| 1 |
/*! |
| 2 |
Types and routines that support the wire format of finite automata. |
| 3 |
|
| 4 |
Currently, this module just exports a few error types and some small helpers |
| 5 |
for deserializing [dense DFAs](crate::dfa::dense::DFA) using correct alignment. |
| 6 |
*/ |
| 7 |
|
| 8 |
/* |
| 9 |
A collection of helper functions, types and traits for serializing automata. |
| 10 |
|
| 11 |
This crate defines its own bespoke serialization mechanism for some structures |
| 12 |
provided in the public API, namely, DFAs. A bespoke mechanism was developed |
| 13 |
primarily because structures like automata demand a specific binary format. |
| 14 |
Attempting to encode their rich structure in an existing serialization |
| 15 |
format is just not feasible. Moreover, the format for each structure is |
| 16 |
generally designed such that deserialization is cheap. More specifically, that |
| 17 |
deserialization can be done in constant time. (The idea being that you can |
| 18 |
embed it into your binary or mmap it, and then use it immediately.) |
| 19 |
|
| 20 |
In order to achieve this, the dense and sparse DFAs in this crate use an |
| 21 |
in-memory representation that very closely corresponds to its binary serialized |
| 22 |
form. This pervades and complicates everything, and in some cases, requires |
| 23 |
dealing with alignment and reasoning about safety. |
| 24 |
|
| 25 |
This technique does have major advantages. In particular, it permits doing |
| 26 |
the potentially costly work of compiling a finite state machine in an offline |
| 27 |
manner, and then loading it at runtime not only without having to re-compile |
| 28 |
the regex, but even without the code required to do the compilation. This, for |
| 29 |
example, permits one to use a pre-compiled DFA not only in environments without |
| 30 |
Rust's standard library, but also in environments without a heap. |
| 31 |
|
| 32 |
In the code below, whenever we insert some kind of padding, it's to enforce a |
| 33 |
4-byte alignment, unless otherwise noted. Namely, u32 is the only state ID type |
| 34 |
supported. (In a previous version of this library, DFAs were generic over the |
| 35 |
state ID representation.) |
| 36 |
|
| 37 |
Also, serialization generally requires the caller to specify endianness, |
| 38 |
where as deserialization always assumes native endianness (otherwise cheap |
| 39 |
deserialization would be impossible). This implies that serializing a structure |
| 40 |
generally requires serializing both its big-endian and little-endian variants, |
| 41 |
and then loading the correct one based on the target's endianness. |
| 42 |
*/ |
| 43 |
|
| 44 |
use core::{cmp, mem::size_of}; |
| 45 |
|
| 46 |
#[cfg(feature = "alloc")] |
| 47 |
use alloc::{vec, vec::Vec}; |
| 48 |
|
| 49 |
use crate::util::{ |
| 50 |
int::Pointer, |
| 51 |
primitives::{PatternID, PatternIDError, StateID, StateIDError}, |
| 52 |
}; |
| 53 |
|
| 54 |
/// A hack to align a smaller type `B` with a bigger type `T`. |
| 55 |
/// |
| 56 |
/// The usual use of this is with `B = [u8]` and `T = u32`. That is, |
| 57 |
/// it permits aligning a sequence of bytes on a 4-byte boundary. This |
| 58 |
/// is useful in contexts where one wants to embed a serialized [dense |
| 59 |
/// DFA](crate::dfa::dense::DFA) into a Rust a program while guaranteeing the |
| 60 |
/// alignment required for the DFA. |
| 61 |
/// |
| 62 |
/// See [`dense::DFA::from_bytes`](crate::dfa::dense::DFA::from_bytes) for an |
| 63 |
/// example of how to use this type. |
| 64 |
#[repr(C)] |
| 65 |
#[derive(Debug)] |
| 66 |
pub struct AlignAs<B: ?Sized, T> { |
| 67 |
/// A zero-sized field indicating the alignment we want. |
| 68 |
pub _align: [T; 0], |
| 69 |
/// A possibly non-sized field containing a sequence of bytes. |
| 70 |
pub bytes: B, |
| 71 |
} |
| 72 |
|
| 73 |
/// An error that occurs when serializing an object from this crate. |
| 74 |
/// |
| 75 |
/// Serialization, as used in this crate, universally refers to the process |
| 76 |
/// of transforming a structure (like a DFA) into a custom binary format |
| 77 |
/// represented by `&[u8]`. To this end, serialization is generally infallible. |
| 78 |
/// However, it can fail when caller provided buffer sizes are too small. When |
| 79 |
/// that occurs, a serialization error is reported. |
| 80 |
/// |
| 81 |
/// A `SerializeError` provides no introspection capabilities. Its only |
| 82 |
/// supported operation is conversion to a human readable error message. |
| 83 |
/// |
| 84 |
/// This error type implements the `std::error::Error` trait only when the |
| 85 |
/// `std` feature is enabled. Otherwise, this type is defined in all |
| 86 |
/// configurations. |
| 87 |
#[derive(Debug)] |
| 88 |
pub struct SerializeError { |
| 89 |
/// The name of the thing that a buffer is too small for. |
| 90 |
/// |
| 91 |
/// Currently, the only kind of serialization error is one that is |
| 92 |
/// committed by a caller: providing a destination buffer that is too |
| 93 |
/// small to fit the serialized object. This makes sense conceptually, |
| 94 |
/// since every valid inhabitant of a type should be serializable. |
| 95 |
/// |
| 96 |
/// This is somewhat exposed in the public API of this crate. For example, |
| 97 |
/// the `to_bytes_{big,little}_endian` APIs return a `Vec<u8>` and are |
| 98 |
/// guaranteed to never panic or error. This is only possible because the |
| 99 |
/// implementation guarantees that it will allocate a `Vec<u8>` that is |
| 100 |
/// big enough. |
| 101 |
/// |
| 102 |
/// In summary, if a new serialization error kind needs to be added, then |
| 103 |
/// it will need careful consideration. |
| 104 |
what: &'static str, |
| 105 |
} |
| 106 |
|
| 107 |
impl SerializeError { |
| 108 |
pub(crate) fn buffer_too_small(what: &'static str) -> SerializeError { |
| 109 |
SerializeError { what } |
| 110 |
} |
| 111 |
} |
| 112 |
|
| 113 |
impl core::fmt::Display for SerializeError { |
| 114 |
fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result { |
| 115 |
write!(f, "destination buffer is too small to write {}", self.what) |
| 116 |
} |
| 117 |
} |
| 118 |
|
| 119 |
#[cfg(feature = "std")] |
| 120 |
impl std::error::Error for SerializeError {} |
| 121 |
|
| 122 |
/// An error that occurs when deserializing an object defined in this crate. |
| 123 |
/// |
| 124 |
/// Serialization, as used in this crate, universally refers to the process |
| 125 |
/// of transforming a structure (like a DFA) into a custom binary format |
| 126 |
/// represented by `&[u8]`. Deserialization, then, refers to the process of |
| 127 |
/// cheaply converting this binary format back to the object's in-memory |
| 128 |
/// representation as defined in this crate. To the extent possible, |
| 129 |
/// deserialization will report this error whenever this process fails. |
| 130 |
/// |
| 131 |
/// A `DeserializeError` provides no introspection capabilities. Its only |
| 132 |
/// supported operation is conversion to a human readable error message. |
| 133 |
/// |
| 134 |
/// This error type implements the `std::error::Error` trait only when the |
| 135 |
/// `std` feature is enabled. Otherwise, this type is defined in all |
| 136 |
/// configurations. |
| 137 |
#[derive(Debug)] |
| 138 |
pub struct DeserializeError(DeserializeErrorKind); |
| 139 |
|
| 140 |
#[derive(Debug)] |
| 141 |
enum DeserializeErrorKind { |
| 142 |
Generic { msg: &'static str }, |
| 143 |
BufferTooSmall { what: &'static str }, |
| 144 |
InvalidUsize { what: &'static str }, |
| 145 |
VersionMismatch { expected: u32, found: u32 }, |
| 146 |
EndianMismatch { expected: u32, found: u32 }, |
| 147 |
AlignmentMismatch { alignment: usize, address: usize }, |
| 148 |
LabelMismatch { expected: &'static str }, |
| 149 |
ArithmeticOverflow { what: &'static str }, |
| 150 |
PatternID { err: PatternIDError, what: &'static str }, |
| 151 |
StateID { err: StateIDError, what: &'static str }, |
| 152 |
} |
| 153 |
|
| 154 |
impl DeserializeError { |
| 155 |
pub(crate) fn generic(msg: &'static str) -> DeserializeError { |
| 156 |
DeserializeError(DeserializeErrorKind::Generic { msg }) |
| 157 |
} |
| 158 |
|
| 159 |
pub(crate) fn buffer_too_small(what: &'static str) -> DeserializeError { |
| 160 |
DeserializeError(DeserializeErrorKind::BufferTooSmall { what }) |
| 161 |
} |
| 162 |
|
| 163 |
fn invalid_usize(what: &'static str) -> DeserializeError { |
| 164 |
DeserializeError(DeserializeErrorKind::InvalidUsize { what }) |
| 165 |
} |
| 166 |
|
| 167 |
fn version_mismatch(expected: u32, found: u32) -> DeserializeError { |
| 168 |
DeserializeError(DeserializeErrorKind::VersionMismatch { |
| 169 |
expected, |
| 170 |
found, |
| 171 |
}) |
| 172 |
} |
| 173 |
|
| 174 |
fn endian_mismatch(expected: u32, found: u32) -> DeserializeError { |
| 175 |
DeserializeError(DeserializeErrorKind::EndianMismatch { |
| 176 |
expected, |
| 177 |
found, |
| 178 |
}) |
| 179 |
} |
| 180 |
|
| 181 |
fn alignment_mismatch( |
| 182 |
alignment: usize, |
| 183 |
address: usize, |
| 184 |
) -> DeserializeError { |
| 185 |
DeserializeError(DeserializeErrorKind::AlignmentMismatch { |
| 186 |
alignment, |
| 187 |
address, |
| 188 |
}) |
| 189 |
} |
| 190 |
|
| 191 |
fn label_mismatch(expected: &'static str) -> DeserializeError { |
| 192 |
DeserializeError(DeserializeErrorKind::LabelMismatch { expected }) |
| 193 |
} |
| 194 |
|
| 195 |
fn arithmetic_overflow(what: &'static str) -> DeserializeError { |
| 196 |
DeserializeError(DeserializeErrorKind::ArithmeticOverflow { what }) |
| 197 |
} |
| 198 |
|
| 199 |
fn pattern_id_error( |
| 200 |
err: PatternIDError, |
| 201 |
what: &'static str, |
| 202 |
) -> DeserializeError { |
| 203 |
DeserializeError(DeserializeErrorKind::PatternID { err, what }) |
| 204 |
} |
| 205 |
|
| 206 |
pub(crate) fn state_id_error( |
| 207 |
err: StateIDError, |
| 208 |
what: &'static str, |
| 209 |
) -> DeserializeError { |
| 210 |
DeserializeError(DeserializeErrorKind::StateID { err, what }) |
| 211 |
} |
| 212 |
} |
| 213 |
|
| 214 |
#[cfg(feature = "std")] |
| 215 |
impl std::error::Error for DeserializeError {} |
| 216 |
|
| 217 |
impl core::fmt::Display for DeserializeError { |
| 218 |
fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result { |
| 219 |
use self::DeserializeErrorKind::*; |
| 220 |
|
| 221 |
match self.0 { |
| 222 |
Generic { msg } => write!(f, "{}", msg), |
| 223 |
BufferTooSmall { what } => { |
| 224 |
write!(f, "buffer is too small to read {}", what) |
| 225 |
} |
| 226 |
InvalidUsize { what } => { |
| 227 |
write!(f, "{} is too big to fit in a usize", what) |
| 228 |
} |
| 229 |
VersionMismatch { expected, found } => write!( |
| 230 |
f, |
| 231 |
"unsupported version: \ |
| 232 |
expected version {} but found version {}", |
| 233 |
expected, found, |
| 234 |
), |
| 235 |
EndianMismatch { expected, found } => write!( |
| 236 |
f, |
| 237 |
"endianness mismatch: expected 0x{:X} but got 0x{:X}. \ |
| 238 |
(Are you trying to load an object serialized with a \ |
| 239 |
different endianness?)", |
| 240 |
expected, found, |
| 241 |
), |
| 242 |
AlignmentMismatch { alignment, address } => write!( |
| 243 |
f, |
| 244 |
"alignment mismatch: slice starts at address \ |
| 245 |
0x{:X}, which is not aligned to a {} byte boundary", |
| 246 |
address, alignment, |
| 247 |
), |
| 248 |
LabelMismatch { expected } => write!( |
| 249 |
f, |
| 250 |
"label mismatch: start of serialized object should \ |
| 251 |
contain a NUL terminated {:?} label, but a different \ |
| 252 |
label was found", |
| 253 |
expected, |
| 254 |
), |
| 255 |
ArithmeticOverflow { what } => { |
| 256 |
write!(f, "arithmetic overflow for {}", what) |
| 257 |
} |
| 258 |
PatternID { ref err, what } => { |
| 259 |
write!(f, "failed to read pattern ID for {}: {}", what, err) |
| 260 |
} |
| 261 |
StateID { ref err, what } => { |
| 262 |
write!(f, "failed to read state ID for {}: {}", what, err) |
| 263 |
} |
| 264 |
} |
| 265 |
} |
| 266 |
} |
| 267 |
|
| 268 |
/// Safely converts a `&[u32]` to `&[StateID]` with zero cost. |
| 269 |
#[cfg_attr(feature = "perf-inline", inline(always))] |
| 270 |
pub(crate) fn u32s_to_state_ids(slice: &[u32]) -> &[StateID] { |
| 271 |
// SAFETY: This is safe because StateID is defined to have the same memory |
| 272 |
// representation as a u32 (it is repr(transparent)). While not every u32 |
| 273 |
// is a "valid" StateID, callers are not permitted to rely on the validity |
| 274 |
// of StateIDs for memory safety. It can only lead to logical errors. (This |
| 275 |
// is why StateID::new_unchecked is safe.) |
| 276 |
unsafe { |
| 277 |
core::slice::from_raw_parts( |
| 278 |
slice.as_ptr().cast::<StateID>(), |
| 279 |
slice.len(), |
| 280 |
) |
| 281 |
} |
| 282 |
} |
| 283 |
|
| 284 |
/// Safely converts a `&mut [u32]` to `&mut [StateID]` with zero cost. |
| 285 |
pub(crate) fn u32s_to_state_ids_mut(slice: &mut [u32]) -> &mut [StateID] { |
| 286 |
// SAFETY: This is safe because StateID is defined to have the same memory |
| 287 |
// representation as a u32 (it is repr(transparent)). While not every u32 |
| 288 |
// is a "valid" StateID, callers are not permitted to rely on the validity |
| 289 |
// of StateIDs for memory safety. It can only lead to logical errors. (This |
| 290 |
// is why StateID::new_unchecked is safe.) |
| 291 |
unsafe { |
| 292 |
core::slice::from_raw_parts_mut( |
| 293 |
slice.as_mut_ptr().cast::<StateID>(), |
| 294 |
slice.len(), |
| 295 |
) |
| 296 |
} |
| 297 |
} |
| 298 |
|
| 299 |
/// Safely converts a `&[u32]` to `&[PatternID]` with zero cost. |
| 300 |
#[cfg_attr(feature = "perf-inline", inline(always))] |
| 301 |
pub(crate) fn u32s_to_pattern_ids(slice: &[u32]) -> &[PatternID] { |
| 302 |
// SAFETY: This is safe because PatternID is defined to have the same |
| 303 |
// memory representation as a u32 (it is repr(transparent)). While not |
| 304 |
// every u32 is a "valid" PatternID, callers are not permitted to rely |
| 305 |
// on the validity of PatternIDs for memory safety. It can only lead to |
| 306 |
// logical errors. (This is why PatternID::new_unchecked is safe.) |
| 307 |
unsafe { |
| 308 |
core::slice::from_raw_parts( |
| 309 |
slice.as_ptr().cast::<PatternID>(), |
| 310 |
slice.len(), |
| 311 |
) |
| 312 |
} |
| 313 |
} |
| 314 |
|
| 315 |
/// Checks that the given slice has an alignment that matches `T`. |
| 316 |
/// |
| 317 |
/// This is useful for checking that a slice has an appropriate alignment |
| 318 |
/// before casting it to a &[T]. Note though that alignment is not itself |
| 319 |
/// sufficient to perform the cast for any `T`. |
| 320 |
pub(crate) fn check_alignment<T>( |
| 321 |
slice: &[u8], |
| 322 |
) -> Result<(), DeserializeError> { |
| 323 |
let alignment = core::mem::align_of::<T>(); |
| 324 |
let address = slice.as_ptr().as_usize(); |
| 325 |
if address % alignment == 0 { |
| 326 |
return Ok(()); |
| 327 |
} |
| 328 |
Err(DeserializeError::alignment_mismatch(alignment, address)) |
| 329 |
} |
| 330 |
|
| 331 |
/// Reads a possibly empty amount of padding, up to 7 bytes, from the beginning |
| 332 |
/// of the given slice. All padding bytes must be NUL bytes. |
| 333 |
/// |
| 334 |
/// This is useful because it can be theoretically necessary to pad the |
| 335 |
/// beginning of a serialized object with NUL bytes to ensure that it starts |
| 336 |
/// at a correctly aligned address. These padding bytes should come immediately |
| 337 |
/// before the label. |
| 338 |
/// |
| 339 |
/// This returns the number of bytes read from the given slice. |
| 340 |
pub(crate) fn skip_initial_padding(slice: &[u8]) -> usize { |
| 341 |
let mut nread = 0; |
| 342 |
while nread < 7 && nread < slice.len() && slice[nread] == 0 { |
| 343 |
nread += 1; |
| 344 |
} |
| 345 |
nread |
| 346 |
} |
| 347 |
|
| 348 |
/// Allocate a byte buffer of the given size, along with some initial padding |
| 349 |
/// such that `buf[padding..]` has the same alignment as `T`, where the |
| 350 |
/// alignment of `T` must be at most `8`. In particular, callers should treat |
| 351 |
/// the first N bytes (second return value) as padding bytes that must not be |
| 352 |
/// overwritten. In all cases, the following identity holds: |
| 353 |
/// |
| 354 |
/// ```ignore |
| 355 |
/// let (buf, padding) = alloc_aligned_buffer::<StateID>(SIZE); |
| 356 |
/// assert_eq!(SIZE, buf[padding..].len()); |
| 357 |
/// ``` |
| 358 |
/// |
| 359 |
/// In practice, padding is often zero. |
| 360 |
/// |
| 361 |
/// The requirement for `8` as a maximum here is somewhat arbitrary. In |
| 362 |
/// practice, we never need anything bigger in this crate, and so this function |
| 363 |
/// does some sanity asserts under the assumption of a max alignment of `8`. |
| 364 |
#[cfg(feature = "alloc")] |
| 365 |
pub(crate) fn alloc_aligned_buffer<T>(size: usize) -> (Vec<u8>, usize) { |
| 366 |
// NOTE: This is a kludge because there's no easy way to allocate a Vec<u8> |
| 367 |
// with an alignment guaranteed to be greater than 1. We could create a |
| 368 |
// Vec<u32>, but this cannot be safely transmuted to a Vec<u8> without |
| 369 |
// concern, since reallocing or dropping the Vec<u8> is UB (different |
| 370 |
// alignment than the initial allocation). We could define a wrapper type |
| 371 |
// to manage this for us, but it seems like more machinery than it's worth. |
| 372 |
let buf = vec![0; size]; |
| 373 |
let align = core::mem::align_of::<T>(); |
| 374 |
let address = buf.as_ptr().as_usize(); |
| 375 |
if address % align == 0 { |
| 376 |
return (buf, 0); |
| 377 |
} |
| 378 |
// Let's try this again. We have to create a totally new alloc with |
| 379 |
// the maximum amount of bytes we might need. We can't just extend our |
| 380 |
// pre-existing 'buf' because that might create a new alloc with a |
| 381 |
// different alignment. |
| 382 |
let extra = align - 1; |
| 383 |
let mut buf = vec![0; size + extra]; |
| 384 |
let address = buf.as_ptr().as_usize(); |
| 385 |
// The code below handles the case where 'address' is aligned to T, so if |
| 386 |
// we got lucky and 'address' is now aligned to T (when it previously |
| 387 |
// wasn't), then we're done. |
| 388 |
if address % align == 0 { |
| 389 |
buf.truncate(size); |
| 390 |
return (buf, 0); |
| 391 |
} |
| 392 |
let padding = ((address & !(align - 1)).checked_add(align).unwrap()) |
| 393 |
.checked_sub(address) |
| 394 |
.unwrap(); |
| 395 |
assert!(padding <= 7, "padding of {} is bigger than 7", padding); |
| 396 |
assert!( |
| 397 |
padding <= extra, |
| 398 |
"padding of {} is bigger than extra {} bytes", |
| 399 |
padding, |
| 400 |
extra |
| 401 |
); |
| 402 |
buf.truncate(size + padding); |
| 403 |
assert_eq!(size + padding, buf.len()); |
| 404 |
assert_eq!( |
| 405 |
0, |
| 406 |
buf[padding..].as_ptr().as_usize() % align, |
| 407 |
"expected end of initial padding to be aligned to {}", |
| 408 |
align, |
| 409 |
); |
| 410 |
(buf, padding) |
| 411 |
} |
| 412 |
|
| 413 |
/// Reads a NUL terminated label starting at the beginning of the given slice. |
| 414 |
/// |
| 415 |
/// If a NUL terminated label could not be found, then an error is returned. |
| 416 |
/// Similarly, if a label is found but doesn't match the expected label, then |
| 417 |
/// an error is returned. |
| 418 |
/// |
| 419 |
/// Upon success, the total number of bytes read (including padding bytes) is |
| 420 |
/// returned. |
| 421 |
pub(crate) fn read_label( |
| 422 |
slice: &[u8], |
| 423 |
expected_label: &'static str, |
| 424 |
) -> Result<usize, DeserializeError> { |
| 425 |
// Set an upper bound on how many bytes we scan for a NUL. Since no label |
| 426 |
// in this crate is longer than 256 bytes, if we can't find one within that |
| 427 |
// range, then we have corrupted data. |
| 428 |
let first_nul = |
| 429 |
slice[..cmp::min(slice.len(), 256)].iter().position(|&b| b == 0); |
| 430 |
let first_nul = match first_nul { |
| 431 |
Some(first_nul) => first_nul, |
| 432 |
None => { |
| 433 |
return Err(DeserializeError::generic( |
| 434 |
"could not find NUL terminated label \ |
| 435 |
at start of serialized object", |
| 436 |
)); |
| 437 |
} |
| 438 |
}; |
| 439 |
let len = first_nul + padding_len(first_nul); |
| 440 |
if slice.len() < len { |
| 441 |
return Err(DeserializeError::generic( |
| 442 |
"could not find properly sized label at start of serialized object" |
| 443 |
)); |
| 444 |
} |
| 445 |
if expected_label.as_bytes() != &slice[..first_nul] { |
| 446 |
return Err(DeserializeError::label_mismatch(expected_label)); |
| 447 |
} |
| 448 |
Ok(len) |
| 449 |
} |
| 450 |
|
| 451 |
/// Writes the given label to the buffer as a NUL terminated string. The label |
| 452 |
/// given must not contain NUL, otherwise this will panic. Similarly, the label |
| 453 |
/// must not be longer than 255 bytes, otherwise this will panic. |
| 454 |
/// |
| 455 |
/// Additional NUL bytes are written as necessary to ensure that the number of |
| 456 |
/// bytes written is always a multiple of 4. |
| 457 |
/// |
| 458 |
/// Upon success, the total number of bytes written (including padding) is |
| 459 |
/// returned. |
| 460 |
pub(crate) fn write_label( |
| 461 |
label: &str, |
| 462 |
dst: &mut [u8], |
| 463 |
) -> Result<usize, SerializeError> { |
| 464 |
let nwrite = write_label_len(label); |
| 465 |
if dst.len() < nwrite { |
| 466 |
return Err(SerializeError::buffer_too_small("label")); |
| 467 |
} |
| 468 |
dst[..label.len()].copy_from_slice(label.as_bytes()); |
| 469 |
for i in 0..(nwrite - label.len()) { |
| 470 |
dst[label.len() + i] = 0; |
| 471 |
} |
| 472 |
assert_eq!(nwrite % 4, 0); |
| 473 |
Ok(nwrite) |
| 474 |
} |
| 475 |
|
| 476 |
/// Returns the total number of bytes (including padding) that would be written |
| 477 |
/// for the given label. This panics if the given label contains a NUL byte or |
| 478 |
/// is longer than 255 bytes. (The size restriction exists so that searching |
| 479 |
/// for a label during deserialization can be done in small bounded space.) |
| 480 |
pub(crate) fn write_label_len(label: &str) -> usize { |
| 481 |
if label.len() > 255 { |
| 482 |
panic!("label must not be longer than 255 bytes"); |
| 483 |
} |
| 484 |
if label.as_bytes().iter().position(|&b| b == 0).is_some() { |
| 485 |
panic!("label must not contain NUL bytes"); |
| 486 |
} |
| 487 |
let label_len = label.len() + 1; // +1 for the NUL terminator |
| 488 |
label_len + padding_len(label_len) |
| 489 |
} |
| 490 |
|
| 491 |
/// Reads the endianness check from the beginning of the given slice and |
| 492 |
/// confirms that the endianness of the serialized object matches the expected |
| 493 |
/// endianness. If the slice is too small or if the endianness check fails, |
| 494 |
/// this returns an error. |
| 495 |
/// |
| 496 |
/// Upon success, the total number of bytes read is returned. |
| 497 |
pub(crate) fn read_endianness_check( |
| 498 |
slice: &[u8], |
| 499 |
) -> Result<usize, DeserializeError> { |
| 500 |
let (n, nr) = try_read_u32(slice, "endianness check")?; |
| 501 |
assert_eq!(nr, write_endianness_check_len()); |
| 502 |
if n != 0xFEFF { |
| 503 |
return Err(DeserializeError::endian_mismatch(0xFEFF, n)); |
| 504 |
} |
| 505 |
Ok(nr) |
| 506 |
} |
| 507 |
|
| 508 |
/// Writes 0xFEFF as an integer using the given endianness. |
| 509 |
/// |
| 510 |
/// This is useful for writing into the header of a serialized object. It can |
| 511 |
/// be read during deserialization as a sanity check to ensure the proper |
| 512 |
/// endianness is used. |
| 513 |
/// |
| 514 |
/// Upon success, the total number of bytes written is returned. |
| 515 |
pub(crate) fn write_endianness_check<E: Endian>( |
| 516 |
dst: &mut [u8], |
| 517 |
) -> Result<usize, SerializeError> { |
| 518 |
let nwrite = write_endianness_check_len(); |
| 519 |
if dst.len() < nwrite { |
| 520 |
return Err(SerializeError::buffer_too_small("endianness check")); |
| 521 |
} |
| 522 |
E::write_u32(0xFEFF, dst); |
| 523 |
Ok(nwrite) |
| 524 |
} |
| 525 |
|
| 526 |
/// Returns the number of bytes written by the endianness check. |
| 527 |
pub(crate) fn write_endianness_check_len() -> usize { |
| 528 |
size_of::<u32>() |
| 529 |
} |
| 530 |
|
| 531 |
/// Reads a version number from the beginning of the given slice and confirms |
| 532 |
/// that is matches the expected version number given. If the slice is too |
| 533 |
/// small or if the version numbers aren't equivalent, this returns an error. |
| 534 |
/// |
| 535 |
/// Upon success, the total number of bytes read is returned. |
| 536 |
/// |
| 537 |
/// N.B. Currently, we require that the version number is exactly equivalent. |
| 538 |
/// In the future, if we bump the version number without a semver bump, then |
| 539 |
/// we'll need to relax this a bit and support older versions. |
| 540 |
pub(crate) fn read_version( |
| 541 |
slice: &[u8], |
| 542 |
expected_version: u32, |
| 543 |
) -> Result<usize, DeserializeError> { |
| 544 |
let (n, nr) = try_read_u32(slice, "version")?; |
| 545 |
assert_eq!(nr, write_version_len()); |
| 546 |
if n != expected_version { |
| 547 |
return Err(DeserializeError::version_mismatch(expected_version, n)); |
| 548 |
} |
| 549 |
Ok(nr) |
| 550 |
} |
| 551 |
|
| 552 |
/// Writes the given version number to the beginning of the given slice. |
| 553 |
/// |
| 554 |
/// This is useful for writing into the header of a serialized object. It can |
| 555 |
/// be read during deserialization as a sanity check to ensure that the library |
| 556 |
/// code supports the format of the serialized object. |
| 557 |
/// |
| 558 |
/// Upon success, the total number of bytes written is returned. |
| 559 |
pub(crate) fn write_version<E: Endian>( |
| 560 |
version: u32, |
| 561 |
dst: &mut [u8], |
| 562 |
) -> Result<usize, SerializeError> { |
| 563 |
let nwrite = write_version_len(); |
| 564 |
if dst.len() < nwrite { |
| 565 |
return Err(SerializeError::buffer_too_small("version number")); |
| 566 |
} |
| 567 |
E::write_u32(version, dst); |
| 568 |
Ok(nwrite) |
| 569 |
} |
| 570 |
|
| 571 |
/// Returns the number of bytes written by writing the version number. |
| 572 |
pub(crate) fn write_version_len() -> usize { |
| 573 |
size_of::<u32>() |
| 574 |
} |
| 575 |
|
| 576 |
/// Reads a pattern ID from the given slice. If the slice has insufficient |
| 577 |
/// length, then this panics. If the deserialized integer exceeds the pattern |
| 578 |
/// ID limit for the current target, then this returns an error. |
| 579 |
/// |
| 580 |
/// Upon success, this also returns the number of bytes read. |
| 581 |
pub(crate) fn read_pattern_id( |
| 582 |
slice: &[u8], |
| 583 |
what: &'static str, |
| 584 |
) -> Result<(PatternID, usize), DeserializeError> { |
| 585 |
let bytes: [u8; PatternID::SIZE] = |
| 586 |
slice[..PatternID::SIZE].try_into().unwrap(); |
| 587 |
let pid = PatternID::from_ne_bytes(bytes) |
| 588 |
.map_err(|err| DeserializeError::pattern_id_error(err, what))?; |
| 589 |
Ok((pid, PatternID::SIZE)) |
| 590 |
} |
| 591 |
|
| 592 |
/// Reads a pattern ID from the given slice. If the slice has insufficient |
| 593 |
/// length, then this panics. Otherwise, the deserialized integer is assumed |
| 594 |
/// to be a valid pattern ID. |
| 595 |
/// |
| 596 |
/// This also returns the number of bytes read. |
| 597 |
pub(crate) fn read_pattern_id_unchecked(slice: &[u8]) -> (PatternID, usize) { |
| 598 |
let pid = PatternID::from_ne_bytes_unchecked( |
| 599 |
slice[..PatternID::SIZE].try_into().unwrap(), |
| 600 |
); |
| 601 |
(pid, PatternID::SIZE) |
| 602 |
} |
| 603 |
|
| 604 |
/// Write the given pattern ID to the beginning of the given slice of bytes |
| 605 |
/// using the specified endianness. The given slice must have length at least |
| 606 |
/// `PatternID::SIZE`, or else this panics. Upon success, the total number of |
| 607 |
/// bytes written is returned. |
| 608 |
pub(crate) fn write_pattern_id<E: Endian>( |
| 609 |
pid: PatternID, |
| 610 |
dst: &mut [u8], |
| 611 |
) -> usize { |
| 612 |
E::write_u32(pid.as_u32(), dst); |
| 613 |
PatternID::SIZE |
| 614 |
} |
| 615 |
|
| 616 |
/// Attempts to read a state ID from the given slice. If the slice has an |
| 617 |
/// insufficient number of bytes or if the state ID exceeds the limit for |
| 618 |
/// the current target, then this returns an error. |
| 619 |
/// |
| 620 |
/// Upon success, this also returns the number of bytes read. |
| 621 |
pub(crate) fn try_read_state_id( |
| 622 |
slice: &[u8], |
| 623 |
what: &'static str, |
| 624 |
) -> Result<(StateID, usize), DeserializeError> { |
| 625 |
if slice.len() < StateID::SIZE { |
| 626 |
return Err(DeserializeError::buffer_too_small(what)); |
| 627 |
} |
| 628 |
read_state_id(slice, what) |
| 629 |
} |
| 630 |
|
| 631 |
/// Reads a state ID from the given slice. If the slice has insufficient |
| 632 |
/// length, then this panics. If the deserialized integer exceeds the state ID |
| 633 |
/// limit for the current target, then this returns an error. |
| 634 |
/// |
| 635 |
/// Upon success, this also returns the number of bytes read. |
| 636 |
pub(crate) fn read_state_id( |
| 637 |
slice: &[u8], |
| 638 |
what: &'static str, |
| 639 |
) -> Result<(StateID, usize), DeserializeError> { |
| 640 |
let bytes: [u8; StateID::SIZE] = |
| 641 |
slice[..StateID::SIZE].try_into().unwrap(); |
| 642 |
let sid = StateID::from_ne_bytes(bytes) |
| 643 |
.map_err(|err| DeserializeError::state_id_error(err, what))?; |
| 644 |
Ok((sid, StateID::SIZE)) |
| 645 |
} |
| 646 |
|
| 647 |
/// Reads a state ID from the given slice. If the slice has insufficient |
| 648 |
/// length, then this panics. Otherwise, the deserialized integer is assumed |
| 649 |
/// to be a valid state ID. |
| 650 |
/// |
| 651 |
/// This also returns the number of bytes read. |
| 652 |
pub(crate) fn read_state_id_unchecked(slice: &[u8]) -> (StateID, usize) { |
| 653 |
let sid = StateID::from_ne_bytes_unchecked( |
| 654 |
slice[..StateID::SIZE].try_into().unwrap(), |
| 655 |
); |
| 656 |
(sid, StateID::SIZE) |
| 657 |
} |
| 658 |
|
| 659 |
/// Write the given state ID to the beginning of the given slice of bytes |
| 660 |
/// using the specified endianness. The given slice must have length at least |
| 661 |
/// `StateID::SIZE`, or else this panics. Upon success, the total number of |
| 662 |
/// bytes written is returned. |
| 663 |
pub(crate) fn write_state_id<E: Endian>( |
| 664 |
sid: StateID, |
| 665 |
dst: &mut [u8], |
| 666 |
) -> usize { |
| 667 |
E::write_u32(sid.as_u32(), dst); |
| 668 |
StateID::SIZE |
| 669 |
} |
| 670 |
|
| 671 |
/// Try to read a u16 as a usize from the beginning of the given slice in |
| 672 |
/// native endian format. If the slice has fewer than 2 bytes or if the |
| 673 |
/// deserialized number cannot be represented by usize, then this returns an |
| 674 |
/// error. The error message will include the `what` description of what is |
| 675 |
/// being deserialized, for better error messages. `what` should be a noun in |
| 676 |
/// singular form. |
| 677 |
/// |
| 678 |
/// Upon success, this also returns the number of bytes read. |
| 679 |
pub(crate) fn try_read_u16_as_usize( |
| 680 |
slice: &[u8], |
| 681 |
what: &'static str, |
| 682 |
) -> Result<(usize, usize), DeserializeError> { |
| 683 |
try_read_u16(slice, what).and_then(|(n, nr)| { |
| 684 |
usize::try_from(n) |
| 685 |
.map(|n| (n, nr)) |
| 686 |
.map_err(|_| DeserializeError::invalid_usize(what)) |
| 687 |
}) |
| 688 |
} |
| 689 |
|
| 690 |
/// Try to read a u32 as a usize from the beginning of the given slice in |
| 691 |
/// native endian format. If the slice has fewer than 4 bytes or if the |
| 692 |
/// deserialized number cannot be represented by usize, then this returns an |
| 693 |
/// error. The error message will include the `what` description of what is |
| 694 |
/// being deserialized, for better error messages. `what` should be a noun in |
| 695 |
/// singular form. |
| 696 |
/// |
| 697 |
/// Upon success, this also returns the number of bytes read. |
| 698 |
pub(crate) fn try_read_u32_as_usize( |
| 699 |
slice: &[u8], |
| 700 |
what: &'static str, |
| 701 |
) -> Result<(usize, usize), DeserializeError> { |
| 702 |
try_read_u32(slice, what).and_then(|(n, nr)| { |
| 703 |
usize::try_from(n) |
| 704 |
.map(|n| (n, nr)) |
| 705 |
.map_err(|_| DeserializeError::invalid_usize(what)) |
| 706 |
}) |
| 707 |
} |
| 708 |
|
| 709 |
/// Try to read a u16 from the beginning of the given slice in native endian |
| 710 |
/// format. If the slice has fewer than 2 bytes, then this returns an error. |
| 711 |
/// The error message will include the `what` description of what is being |
| 712 |
/// deserialized, for better error messages. `what` should be a noun in |
| 713 |
/// singular form. |
| 714 |
/// |
| 715 |
/// Upon success, this also returns the number of bytes read. |
| 716 |
pub(crate) fn try_read_u16( |
| 717 |
slice: &[u8], |
| 718 |
what: &'static str, |
| 719 |
) -> Result<(u16, usize), DeserializeError> { |
| 720 |
check_slice_len(slice, size_of::<u16>(), what)?; |
| 721 |
Ok((read_u16(slice), size_of::<u16>())) |
| 722 |
} |
| 723 |
|
| 724 |
/// Try to read a u32 from the beginning of the given slice in native endian |
| 725 |
/// format. If the slice has fewer than 4 bytes, then this returns an error. |
| 726 |
/// The error message will include the `what` description of what is being |
| 727 |
/// deserialized, for better error messages. `what` should be a noun in |
| 728 |
/// singular form. |
| 729 |
/// |
| 730 |
/// Upon success, this also returns the number of bytes read. |
| 731 |
pub(crate) fn try_read_u32( |
| 732 |
slice: &[u8], |
| 733 |
what: &'static str, |
| 734 |
) -> Result<(u32, usize), DeserializeError> { |
| 735 |
check_slice_len(slice, size_of::<u32>(), what)?; |
| 736 |
Ok((read_u32(slice), size_of::<u32>())) |
| 737 |
} |
| 738 |
|
| 739 |
/// Try to read a u128 from the beginning of the given slice in native endian |
| 740 |
/// format. If the slice has fewer than 16 bytes, then this returns an error. |
| 741 |
/// The error message will include the `what` description of what is being |
| 742 |
/// deserialized, for better error messages. `what` should be a noun in |
| 743 |
/// singular form. |
| 744 |
/// |
| 745 |
/// Upon success, this also returns the number of bytes read. |
| 746 |
pub(crate) fn try_read_u128( |
| 747 |
slice: &[u8], |
| 748 |
what: &'static str, |
| 749 |
) -> Result<(u128, usize), DeserializeError> { |
| 750 |
check_slice_len(slice, size_of::<u128>(), what)?; |
| 751 |
Ok((read_u128(slice), size_of::<u128>())) |
| 752 |
} |
| 753 |
|
| 754 |
/// Read a u16 from the beginning of the given slice in native endian format. |
| 755 |
/// If the slice has fewer than 2 bytes, then this panics. |
| 756 |
/// |
| 757 |
/// Marked as inline to speed up sparse searching which decodes integers from |
| 758 |
/// its automaton at search time. |
| 759 |
#[cfg_attr(feature = "perf-inline", inline(always))] |
| 760 |
pub(crate) fn read_u16(slice: &[u8]) -> u16 { |
| 761 |
let bytes: [u8; 2] = slice[..size_of::<u16>()].try_into().unwrap(); |
| 762 |
u16::from_ne_bytes(bytes) |
| 763 |
} |
| 764 |
|
| 765 |
/// Read a u32 from the beginning of the given slice in native endian format. |
| 766 |
/// If the slice has fewer than 4 bytes, then this panics. |
| 767 |
/// |
| 768 |
/// Marked as inline to speed up sparse searching which decodes integers from |
| 769 |
/// its automaton at search time. |
| 770 |
#[cfg_attr(feature = "perf-inline", inline(always))] |
| 771 |
pub(crate) fn read_u32(slice: &[u8]) -> u32 { |
| 772 |
let bytes: [u8; 4] = slice[..size_of::<u32>()].try_into().unwrap(); |
| 773 |
u32::from_ne_bytes(bytes) |
| 774 |
} |
| 775 |
|
| 776 |
/// Read a u128 from the beginning of the given slice in native endian format. |
| 777 |
/// If the slice has fewer than 16 bytes, then this panics. |
| 778 |
pub(crate) fn read_u128(slice: &[u8]) -> u128 { |
| 779 |
let bytes: [u8; 16] = slice[..size_of::<u128>()].try_into().unwrap(); |
| 780 |
u128::from_ne_bytes(bytes) |
| 781 |
} |
| 782 |
|
| 783 |
/// Checks that the given slice has some minimal length. If it's smaller than |
| 784 |
/// the bound given, then a "buffer too small" error is returned with `what` |
| 785 |
/// describing what the buffer represents. |
| 786 |
pub(crate) fn check_slice_len<T>( |
| 787 |
slice: &[T], |
| 788 |
at_least_len: usize, |
| 789 |
what: &'static str, |
| 790 |
) -> Result<(), DeserializeError> { |
| 791 |
if slice.len() < at_least_len { |
| 792 |
return Err(DeserializeError::buffer_too_small(what)); |
| 793 |
} |
| 794 |
Ok(()) |
| 795 |
} |
| 796 |
|
| 797 |
/// Multiply the given numbers, and on overflow, return an error that includes |
| 798 |
/// 'what' in the error message. |
| 799 |
/// |
| 800 |
/// This is useful when doing arithmetic with untrusted data. |
| 801 |
pub(crate) fn mul( |
| 802 |
a: usize, |
| 803 |
b: usize, |
| 804 |
what: &'static str, |
| 805 |
) -> Result<usize, DeserializeError> { |
| 806 |
match a.checked_mul(b) { |
| 807 |
Some(c) => Ok(c), |
| 808 |
None => Err(DeserializeError::arithmetic_overflow(what)), |
| 809 |
} |
| 810 |
} |
| 811 |
|
| 812 |
/// Add the given numbers, and on overflow, return an error that includes |
| 813 |
/// 'what' in the error message. |
| 814 |
/// |
| 815 |
/// This is useful when doing arithmetic with untrusted data. |
| 816 |
pub(crate) fn add( |
| 817 |
a: usize, |
| 818 |
b: usize, |
| 819 |
what: &'static str, |
| 820 |
) -> Result<usize, DeserializeError> { |
| 821 |
match a.checked_add(b) { |
| 822 |
Some(c) => Ok(c), |
| 823 |
None => Err(DeserializeError::arithmetic_overflow(what)), |
| 824 |
} |
| 825 |
} |
| 826 |
|
| 827 |
/// Shift `a` left by `b`, and on overflow, return an error that includes |
| 828 |
/// 'what' in the error message. |
| 829 |
/// |
| 830 |
/// This is useful when doing arithmetic with untrusted data. |
| 831 |
pub(crate) fn shl( |
| 832 |
a: usize, |
| 833 |
b: usize, |
| 834 |
what: &'static str, |
| 835 |
) -> Result<usize, DeserializeError> { |
| 836 |
let amount = u32::try_from(b) |
| 837 |
.map_err(|_| DeserializeError::arithmetic_overflow(what))?; |
| 838 |
match a.checked_shl(amount) { |
| 839 |
Some(c) => Ok(c), |
| 840 |
None => Err(DeserializeError::arithmetic_overflow(what)), |
| 841 |
} |
| 842 |
} |
| 843 |
|
| 844 |
/// Returns the number of additional bytes required to add to the given length |
| 845 |
/// in order to make the total length a multiple of 4. The return value is |
| 846 |
/// always less than 4. |
| 847 |
pub(crate) fn padding_len(non_padding_len: usize) -> usize { |
| 848 |
(4 - (non_padding_len & 0b11)) & 0b11 |
| 849 |
} |
| 850 |
|
| 851 |
/// A simple trait for writing code generic over endianness. |
| 852 |
/// |
| 853 |
/// This is similar to what byteorder provides, but we only need a very small |
| 854 |
/// subset. |
| 855 |
pub(crate) trait Endian { |
| 856 |
/// Writes a u16 to the given destination buffer in a particular |
| 857 |
/// endianness. If the destination buffer has a length smaller than 2, then |
| 858 |
/// this panics. |
| 859 |
fn write_u16(n: u16, dst: &mut [u8]); |
| 860 |
|
| 861 |
/// Writes a u32 to the given destination buffer in a particular |
| 862 |
/// endianness. If the destination buffer has a length smaller than 4, then |
| 863 |
/// this panics. |
| 864 |
fn write_u32(n: u32, dst: &mut [u8]); |
| 865 |
|
| 866 |
/// Writes a u128 to the given destination buffer in a particular |
| 867 |
/// endianness. If the destination buffer has a length smaller than 16, |
| 868 |
/// then this panics. |
| 869 |
fn write_u128(n: u128, dst: &mut [u8]); |
| 870 |
} |
| 871 |
|
| 872 |
/// Little endian writing. |
| 873 |
pub(crate) enum LE {} |
| 874 |
/// Big endian writing. |
| 875 |
pub(crate) enum BE {} |
| 876 |
|
| 877 |
#[cfg(target_endian = "little")] |
| 878 |
pub(crate) type NE = LE; |
| 879 |
#[cfg(target_endian = "big")] |
| 880 |
pub(crate) type NE = BE; |
| 881 |
|
| 882 |
impl Endian for LE { |
| 883 |
fn write_u16(n: u16, dst: &mut [u8]) { |
| 884 |
dst[..2].copy_from_slice(&n.to_le_bytes()); |
| 885 |
} |
| 886 |
|
| 887 |
fn write_u32(n: u32, dst: &mut [u8]) { |
| 888 |
dst[..4].copy_from_slice(&n.to_le_bytes()); |
| 889 |
} |
| 890 |
|
| 891 |
fn write_u128(n: u128, dst: &mut [u8]) { |
| 892 |
dst[..16].copy_from_slice(&n.to_le_bytes()); |
| 893 |
} |
| 894 |
} |
| 895 |
|
| 896 |
impl Endian for BE { |
| 897 |
fn write_u16(n: u16, dst: &mut [u8]) { |
| 898 |
dst[..2].copy_from_slice(&n.to_be_bytes()); |
| 899 |
} |
| 900 |
|
| 901 |
fn write_u32(n: u32, dst: &mut [u8]) { |
| 902 |
dst[..4].copy_from_slice(&n.to_be_bytes()); |
| 903 |
} |
| 904 |
|
| 905 |
fn write_u128(n: u128, dst: &mut [u8]) { |
| 906 |
dst[..16].copy_from_slice(&n.to_be_bytes()); |
| 907 |
} |
| 908 |
} |
| 909 |
|
| 910 |
#[cfg(all(test, feature = "alloc"))] |
| 911 |
mod tests { |
| 912 |
use super::*; |
| 913 |
|
| 914 |
#[test] |
| 915 |
fn labels() { |
| 916 |
let mut buf = [0; 1024]; |
| 917 |
|
| 918 |
let nwrite = write_label("fooba", &mut buf).unwrap(); |
| 919 |
assert_eq!(nwrite, 8); |
| 920 |
assert_eq!(&buf[..nwrite], b"fooba\x00\x00\x00"); |
| 921 |
|
| 922 |
let nread = read_label(&buf, "fooba").unwrap(); |
| 923 |
assert_eq!(nread, 8); |
| 924 |
} |
| 925 |
|
| 926 |
#[test] |
| 927 |
#[should_panic] |
| 928 |
fn bad_label_interior_nul() { |
| 929 |
// interior NULs are not allowed |
| 930 |
write_label("foo\x00bar", &mut [0; 1024]).unwrap(); |
| 931 |
} |
| 932 |
|
| 933 |
#[test] |
| 934 |
fn bad_label_almost_too_long() { |
| 935 |
// ok |
| 936 |
write_label(&"z".repeat(255), &mut [0; 1024]).unwrap(); |
| 937 |
} |
| 938 |
|
| 939 |
#[test] |
| 940 |
#[should_panic] |
| 941 |
fn bad_label_too_long() { |
| 942 |
// labels longer than 255 bytes are banned |
| 943 |
write_label(&"z".repeat(256), &mut [0; 1024]).unwrap(); |
| 944 |
} |
| 945 |
|
| 946 |
#[test] |
| 947 |
fn padding() { |
| 948 |
assert_eq!(0, padding_len(8)); |
| 949 |
assert_eq!(3, padding_len(9)); |
| 950 |
assert_eq!(2, padding_len(10)); |
| 951 |
assert_eq!(1, padding_len(11)); |
| 952 |
assert_eq!(0, padding_len(12)); |
| 953 |
assert_eq!(3, padding_len(13)); |
| 954 |
assert_eq!(2, padding_len(14)); |
| 955 |
assert_eq!(1, padding_len(15)); |
| 956 |
assert_eq!(0, padding_len(16)); |
| 957 |
} |
| 958 |
} |