use failure::*; use std::convert::TryInto; use proxmox::tools::io::{ReadExt, WriteExt}; const MAX_BLOB_SIZE: usize = 128*1024*1024; use super::*; /// Data blob binary storage format /// /// Data blobs store arbitrary binary data (< 128MB), and can be /// compressed and encrypted. A simply binary format is used to store /// them on disk or transfer them over the network. Please use index /// files to store large data files (".fidx" of ".didx"). /// pub struct DataBlob { raw_data: Vec, // tagged, compressed, encryped data } impl DataBlob { /// accessor to raw_data field pub fn raw_data(&self) -> &[u8] { &self.raw_data } /// Consume self and returns raw_data pub fn into_inner(self) -> Vec { self.raw_data } /// accessor to chunk type (magic number) pub fn magic(&self) -> &[u8; 8] { self.raw_data[0..8].try_into().unwrap() } /// accessor to crc32 checksum pub fn crc(&self) -> u32 { let crc_o = proxmox::tools::offsetof!(DataBlobHeader, crc); u32::from_le_bytes(self.raw_data[crc_o..crc_o+4].try_into().unwrap()) } // set the CRC checksum field pub fn set_crc(&mut self, crc: u32) { let crc_o = proxmox::tools::offsetof!(DataBlobHeader, crc); self.raw_data[crc_o..crc_o+4].copy_from_slice(&crc.to_le_bytes()); } /// compute the CRC32 checksum pub fn compute_crc(&self) -> u32 { let mut hasher = crc32fast::Hasher::new(); let start = std::mem::size_of::(); // start after HEAD hasher.update(&self.raw_data[start..]); hasher.finalize() } /// verify the CRC32 checksum pub fn verify_crc(&self) -> Result<(), Error> { let expected_crc = self.compute_crc(); if expected_crc != self.crc() { bail!("Data blob has wrong CRC checksum."); } Ok(()) } /// Create a DataBlob, optionally compressed and/or encrypted pub fn encode( data: &[u8], config: Option<&CryptConfig>, compress: bool, ) -> Result { if data.len() > MAX_BLOB_SIZE { bail!("data blob too large ({} bytes).", data.len()); } let mut blob = if let Some(config) = config { let compr_data; let (_compress, data, magic) = if compress { compr_data = zstd::block::compress(data, 1)?; // Note: We only use compression if result is shorter if compr_data.len() < data.len() { (true, &compr_data[..], ENCR_COMPR_BLOB_MAGIC_1_0) } else { (false, data, ENCRYPTED_BLOB_MAGIC_1_0) } } else { (false, data, ENCRYPTED_BLOB_MAGIC_1_0) }; let header_len = std::mem::size_of::(); let mut raw_data = Vec::with_capacity(data.len() + header_len); let dummy_head = EncryptedDataBlobHeader { head: DataBlobHeader { magic: [0u8; 8], crc: [0; 4] }, iv: [0u8; 16], tag: [0u8; 16], }; unsafe { raw_data.write_le_value(dummy_head)?; } let (iv, tag) = config.encrypt_to(data, &mut raw_data)?; let head = EncryptedDataBlobHeader { head: DataBlobHeader { magic, crc: [0; 4] }, iv, tag, }; unsafe { (&mut raw_data[0..header_len]).write_le_value(head)?; } DataBlob { raw_data } } else { let max_data_len = data.len() + std::mem::size_of::(); if compress { let mut comp_data = Vec::with_capacity(max_data_len); let head = DataBlobHeader { magic: COMPRESSED_BLOB_MAGIC_1_0, crc: [0; 4], }; unsafe { comp_data.write_le_value(head)?; } zstd::stream::copy_encode(data, &mut comp_data, 1)?; if comp_data.len() < max_data_len { let mut blob = DataBlob { raw_data: comp_data }; blob.set_crc(blob.compute_crc()); return Ok(blob); } } let mut raw_data = Vec::with_capacity(max_data_len); let head = DataBlobHeader { magic: UNCOMPRESSED_BLOB_MAGIC_1_0, crc: [0; 4], }; unsafe { raw_data.write_le_value(head)?; } raw_data.extend_from_slice(data); DataBlob { raw_data } }; blob.set_crc(blob.compute_crc()); Ok(blob) } /// Decode blob data pub fn decode(self, config: Option<&CryptConfig>) -> Result, Error> { let magic = self.magic(); if magic == &UNCOMPRESSED_BLOB_MAGIC_1_0 { let data_start = std::mem::size_of::(); return Ok(self.raw_data[data_start..].to_vec()); } else if magic == &COMPRESSED_BLOB_MAGIC_1_0 { let data_start = std::mem::size_of::(); let data = zstd::block::decompress(&self.raw_data[data_start..], MAX_BLOB_SIZE)?; return Ok(data); } else if magic == &ENCR_COMPR_BLOB_MAGIC_1_0 || magic == &ENCRYPTED_BLOB_MAGIC_1_0 { let header_len = std::mem::size_of::(); let head = unsafe { (&self.raw_data[..header_len]).read_le_value::()? }; if let Some(config) = config { let data = if magic == &ENCR_COMPR_BLOB_MAGIC_1_0 { config.decode_compressed_chunk(&self.raw_data[header_len..], &head.iv, &head.tag)? } else { config.decode_uncompressed_chunk(&self.raw_data[header_len..], &head.iv, &head.tag)? }; return Ok(data); } else { bail!("unable to decrypt blob - missing CryptConfig"); } } else if magic == &AUTH_COMPR_BLOB_MAGIC_1_0 || magic == &AUTHENTICATED_BLOB_MAGIC_1_0 { let header_len = std::mem::size_of::(); let head = unsafe { (&self.raw_data[..header_len]).read_le_value::()? }; let data_start = std::mem::size_of::(); // Note: only verify if we have a crypt config if let Some(config) = config { let signature = config.compute_auth_tag(&self.raw_data[data_start..]); if signature != head.tag { bail!("verifying blob signature failed"); } } if magic == &AUTH_COMPR_BLOB_MAGIC_1_0 { let data = zstd::block::decompress(&self.raw_data[data_start..], 16*1024*1024)?; return Ok(data); } else { return Ok(self.raw_data[data_start..].to_vec()); } } else { bail!("Invalid blob magic number."); } } /// Create a signed DataBlob, optionally compressed pub fn create_signed( data: &[u8], config: &CryptConfig, compress: bool, ) -> Result { if data.len() > MAX_BLOB_SIZE { bail!("data blob too large ({} bytes).", data.len()); } let compr_data; let (_compress, data, magic) = if compress { compr_data = zstd::block::compress(data, 1)?; // Note: We only use compression if result is shorter if compr_data.len() < data.len() { (true, &compr_data[..], AUTH_COMPR_BLOB_MAGIC_1_0) } else { (false, data, AUTHENTICATED_BLOB_MAGIC_1_0) } } else { (false, data, AUTHENTICATED_BLOB_MAGIC_1_0) }; let header_len = std::mem::size_of::(); let mut raw_data = Vec::with_capacity(data.len() + header_len); let head = AuthenticatedDataBlobHeader { head: DataBlobHeader { magic, crc: [0; 4] }, tag: config.compute_auth_tag(data), }; unsafe { raw_data.write_le_value(head)?; } raw_data.extend_from_slice(data); let mut blob = DataBlob { raw_data }; blob.set_crc(blob.compute_crc()); return Ok(blob); } /// Create Instance from raw data pub fn from_raw(data: Vec) -> Result { if data.len() < std::mem::size_of::() { bail!("blob too small ({} bytes).", data.len()); } let magic = &data[0..8]; if magic == ENCR_COMPR_BLOB_MAGIC_1_0 || magic == ENCRYPTED_BLOB_MAGIC_1_0 { if data.len() < std::mem::size_of::() { bail!("encrypted blob too small ({} bytes).", data.len()); } let blob = DataBlob { raw_data: data }; Ok(blob) } else if magic == COMPRESSED_BLOB_MAGIC_1_0 || magic == UNCOMPRESSED_BLOB_MAGIC_1_0 { let blob = DataBlob { raw_data: data }; Ok(blob) } else if magic == AUTH_COMPR_BLOB_MAGIC_1_0 || magic == AUTHENTICATED_BLOB_MAGIC_1_0 { if data.len() < std::mem::size_of::() { bail!("authenticated blob too small ({} bytes).", data.len()); } let blob = DataBlob { raw_data: data }; Ok(blob) } else { bail!("unable to parse raw blob - wrong magic"); } } } // TODO: impl. other blob types use std::io::{Read, BufRead, Write, Seek, SeekFrom}; enum BlobWriterState<'a, W: Write> { Uncompressed { writer: W, hasher: crc32fast::Hasher }, Compressed { compr: zstd::stream::write::Encoder, hasher: crc32fast::Hasher }, Signed { writer: W, hasher: crc32fast::Hasher, signer: openssl::sign::Signer<'a>, }, } /// Write compressed data blobs pub struct DataBlobWriter<'a, W: Write> { state: BlobWriterState<'a, W>, } impl <'a, W: Write + Seek> DataBlobWriter<'a, W> { pub fn new_uncompressed(mut writer: W) -> Result { let hasher = crc32fast::Hasher::new(); writer.seek(SeekFrom::Start(0))?; let head = DataBlobHeader { magic: UNCOMPRESSED_BLOB_MAGIC_1_0, crc: [0; 4] }; unsafe { writer.write_le_value(head)?; } let state = BlobWriterState::Uncompressed { writer, hasher }; Ok(Self { state }) } pub fn new_compressed(mut writer: W) -> Result { let hasher = crc32fast::Hasher::new(); writer.seek(SeekFrom::Start(0))?; let head = DataBlobHeader { magic: COMPRESSED_BLOB_MAGIC_1_0, crc: [0; 4] }; unsafe { writer.write_le_value(head)?; } let compr = zstd::stream::write::Encoder::new(writer, 1)?; let state = BlobWriterState::Compressed { compr, hasher }; Ok(Self { state }) } pub fn new_signed(mut writer: W, config: &'a CryptConfig) -> Result { let hasher = crc32fast::Hasher::new(); writer.seek(SeekFrom::Start(0))?; let head = AuthenticatedDataBlobHeader { head: DataBlobHeader { magic: AUTHENTICATED_BLOB_MAGIC_1_0, crc: [0; 4] }, tag: [0u8; 32], }; unsafe { writer.write_le_value(head)?; } let signer = config.data_signer(); let state = BlobWriterState::Signed { writer, hasher, signer }; Ok(Self { state }) } pub fn finish(self) -> Result { match self.state { BlobWriterState::Uncompressed { mut writer, hasher } => { // write CRC let crc = hasher.finalize(); let head = DataBlobHeader { magic: COMPRESSED_BLOB_MAGIC_1_0, crc: crc.to_le_bytes() }; writer.seek(SeekFrom::Start(0))?; unsafe { writer.write_le_value(head)?; } return Ok(writer) } BlobWriterState::Compressed { compr, hasher } => { let mut writer = compr.finish()?; // write CRC let crc = hasher.finalize(); let head = DataBlobHeader { magic: COMPRESSED_BLOB_MAGIC_1_0, crc: crc.to_le_bytes() }; writer.seek(SeekFrom::Start(0))?; unsafe { writer.write_le_value(head)?; } return Ok(writer) } BlobWriterState::Signed { mut writer, hasher, signer, .. } => { // write CRC and hmac let crc = hasher.finalize(); let mut head = AuthenticatedDataBlobHeader { head: DataBlobHeader { magic: AUTHENTICATED_BLOB_MAGIC_1_0, crc: crc.to_le_bytes() }, tag: [0u8; 32], }; signer.sign(&mut head.tag)?; writer.seek(SeekFrom::Start(0))?; unsafe { writer.write_le_value(head)?; } return Ok(writer) } } } } impl <'a, W: Write + Seek> Write for DataBlobWriter<'a, W> { fn write(&mut self, buf: &[u8]) -> Result { match self.state { BlobWriterState::Uncompressed { ref mut writer, ref mut hasher } => { hasher.update(buf); writer.write(buf) } BlobWriterState::Compressed { ref mut compr, ref mut hasher } => { hasher.update(buf); compr.write(buf) } BlobWriterState::Signed { ref mut writer, ref mut hasher, ref mut signer, .. } => { hasher.update(buf); signer.update(buf). map_err(|err| { std::io::Error::new( std::io::ErrorKind::Other, format!("hmac update failed - {}", err)) })?; writer.write(buf) } } } fn flush(&mut self) -> Result<(), std::io::Error> { match self.state { BlobWriterState::Uncompressed { ref mut writer, .. } => { writer.flush() } BlobWriterState::Compressed { ref mut compr, .. } => { compr.flush() } BlobWriterState::Signed { ref mut writer, .. } => { writer.flush() } } } } /// Read compressed data blobs pub struct CompressedDataBlobReader { decompr: zstd::stream::read::Decoder, hasher: Option, expected_crc: u32, } impl CompressedDataBlobReader { pub fn new(mut reader: R) -> Result { let head: DataBlobHeader = unsafe { reader.read_le_value()? }; if head.magic != COMPRESSED_BLOB_MAGIC_1_0 { bail!("got wrong magic number"); } let expected_crc = u32::from_le_bytes(head.crc); let decompr = zstd::stream::read::Decoder::with_buffer(reader)?; Ok(Self { decompr: decompr, hasher: Some(crc32fast::Hasher::new()), expected_crc }) } } impl Read for CompressedDataBlobReader { fn read(&mut self, buf: &mut [u8]) -> Result { let count = self.decompr.read(buf)?; if count == 0 { // EOF, verify crc let hasher = self.hasher.take().expect("blob reader already finished"); let crc = hasher.finalize(); if crc != self.expected_crc { return Err(std::io::Error::new(std::io::ErrorKind::Other, "blob reader crc error")); } } else { let hasher = self.hasher.as_mut().expect("blob reader already finished"); hasher.update(buf); } Ok(count) } }