The .lzma File Format --------------------- 0. Preface 0.1. Copyright Notices 0.2. Changes 1. Conventions 1.1. Byte and Its Representation 1.2. Multibyte Integers 2. Stream 2.1. Stream Types 2.1.1. Single-Block Stream 2.1.2. Multi-Block Stream 2.2. Stream Header 2.2.1. Header Magic Bytes 2.2.2. Stream Flags 2.2.3. CRC32 3. Block 3.1. Block Header 3.1.1. Block Flags 3.1.2. Compressed Size 3.1.3. Uncompressed Size 3.1.4. List of Filter Flags 3.1.4.1. Misc 3.1.4.2. External ID 3.1.4.3. External Size of Properties 3.1.4.4. Filter Properties 3.1.5. CRC32 3.1.6. Header Padding 3.2. Compressed Data 3.3. Block Footer 3.3.1. Check 3.3.2. Stream Footer 3.3.2.1. Uncompressed Size 3.3.2.2. Backward Size 3.3.2.3. Stream Flags 3.3.2.4. Footer Magic Bytes 3.3.3. Footer Padding 4. Filters 4.1. Detecting when All Data Has Been Decoded 4.1.1. With Uncompressed Size 4.1.2. With End of Input 4.1.3. With End of Payload Marker 4.2. Alignment 4.3. Filters 4.3.1. Copy 4.3.2. Subblock 4.3.2.1. Format of the Encoded Output 4.3.3. Delta 4.3.3.1. Format of the Encoded Output 4.3.4. LZMA 4.3.4.1. LZMA Properties 4.3.4.2. Dictionary Flags 4.3.5. Branch/Call/Jump Filters for Executables 5. Metadata 5.1. Metadata Flags 5.2. Size of Header Metadata Block 5.3. Total Size 5.4. Uncompressed Size 5.5. Index 5.5.1. Number of Data Blocks 5.5.2. Total Sizes 5.5.3. Uncompressed Sizes 5.6. Extra 5.6.1. 0x00: Dummy/Padding 5.6.2. 0x01: OpenPGP Signature 5.6.3. 0x02: Filter Information 5.6.4. 0x03: Comment 5.6.5. 0x04: List of Checks 5.6.6. 0x05: Original Filename 5.6.7. 0x07: Modification Time 5.6.8. 0x09: High-Resolution Modification Time 5.6.9. 0x0B: MIME Type 5.6.10. 0x0D: Homepage URL 6. Custom Filter and Extra Record IDs 6.1. Reserved Custom Filter ID Ranges 7. Cyclic Redundancy Checks 8. References 8.1. Normative References 8.2. Informative References 0. Preface This document describes the .lzma file format (filename suffix `.lzma', MIME type `application/x-lzma'). It is intended that this format replace the format used by the LZMA_Alone tool included in LZMA SDK up to and including version 4.57. IMPORTANT: The version described in this document is a draft, NOT a final, official version. Changes are possible. 0.1. Copyright Notices Copyright (C) 2006, 2007 Lasse Collin Copyright (C) 2006 Ville Koskinen Copying and distribution of this file, with or without modification, are permitted in any medium without royalty provided the copyright notice and this notice are preserved. Modified versions must be marked as such. All source code examples given in this document are put into the public domain by the authors of this document. Thanks for helping with this document goes to Igor Pavlov, Mark Adler and Mikko Pouru. 0.2. Changes Last modified: 2008-02-01 19:25+0200 (A changelog will be kept once the first official version is made.) 1. Conventions The keywords `must', `must not', `required', `should', `should not', `recommended', `may', and `optional' in this document are to be interpreted as described in [RFC-2119]. These words are not capitalized in this document. Indicating a warning means displaying a message, returning appropriate exit status, or something else to let the user know that something worth warning occurred. The operation should still finish if a warning is indicated. Indicating an error means displaying a message, returning appropriate exit status, or something else to let the user know that something prevented successfully finishing the operation. The operation must be aborted once an error has been indicated. 1.1. Byte and Its Representation In this document, byte is always 8 bits. A `nul byte' has all bits unset. That is, the value of a nul byte is 0x00. To represent byte blocks, this document uses notation that is similar to the notation used in [RFC-1952]: +-------+ | Foo | One byte. +-------+ +---+---+ | Foo | Two bytes; that is, some of the vertical bars +---+---+ can be missing. +=======+ | Foo | Zero or more bytes. +=======+ In this document, a boxed byte or a byte sequence declared using this notation is called `a field'. The example field above would be called called `the Foo field' or plain `Foo'. 1.2. Multibyte Integers Multibyte integers of static length, such as CRC values, are stored in little endian byte order (least significant byte first). When smaller values are more likely than bigger values (e.g. file sizes), multibyte integers are encoded in a simple variable-length representation: - Numbers in the range [0, 127] are copied as is, and take one byte of space. - Bigger numbers will occupy two or more bytes. The lowest seven bits of every byte are used for data; the highest (eighth) bit indicates either that 0) the byte is in the middle of the byte sequence, or 1) the byte is the first or the last byte. For now, the value of the variable-length integers is limited to 63 bits, which limits the encoded size of the integer to nine bytes. These limits may be increased in future if needed. Note that the encoding is not as optimal as it could be. For example, it is possible to encode the number 42 using any number of bytes between one and nine. This is convenient for non-streamed encoders, that write Compressed Size or Uncompressed Size fields to the Block Header (see Section 3.1) after the Compressed Data field is written to the disk. In several situations, the decoder needs to compare that two fields contain identical information. When comparing fields using the encoding described in this Section, the decoder must consider two fields identical if their decoded values are identical; it does not matter if the encoded variable-length representations differ. The following C code illustrates encoding and decoding 63-bit variables; the highest bit of uint64_t must be unset. The functions return the number of bytes occupied by the integer (1-9), or zero on error. #include #include size_t encode(uint8_t buf[static 9], uint64_t num) { if (num >= (UINT64_C(1) << (9 * 7))) return 0; if (num <= 0x7F) { buf[0] = num; return 1; } buf[0] = (num & 0x7F) | 0x80; num >>= 7; size_t i = 1; while (num >= 0x80) { buf[i++] = num & 0x7F; num >>= 7; } buf[i++] = num | 0x80; return i; } size_t decode(const uint8_t buf[], size_t size_max, uint64_t *num) { if (size_max == 0) return 0; if (size_max > 9) size_max = 9; *num = buf[0] & 0x7F; if (!(buf[0] & 0x80)) return 1; size_t i = 1; do { if (i == size_max) return 0; *num |= (uint64_t)(buf[i] & 0x7F) << (7 * i); } while (!(buf[i++] & 0x80)); return i; } size_t decode_reverse(const uint8_t buf[], size_t size_max, uint64_t *num) { if (size_max == 0) return 0; const size_t end = size_max > 9 ? size_max - 9 : 0; size_t i = size_max - 1; *num = buf[i] & 0x7F; if (!(buf[i] & 0x80)) return 1; do { if (i-- == end) return 0; *num <<= 7; *num |= buf[i] & 0x7F; } while (!(buf[i] & 0x80)); return size_max - i; } 2. Stream +========+========+========+ | Stream | Stream | Stream | ... +========+========+========+ A file contains usually only one Stream. However, it is possible to concatenate multiple Streams together with no additional processing. It is up to the implementation to decide if the decoder will continue decoding from the next Stream once the end of the first Stream has been reached. 2.1. Stream Types There are two types of Streams: Single-Block Streams and Multi-Block Streams. Decoders conforming to this specification must support at least Single-Block Streams. Supporting Multi-Block Streams is optional. If the decoder supports only Single-Block Streams, the documentation of the decoder should mention this fact clearly. 2.1.1. Single-Block Stream +===============+============+ | Stream Header | Data Block | +===============+============+ As the name says, a Single-Block Stream has exactly one Block. The Block must be a Data Block; Metadata Blocks are not allowed in Single-Block Streams. 2.1.2. Multi-Block Stream +===============+=======================+ | Stream Header | Header Metadata Block | +===============+=======================+ +============+ +============+=======================+ ---> | Data Block | ... | Data Block | Footer Metadata Block | +============+ +============+=======================+ Notes: - Stream Header is mandatory. - Header Metadata Block is optional. - Each Multi-Block Stream has at least one Data Block. The maximum number of Data Blocks is not limited. - Footer Metadata Block is mandatory. 2.2. Stream Header +---+---+---+---+---+---+--------------+--+--+--+--+ | Header Magic Bytes | Stream Flags | CRC32 | +---+---+---+---+---+---+--------------+--+--+--+--+ 2.2.1. Header Magic Bytes The first six (6) bytes of the Stream are so called Header Magic Bytes. They can be used to identify the file type. Using a C array and ASCII: const uint8_t HEADER_MAGIC[6] = { 0xFF, 'L', 'Z', 'M', 'A', 0x00 }; In plain hexadecimal: FF 4C 5A 4D 41 00 Notes: - The first byte (0xFF) was chosen so that the files cannot be erroneously detected as being in LZMA_Alone format, in which the first byte is in the the range [0x00, 0xE0]. - The sixth byte (0x00) was chosen to prevent applications from misdetecting the file as a text file. 2.2.2. Stream Flags Bit(s) Mask Description 0-2 0x07 Type of Check (see Section 3.3.1): ID Size Check name 0x00 0 bytes None 0x01 4 bytes CRC32 0x02 4 bytes (Reserved) 0x03 8 bytes CRC64 0x04 16 bytes (Reserved) 0x05 32 bytes SHA-256 0x06 32 bytes (Reserved) 0x07 64 bytes (Reserved) 3 0x08 The CRC32 field is present in Block Headers. 4 0x10 If unset, this is a Single-Block Stream; if set, this is a Multi-Block Stream. 5-7 0xE0 Reserved for future use; must be zero for now. Implementations must support at least the Check IDs 0x00 (None) and 0x01 (CRC32). Supporting other Check IDs is optional. If an unsupported Check is used, the decoder must indicate a warning or error. If any reserved bit is set, the decoder must indicate an error. It is possible that there is a new field present which the decoder is not aware of, and can thus parse the Stream Header incorrectly. 2.2.3. CRC32 The CRC32 is calculated from the Stream Flags field. It is stored as an unsigned 32-bit little endian integer. If the calculated value does not match the stored one, the decoder must indicate an error. Note that this field is always present; the bit in Stream Flags controls only presence of CRC32 in Block Headers. 3. Block +==============+=================+==============+ | Block Header | Compressed Data | Block Footer | +==============+=================+==============+ There are two types of Blocks: - Data Blocks hold the actual compressed data. - Metadata Blocks hold the Index, Extra, and a few other non-data fields (see Section 5). The type of the Block is indicated by the corresponding bit in the Block Flags field (see Section 3.1.1). 3.1. Block Header +------+------+=================+===================+ | Block Flags | Compressed Size | Uncompressed Size | +------+------+=================+===================+ +======================+--+--+--+--+================+ ---> | List of Filter Flags | CRC32 | Header Padding | +======================+--+--+--+--+================+ 3.1.1. Block Flags The first byte of the Block Flags field is a bit field: Bit(s) Mask Description 0-2 0x07 Number of filters (0-7) 3 0x08 Use End of Payload Marker (even if Uncompressed Size is stored to Block Header). 4 0x10 The Compressed Size field is present. 5 0x20 The Uncompressed Size field is present. 6 0x40 Reserved for future use; must be zero for now. 7 0x80 This is a Metadata Block. The second byte of the Block Flags field is also a bit field: Bit(s) Mask Description 0-4 0x1F Size of the Header Padding field (0-31 bytes) 5-7 0xE0 Reserved for future use; must be zero for now. The decoder must indicate an error if End of Payload Marker is not used and Uncompressed Size is not stored to the Block Header. Because of this, the first byte of Block Flags can never be a nul byte. This is useful when detecting beginning of the Block after Footer Padding (see Section 3.3.3). If any reserved bit is set, the decoder must indicate an error. It is possible that there is a new field present which the decoder is not aware of, and can thus parse the Block Header incorrectly. 3.1.2. Compressed Size This field is present only if the appropriate bit is set in the Block Flags field (see Section 3.1.1). This field contains the size of the Compressed Data field. The size is stored using the encoding described in Section 1.2. If the Compressed Size does not match the real size of the Compressed Data field, the decoder must indicate an error. Having the Compressed Size field in the Block Header can be useful for multithreaded decoding when seeking is not possible. If the Blocks are small enough, the decoder can read multiple Blocks into its internal buffer, and decode the Blocks in parallel. Compressed Size can also be useful when seeking forwards to a specific location in streamed mode: the decoder can quickly skip over irrelevant Blocks, without decoding them. 3.1.3. Uncompressed Size This field is present only if the appropriate bit is set in the Block Flags field (see Section 3.1.1). The Uncompressed Size field contains the size of the Block after uncompressing. Storing Uncompressed Size serves several purposes: - The decoder will know when all of the data has been decoded without an explicit End of Payload Marker. - The decoder knows how much memory it needs to allocate for a temporary buffer in multithreaded mode. - Simple error detection: wrong size indicates a broken file. - Sometimes it is useful to know the file size without uncompressing the file. It should be noted that the only reliable way to find out what the real uncompressed size is is to uncompress the Block, because the Block Header and Metadata Block fields may contain (intentionally or unintentionally) invalid information. Uncompressed Size is stored using the encoding described in Section 1.2. If the Uncompressed Size does not match the real uncompressed size, the decoder must indicate an error. 3.1.4. List of Filter Flags +================+================+ +================+ | Filter 0 Flags | Filter 1 Flags | ... | Filter n Flags | +================+================+ +================+ The number of Filter Flags fields is stored in the Block Flags field (see Section 3.1.1). As a special case, if the number of Filter Flags fields is zero, it is equivalent to having the Copy filter as the only filter. The format of each Filter Flags field is as follows: +------+=============+=============================+ | Misc | External ID | External Size of Properties | +------+=============+=============================+ +===================+ ---> | Filter Properties | +===================+ The list of officially defined Filter IDs and the formats of their Filter Properties are described in Section 4.3. 3.1.4.1. Misc To save space, the most commonly used Filter IDs and the Size of Filter Properties are encoded in a single byte. Depending on the contents of the Misc field, Filter ID is the value of the Misc or External ID field. Value Filter ID Size of Filter Properties 0x00 - 0x1F Misc 0 bytes 0x20 - 0x3F Misc 1 byte 0x40 - 0x5F Misc 2 bytes 0x60 - 0x7F Misc 3 bytes 0x80 - 0x9F Misc 4 bytes 0xA0 - 0xBF Misc 5 bytes 0xC0 - 0xDF Misc 6 bytes 0xE0 - 0xFE External ID 0-30 bytes 0xFF External ID External Size of Properties The following code demonstrates parsing the Misc field and, when needed, the External ID and External Size of Properties fields. uint64_t id; uint64_t properties_size; uint8_t misc = read_byte(); if (misc >= 0xE0) { id = read_variable_length_integer(); if (misc == 0xFF) properties_size = read_variable_length_integer(); else properties_size = misc - 0xE0; } else { id = misc; properties_size = misc / 0x20; } 3.1.4.2. External ID This field is present only if the Misc field contains a value that indicates usage of External ID. The External ID is stored using the encoding described in Section 1.2. 3.1.4.3. External Size of Properties This field is present only if the Misc field contains a value that indicates usage of External Size of Properties. The size of Filter Properties is stored using the encoding described in Section 1.2. 3.1.4.4. Filter Properties Size of this field depends on the Misc field (Section 3.1.4.1) and, if present, External Size of Properties field (Section 3.1.4.3). The format of this field is depends on the selected filter; see Section 4.3 for details. 3.1.5. CRC32 This field is present only if the appropriate bit is set in the Stream Flags field (see Section 2.2.2). The CRC32 is calculated over everything in the Block Header field except the Header Padding field and the CRC32 field itself. It is stored as an unsigned 32-bit little endian integer. If the calculated value does not match the stored one, the decoder must indicate an error. 3.1.6. Header Padding This field contains as many nul bytes as indicated by the value stored in the Header Flags field. If the Header Padding field contains any non-nul bytes, the decoder must indicate an error. The intent of the Header Padding field is to allow alignment of Compressed Data. The usefulness of alignment is described in Section 4.3. 3.2. Compressed Data The format of Compressed Data depends on Block Flags and List of Filter Flags. Excluding the descriptions of the simplest filters in Section 4, the format of the filter-specific encoded data is out of scope of this document. Note a special case: if End of Payload Marker (see Section 3.1.1) is not used and Uncompressed Size is zero, the size of the Compressed Data field is always zero. 3.3. Block Footer +=======+===============+================+ | Check | Stream Footer | Footer Padding | +=======+===============+================+ 3.3.1. Check The type and size of the Check field depends on which bits are set in the Stream Flags field (see Section 2.2.2). The Check, when used, is calculated from the original uncompressed data. If the calculated Check does not match the stored one, the decoder must indicate an error. If the selected type of Check is not supported by the decoder, it must indicate a warning or error. 3.3.2. Stream Footer +===================+===============+--------------+ | Uncompressed Size | Backward Size | Stream Flags | +===================+===============+--------------+ +----------+---------+ ---> | Footer Magic Bytes | +----------+---------+ Stream Footer is present only in - Data Block of a Single-Block Stream; and - Footer Metadata Block of a Multi-Block Stream. The Stream Footer field is placed inside Block Footer, because no padding is allowed between Check and Stream Footer. 3.3.2.1. Uncompressed Size This field is present only in the Data Block of a Single-Block Stream if Uncompressed Size is not stored to the Block Header (see Section 3.1.1). Without the Uncompressed Size field in Stream Footer it would not be possible to quickly find out the Uncompressed Size of the Stream in all cases. Uncompressed Size is stored using the encoding described in Section 1.2. If the stored value does not match the real uncompressed size of the Single-Block Stream, the decoder must indicate an error. 3.3.2.2. Backward Size This field contains the total size of the Block Header, Compressed Data, Check, and Uncompressed Size fields. The value is stored using the encoding described in Section 1.2. If the Backward Size does not match the real total size of the appropriate fields, the decoder must indicate an error. Implementations reading the Stream backwards should notice that the value in this field can never be zero. 3.3.2.3. Stream Flags This is a copy of the Stream Flags field from the Stream Header. The information stored to Stream Flags is needed when parsing the Stream backwards. 3.3.2.4. Footer Magic Bytes As the last step of the decoding process, the decoder must verify the existence of Footer Magic Bytes. If they are not found, an error must be indicated. Using a C array and ASCII: const uint8_t FOOTER_MAGIC[2] = { 'Y', 'Z' }; In hexadecimal: 59 5A The primary reason to have Footer Magic Bytes is to make it easier to detect incomplete files quickly, without uncompressing. If the file does not end with Footer Magic Bytes (excluding Footer Padding described in Section 3.3.3), it cannot be undamaged, unless someone has intentionally appended garbage after the end of the Stream. (Appending garbage at the end of the file does not prevent uncompressing the file, but may give a warning or error depending on the decoder implementation.) 3.3.3. Footer Padding In certain situations it is convenient to be able to pad Blocks or Streams to be multiples of, for example, 512 bytes. Footer Padding makes this possible. Note that this is in no way required to enforce alignment in the way described in Section 4.3; the Header Padding field is enough for that. When Footer Padding is used, it must contain only nul bytes. Any non-nul byte should be considered as the beginning of a new Block or Stream. The possibility of Padding should be taken into account when designing an application that wants to find out information about a Stream by parsing Footer Metadata Block. Support for Padding was inspired by a related note in [GNU-tar]. 4. Filters The Block Flags field defines how many filters are used. When more than one filter is used, the filters are chained; that is, the output of one filter is the input of another filter. The following figure illustrates the direction of data flow. v Uncompressed Data ^ | Filter 0 | Encoder | Filter 1 | Decoder | ... | | Filter n | v Compressed Data ^ The filters are independent from each other, except that they must cooperate a little to make it possible, in all cases, to detect when all of the data has been decoded. In addition, the filters should cooperate in the encoder to keep the alignment optimal. 4.1. Detecting when All Data Has Been Decoded There must be a way for the decoder to detect when all of the Compressed Data has been decoded. This is simple when only one filter is used, but a bit more complex when multiple filters are chained. This file format supports three methods to detect when all of the data has been decoded: - Uncompressed size - End of Input - End of Payload Marker In both encoder and decoder, filters are initialized starting from the first filter in the chain. For each filter, one of these three methods is used. 4.1.1. With Uncompressed Size This method is the only method supported by all filters. It must be used when uncompressed size is known by the filter-specific encoder or decoder. In practice this means that Uncompressed Size has been stored to the Block Header. In case of the first filter in the chain, the uncompressed size given to the filter-specific encoder or decoder equals the Uncompressed Size stored in the Block Header. For the rest of the filters in the chain, uncompressed size is the size of the output data of the previous filter in the chain. Note that when Use End of Payload Marker bit is set in Block Flags, Uncompressed Size is considered to be unknown even if it was present in the Block Header. Thus, if End of Payload Marker is used, uncompressed size of all of the filters in the chain is unknown, and can never be used to detect when all of the data has been decoded. Once the correct number of bytes has been written out, the filter-specific decoder indicates to its caller that all of the data has been decoded. If the filter-specific decoder detects End of Input or End of Payload Marker before the correct number of bytes is decoded, the decoder must indicate an error. 4.1.2. With End of Input Most filters will know that all of the data has been decoded when the End of Input data has been reached. Once the filter knows that it has received the input data in its entirety, it finishes its job, and indicates to its caller that all of the data has been decoded. The filter-specific decoder must indicate an error if it detects End of Payload Marker. Note that this method can work only when the filter is not the last filter in the chain, because only another filter can indicate the End of Input data. In practice this means, that a filter later in the chain must support embedding End of Payload Marker. When a filter that cannot embed End of Payload Marker is the last filter in the chain, Subblock filter is appended to the chain as an implicit filter. In the simplest case, this occurs when no filters are specified, and the End of Payload Marker bit is set in Block Flags. 4.1.3. With End of Payload Marker End of Payload Marker is a filter-specific bit sequence that indicates the end of data. It is supported by only a few filters. It is used when uncompressed size is unknown, and the filter - doesn't support End of Input; or - is the last filter in the chain. End of Payload Marker is embedded at the end of the encoded data by the filter-specific encoder. When the filter-specific decoder detects the embedded End of Payload Marker, the decoder knows that all of the data has been decoded. Then it finishes its job, and indicates to its caller that all of the data has been decoded. If the filter-specific decoder detects End of Input before End of Payload Marker, the decoder must indicate an error. If the filter supports both End of Input and End of Payload Marker, the former is used, unless the filter is the last filter in the chain. 4.2. Alignment Some filters give better compression ratio or are faster when the input or output data is aligned. For optimal results, the encoder should try to enforce proper alignment when possible. Not enforcing alignment in the encoder is not an error. Thus, the decoder must be able to handle files with suboptimal alignment. Alignment of uncompressed input data is usually the job of the application producing the data. For example, to get the best results, an archiver tool should make sure that all PowerPC executable files in the archive stream start at offsets that are multiples of four bytes. Some filters, for example LZMA, can be configured to take advantage of specified alignment of input data. Note that taking advantage of aligned input can be benefical also when a filter is not the first filter in the chain. For example, if you compress PowerPC executables, you may want to use the PowerPC filter and chain that with the LZMA filter. Because not only the input but also the output alignment of the PowerPC filter is four bytes, it is now benefical to set LZMA settings so that the LZMA encoder can take advantage of its four-byte-aligned input data. The output of the last filter in the chain is stored to the Compressed Data field. Aligning Compressed Data appropriately can increase - speed, if the filtered data is handled multiple bytes at a time by the filter-specific encoder and decoder, because accessing aligned data in computer memory is usually faster; and - compression ratio, if the output data is later compressed with an external compression tool. Compressed Data in a Stream can be aligned by using the Header Padding field in the Block Header. 4.3. Filters 4.3.1. Copy This is a dummy filter that simply copies all data from input to output unmodified. Filter ID: 0x00 Size of Filter Properties: 0 bytes Changes size of data: No Detecting when all of the data has been decoded: Uncompressed size: Yes End of Payload Marker: No End of Input: Yes Preferred alignment: Input data: 1 byte Output data: 1 byte 4.3.2. Subblock The Subblock filter can be used to - embed End of Payload Marker when the otherwise last filter in the chain does not support embedding it; and - apply additional filters in the middle of a Block. Filter ID: 0x01 Size of Filter Properties: 0 bytes Changes size of data: Yes, unpredictably Detecting when all of the data has been decoded: Uncompressed size: Yes End of Payload Marker: Yes End of Input: Yes Preferred alignment: Input data: 1 byte Output data: Freely adjustable 4.3.2.1. Format of the Encoded Output The encoded data from the Subblock filter consist of zero or more Subblocks: +==========+==========+ | Subblock | Subblock | ... +==========+==========+ Each Subblock contains two fields: +----------------+===============+ | Subblock Flags | Subblock Data | +----------------+===============+ Subblock Flags is a bitfield: Bits Mask Description 0-3 0x0F The interpretation of these bits depend on the Subblock Type: - 0x20 Bits 0-3 for Size - 0x30 Bits 0-3 for Repeat Count - Other These bits must be zero. 4-7 0xF0 Subblock Type: - 0x00: Padding - 0x10: End of Payload Marker - 0x20: Data - 0x30: Repeating Data - 0x40: Set Subfilter - 0x50: Unset Subfilter If some other value is detected, the decoder must indicate an error. The format of the Subblock Data field depends on Subblock Type. Subblocks with the Subblock Type 0x00 (Padding) don't have a Subblock Data field. These Subblocks can be useful for fixing alignment. There can be at maximum of 31 consecutive Subblocks with this Subblock Type; if there are more, the decoder must indicate an error. Subblock with the Subblock Type 0x10 (End of Payload Marker) doesn't have a Subblock Data field. The decoder must indicate an error if this Subblock Type is detected when Subfilter is enabled, or when the Subblock filter is not supposed to embed the End of Payload Marker. Subblocks with the Subblock Type 0x20 (Data) contain the rest of the Size, which is followed by Size + 1 bytes in the Data field (that is, Data can never be empty): +------+------+------+======+ | Bits 4-27 for Size | Data | +------+------+------+======+ Subblocks with the Subblock Type 0x30 (Repeating Data) contain the rest of the Repeat Count, the Size of the Data, and finally the actual Data to be repeated: +---------+---------+--------+------+======+ | Bits 4-27 for Repeat Count | Size | Data | +---------+---------+--------+------+======+ The size of the Data field is Size + 1. It is repeated Repeat Count + 1 times. That is, the minimum size of Data is one byte; the maximum size of Data is 256 bytes. The minimum number of repeats is one; the maximum number of repeats is 2^28. If Subfilter is not used, the Data field of Subblock Types 0x20 and 0x30 is the output of the decoded Subblock filter. If Subfilter is used, Data is the input of the Subfilter, and the decoded output of the Subfilter is the decoded output of the Subblock filter. Subblocks with the Subblock Type 0x40 (Set Subfilter) contain a Filter Flags field in Subblock Data: +==============+ | Filter Flags | +==============+ It is an error to set the Subfilter to Filter ID 0x00 (Copy) or 0x01 (Subblock). All the other Filter IDs are allowed. The decoder must indicate an error if this Subblock Type is detected when a Subfilter is already enabled. Subblocks with the Subblock Type 0x50 (Unset Subfilter) don't have a Subblock Data field. There must be at least one Subblock with Subblock Type 0x20 or 0x30 between Subblocks with Subblock Type 0x40 and 0x50; if there isn't, the decoder must indicate an error. Subblock Types 0x40 and 0x50 are always used as a pair: If the Subblock filter has been enabled with Subblock Type 0x40, it must always be disabled later with Subblock Type 0x50. Disabling must be done even if the Subfilter used End of Payload Marker; after the Subfilter has detected End of Payload Marker, the next Subblock that is not Padding must unset the Subfilter. When the Subblock filter is used as an implicit filter to embed End of Payload marker, the Subblock Types 0x40 and 0x50 (Set or Unset Subfilter) must not be used. The decoder must indicate an error if it detects any of these Subblock Types in an implicit Subblock filter. The following code illustrates the basic structure of a Subblock decoder. uint32_t consecutive_padding = 0; bool got_output_with_subfilter = false; while (true) { uint32_t size; uint32_t repeat; uint8_t flags = read_byte(); if (flags != 0) consecutive_padding = 0; switch (flags >> 4) { case 0: // Padding if (flags & 0x0F) return DATA_ERROR; if (++consecutive_padding == 32) return DATA_ERROR; break; case 1: // End of Payload Marker if (flags & 0x0F) return DATA_ERROR; if (subfilter_enabled || !allow_eopm) return DATA_ERROR; break; case 2: // Data size = flags & 0x0F; for (size_t i = 4; i < 28; i += 8) size |= (uint32_t)(read_byte()) << i; // If any output is produced, this will // set got_output_with_subfilter to true. copy_data(size); break; case 3: // Repeating Data repeat = flags & 0x0F; for (size_t i = 4; i < 28; i += 8) repeat |= (uint32_t)(read_byte()) << i; size = read_byte(); // If any output is produced, this will // set got_output_with_subfilter to true. copy_repeating_data(size, repeat); break; case 4: // Set Subfilter if (flags & 0x0F) return DATA_ERROR; if (subfilter_enabled) return DATA_ERROR; got_output_with_subfilter = false; set_subfilter(); break; case 5: // Unset Subfilter if (flags & 0x0F) return DATA_ERROR; if (!subfilter_enabled) return DATA_ERROR; if (!got_output_with_subfilter) return DATA_ERROR; unset_subfilter(); break; default: return DATA_ERROR; } } 4.3.3. Delta The Delta filter may increase compression ratio when the value of the next byte correlates with the value of an earlier byte at specified distance. Filter ID: 0x20 Size of Filter Properties: 1 byte Changes size of data: No Detecting when all of the data has been decoded: Uncompressed size: Yes End of Payload Marker: No End of Input: Yes Preferred alignment: Input data: 1 byte Output data: Same as the original input data The Properties byte indicates the delta distance, which can be 1-256 bytes backwards from the current byte: 0x00 indicates distance of 1 byte and 0xFF distance of 256 bytes. 4.3.3.1. Format of the Encoded Output The code below illustrates both encoding and decoding with the Delta filter. // Distance is in the range [1, 256]. const unsigned int distance = get_properties_byte() + 1; uint8_t pos = 0; uint8_t delta[256]; memset(delta, 0, sizeof(delta)); while (1) { const int byte = read_byte(); if (byte == EOF) break; uint8_t tmp = delta[(uint8_t)(distance + pos)]; if (is_encoder) { tmp = (uint8_t)(byte) - tmp; delta[pos] = (uint8_t)(byte); } else { tmp = (uint8_t)(byte) + tmp; delta[pos] = tmp; } write_byte(tmp); --pos; } 4.3.4. LZMA LZMA (Lempel-Ziv-Markov chain-Algorithm) is a general-purporse compression algorithm with high compression ratio and fast decompression. LZMA based on LZ77 and range coding algorithms. Filter ID: 0x40 Size of Filter Properties: 2 bytes Changes size of data: Yes, unpredictably Detecting when all of the data has been decoded: Uncompressed size: Yes End of Payload Marker: Yes End of Input: No Preferred alignment: Input data: Adjustable to 1/2/4/8/16 byte(s) Output data: 1 byte At the time of writing, there is no other documentation about how LZMA works than the source code in LZMA SDK. Once such documentation gets written, it will probably be published as a separate document, because including the documentation here would lengthen this document considerably. The format of the Filter Properties field is as follows: +-----------------+------------------+ | LZMA Properties | Dictionary Flags | +-----------------+------------------+ 4.3.4.1. LZMA Properties The LZMA Properties field contains three properties. An abbreviation is given in parentheses, followed by the value range of the property. The field consists of 1) the number of literal context bits (lc, [0, 8]); 2) the number of literal position bits (lp, [0, 4]); and 3) the number of position bits (pb, [0, 4]). They are encoded using the following formula: LZMA Properties = (pb * 5 + lp) * 9 + lc The following C code illustrates a straightforward way to decode the properties: uint8_t lc, lp, pb; uint8_t prop = get_lzma_properties() & 0xFF; if (prop > (4 * 5 + 4) * 9 + 8) return LZMA_PROPERTIES_ERROR; pb = prop / (9 * 5); prop -= pb * 9 * 5; lp = prop / 9; lc = prop - lp * 9; 4.3.4.2. Dictionary Flags Currently the lowest six bits of the Dictionary Flags field are in use: Bits Mask Description 0-5 0x3F Dictionary Size 6-7 0xC0 Reserved for future use; must be zero for now. Dictionary Size is encoded with one-bit mantissa and five-bit exponent. To avoid wasting space, one-byte dictionary has its own special value. Raw value Mantissa Exponent Dictionary size 0 1 0 1 byte 1 2 0 2 bytes 2 3 0 3 bytes 3 2 1 4 bytes 4 3 1 6 bytes 5 2 2 8 bytes 6 3 2 12 bytes 7 2 3 16 bytes 8 3 3 24 bytes 9 2 4 32 bytes ... ... ... ... 61 2 30 2 GiB 62 3 30 3 GiB 63 2 31 4 GiB (*) (*) The real maximum size of the dictionary is one byte less than 4 GiB, because the distance of 4 GiB is reserved for End of Payload Marker. Instead of having a table in the decoder, the dictionary size can be decoded using the following C code: uint64_t dictionary_size; const uint8_t bits = get_dictionary_flags() & 0x3F; if (bits == 0) { dictionary_size = 1; } else { dictionary_size = 2 | ((bits + 1) & 1); dictionary_size = dictionary_size << ((bits - 1) / 2); } 4.3.5. Branch/Call/Jump Filters for Executables These filters convert relative branch, call, and jump instructions to their absolute counterparts in executable files. This conversion increases redundancy and thus compression ratio. Size of Filter Properties: 0 or 4 bytes Changes size of data: No Detecting when all of the data has been decoded: Uncompressed size: Yes End of Payload Marker: No End of Input: Yes Below is the list of filters in this category. The alignment is the same for both input and output data. Filter ID Alignment Description 0x04 1 byte x86 filter (BCJ) 0x05 4 bytes PowerPC (big endian) filter 0x06 16 bytes IA64 filter 0x07 4 bytes ARM (little endian) filter 0x08 2 bytes ARM Thumb (little endian) filter 0x09 4 bytes SPARC filter If the size of Filter Properties is four bytes, the Filter Properties field contains the start offset used for address conversions. It is stored as an unsigned 32-bit little endian integer. If the size of Filter Properties is zero, the start offset is zero. Setting the start offset may be useful if an executable has multiple sections, and there are many cross-section calls. Taking advantage of this feature usually requires usage of the Subblock filter. 5. Metadata Metadata is stored in Metadata Blocks, which can be in the beginning or at the end of a Multi-Block Stream. Because of Blocks, it is possible to compress Metadata in the same way as the actual data is compressed. This Section describes the format of the data stored in Metadata Blocks. +----------------+===============================+ | Metadata Flags | Size of Header Metadata Block | +----------------+===============================+ +============+===================+=======+=======+ ---> | Total Size | Uncompressed Size | Index | Extra | +============+===================+=======+=======+ Stream must be parseable backwards. That is, there must be a way to locate the beginning of the Stream by starting from the end of the Stream. Thus, the Footer Metadata Block must contain the Total Size field or the Index field. If the Stream has Header Metadata Block, also the Size of Header Metadata Block field must be present in Footer Metadata Block. It must be possible to quickly locate the Blocks in non-streamed mode. Thus, the Index field must be present at least in one Metadata Block. If the above conditions are not met, the decoder must indicate an error. There should be no additional data after the last field. If there is, the the decoder should indicate an error. 5.1. Metadata Flags This field describes which fields are present in a Metadata Block: Bit(s) Mask Desription 0 0x01 Size of Header Metadata Block is present. 1 0x02 Total Size is present. 2 0x04 Uncompressed Size is present. 3 0x08 Index is present. 4-6 0x70 Reserve for future use; must be zero for now. 7 0x80 Extra is present. If any reserved bit is set, the decoder must indicate an error. It is possible that there is a new field present which the decoder is not aware of, and can thus parse the Metadata incorrectly. 5.2. Size of Header Metadata Block This field is present only if the appropriate bit is set in the Metadata Flags field (see Section 5.1). Size of Header Metadata Block is needed to make it possible to parse the Stream backwards. The size is stored using the encoding described in Section 1.2. The decoder must verify that that the value stored in this field is non-zero. In Footer Metadata Block, the decoder must also verify that the stored size matches the real size of Header Metadata Block. In the Header Meatadata Block, the value of this field is ignored as long as it is not zero. 5.3. Total Size This field is present only if the appropriate bit is set in the Metadata Flags field (see Section 5.1). This field contains the total size of the Data Blocks in the Stream. Total Size is stored using the encoding described in Section 1.2. If the stored value does not match the real total size of the Data Blocks, the decoder must indicate an error. The value of this field must be non-zero. Total Size can be used to quickly locate the beginning or end of the Stream. This can be useful for example when doing random-access reading, and the Index field is not in the Metadata Block currently being read. It is useless to have both Total Size and Index in the same Metadata Block, because Total Size can be calculated from the Index field. 5.4. Uncompressed Size This field is present only if the appropriate bit is set in the Metadata Flags field (see Section 5.1). This field contains the total uncompressed size of the Data Blocks in the Stream. Uncompresssed Size is stored using the encoding described in Section 1.2. If the stored value does not match the real uncompressed size of the Data Blocks, the decoder must indicate an error. It is useless to have both Uncompressed Size and Index in the same Metadata Block, because Uncompressed Size can be calculated from the Index field. 5.5. Index +=======================+=============+====================+ | Number of Data Blocks | Total Sizes | Uncompressed Sizes | +=======================+=============+====================+ Index serves several purporses. Using it, one can - verify that all Blocks in a Stream have been processed; - find out the Uncompressed Size of a Stream; and - quickly access the beginning of any Block (random access). 5.5.1. Number of Data Blocks This field contains the number of Data Blocks in the Stream. The value is stored using the encoding described in Section 1.2. If the decoder has decoded all the Data Blocks of the Stream, and then notices that the Number of Records doesn't match the real number of Data Blocks, the decoder must indicate an error. The value of this field must be non-zero. 5.5.2. Total Sizes +============+============+ | Total Size | Total Size | ... +============+============+ This field lists the Total Sizes of every Data Block in the Stream. There are as many Total Size fields as indicated by the Number of Data Blocks field. Total Size is the size of Block Header, Compressed Data, and Block Footer. It is stored using the encoding described in Section 1.2. If the Total Sizes do not match the real sizes of respective Blocks, the decoder should indicate an error. All the Total Size fields must have a non-zero value. 5.5.3. Uncompressed Sizes +===================+===================+ | Uncompressed Size | Uncompressed Size | ... +===================+===================+ This field lists the Uncompressed Sizes of every Data Block in the Stream. There are as many Uncompressed Size fields as indicated by the Number of Records field. Uncompressed Sizes are stored using the encoding described in Section 1.2. If the Uncompressed Sizes do not match the real sizes of respective Blocks, the decoder shoud indicate an error. 5.6. Extra This field is present only if the appropriate bit is set in the Metadata Flags field (see Section 5.1). Note that the bit does not indicate that there is any data in the Extra field; it only indicates that Extra may be non-empty. The Extra field contains only information that is not required to properly uncompress the Stream or to do random-access reading. Supporting the Extra field is optional. In case the decoder doesn't support the Extra field, it should silently ignore it. Extra consists of zero or more Records: +========+========+ | Record | Record | ... +========+========+ Excluding Records with Record ID 0x00, each Record contains three fields: +==========+==============+======+ | Reord ID | Size of Data | Data | +==========+==============+======+ The Record ID and Size of Data are stored using the encoding described in Section 1.2. Data can be binary or UTF-8 [RFC-3629] strings. Non-UTF-8 strings should be avoided. Because the Size of Data is known, there is no need to terminate strings with a nul byte, although doing so should not be considered an error. The Record IDs are divided in two categories: - Safe-to-Copy Records may be preserved as is when the Stream is modified in ways that don't change the actual uncompressed data. Examples of such operatings include recompressing and adding, modifying, or deleting unrelated Extra Records. - Unsafe-to-Copy Records should be removed (and possibly recreated) when any kind of changes are made to the Stream. When the actual uncompressed data is modified, all Records should be removed (and possibly recreated), unless the application knows that the Data stored to the Record(s) is still valid. The following subsections describe the standard Record IDs and the format of their Data fields. Safe-to-Copy Records have an odd ID, while Unsafe-to-Copy Records have an even ID. 5.6.1. 0x00: Dummy/Padding This Record is special, because it doesn't have the Size of Data or Data fields. Dummy Records can be used, for example, to fill Metadata Block when a few bytes of extra space has been reserved for it. There can be any number of Dummy Records. 5.6.2. 0x01: OpenPGP Signature OpenPGP signature is computed from uncompressed data. The signature can be used to verify that the contents of a Stream has been created by a trustworthy source. If the decoder supports decoding concatenated Streams, it must indicate an error when verifying OpenPGP signatures if there is more than one Stream. OpenPGP format is documented in [RFC-2440]. 5.6.3. 0x02: Filter Information The Filter Information Record contains information about the filters used in the Stream. This field can be used to quickly - display which filters are used in each Block; - check if all the required filters are supported by the current decoder version; and - check how much memory is required to decode each Block. The format of the Filter Information field is as follows: +=================+=================+ | Block 0 Filters | Block 1 Filters | ... +=================+=================+ There can be at maximum of as many Block Filters fields as there are Data Blocks in the Stream. The format of the Block Filters field is as follows: +------------------+======================+============+ | Block Properties | List of Filter Flags | Subfilters | +------------------+======================+============+ Block Properties is a bitfield: Bit(s) Mask Description 0-2 0x07 Number of filters (0-7) 3 0x08 End of Payload Marker is used. 4 0x10 The Subfilters field is present. 5-7 0xE0 Reserved for future use; must be zero for now. The contents of the List of Filter Flags field must match the List of Filter Flags field in the respective Block Header. The Subfilters field may be present only if the List of Filter Flags contains a Filter Flags field for a Subblock filter. The format of the Subfilters field is as follows: +======================+=========================+ | Number of Subfilters | List of Subfilter Flags | +======================+=========================+ The value stored in the Number of Subfilters field is stored using the encoding described in Section 1.2. The List of Subfilter Flags field contains as many Filter Flags fields as indicated by the Number of Subfilters field. These Filter Flags fields list some or all the Subfilters used via the Subblock filter. The order of the listed Subfilters is not significant. Decoders supporting this Record should indicate a warning or error if this Record contains Filter Flags that are not actually used by the respective Blocks. 5.6.4. 0x03: Comment Free-form comment is stored in UTF-8 [RFC-3629] encoding. The beginning of a new line should be indicated using the ASCII Line Feed character (0x0A). When the Line Feed character is not the native way to indicate new line in the underlying operating system, the encoder and decoder should convert the newline characters to and from Line Feeds. 5.6.5. 0x04: List of Checks +=======+=======+ | Check | Check | ... +=======+=======+ There are as many Check fields as there are Blocks in the Stream. The size of Check fields depend on Stream Flags (see Section 2.2.2). Decoders supporting this Record should indicate a warning or error if the Checks don't match the respective Blocks. 5.6.6. 0x05: Original Filename Original filename is stored in UTF-8 [RFC-3629] encoding. The filename must not include any path, only the filename itself. Special care must be taken to prevent directory traversal vulnerabilities. When files are moved between different operating systems, it is possible that filename valid in the source system is not valid in the target system. It is implementation defined how the decoder handles this kind of situations. 5.6.7. 0x07: Modification Time Modification time is stored as POSIX time, as an unsigned little endian integer. The number of bits depends on the Size of Data field. Note that the usage of unsigned integer limits the earliest representable time to 1970-01-01T00:00:00. 5.6.8. 0x09: High-Resolution Modification Time This Record extends the `0x04: Modification time' Record with a subsecond time information. There are two supported formats of this field, which can be distinguished by looking at the Size of Data field. Size Data 3 [0; 9,999,999] times 100 nanoseconds 4 [0; 999,999,999] nanoseconds The value is stored as an unsigned 24-bit or 32-bit little endian integer. 5.6.9. 0x0B: MIME Type MIME type of the uncompressed Stream. This can be used to detect the content type. [IANA-MIME] 5.6.10. 0x0D: Homepage URL This field can be used, for example, when distributing software packages (sources or binaries). The field would indicate the homepage of the program. For details on how to encode URLs, see [RFC-1738]. 6. Custom Filter and Extra Record IDs If a developer wants to use custom Filter or Extra Record IDs, he has two choices. The first choice is to contact Lasse Collin and ask him to allocate a range of IDs for the developer. The second choice is to generate a 40-bit random integer, which the developer can use as his personal Developer ID. To minimalize the risk of collisions, Developer ID has to be a randomly generated integer, not manually selected "hex word". The following command, which works on many free operating systems, can be used to generate Developer ID: dd if=/dev/urandom bs=5 count=1 | hexdump The developer can then use his Developer ID to create unique (well, hopefully unique) Filter and Extra Record IDs. Bits Mask Description 0-15 0x0000_0000_0000_FFFF Filter or Extra Record ID 16-55 0x00FF_FFFF_FFFF_0000 Developer ID 56-62 0x7F00_0000_0000_0000 Static prefix: 0x7F The resulting 63-bit integer will use 9 bytes of space when stored using the encoding described in Section 1.2. To get a shorter ID, see the beginning of this Section how to request a custom ID range. Note that Filter and Metadata Record IDs are in their own namespaces. That is, you can use the same ID value as Filter ID and Metadata Record ID, and the meanings of the IDs do not need to be related to each other. 6.1. Reserved Custom Filter ID Ranges Range Description 0x0000_0000 - 0x0000_00DF IDs fitting into the Misc field 0x0002_0000 - 0x0007_FFFF Reserved to ease .7z compatibility 0x0200_0000 - 0x07FF_FFFF Reserved to ease .7z compatibility 7. Cyclic Redundancy Checks There are several incompatible variations to calculate CRC32 and CRC64. For simplicity and clarity, complete examples are provided to calculate the checks as they are used in this file format. Implementations may use different code as long as it gives identical results. The program below reads data from standard input, calculates the CRC32 and CRC64 values, and prints the calculated values as big endian hexadecimal strings to standard output. #include #include #include uint32_t crc32_table[256]; uint64_t crc64_table[256]; void init(void) { static const uint32_t poly32 = UINT32_C(0xEDB88320); static const uint64_t poly64 = UINT64_C(0xC96C5795D7870F42); for (size_t i = 0; i < 256; ++i) { uint32_t crc32 = i; uint64_t crc64 = i; for (size_t j = 0; j < 8; ++j) { if (crc32 & 1) crc32 = (crc32 >> 1) ^ poly32; else crc32 >>= 1; if (crc64 & 1) crc64 = (crc64 >> 1) ^ poly64; else crc64 >>= 1; } crc32_table[i] = crc32; crc64_table[i] = crc64; } } uint32_t crc32(const uint8_t *buf, size_t size, uint32_t crc) { crc = ~crc; for (size_t i = 0; i < size; ++i) crc = crc32_table[buf[i] ^ (crc & 0xFF)] ^ (crc >> 8); return ~crc; } uint64_t crc64(const uint8_t *buf, size_t size, uint64_t crc) { crc = ~crc; for (size_t i = 0; i < size; ++i) crc = crc64_table[buf[i] ^ (crc & 0xFF)] ^ (crc >> 8); return ~crc; } int main() { init(); uint32_t value32 = 0; uint64_t value64 = 0; uint64_t total_size = 0; uint8_t buf[8192]; while (1) { const size_t buf_size = fread(buf, 1, 8192, stdin); if (buf_size == 0) break; total_size += buf_size; value32 = crc32(buf, buf_size, value32); value64 = crc64(buf, buf_size, value64); } printf("Bytes: %" PRIu64 "\n", total_size); printf("CRC-32: 0x%08" PRIX32 "\n", value32); printf("CRC-64: 0x%016" PRIX64 "\n", value64); return 0; } 8. References 8.1. Normative References [RFC-1738] Uniform Resource Locators (URL) http://www.ietf.org/rfc/rfc1738.txt [RFC-2119] Key words for use in RFCs to Indicate Requirement Levels http://www.ietf.org/rfc/rfc2119.txt [RFC-2440] OpenPGP Message Format http://www.ietf.org/rfc/rfc2440.txt [RFC-3629] UTF-8, a transformation format of ISO 10646 http://www.ietf.org/rfc/rfc3629.txt [IANA-MIME] MIME Media Types http://www.iana.org/assignments/media-types/ 8.2. Informative References LZMA SDK - The original LZMA implementation http://7-zip.org/sdk.html LZMA Utils - LZMA adapted to POSIX-like systems http://tukaani.org/lzma/ [RFC-1952] GZIP file format specification version 4.3 http://www.ietf.org/rfc/rfc1952.txt - Notation of byte boxes in section `2.1. Overall conventions' [GNU-tar] GNU tar 1.16.1 manual http://www.gnu.org/software/tar/manual/html_node/Blocking-Factor.html - Node 9.4.2 `Blocking Factor', paragraph that begins `gzip will complain about trailing garbage' - Note that this URL points to the latest version of the manual, and may some day not contain the note which is in 1.16.1. For the exact version of the manual, download GNU tar 1.16.1: ftp://ftp.gnu.org/pub/gnu/tar/tar-1.16.1.tar.gz