Move performance-critical Python functions to Rust: streaming I/O, multipart assembly, and AES-256-GCM encryption
This commit is contained in:
@@ -19,3 +19,6 @@ regex = "1"
|
||||
lru = "0.14"
|
||||
parking_lot = "0.12"
|
||||
percent-encoding = "2"
|
||||
aes-gcm = "0.10"
|
||||
hkdf = "0.12"
|
||||
uuid = { version = "1", features = ["v4"] }
|
||||
|
||||
192
myfsio_core/src/crypto.rs
Normal file
192
myfsio_core/src/crypto.rs
Normal file
@@ -0,0 +1,192 @@
|
||||
use aes_gcm::aead::Aead;
|
||||
use aes_gcm::{Aes256Gcm, KeyInit, Nonce};
|
||||
use hkdf::Hkdf;
|
||||
use pyo3::exceptions::{PyIOError, PyValueError};
|
||||
use pyo3::prelude::*;
|
||||
use sha2::Sha256;
|
||||
use std::fs::File;
|
||||
use std::io::{Read, Seek, SeekFrom, Write};
|
||||
|
||||
const DEFAULT_CHUNK_SIZE: usize = 65536;
|
||||
const HEADER_SIZE: usize = 4;
|
||||
|
||||
fn read_exact_chunk(reader: &mut impl Read, buf: &mut [u8]) -> std::io::Result<usize> {
|
||||
let mut filled = 0;
|
||||
while filled < buf.len() {
|
||||
match reader.read(&mut buf[filled..]) {
|
||||
Ok(0) => break,
|
||||
Ok(n) => filled += n,
|
||||
Err(ref e) if e.kind() == std::io::ErrorKind::Interrupted => continue,
|
||||
Err(e) => return Err(e),
|
||||
}
|
||||
}
|
||||
Ok(filled)
|
||||
}
|
||||
|
||||
fn derive_chunk_nonce(base_nonce: &[u8], chunk_index: u32) -> Result<[u8; 12], String> {
|
||||
let hkdf = Hkdf::<Sha256>::new(Some(base_nonce), b"chunk_nonce");
|
||||
let mut okm = [0u8; 12];
|
||||
hkdf.expand(&chunk_index.to_be_bytes(), &mut okm)
|
||||
.map_err(|e| format!("HKDF expand failed: {}", e))?;
|
||||
Ok(okm)
|
||||
}
|
||||
|
||||
#[pyfunction]
|
||||
#[pyo3(signature = (input_path, output_path, key, base_nonce, chunk_size=DEFAULT_CHUNK_SIZE))]
|
||||
pub fn encrypt_stream_chunked(
|
||||
py: Python<'_>,
|
||||
input_path: &str,
|
||||
output_path: &str,
|
||||
key: &[u8],
|
||||
base_nonce: &[u8],
|
||||
chunk_size: usize,
|
||||
) -> PyResult<u32> {
|
||||
if key.len() != 32 {
|
||||
return Err(PyValueError::new_err(format!(
|
||||
"Key must be 32 bytes, got {}",
|
||||
key.len()
|
||||
)));
|
||||
}
|
||||
if base_nonce.len() != 12 {
|
||||
return Err(PyValueError::new_err(format!(
|
||||
"Base nonce must be 12 bytes, got {}",
|
||||
base_nonce.len()
|
||||
)));
|
||||
}
|
||||
|
||||
let chunk_size = if chunk_size == 0 {
|
||||
DEFAULT_CHUNK_SIZE
|
||||
} else {
|
||||
chunk_size
|
||||
};
|
||||
|
||||
let inp = input_path.to_owned();
|
||||
let out = output_path.to_owned();
|
||||
let key_arr: [u8; 32] = key.try_into().unwrap();
|
||||
let nonce_arr: [u8; 12] = base_nonce.try_into().unwrap();
|
||||
|
||||
py.detach(move || {
|
||||
let cipher = Aes256Gcm::new(&key_arr.into());
|
||||
|
||||
let mut infile = File::open(&inp)
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to open input: {}", e)))?;
|
||||
let mut outfile = File::create(&out)
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to create output: {}", e)))?;
|
||||
|
||||
outfile
|
||||
.write_all(&[0u8; 4])
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to write header: {}", e)))?;
|
||||
|
||||
let mut buf = vec![0u8; chunk_size];
|
||||
let mut chunk_index: u32 = 0;
|
||||
|
||||
loop {
|
||||
let n = read_exact_chunk(&mut infile, &mut buf)
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to read: {}", e)))?;
|
||||
if n == 0 {
|
||||
break;
|
||||
}
|
||||
|
||||
let nonce_bytes = derive_chunk_nonce(&nonce_arr, chunk_index)
|
||||
.map_err(|e| PyValueError::new_err(e))?;
|
||||
let nonce = Nonce::from_slice(&nonce_bytes);
|
||||
|
||||
let encrypted = cipher
|
||||
.encrypt(nonce, &buf[..n])
|
||||
.map_err(|e| PyValueError::new_err(format!("Encrypt failed: {}", e)))?;
|
||||
|
||||
let size = encrypted.len() as u32;
|
||||
outfile
|
||||
.write_all(&size.to_be_bytes())
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to write chunk size: {}", e)))?;
|
||||
outfile
|
||||
.write_all(&encrypted)
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to write chunk: {}", e)))?;
|
||||
|
||||
chunk_index += 1;
|
||||
}
|
||||
|
||||
outfile
|
||||
.seek(SeekFrom::Start(0))
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to seek: {}", e)))?;
|
||||
outfile
|
||||
.write_all(&chunk_index.to_be_bytes())
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to write chunk count: {}", e)))?;
|
||||
|
||||
Ok(chunk_index)
|
||||
})
|
||||
}
|
||||
|
||||
#[pyfunction]
|
||||
pub fn decrypt_stream_chunked(
|
||||
py: Python<'_>,
|
||||
input_path: &str,
|
||||
output_path: &str,
|
||||
key: &[u8],
|
||||
base_nonce: &[u8],
|
||||
) -> PyResult<u32> {
|
||||
if key.len() != 32 {
|
||||
return Err(PyValueError::new_err(format!(
|
||||
"Key must be 32 bytes, got {}",
|
||||
key.len()
|
||||
)));
|
||||
}
|
||||
if base_nonce.len() != 12 {
|
||||
return Err(PyValueError::new_err(format!(
|
||||
"Base nonce must be 12 bytes, got {}",
|
||||
base_nonce.len()
|
||||
)));
|
||||
}
|
||||
|
||||
let inp = input_path.to_owned();
|
||||
let out = output_path.to_owned();
|
||||
let key_arr: [u8; 32] = key.try_into().unwrap();
|
||||
let nonce_arr: [u8; 12] = base_nonce.try_into().unwrap();
|
||||
|
||||
py.detach(move || {
|
||||
let cipher = Aes256Gcm::new(&key_arr.into());
|
||||
|
||||
let mut infile = File::open(&inp)
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to open input: {}", e)))?;
|
||||
let mut outfile = File::create(&out)
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to create output: {}", e)))?;
|
||||
|
||||
let mut header = [0u8; HEADER_SIZE];
|
||||
infile
|
||||
.read_exact(&mut header)
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to read header: {}", e)))?;
|
||||
let chunk_count = u32::from_be_bytes(header);
|
||||
|
||||
let mut size_buf = [0u8; HEADER_SIZE];
|
||||
for chunk_index in 0..chunk_count {
|
||||
infile
|
||||
.read_exact(&mut size_buf)
|
||||
.map_err(|e| {
|
||||
PyIOError::new_err(format!(
|
||||
"Failed to read chunk {} size: {}",
|
||||
chunk_index, e
|
||||
))
|
||||
})?;
|
||||
let chunk_size = u32::from_be_bytes(size_buf) as usize;
|
||||
|
||||
let mut encrypted = vec![0u8; chunk_size];
|
||||
infile.read_exact(&mut encrypted).map_err(|e| {
|
||||
PyIOError::new_err(format!("Failed to read chunk {}: {}", chunk_index, e))
|
||||
})?;
|
||||
|
||||
let nonce_bytes = derive_chunk_nonce(&nonce_arr, chunk_index)
|
||||
.map_err(|e| PyValueError::new_err(e))?;
|
||||
let nonce = Nonce::from_slice(&nonce_bytes);
|
||||
|
||||
let decrypted = cipher.decrypt(nonce, encrypted.as_ref()).map_err(|e| {
|
||||
PyValueError::new_err(format!("Decrypt chunk {} failed: {}", chunk_index, e))
|
||||
})?;
|
||||
|
||||
outfile.write_all(&decrypted).map_err(|e| {
|
||||
PyIOError::new_err(format!("Failed to write chunk {}: {}", chunk_index, e))
|
||||
})?;
|
||||
}
|
||||
|
||||
Ok(chunk_count)
|
||||
})
|
||||
}
|
||||
@@ -1,7 +1,9 @@
|
||||
mod crypto;
|
||||
mod hashing;
|
||||
mod metadata;
|
||||
mod sigv4;
|
||||
mod storage;
|
||||
mod streaming;
|
||||
mod validation;
|
||||
|
||||
use pyo3::prelude::*;
|
||||
@@ -38,6 +40,12 @@ mod myfsio_core {
|
||||
m.add_function(wrap_pyfunction!(storage::search_objects_scan, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(storage::build_object_cache, m)?)?;
|
||||
|
||||
m.add_function(wrap_pyfunction!(streaming::stream_to_file_with_md5, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(streaming::assemble_parts_with_md5, m)?)?;
|
||||
|
||||
m.add_function(wrap_pyfunction!(crypto::encrypt_stream_chunked, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(crypto::decrypt_stream_chunked, m)?)?;
|
||||
|
||||
Ok(())
|
||||
}
|
||||
}
|
||||
|
||||
107
myfsio_core/src/streaming.rs
Normal file
107
myfsio_core/src/streaming.rs
Normal file
@@ -0,0 +1,107 @@
|
||||
use md5::{Digest, Md5};
|
||||
use pyo3::exceptions::{PyIOError, PyValueError};
|
||||
use pyo3::prelude::*;
|
||||
use std::fs::{self, File};
|
||||
use std::io::{Read, Write};
|
||||
use uuid::Uuid;
|
||||
|
||||
const DEFAULT_CHUNK_SIZE: usize = 262144;
|
||||
|
||||
#[pyfunction]
|
||||
#[pyo3(signature = (stream, tmp_dir, chunk_size=DEFAULT_CHUNK_SIZE))]
|
||||
pub fn stream_to_file_with_md5(
|
||||
py: Python<'_>,
|
||||
stream: &Bound<'_, PyAny>,
|
||||
tmp_dir: &str,
|
||||
chunk_size: usize,
|
||||
) -> PyResult<(String, String, u64)> {
|
||||
let chunk_size = if chunk_size == 0 {
|
||||
DEFAULT_CHUNK_SIZE
|
||||
} else {
|
||||
chunk_size
|
||||
};
|
||||
|
||||
fs::create_dir_all(tmp_dir)
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to create tmp dir: {}", e)))?;
|
||||
|
||||
let tmp_name = format!("{}.tmp", Uuid::new_v4().as_hyphenated());
|
||||
let tmp_path_buf = std::path::PathBuf::from(tmp_dir).join(&tmp_name);
|
||||
let tmp_path = tmp_path_buf.to_string_lossy().into_owned();
|
||||
|
||||
let mut file = File::create(&tmp_path)
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to create temp file: {}", e)))?;
|
||||
let mut hasher = Md5::new();
|
||||
let mut total_bytes: u64 = 0;
|
||||
|
||||
let result: PyResult<()> = (|| {
|
||||
loop {
|
||||
let chunk: Vec<u8> = stream.call_method1("read", (chunk_size,))?.extract()?;
|
||||
if chunk.is_empty() {
|
||||
break;
|
||||
}
|
||||
hasher.update(&chunk);
|
||||
file.write_all(&chunk)
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to write: {}", e)))?;
|
||||
total_bytes += chunk.len() as u64;
|
||||
|
||||
py.check_signals()?;
|
||||
}
|
||||
Ok(())
|
||||
})();
|
||||
|
||||
if let Err(e) = result {
|
||||
drop(file);
|
||||
let _ = fs::remove_file(&tmp_path);
|
||||
return Err(e);
|
||||
}
|
||||
|
||||
drop(file);
|
||||
|
||||
let md5_hex = format!("{:x}", hasher.finalize());
|
||||
Ok((tmp_path, md5_hex, total_bytes))
|
||||
}
|
||||
|
||||
#[pyfunction]
|
||||
pub fn assemble_parts_with_md5(
|
||||
py: Python<'_>,
|
||||
part_paths: Vec<String>,
|
||||
dest_path: &str,
|
||||
) -> PyResult<String> {
|
||||
if part_paths.is_empty() {
|
||||
return Err(PyValueError::new_err("No parts to assemble"));
|
||||
}
|
||||
|
||||
let dest = dest_path.to_owned();
|
||||
let parts = part_paths;
|
||||
|
||||
py.detach(move || {
|
||||
if let Some(parent) = std::path::Path::new(&dest).parent() {
|
||||
fs::create_dir_all(parent)
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to create dest dir: {}", e)))?;
|
||||
}
|
||||
|
||||
let mut target = File::create(&dest)
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to create dest file: {}", e)))?;
|
||||
let mut hasher = Md5::new();
|
||||
let mut buf = vec![0u8; 1024 * 1024];
|
||||
|
||||
for part_path in &parts {
|
||||
let mut part = File::open(part_path)
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to open part {}: {}", part_path, e)))?;
|
||||
loop {
|
||||
let n = part
|
||||
.read(&mut buf)
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to read part: {}", e)))?;
|
||||
if n == 0 {
|
||||
break;
|
||||
}
|
||||
hasher.update(&buf[..n]);
|
||||
target
|
||||
.write_all(&buf[..n])
|
||||
.map_err(|e| PyIOError::new_err(format!("Failed to write: {}", e)))?;
|
||||
}
|
||||
}
|
||||
|
||||
Ok(format!("{:x}", hasher.finalize()))
|
||||
})
|
||||
}
|
||||
Reference in New Issue
Block a user