binman: Add a utility library for coreboot CBFS

Coreboot uses a simple flash-based filesystem called Coreboot Filesystem
(CBFS) to organise files used during boot. This allows files to be named
and their position in the flash to be set. It has special features for
dealing with x86 devices which typically memory-map their SPI flash to the
top of 32-bit address space and need a 'boot block' ending there.

Create a library to help create and read CBFS files. This includes a
writer class, a reader class and associated other helpers. Only a subset
of features are currently supported.

Signed-off-by: Simon Glass <sjg@chromium.org>

.travis.yml

@@ -32,6 +32,7 @@ addons:
     - device-tree-compiler
     - lzop
     - liblz4-tool
+    - lzma-alone
     - libisl15
     - clang-7
     - srecord

tools/binman/binman.py

@@ -62,6 +62,7 @@ def RunTests(debug, verbosity, processes, test_preserve_dirs, args, toolpath):
             name to execute (as in 'binman -t testSections', for example)
         toolpath: List of paths to use for tools
     """
+    import cbfs_util_test
     import elf_test
     import entry_test
     import fdt_test
@@ -90,7 +91,8 @@ def RunTests(debug, verbosity, processes, test_preserve_dirs, args, toolpath):
     suite = unittest.TestSuite()
     loader = unittest.TestLoader()
     for module in (entry_test.TestEntry, ftest.TestFunctional, fdt_test.TestFdt,
-                   elf_test.TestElf, image_test.TestImage):
+                   elf_test.TestElf, image_test.TestImage,
+                   cbfs_util_test.TestCbfs):
         # Test the test module about our arguments, if it is interested
         if hasattr(module, 'setup_test_args'):
             setup_test_args = getattr(module, 'setup_test_args')

tools/binman/cbfs_util.py

@@ -0,0 +1,720 @@
+# SPDX-License-Identifier: GPL-2.0+
+# Copyright 2019 Google LLC
+# Written by Simon Glass <sjg@chromium.org>
+
+"""Support for coreboot's CBFS format
+
+CBFS supports a header followed by a number of files, generally targeted at SPI
+flash.
+
+The format is somewhat defined by documentation in the coreboot tree although
+it is necessary to rely on the C structures and source code (mostly cbfstool)
+to fully understand it.
+
+Currently supported: raw and stage types with compression
+"""
+
+from __future__ import print_function
+
+from collections import OrderedDict
+import io
+import struct
+import sys
+
+import command
+import elf
+import tools
+
+# Set to True to enable printing output while working
+DEBUG = False
+
+# Set to True to enable output from running cbfstool for debugging
+VERBOSE = False
+
+# The master header, at the start of the CBFS
+HEADER_FORMAT      = '>IIIIIIII'
+HEADER_LEN         = 0x20
+HEADER_MAGIC       = 0x4f524243
+HEADER_VERSION1    = 0x31313131
+HEADER_VERSION2    = 0x31313132
+
+# The file header, at the start of each file in the CBFS
+FILE_HEADER_FORMAT = b'>8sIIII'
+FILE_HEADER_LEN    = 0x18
+FILE_MAGIC         = b'LARCHIVE'
+FILENAME_ALIGN     = 16  # Filename lengths are aligned to this
+
+# A stage header containing information about 'stage' files
+# Yes this is correct: this header is in litte-endian format
+STAGE_FORMAT       = '<IQQII'
+STAGE_LEN          = 0x1c
+
+# An attribute describring the compression used in a file
+ATTR_COMPRESSION_FORMAT = '>IIII'
+ATTR_COMPRESSION_LEN = 0x10
+
+# Attribute tags
+# Depending on how the header was initialised, it may be backed with 0x00 or
+# 0xff. Support both.
+FILE_ATTR_TAG_UNUSED        = 0
+FILE_ATTR_TAG_UNUSED2       = 0xffffffff
+FILE_ATTR_TAG_COMPRESSION   = 0x42435a4c
+FILE_ATTR_TAG_HASH          = 0x68736148
+FILE_ATTR_TAG_POSITION      = 0x42435350  # PSCB
+FILE_ATTR_TAG_ALIGNMENT     = 0x42434c41  # ALCB
+FILE_ATTR_TAG_PADDING       = 0x47444150  # PDNG
+
+# This is 'the size of bootblock reserved in firmware image (cbfs.txt)'
+# Not much more info is available, but we set it to 4, due to this comment in
+# cbfstool.c:
+# This causes 4 bytes to be left out at the end of the image, for two reasons:
+# 1. The cbfs master header pointer resides there
+# 2. Ssme cbfs implementations assume that an image that resides below 4GB has
+#    a bootblock and get confused when the end of the image is at 4GB == 0.
+MIN_BOOTBLOCK_SIZE     = 4
+
+# Files start aligned to this boundary in the CBFS
+ENTRY_ALIGN    = 0x40
+
+# CBFSs must declare an architecture since much of the logic is designed with
+# x86 in mind. The effect of setting this value is not well documented, but in
+# general x86 is used and this makes use of a boot block and an image that ends
+# at the end of 32-bit address space.
+ARCHITECTURE_UNKNOWN  = 0xffffffff
+ARCHITECTURE_X86      = 0x00000001
+ARCHITECTURE_ARM      = 0x00000010
+ARCHITECTURE_AARCH64  = 0x0000aa64
+ARCHITECTURE_MIPS     = 0x00000100
+ARCHITECTURE_RISCV    = 0xc001d0de
+ARCHITECTURE_PPC64    = 0x407570ff
+
+ARCH_NAMES = {
+    ARCHITECTURE_UNKNOWN  : 'unknown',
+    ARCHITECTURE_X86      : 'x86',
+    ARCHITECTURE_ARM      : 'arm',
+    ARCHITECTURE_AARCH64  : 'arm64',
+    ARCHITECTURE_MIPS     : 'mips',
+    ARCHITECTURE_RISCV    : 'riscv',
+    ARCHITECTURE_PPC64    : 'ppc64',
+    }
+
+# File types. Only supported ones are included here
+TYPE_CBFSHEADER     = 0x02   # Master header, HEADER_FORMAT
+TYPE_STAGE          = 0x10   # Stage, holding an executable, see STAGE_FORMAT
+TYPE_RAW            = 0x50   # Raw file, possibly compressed
+
+# Compression types
+COMPRESS_NONE, COMPRESS_LZMA, COMPRESS_LZ4 = range(3)
+
+COMPRESS_NAMES = {
+    COMPRESS_NONE : 'none',
+    COMPRESS_LZMA : 'lzma',
+    COMPRESS_LZ4  : 'lz4',
+    }
+
+def find_arch(find_name):
+    """Look up an architecture name
+
+    Args:
+        find_name: Architecture name to find
+
+    Returns:
+        ARCHITECTURE_... value or None if not found
+    """
+    for arch, name in ARCH_NAMES.items():
+        if name == find_name:
+            return arch
+    return None
+
+def find_compress(find_name):
+    """Look up a compression algorithm name
+
+    Args:
+        find_name: Compression algorithm name to find
+
+    Returns:
+        COMPRESS_... value or None if not found
+    """
+    for compress, name in COMPRESS_NAMES.items():
+        if name == find_name:
+            return compress
+    return None
+
+def align_int(val, align):
+    """Align a value up to the given alignment
+
+    Args:
+        val: Integer value to align
+        align: Integer alignment value (e.g. 4 to align to 4-byte boundary)
+
+    Returns:
+        integer value aligned to the required boundary, rounding up if necessary
+    """
+    return int((val + align - 1) / align) * align
+
+def _pack_string(instr):
+    """Pack a string to the required aligned size by adding padding
+
+    Args:
+        instr: String to process
+
+    Returns:
+        String with required padding (at least one 0x00 byte) at the end
+    """
+    val = tools.ToBytes(instr)
+    pad_len = align_int(len(val) + 1, FILENAME_ALIGN)
+    return val + tools.GetBytes(0, pad_len - len(val))
+
+
+class CbfsFile(object):
+    """Class to represent a single CBFS file
+
+    This is used to hold the information about a file, including its contents.
+    Use the get_data() method to obtain the raw output for writing to CBFS.
+
+    Properties:
+        name: Name of file
+        offset: Offset of file data from start of file header
+        data: Contents of file, uncompressed
+        data_len: Length of (possibly compressed) data in bytes
+        ftype: File type (TYPE_...)
+        compression: Compression type (COMPRESS_...)
+        memlen: Length of data in memory (typically the uncompressed length)
+        load: Load address in memory if known, else None
+        entry: Entry address in memory if known, else None. This is where
+            execution starts after the file is loaded
+        base_address: Base address to use for 'stage' files
+    """
+    def __init__(self, name, ftype, data, compress=COMPRESS_NONE):
+        self.name = name
+        self.offset = None
+        self.data = data
+        self.ftype = ftype
+        self.compress = compress
+        self.memlen = len(data)
+        self.load = None
+        self.entry = None
+        self.base_address = None
+        self.data_len = 0
+
+    def decompress(self):
+        """Handle decompressing data if necessary"""
+        indata = self.data
+        if self.compress == COMPRESS_LZ4:
+            data = tools.Decompress(indata, 'lz4')
+        elif self.compress == COMPRESS_LZMA:
+            data = tools.Decompress(indata, 'lzma')
+        else:
+            data = indata
+        self.memlen = len(data)
+        self.data = data
+        self.data_len = len(indata)
+
+    @classmethod
+    def stage(cls, base_address, name, data):
+        """Create a new stage file
+
+        Args:
+            base_address: Int base address for memory-mapping of ELF file
+            name: String file name to put in CBFS (does not need to correspond
+                to the name that the file originally came from)
+            data: Contents of file
+
+        Returns:
+            CbfsFile object containing the file information
+        """
+        cfile = CbfsFile(name, TYPE_STAGE, data)
+        cfile.base_address = base_address
+        return cfile
+
+    @classmethod
+    def raw(cls, name, data, compress):
+        """Create a new raw file
+
+        Args:
+            name: String file name to put in CBFS (does not need to correspond
+                to the name that the file originally came from)
+            data: Contents of file
+            compress: Compression algorithm to use (COMPRESS_...)
+
+        Returns:
+            CbfsFile object containing the file information
+        """
+        return CbfsFile(name, TYPE_RAW, data, compress)
+
+    def get_data(self):
+        """Obtain the contents of the file, in CBFS format
+
+        Returns:
+            bytes representing the contents of this file, packed and aligned
+                for directly inserting into the final CBFS output
+        """
+        name = _pack_string(self.name)
+        hdr_len = len(name) + FILE_HEADER_LEN
+        attr_pos = 0
+        content = b''
+        attr = b''
+        data = self.data
+        if self.ftype == TYPE_STAGE:
+            elf_data = elf.DecodeElf(data, self.base_address)
+            content = struct.pack(STAGE_FORMAT, self.compress,
+                                  elf_data.entry, elf_data.load,
+                                  len(elf_data.data), elf_data.memsize)
+            data = elf_data.data
+        elif self.ftype == TYPE_RAW:
+            orig_data = data
+            if self.compress == COMPRESS_LZ4:
+                data = tools.Compress(orig_data, 'lz4')
+            elif self.compress == COMPRESS_LZMA:
+                data = tools.Compress(orig_data, 'lzma')
+            attr = struct.pack(ATTR_COMPRESSION_FORMAT,
+                               FILE_ATTR_TAG_COMPRESSION, ATTR_COMPRESSION_LEN,
+                               self.compress, len(orig_data))
+        else:
+            raise ValueError('Unknown type %#x when writing\n' % self.ftype)
+        if attr:
+            attr_pos = hdr_len
+            hdr_len += len(attr)
+        hdr = struct.pack(FILE_HEADER_FORMAT, FILE_MAGIC,
+                          len(content) + len(data),
+                          self.ftype, attr_pos, hdr_len)
+        return hdr + name + attr + content + data
+
+
+class CbfsWriter(object):
+    """Class to handle writing a Coreboot File System (CBFS)
+
+    Usage is something like:
+
+        cbw = CbfsWriter(size)
+        cbw.add_file_raw('u-boot', tools.ReadFile('u-boot.bin'))
+        ...
+        data = cbw.get_data()
+
+    Attributes:
+        _master_name: Name of the file containing the master header
+        _size: Size of the filesystem, in bytes
+        _files: Ordered list of files in the CBFS, each a CbfsFile
+        _arch: Architecture of the CBFS (ARCHITECTURE_...)
+        _bootblock_size: Size of the bootblock, typically at the end of the CBFS
+        _erase_byte: Byte to use for empty space in the CBFS
+        _align: Alignment to use for files, typically ENTRY_ALIGN
+        _base_address: Boot block offset in bytes from the start of CBFS.
+            Typically this is located at top of the CBFS. It is 0 when there is
+            no boot block
+        _header_offset: Offset of master header in bytes from start of CBFS
+        _contents_offset: Offset of first file header
+        _hdr_at_start: True if the master header is at the start of the CBFS,
+            instead of the end as normal for x86
+        _add_fileheader: True to add a fileheader around the master header
+    """
+    def __init__(self, size, arch=ARCHITECTURE_X86):
+        """Set up a new CBFS
+
+        This sets up all properties to default values. Files can be added using
+        add_file_raw(), etc.
+
+        Args:
+            size: Size of CBFS in bytes
+            arch: Architecture to declare for CBFS
+        """
+        self._master_name = 'cbfs master header'
+        self._size = size
+        self._files = OrderedDict()
+        self._arch = arch
+        self._bootblock_size = 0
+        self._erase_byte = 0xff
+        self._align = ENTRY_ALIGN
+        self._add_fileheader = False
+        if self._arch == ARCHITECTURE_X86:
+            # Allow 4 bytes for the header pointer. That holds the
+            # twos-compliment negative offset of the master header in bytes
+            # measured from one byte past the end of the CBFS
+            self._base_address = self._size - max(self._bootblock_size,
+                                                  MIN_BOOTBLOCK_SIZE)
+            self._header_offset = self._base_address - HEADER_LEN
+            self._contents_offset = 0
+            self._hdr_at_start = False
+        else:
+            # For non-x86, different rules apply
+            self._base_address = 0
+            self._header_offset = align_int(self._base_address +
+                                            self._bootblock_size, 4)
+            self._contents_offset = align_int(self._header_offset +
+                                              FILE_HEADER_LEN +
+                                              self._bootblock_size, self._align)
+            self._hdr_at_start = True
+
+    def _skip_to(self, fd, offset):
+        """Write out pad bytes until a given offset
+
+        Args:
+            fd: File objext to write to
+            offset: Offset to write to
+        """
+        if fd.tell() > offset:
+            raise ValueError('No space for data before offset %#x (current offset %#x)' %
+                             (offset, fd.tell()))
+        fd.write(tools.GetBytes(self._erase_byte, offset - fd.tell()))
+
+    def _align_to(self, fd, align):
+        """Write out pad bytes until a given alignment is reached
+
+        This only aligns if the resulting output would not reach the end of the
+        CBFS, since we want to leave the last 4 bytes for the master-header
+        pointer.
+
+        Args:
+            fd: File objext to write to
+            align: Alignment to require (e.g. 4 means pad to next 4-byte
+                boundary)
+        """
+        offset = align_int(fd.tell(), align)
+        if offset < self._size:
+            self._skip_to(fd, offset)
+
+    def add_file_stage(self, name, data):
+        """Add a new stage file to the CBFS
+
+        Args:
+            name: String file name to put in CBFS (does not need to correspond
+                to the name that the file originally came from)
+            data: Contents of file
+
+        Returns:
+            CbfsFile object created
+        """
+        cfile = CbfsFile.stage(self._base_address, name, data)
+        self._files[name] = cfile
+        return cfile
+
+    def add_file_raw(self, name, data, compress=COMPRESS_NONE):
+        """Create a new raw file
+
+        Args:
+            name: String file name to put in CBFS (does not need to correspond
+                to the name that the file originally came from)
+            data: Contents of file
+            compress: Compression algorithm to use (COMPRESS_...)
+
+        Returns:
+            CbfsFile object created
+        """
+        cfile = CbfsFile.raw(name, data, compress)
+        self._files[name] = cfile
+        return cfile
+
+    def _write_header(self, fd, add_fileheader):
+        """Write out the master header to a CBFS
+
+        Args:
+            fd: File object
+            add_fileheader: True to place the master header in a file header
+                record
+        """
+        if fd.tell() > self._header_offset:
+            raise ValueError('No space for header at offset %#x (current offset %#x)' %
+                             (self._header_offset, fd.tell()))
+        if not add_fileheader:
+            self._skip_to(fd, self._header_offset)
+        hdr = struct.pack(HEADER_FORMAT, HEADER_MAGIC, HEADER_VERSION2,
+                          self._size, self._bootblock_size, self._align,
+                          self._contents_offset, self._arch, 0xffffffff)
+        if add_fileheader:
+            name = _pack_string(self._master_name)
+            fd.write(struct.pack(FILE_HEADER_FORMAT, FILE_MAGIC, len(hdr),
+                                 TYPE_CBFSHEADER, 0,
+                                 FILE_HEADER_LEN + len(name)))
+            fd.write(name)
+            self._header_offset = fd.tell()
+            fd.write(hdr)
+            self._align_to(fd, self._align)
+        else:
+            fd.write(hdr)
+
+    def get_data(self):
+        """Obtain the full contents of the CBFS
+
+        Thhis builds the CBFS with headers and all required files.
+
+        Returns:
+            'bytes' type containing the data
+        """
+        fd = io.BytesIO()
+
+        # THe header can go at the start in some cases
+        if self._hdr_at_start:
+            self._write_header(fd, add_fileheader=self._add_fileheader)
+        self._skip_to(fd, self._contents_offset)
+
+        # Write out each file
+        for cbf in self._files.values():
+            fd.write(cbf.get_data())
+            self._align_to(fd, self._align)
+        if not self._hdr_at_start:
+            self._write_header(fd, add_fileheader=self._add_fileheader)
+
+        # Pad to the end and write a pointer to the CBFS master header
+        self._skip_to(fd, self._base_address or self._size - 4)
+        rel_offset = self._header_offset - self._size
+        fd.write(struct.pack('<I', rel_offset & 0xffffffff))
+
+        return fd.getvalue()
+
+
+class CbfsReader(object):
+    """Class to handle reading a Coreboot File System (CBFS)
+
+    Usage is something like:
+        cbfs = cbfs_util.CbfsReader(data)
+        cfile = cbfs.files['u-boot']
+        self.WriteFile('u-boot.bin', cfile.data)
+
+    Attributes:
+        files: Ordered list of CbfsFile objects
+        align: Alignment to use for files, typically ENTRT_ALIGN
+        stage_base_address: Base address to use when mapping ELF files into the
+            CBFS for TYPE_STAGE files. If this is larger than the code address
+            of the ELF file, then data at the start of the ELF file will not
+            appear in the CBFS. Currently there are no tests for behaviour as
+            documentation is sparse
+        magic: Integer magic number from master header (HEADER_MAGIC)
+        version: Version number of CBFS (HEADER_VERSION2)
+        rom_size: Size of CBFS
+        boot_block_size: Size of boot block
+        cbfs_offset: Offset of the first file in bytes from start of CBFS
+        arch: Architecture of CBFS file (ARCHITECTURE_...)
+    """
+    def __init__(self, data, read=True):
+        self.align = ENTRY_ALIGN
+        self.arch = None
+        self.boot_block_size = None
+        self.cbfs_offset = None
+        self.files = OrderedDict()
+        self.magic = None
+        self.rom_size = None
+        self.stage_base_address = 0
+        self.version = None
+        self.data = data
+        if read:
+            self.read()
+
+    def read(self):
+        """Read all the files in the CBFS and add them to self.files"""
+        with io.BytesIO(self.data) as fd:
+            # First, get the master header
+            if not self._find_and_read_header(fd, len(self.data)):
+                raise ValueError('Cannot find master header')
+            fd.seek(self.cbfs_offset)
+
+            # Now read in the files one at a time
+            while True:
+                cfile = self._read_next_file(fd)
+                if cfile:
+                    self.files[cfile.name] = cfile
+                elif cfile is False:
+                    break
+
+    def _find_and_read_header(self, fd, size):
+        """Find and read the master header in the CBFS
+
+        This looks at the pointer word at the very end of the CBFS. This is an
+        offset to the header relative to the size of the CBFS, which is assumed
+        to be known. Note that the offset is in *little endian* format.
+
+        Args:
+            fd: File to read from
+            size: Size of file
+
+        Returns:
+            True if header was found, False if not
+        """
+        orig_pos = fd.tell()
+        fd.seek(size - 4)
+        rel_offset, = struct.unpack('<I', fd.read(4))
+        pos = (size + rel_offset) & 0xffffffff
+        fd.seek(pos)
+        found = self._read_header(fd)
+        if not found:
+            print('Relative offset seems wrong, scanning whole image')
+            for pos in range(0, size - HEADER_LEN, 4):
+                fd.seek(pos)
+                found = self._read_header(fd)
+                if found:
+                    break
+        fd.seek(orig_pos)
+        return found
+
+    def _read_next_file(self, fd):
+        """Read the next file from a CBFS
+
+        Args:
+            fd: File to read from
+
+        Returns:
+            CbfsFile object, if found
+            None if no object found, but data was parsed (e.g. TYPE_CBFSHEADER)
+            False if at end of CBFS and reading should stop
+        """
+        file_pos = fd.tell()
+        data = fd.read(FILE_HEADER_LEN)
+        if len(data) < FILE_HEADER_LEN:
+            print('File header at %x ran out of data' % file_pos)
+            return False
+        magic, size, ftype, attr, offset = struct.unpack(FILE_HEADER_FORMAT,
+                                                         data)
+        if magic != FILE_MAGIC:
+            return False
+        pos = fd.tell()
+        name = self._read_string(fd)
+        if name is None:
+            print('String at %x ran out of data' % pos)
+            return False
+
+        if DEBUG:
+            print('name', name)
+
+        # If there are attribute headers present, read those
+        compress = self._read_attr(fd, file_pos, attr, offset)
+        if compress is None:
+            return False
+
+        # Create the correct CbfsFile object depending on the type
+        cfile = None
+        fd.seek(file_pos + offset, io.SEEK_SET)
+        if ftype == TYPE_CBFSHEADER:
+            self._read_header(fd)
+        elif ftype == TYPE_STAGE:
+            data = fd.read(STAGE_LEN)
+            cfile = CbfsFile.stage(self.stage_base_address, name, b'')
+            (cfile.compress, cfile.entry, cfile.load, cfile.data_len,
+             cfile.memlen) = struct.unpack(STAGE_FORMAT, data)
+            cfile.data = fd.read(cfile.data_len)
+        elif ftype == TYPE_RAW:
+            data = fd.read(size)
+            cfile = CbfsFile.raw(name, data, compress)
+            cfile.decompress()
+            if DEBUG:
+                print('data', data)
+        else:
+            raise ValueError('Unknown type %#x when reading\n' % ftype)
+        if cfile:
+            cfile.offset = offset
+
+        # Move past the padding to the start of a possible next file. If we are
+        # already at an alignment boundary, then there is no padding.
+        pad = (self.align - fd.tell() % self.align) % self.align
+        fd.seek(pad, io.SEEK_CUR)
+        return cfile
+
+    @classmethod
+    def _read_attr(cls, fd, file_pos, attr, offset):
+        """Read attributes from the file
+
+        CBFS files can have attributes which are things that cannot fit into the
+        header. The only attribute currently supported is compression.
+
+        Args:
+            fd: File to read from
+            file_pos: Position of file in fd
+            attr: Offset of attributes, 0 if none
+            offset: Offset of file data (used to indicate the end of the
+                                         attributes)
+
+        Returns:
+            Compression to use for the file (COMPRESS_...)
+        """
+        compress = COMPRESS_NONE
+        if not attr:
+            return compress
+        attr_size = offset - attr
+        fd.seek(file_pos + attr, io.SEEK_SET)
+        while attr_size:
+            pos = fd.tell()
+            hdr = fd.read(8)
+            if len(hdr) < 8:
+                print('Attribute tag at %x ran out of data' % pos)
+                return None
+            atag, alen = struct.unpack(">II", hdr)
+            data = hdr + fd.read(alen - 8)
+            if atag == FILE_ATTR_TAG_COMPRESSION:
+                # We don't currently use this information
+                atag, alen, compress, _decomp_size = struct.unpack(
+                    ATTR_COMPRESSION_FORMAT, data)
+            else:
+                print('Unknown attribute tag %x' % atag)
+            attr_size -= len(data)
+        return compress
+
+    def _read_header(self, fd):
+        """Read the master header
+
+        Reads the header and stores the information obtained into the member
+        variables.
+
+        Args:
+            fd: File to read from
+
+        Returns:
+            True if header was read OK, False if it is truncated or has the
+                wrong magic or version
+        """
+        pos = fd.tell()
+        data = fd.read(HEADER_LEN)
+        if len(data) < HEADER_LEN:
+            print('Header at %x ran out of data' % pos)
+            return False
+        (self.magic, self.version, self.rom_size, self.boot_block_size,
+         self.align, self.cbfs_offset, self.arch, _) = struct.unpack(
+             HEADER_FORMAT, data)
+        return self.magic == HEADER_MAGIC and (
+            self.version == HEADER_VERSION1 or
+            self.version == HEADER_VERSION2)
+
+    @classmethod
+    def _read_string(cls, fd):
+        """Read a string from a file
+
+        This reads a string and aligns the data to the next alignment boundary
+
+        Args:
+            fd: File to read from
+
+        Returns:
+            string read ('str' type) encoded to UTF-8, or None if we ran out of
+                data
+        """
+        val = b''
+        while True:
+            data = fd.read(FILENAME_ALIGN)
+            if len(data) < FILENAME_ALIGN:
+                return None
+            pos = data.find(b'\0')
+            if pos == -1:
+                val += data
+            else:
+                val += data[:pos]
+                break
+        return val.decode('utf-8')
+
+
+def cbfstool(fname, *cbfs_args):
+    """Run cbfstool with provided arguments
+
+    If the tool fails then this function raises an exception and prints out the
+    output and stderr.
+
+    Args:
+        fname: Filename of CBFS
+        *cbfs_args: List of arguments to pass to cbfstool
+
+    Returns:
+        CommandResult object containing the results
+    """
+    args = ('cbfstool', fname) + cbfs_args
+    result = command.RunPipe([args], capture=not VERBOSE,
+                             capture_stderr=not VERBOSE, raise_on_error=False)
+    if result.return_code:
+        print(result.stderr, file=sys.stderr)
+        raise Exception("Failed to run (error %d): '%s'" %
+                        (result.return_code, ' '.join(args)))

tools/binman/cbfs_util_test.py

@@ -0,0 +1,540 @@
+#!/usr/bin/env python
+# SPDX-License-Identifier: GPL-2.0+
+# Copyright 2019 Google LLC
+# Written by Simon Glass <sjg@chromium.org>
+
+"""Tests for cbfs_util
+
+These create and read various CBFSs and compare the results with expected
+values and with cbfstool
+"""
+
+from __future__ import print_function
+
+import io
+import os
+import shutil
+import struct
+import tempfile
+import unittest
+
+import cbfs_util
+from cbfs_util import CbfsWriter
+import elf
+import test_util
+import tools
+
+U_BOOT_DATA           = b'1234'
+U_BOOT_DTB_DATA       = b'udtb'
+COMPRESS_DATA         = b'compress xxxxxxxxxxxxxxxxxxxxxx data'
+
+
+class TestCbfs(unittest.TestCase):
+    """Test of cbfs_util classes"""
+    #pylint: disable=W0212
+    @classmethod
+    def setUpClass(cls):
+        # Create a temporary directory for test files
+        cls._indir = tempfile.mkdtemp(prefix='cbfs_util.')
+        tools.SetInputDirs([cls._indir])
+
+        # Set up some useful data files
+        TestCbfs._make_input_file('u-boot.bin', U_BOOT_DATA)
+        TestCbfs._make_input_file('u-boot.dtb', U_BOOT_DTB_DATA)
+        TestCbfs._make_input_file('compress', COMPRESS_DATA)
+
+        # Set up a temporary output directory, used by the tools library when
+        # compressing files
+        tools.PrepareOutputDir(None)
+
+        cls.have_cbfstool = True
+        try:
+            tools.Run('which', 'cbfstool')
+        except:
+            cls.have_cbfstool = False
+
+        cls.have_lz4 = True
+        try:
+            tools.Run('lz4', '--no-frame-crc', '-c',
+                      tools.GetInputFilename('u-boot.bin'))
+        except:
+            cls.have_lz4 = False
+
+    @classmethod
+    def tearDownClass(cls):
+        """Remove the temporary input directory and its contents"""
+        if cls._indir:
+            shutil.rmtree(cls._indir)
+        cls._indir = None
+        tools.FinaliseOutputDir()
+
+    @classmethod
+    def _make_input_file(cls, fname, contents):
+        """Create a new test input file, creating directories as needed
+
+        Args:
+            fname: Filename to create
+            contents: File contents to write in to the file
+        Returns:
+            Full pathname of file created
+        """
+        pathname = os.path.join(cls._indir, fname)
+        tools.WriteFile(pathname, contents)
+        return pathname
+
+    def _check_hdr(self, data, size, offset=0, arch=cbfs_util.ARCHITECTURE_X86):
+        """Check that the CBFS has the expected header
+
+        Args:
+            data: Data to check
+            size: Expected ROM size
+            offset: Expected offset to first CBFS file
+            arch: Expected architecture
+
+        Returns:
+            CbfsReader object containing the CBFS
+        """
+        cbfs = cbfs_util.CbfsReader(data)
+        self.assertEqual(cbfs_util.HEADER_MAGIC, cbfs.magic)
+        self.assertEqual(cbfs_util.HEADER_VERSION2, cbfs.version)
+        self.assertEqual(size, cbfs.rom_size)
+        self.assertEqual(0, cbfs.boot_block_size)
+        self.assertEqual(cbfs_util.ENTRY_ALIGN, cbfs.align)
+        self.assertEqual(offset, cbfs.cbfs_offset)
+        self.assertEqual(arch, cbfs.arch)
+        return cbfs
+
+    def _check_uboot(self, cbfs, ftype=cbfs_util.TYPE_RAW, offset=0x38,
+                     data=U_BOOT_DATA):
+        """Check that the U-Boot file is as expected
+
+        Args:
+            cbfs: CbfsReader object to check
+            ftype: Expected file type
+            offset: Expected offset of file
+            data: Expected data in file
+
+        Returns:
+            CbfsFile object containing the file
+        """
+        self.assertIn('u-boot', cbfs.files)
+        cfile = cbfs.files['u-boot']
+        self.assertEqual('u-boot', cfile.name)
+        self.assertEqual(offset, cfile.offset)
+        self.assertEqual(data, cfile.data)
+        self.assertEqual(ftype, cfile.ftype)
+        self.assertEqual(cbfs_util.COMPRESS_NONE, cfile.compress)
+        self.assertEqual(len(data), cfile.memlen)
+        return cfile
+
+    def _check_dtb(self, cbfs, offset=0x38, data=U_BOOT_DTB_DATA):
+        """Check that the U-Boot dtb file is as expected
+
+        Args:
+            cbfs: CbfsReader object to check
+            offset: Expected offset of file
+            data: Expected data in file
+        """
+        self.assertIn('u-boot-dtb', cbfs.files)
+        cfile = cbfs.files['u-boot-dtb']
+        self.assertEqual('u-boot-dtb', cfile.name)
+        self.assertEqual(offset, cfile.offset)
+        self.assertEqual(U_BOOT_DTB_DATA, cfile.data)
+        self.assertEqual(cbfs_util.TYPE_RAW, cfile.ftype)
+        self.assertEqual(cbfs_util.COMPRESS_NONE, cfile.compress)
+        self.assertEqual(len(U_BOOT_DTB_DATA), cfile.memlen)
+
+    def _check_raw(self, data, size, offset=0, arch=cbfs_util.ARCHITECTURE_X86):
+        """Check that two raw files are added as expected
+
+        Args:
+            data: Data to check
+            size: Expected ROM size
+            offset: Expected offset to first CBFS file
+            arch: Expected architecture
+        """
+        cbfs = self._check_hdr(data, size, offset=offset, arch=arch)
+        self._check_uboot(cbfs)
+        self._check_dtb(cbfs)
+
+    def _get_expected_cbfs(self, size, arch='x86', compress=None):
+        """Get the file created by cbfstool for a particular scenario
+
+        Args:
+            size: Size of the CBFS in bytes
+            arch: Architecture of the CBFS, as a string
+            compress: Compression to use, e.g. cbfs_util.COMPRESS_LZMA
+
+        Returns:
+            Resulting CBFS file, or None if cbfstool is not available
+        """
+        if not self.have_cbfstool or not self.have_lz4:
+            return None
+        cbfs_fname = os.path.join(self._indir, 'test.cbfs')
+        cbfs_util.cbfstool(cbfs_fname, 'create', '-m', arch, '-s', '%#x' % size)
+        cbfs_util.cbfstool(cbfs_fname, 'add', '-n', 'u-boot', '-t', 'raw',
+                           '-c', compress and compress[0] or 'none',
+                           '-f', tools.GetInputFilename(
+                               compress and 'compress' or 'u-boot.bin'))
+        cbfs_util.cbfstool(cbfs_fname, 'add', '-n', 'u-boot-dtb', '-t', 'raw',
+                           '-c', compress and compress[1] or 'none',
+                           '-f', tools.GetInputFilename(
+                               compress and 'compress' or 'u-boot.dtb'))
+        return cbfs_fname
+
+    def _compare_expected_cbfs(self, data, cbfstool_fname):
+        """Compare against what cbfstool creates
+
+        This compares what binman creates with what cbfstool creates for what
+        is proportedly the same thing.
+
+        Args:
+            data: CBFS created by binman
+            cbfstool_fname: CBFS created by cbfstool
+        """
+        if not self.have_cbfstool or not self.have_lz4:
+            return
+        expect = tools.ReadFile(cbfstool_fname)
+        if expect != data:
+            tools.WriteFile('/tmp/expect', expect)
+            tools.WriteFile('/tmp/actual', data)
+            print('diff -y <(xxd -g1 /tmp/expect) <(xxd -g1 /tmp/actual) | colordiff')
+            self.fail('cbfstool produced a different result')
+
+    def test_cbfs_functions(self):
+        """Test global functions of cbfs_util"""
+        self.assertEqual(cbfs_util.ARCHITECTURE_X86, cbfs_util.find_arch('x86'))
+        self.assertIsNone(cbfs_util.find_arch('bad-arch'))
+
+        self.assertEqual(cbfs_util.COMPRESS_LZMA, cbfs_util.find_compress('lzma'))
+        self.assertIsNone(cbfs_util.find_compress('bad-comp'))
+
+    def test_cbfstool_failure(self):
+        """Test failure to run cbfstool"""
+        if not self.have_cbfstool:
+            self.skipTest('No cbfstool available')
+        try:
+            # In verbose mode this test fails since stderr is not captured. Fix
+            # this by turning off verbosity.
+            old_verbose = cbfs_util.VERBOSE
+            cbfs_util.VERBOSE = False
+            with test_util.capture_sys_output() as (_stdout, stderr):
+                with self.assertRaises(Exception) as e:
+                    cbfs_util.cbfstool('missing-file', 'bad-command')
+        finally:
+            cbfs_util.VERBOSE = old_verbose
+        self.assertIn('Unknown command', stderr.getvalue())
+        self.assertIn('Failed to run', str(e.exception))
+
+    def test_cbfs_raw(self):
+        """Test base handling of a Coreboot Filesystem (CBFS)"""
+        size = 0xb0
+        cbw = CbfsWriter(size)
+        cbw.add_file_raw('u-boot', U_BOOT_DATA)
+        cbw.add_file_raw('u-boot-dtb', U_BOOT_DTB_DATA)
+        data = cbw.get_data()
+        self._check_raw(data, size)
+        cbfs_fname = self._get_expected_cbfs(size=size)
+        self._compare_expected_cbfs(data, cbfs_fname)
+
+    def test_cbfs_invalid_file_type(self):
+        """Check handling of an invalid file type when outputiing a CBFS"""
+        size = 0xb0
+        cbw = CbfsWriter(size)
+        cfile = cbw.add_file_raw('u-boot', U_BOOT_DATA)
+
+        # Change the type manually before generating the CBFS, and make sure
+        # that the generator complains
+        cfile.ftype = 0xff
+        with self.assertRaises(ValueError) as e:
+            cbw.get_data()
+        self.assertIn('Unknown type 0xff when writing', str(e.exception))
+
+    def test_cbfs_invalid_file_type_on_read(self):
+        """Check handling of an invalid file type when reading the CBFS"""
+        size = 0xb0
+        cbw = CbfsWriter(size)
+        cbw.add_file_raw('u-boot', U_BOOT_DATA)
+
+        data = cbw.get_data()
+
+        # Read in the first file header
+        cbr = cbfs_util.CbfsReader(data, read=False)
+        with io.BytesIO(data) as fd:
+            self.assertTrue(cbr._find_and_read_header(fd, len(data)))
+            pos = fd.tell()
+            hdr_data = fd.read(cbfs_util.FILE_HEADER_LEN)
+            magic, size, ftype, attr, offset = struct.unpack(
+                cbfs_util.FILE_HEADER_FORMAT, hdr_data)
+
+        # Create a new CBFS with a change to the file type
+        ftype = 0xff
+        newdata = data[:pos]
+        newdata += struct.pack(cbfs_util.FILE_HEADER_FORMAT, magic, size, ftype,
+                               attr, offset)
+        newdata += data[pos + cbfs_util.FILE_HEADER_LEN:]
+
+        # Read in this CBFS and make sure that the reader complains
+        with self.assertRaises(ValueError) as e:
+            cbfs_util.CbfsReader(newdata)
+        self.assertIn('Unknown type 0xff when reading', str(e.exception))
+
+    def test_cbfs_no_space(self):
+        """Check handling of running out of space in the CBFS"""
+        size = 0x60
+        cbw = CbfsWriter(size)
+        cbw.add_file_raw('u-boot', U_BOOT_DATA)
+        with self.assertRaises(ValueError) as e:
+            cbw.get_data()
+        self.assertIn('No space for header', str(e.exception))
+
+    def test_cbfs_no_space_skip(self):
+        """Check handling of running out of space in CBFS with file header"""
+        size = 0x70
+        cbw = CbfsWriter(size)
+        cbw._add_fileheader = True
+        cbw.add_file_raw('u-boot', U_BOOT_DATA)
+        with self.assertRaises(ValueError) as e:
+            cbw.get_data()
+        self.assertIn('No space for data before offset', str(e.exception))
+
+    def test_cbfs_bad_header_ptr(self):
+        """Check handling of a bad master-header pointer"""
+        size = 0x70
+        cbw = CbfsWriter(size)
+        cbw.add_file_raw('u-boot', U_BOOT_DATA)
+        data = cbw.get_data()
+
+        # Add one to the pointer to make it invalid
+        newdata = data[:-4] + struct.pack('<I', cbw._header_offset + 1)
+
+        # We should still be able to find the master header by searching
+        with test_util.capture_sys_output() as (stdout, _stderr):
+            cbfs = cbfs_util.CbfsReader(newdata)
+        self.assertIn('Relative offset seems wrong', stdout.getvalue())
+        self.assertIn('u-boot', cbfs.files)
+        self.assertEqual(size, cbfs.rom_size)
+
+    def test_cbfs_bad_header(self):
+        """Check handling of a bad master header"""
+        size = 0x70
+        cbw = CbfsWriter(size)
+        cbw.add_file_raw('u-boot', U_BOOT_DATA)
+        data = cbw.get_data()
+
+        # Drop most of the header and try reading the modified CBFS
+        newdata = data[:cbw._header_offset + 4]
+
+        with test_util.capture_sys_output() as (stdout, _stderr):
+            with self.assertRaises(ValueError) as e:
+                cbfs_util.CbfsReader(newdata)
+        self.assertIn('Relative offset seems wrong', stdout.getvalue())
+        self.assertIn('Cannot find master header', str(e.exception))
+
+    def test_cbfs_bad_file_header(self):
+        """Check handling of a bad file header"""
+        size = 0x70
+        cbw = CbfsWriter(size)
+        cbw.add_file_raw('u-boot', U_BOOT_DATA)
+        data = cbw.get_data()
+
+        # Read in the CBFS master header (only), then stop
+        cbr = cbfs_util.CbfsReader(data, read=False)
+        with io.BytesIO(data) as fd:
+            self.assertTrue(cbr._find_and_read_header(fd, len(data)))
+            pos = fd.tell()
+
+        # Remove all but 4 bytes of the file headerm and try to read the file
+        newdata = data[:pos + 4]
+        with test_util.capture_sys_output() as (stdout, _stderr):
+            with io.BytesIO(newdata) as fd:
+                fd.seek(pos)
+                self.assertEqual(False, cbr._read_next_file(fd))
+        self.assertIn('File header at 0 ran out of data', stdout.getvalue())
+
+    def test_cbfs_bad_file_string(self):
+        """Check handling of an incomplete filename string"""
+        size = 0x70
+        cbw = CbfsWriter(size)
+        cbw.add_file_raw('16-characters xx', U_BOOT_DATA)
+        data = cbw.get_data()
+
+        # Read in the CBFS master header (only), then stop
+        cbr = cbfs_util.CbfsReader(data, read=False)
+        with io.BytesIO(data) as fd:
+            self.assertTrue(cbr._find_and_read_header(fd, len(data)))
+            pos = fd.tell()
+
+        # Create a new CBFS with only the first 16 bytes of the file name, then
+        # try to read the file
+        newdata = data[:pos + cbfs_util.FILE_HEADER_LEN + 16]
+        with test_util.capture_sys_output() as (stdout, _stderr):
+            with io.BytesIO(newdata) as fd:
+                fd.seek(pos)
+                self.assertEqual(False, cbr._read_next_file(fd))
+        self.assertIn('String at %x ran out of data' %
+                      cbfs_util.FILE_HEADER_LEN, stdout.getvalue())
+
+    def test_cbfs_debug(self):
+        """Check debug output"""
+        size = 0x70
+        cbw = CbfsWriter(size)
+        cbw.add_file_raw('u-boot', U_BOOT_DATA)
+        data = cbw.get_data()
+
+        try:
+            cbfs_util.DEBUG = True
+            with test_util.capture_sys_output() as (stdout, _stderr):
+                cbfs_util.CbfsReader(data)
+            self.assertEqual('name u-boot\ndata %s\n' % U_BOOT_DATA,
+                             stdout.getvalue())
+        finally:
+            cbfs_util.DEBUG = False
+
+    def test_cbfs_bad_attribute(self):
+        """Check handling of bad attribute tag"""
+        if not self.have_lz4:
+            self.skipTest('lz4 --no-frame-crc not available')
+        size = 0x140
+        cbw = CbfsWriter(size)
+        cbw.add_file_raw('u-boot', COMPRESS_DATA,
+                         compress=cbfs_util.COMPRESS_LZ4)
+        data = cbw.get_data()
+
+        # Search the CBFS for the expected compression tag
+        with io.BytesIO(data) as fd:
+            while True:
+                pos = fd.tell()
+                tag, = struct.unpack('>I', fd.read(4))
+                if tag == cbfs_util.FILE_ATTR_TAG_COMPRESSION:
+                    break
+
+        # Create a new CBFS with the tag changed to something invalid
+        newdata = data[:pos] + struct.pack('>I', 0x123) + data[pos + 4:]
+        with test_util.capture_sys_output() as (stdout, _stderr):
+            cbfs_util.CbfsReader(newdata)
+        self.assertEqual('Unknown attribute tag 123\n', stdout.getvalue())
+
+    def test_cbfs_missing_attribute(self):
+        """Check handling of an incomplete attribute tag"""
+        if not self.have_lz4:
+            self.skipTest('lz4 --no-frame-crc not available')
+        size = 0x140
+        cbw = CbfsWriter(size)
+        cbw.add_file_raw('u-boot', COMPRESS_DATA,
+                         compress=cbfs_util.COMPRESS_LZ4)
+        data = cbw.get_data()
+
+        # Read in the CBFS master header (only), then stop
+        cbr = cbfs_util.CbfsReader(data, read=False)
+        with io.BytesIO(data) as fd:
+            self.assertTrue(cbr._find_and_read_header(fd, len(data)))
+            pos = fd.tell()
+
+        # Create a new CBFS with only the first 4 bytes of the compression tag,
+        # then try to read the file
+        tag_pos = pos + cbfs_util.FILE_HEADER_LEN + cbfs_util.FILENAME_ALIGN
+        newdata = data[:tag_pos + 4]
+        with test_util.capture_sys_output() as (stdout, _stderr):
+            with io.BytesIO(newdata) as fd:
+                fd.seek(pos)
+                self.assertEqual(False, cbr._read_next_file(fd))
+        self.assertIn('Attribute tag at %x ran out of data' % tag_pos,
+                      stdout.getvalue())
+
+    def test_cbfs_file_master_header(self):
+        """Check handling of a file containing a master header"""
+        size = 0x100
+        cbw = CbfsWriter(size)
+        cbw._add_fileheader = True
+        cbw.add_file_raw('u-boot', U_BOOT_DATA)
+        data = cbw.get_data()
+
+        cbr = cbfs_util.CbfsReader(data)
+        self.assertIn('u-boot', cbr.files)
+        self.assertEqual(size, cbr.rom_size)
+
+    def test_cbfs_arch(self):
+        """Test on non-x86 architecture"""
+        size = 0x100
+        cbw = CbfsWriter(size, arch=cbfs_util.ARCHITECTURE_PPC64)
+        cbw.add_file_raw('u-boot', U_BOOT_DATA)
+        cbw.add_file_raw('u-boot-dtb', U_BOOT_DTB_DATA)
+        data = cbw.get_data()
+        self._check_raw(data, size, offset=0x40,
+                        arch=cbfs_util.ARCHITECTURE_PPC64)
+
+        # Compare against what cbfstool creates
+        cbfs_fname = self._get_expected_cbfs(size=size, arch='ppc64')
+        self._compare_expected_cbfs(data, cbfs_fname)
+
+    def test_cbfs_stage(self):
+        """Tests handling of a Coreboot Filesystem (CBFS)"""
+        if not elf.ELF_TOOLS:
+            self.skipTest('Python elftools not available')
+        elf_fname = os.path.join(self._indir, 'cbfs-stage.elf')
+        elf.MakeElf(elf_fname, U_BOOT_DATA, U_BOOT_DTB_DATA)
+
+        size = 0xb0
+        cbw = CbfsWriter(size)
+        cbw.add_file_stage('u-boot', tools.ReadFile(elf_fname))
+
+        data = cbw.get_data()
+        cbfs = self._check_hdr(data, size)
+        load = 0xfef20000
+        entry = load + 2
+
+        cfile = self._check_uboot(cbfs, cbfs_util.TYPE_STAGE, offset=0x28,
+                                  data=U_BOOT_DATA + U_BOOT_DTB_DATA)
+
+        self.assertEqual(entry, cfile.entry)
+        self.assertEqual(load, cfile.load)
+        self.assertEqual(len(U_BOOT_DATA) + len(U_BOOT_DTB_DATA),
+                         cfile.data_len)
+
+        # Compare against what cbfstool creates
+        if self.have_cbfstool:
+            cbfs_fname = os.path.join(self._indir, 'test.cbfs')
+            cbfs_util.cbfstool(cbfs_fname, 'create', '-m', 'x86', '-s',
+                               '%#x' % size)
+            cbfs_util.cbfstool(cbfs_fname, 'add-stage', '-n', 'u-boot',
+                               '-f', elf_fname)
+            self._compare_expected_cbfs(data, cbfs_fname)
+
+    def test_cbfs_raw_compress(self):
+        """Test base handling of compressing raw files"""
+        if not self.have_lz4:
+            self.skipTest('lz4 --no-frame-crc not available')
+        size = 0x140
+        cbw = CbfsWriter(size)
+        cbw.add_file_raw('u-boot', COMPRESS_DATA,
+                         compress=cbfs_util.COMPRESS_LZ4)
+        cbw.add_file_raw('u-boot-dtb', COMPRESS_DATA,
+                         compress=cbfs_util.COMPRESS_LZMA)
+        data = cbw.get_data()
+
+        cbfs = self._check_hdr(data, size)
+        self.assertIn('u-boot', cbfs.files)
+        cfile = cbfs.files['u-boot']
+        self.assertEqual(cfile.name, 'u-boot')
+        self.assertEqual(cfile.offset, 56)
+        self.assertEqual(cfile.data, COMPRESS_DATA)
+        self.assertEqual(cfile.ftype, cbfs_util.TYPE_RAW)
+        self.assertEqual(cfile.compress, cbfs_util.COMPRESS_LZ4)
+        self.assertEqual(cfile.memlen, len(COMPRESS_DATA))
+
+        self.assertIn('u-boot-dtb', cbfs.files)
+        cfile = cbfs.files['u-boot-dtb']
+        self.assertEqual(cfile.name, 'u-boot-dtb')
+        self.assertEqual(cfile.offset, 56)
+        self.assertEqual(cfile.data, COMPRESS_DATA)
+        self.assertEqual(cfile.ftype, cbfs_util.TYPE_RAW)
+        self.assertEqual(cfile.compress, cbfs_util.COMPRESS_LZMA)
+        self.assertEqual(cfile.memlen, len(COMPRESS_DATA))
+
+        cbfs_fname = self._get_expected_cbfs(size=size, compress=['lz4', 'lzma'])
+        self._compare_expected_cbfs(data, cbfs_fname)
+
+
+if __name__ == '__main__':
+    unittest.main()