error {
	var body, dir bytes.Buffer
	offset := int64(headerSize)
	for _, e := range entries {
		name, err := Latin1(e.Name)
		if err != nil {
			return err
		}
		if len(name) == 0 {
			return errors.New("cfl: empty entry name")
		}
		if len(name) > math.MaxInt16 {
			return fmt.Errorf("cfl: entry name %q is too long", e.Name)
		}
		if offset+4+int64(len(e.Data)) > math.MaxInt32 {
			return errors.New("cfl: entries over 2 GiB")
		}
		le32(&dir, uint32(len(e.Data)))
		le32(&dir, uint32(offset))
		le32(&dir, None)
		binary.Write(&dir, binary.LittleEndian, int16(len(name)))
		dir.Write(name)

		le32(&body, uint32(len(e.Data)))
		body.Write(e.Data)
		offset += 4 + int64(len(e.Data))
	}
	if offset+8+int64(dir.Len()) > math.MaxInt32 {
		return errors.New("cfl: entries over 2 GiB")
	}

	var head bytes.Buffer
	head.WriteString("CFL3")
	le32(&head, uint32(offset))
	le32(&head, uint32(dir.Len()))
	le32(&body, None)
	le32(&body, uint32(dir.Len()))
	for _, b := range [][]byte{head.Bytes(), body.Bytes(), dir.Bytes()} {
		if _, err := w.Write(b); err != nil {
			return err
		}
	}
	return nil
}

func le32(b *bytes.Buffer, v uint32) {
	b.Write(binary.LittleEndian.AppendUint32(nil, v))
}

// Read reads a CFL (or DFL) whose entries and directory are uncompressed. Entries' Data share memory with data.
func Read(data []byte) ([]Entry, error) {
	if len(data) < headerSize {
		return nil, errors.New("cfl: too short")
	}
	var hashes bool
	switch string(data[:4]) {
	case "CFL3":
	case "DFL3":
		hashes = true
	default:
		return nil, fmt.Errorf("cfl: unknown header %q", data[:4])
	}
	dirOffset := int64(binary.LittleEndian.Uint32(data[4:]))
	dirSize := int64(binary.LittleEndian.Uint32(data[8:]))

	r := reader{data: data, pos: dirOffset}
	flag := r.u32()
	stored := int64(r.u32())
	if r.err != nil {
		return nil, fmt.Errorf("cfl: directory offset %d is past the end", dirOffset)
	}
	switch flag {
	case None:
	case LZMA:
		return nil, ErrLZMA
	default:
		return nil, fmt.Errorf("cfl: directory has unknown compression %d", flag)
	}
	if stored != dirSize {
		return nil, fmt.Errorf("cfl: directory is %d bytes, header says %d", stored, dirSize)
	}
	dir := r.bytes(stored)
	if r.err != nil {
		return nil, errors.New("cfl: directory runs past the end")
	}

	var entries []Entry
	d := reader{data: dir}
	for d.pos < int64(len(dir)) {
		size := int64(int32(d.u32()))
		offset := int64(int32(d.u32()))
		flag := d.u32()
		nameLen := int64(int16(d.u16()))
		if nameLen < 0 {
			return nil, fmt.Errorf("cfl: directory entry %d has a negative name length", len(entries))
		}
		name := d.bytes(nameLen)
		if hashes {
			hashLen := int64(int32(d.u32()))
			if hashLen < 0 {
				return nil, fmt.Errorf("cfl: directory entry %d has a negative hash length", len(entries))
			}
			d.bytes(hashLen)
		}
		if d.err != nil {
			return nil, fmt.Errorf("cfl: directory entry %d is cut off", len(entries))
		}
		if len(name) == 0 {
			return nil, fmt.Errorf("cfl: directory entry %d has no name", len(entries))
		}
		n := latin1String(name)
		switch flag {
		case None:
		case LZMA:
			return nil, ErrLZMA
		default:
			return nil, fmt.Errorf("cfl: %q has unknown compression %d", n, flag)
		}
		if size < 0 || offset < 0 {
			return nil, fmt.Errorf("cfl: %q has a negative size or offset", n)
		}
		e := reader{data: data, pos: offset}
		stored := int64(e.u32())
		body := e.bytes(stored)
		if e.err != nil {
			return nil, fmt.Errorf("cfl: %q runs past the end", n)
		}
		if stored != size {
			return nil, fmt.Errorf("cfl: %q is %d bytes, directory says %d", n, stored, size)
		}
		entries = append(entries, Entry{Name: n, Data: body})
	}
	return entries, nil
}

// reader reads little-endian values from data, setting err instead of reading past the end.
type reader struct {
	data []byte
	pos  int64
	err  error
}

func (r *reader) bytes(n int64) []byte {
	if r.err != nil || n < 0 || r.pos < 0 || r.pos > int64(len(r.data)) || n > int64(len(r.data))-r.pos {
		r.err = io.ErrUnexpectedEOF
		return nil
	}
	b := r.data[r.pos : r.pos+n : r.pos+n]
	r.pos += n
	return b
}

func (r *reader) u32() uint32 {
	b := r.bytes(4)
	if b == nil {
		return 0
	}
	return binary.LittleEndian.Uint32(b)
}

func (r *reader) u16() uint16 {
	b := r.bytes(2)
	if b == nil {
		return 0
	}
	return binary.LittleEndian.Uint16(b)
}

// Latin1 encodes s as Latin-1, as CFLs store names.
func Latin1(s string) ([]byte, error) {
	if !utf8.ValidString(s) {
		return nil, fmt.Errorf("cfl: name %q isn't valid UTF-8", s)
	}
	b := make([]byte, 0, len(s))
	for _, r := range s {
		if r > 0xff {
			return nil, fmt.Errorf("cfl: name %q has characters outside Latin-1", s)
		}
		b = append(b, byte(r))
	}
	return b, nil
}

func latin1String(b []byte) string {
	r := make([]rune, len(b))
	for i, c := range b {
		r[i] = rune(c)
	}
	return string(r)
}