| 1 | //! Extra streaming decompression functionality. |
| 2 | //! |
| 3 | //! As of now this is mainly intended for use to build a higher-level wrapper. |
| 4 | #[cfg (feature = "with-alloc" )] |
| 5 | use crate::alloc::boxed::Box; |
| 6 | use core::{cmp, mem}; |
| 7 | |
| 8 | use crate::inflate::core::{decompress, inflate_flags, DecompressorOxide, TINFL_LZ_DICT_SIZE}; |
| 9 | use crate::inflate::TINFLStatus; |
| 10 | use crate::{DataFormat, MZError, MZFlush, MZResult, MZStatus, StreamResult}; |
| 11 | |
| 12 | /// Tag that determines reset policy of [InflateState](struct.InflateState.html) |
| 13 | pub trait ResetPolicy { |
| 14 | /// Performs reset |
| 15 | fn reset(&self, state: &mut InflateState); |
| 16 | } |
| 17 | |
| 18 | /// Resets state, without performing expensive ops (e.g. zeroing buffer) |
| 19 | /// |
| 20 | /// Note that not zeroing buffer can lead to security issues when dealing with untrusted input. |
| 21 | pub struct MinReset; |
| 22 | |
| 23 | impl ResetPolicy for MinReset { |
| 24 | fn reset(&self, state: &mut InflateState) { |
| 25 | state.decompressor().init(); |
| 26 | state.dict_ofs = 0; |
| 27 | state.dict_avail = 0; |
| 28 | state.first_call = true; |
| 29 | state.has_flushed = false; |
| 30 | state.last_status = TINFLStatus::NeedsMoreInput; |
| 31 | } |
| 32 | } |
| 33 | |
| 34 | /// Resets state and zero memory, continuing to use the same data format. |
| 35 | pub struct ZeroReset; |
| 36 | |
| 37 | impl ResetPolicy for ZeroReset { |
| 38 | #[inline ] |
| 39 | fn reset(&self, state: &mut InflateState) { |
| 40 | MinReset.reset(state); |
| 41 | state.dict = [0; TINFL_LZ_DICT_SIZE]; |
| 42 | } |
| 43 | } |
| 44 | |
| 45 | /// Full reset of the state, including zeroing memory. |
| 46 | /// |
| 47 | /// Requires to provide new data format. |
| 48 | pub struct FullReset(pub DataFormat); |
| 49 | |
| 50 | impl ResetPolicy for FullReset { |
| 51 | #[inline ] |
| 52 | fn reset(&self, state: &mut InflateState) { |
| 53 | ZeroReset.reset(state); |
| 54 | state.data_format = self.0; |
| 55 | } |
| 56 | } |
| 57 | |
| 58 | /// A struct that compbines a decompressor with extra data for streaming decompression. |
| 59 | /// |
| 60 | pub struct InflateState { |
| 61 | /// Inner decompressor struct |
| 62 | decomp: DecompressorOxide, |
| 63 | |
| 64 | /// Buffer of input bytes for matches. |
| 65 | /// TODO: Could probably do this a bit cleaner with some |
| 66 | /// Cursor-like class. |
| 67 | /// We may also look into whether we need to keep a buffer here, or just one in the |
| 68 | /// decompressor struct. |
| 69 | dict: [u8; TINFL_LZ_DICT_SIZE], |
| 70 | /// Where in the buffer are we currently at? |
| 71 | dict_ofs: usize, |
| 72 | /// How many bytes of data to be flushed is there currently in the buffer? |
| 73 | dict_avail: usize, |
| 74 | |
| 75 | first_call: bool, |
| 76 | has_flushed: bool, |
| 77 | |
| 78 | /// Whether the input data is wrapped in a zlib header and checksum. |
| 79 | /// TODO: This should be stored in the decompressor. |
| 80 | data_format: DataFormat, |
| 81 | last_status: TINFLStatus, |
| 82 | } |
| 83 | |
| 84 | impl Default for InflateState { |
| 85 | fn default() -> Self { |
| 86 | InflateState { |
| 87 | decomp: DecompressorOxide::default(), |
| 88 | dict: [0; TINFL_LZ_DICT_SIZE], |
| 89 | dict_ofs: 0, |
| 90 | dict_avail: 0, |
| 91 | first_call: true, |
| 92 | has_flushed: false, |
| 93 | data_format: DataFormat::Raw, |
| 94 | last_status: TINFLStatus::NeedsMoreInput, |
| 95 | } |
| 96 | } |
| 97 | } |
| 98 | impl InflateState { |
| 99 | /// Create a new state. |
| 100 | /// |
| 101 | /// Note that this struct is quite large due to internal buffers, and as such storing it on |
| 102 | /// the stack is not recommended. |
| 103 | /// |
| 104 | /// # Parameters |
| 105 | /// `data_format`: Determines whether the compressed data is assumed to wrapped with zlib |
| 106 | /// metadata. |
| 107 | pub fn new(data_format: DataFormat) -> InflateState { |
| 108 | InflateState { |
| 109 | data_format, |
| 110 | ..Default::default() |
| 111 | } |
| 112 | } |
| 113 | |
| 114 | /// Create a new state on the heap. |
| 115 | /// |
| 116 | /// # Parameters |
| 117 | /// `data_format`: Determines whether the compressed data is assumed to wrapped with zlib |
| 118 | /// metadata. |
| 119 | #[cfg (feature = "with-alloc" )] |
| 120 | pub fn new_boxed(data_format: DataFormat) -> Box<InflateState> { |
| 121 | let mut b: Box<InflateState> = Box::default(); |
| 122 | b.data_format = data_format; |
| 123 | b |
| 124 | } |
| 125 | |
| 126 | /// Access the innner decompressor. |
| 127 | pub fn decompressor(&mut self) -> &mut DecompressorOxide { |
| 128 | &mut self.decomp |
| 129 | } |
| 130 | |
| 131 | /// Return the status of the last call to `inflate` with this `InflateState`. |
| 132 | pub const fn last_status(&self) -> TINFLStatus { |
| 133 | self.last_status |
| 134 | } |
| 135 | |
| 136 | /// Create a new state using miniz/zlib style window bits parameter. |
| 137 | /// |
| 138 | /// The decompressor does not support different window sizes. As such, |
| 139 | /// any positive (>0) value will set the zlib header flag, while a negative one |
| 140 | /// will not. |
| 141 | #[cfg (feature = "with-alloc" )] |
| 142 | pub fn new_boxed_with_window_bits(window_bits: i32) -> Box<InflateState> { |
| 143 | let mut b: Box<InflateState> = Box::default(); |
| 144 | b.data_format = DataFormat::from_window_bits(window_bits); |
| 145 | b |
| 146 | } |
| 147 | |
| 148 | #[inline ] |
| 149 | /// Reset the decompressor without re-allocating memory, using the given |
| 150 | /// data format. |
| 151 | pub fn reset(&mut self, data_format: DataFormat) { |
| 152 | self.reset_as(FullReset(data_format)); |
| 153 | } |
| 154 | |
| 155 | #[inline ] |
| 156 | /// Resets the state according to specified policy. |
| 157 | pub fn reset_as<T: ResetPolicy>(&mut self, policy: T) { |
| 158 | policy.reset(self) |
| 159 | } |
| 160 | } |
| 161 | |
| 162 | /// Try to decompress from `input` to `output` with the given [`InflateState`] |
| 163 | /// |
| 164 | /// # `flush` |
| 165 | /// |
| 166 | /// Generally, the various [`MZFlush`] flags have meaning only on the compression side. They can be |
| 167 | /// supplied here, but the only one that has any semantic meaning is [`MZFlush::Finish`], which is a |
| 168 | /// signal that the stream is expected to finish, and failing to do so is an error. It isn't |
| 169 | /// necessary to specify it when the stream ends; you'll still get returned a |
| 170 | /// [`MZStatus::StreamEnd`] anyway. Other values either have no effect or cause errors. It's |
| 171 | /// likely that you'll almost always just want to use [`MZFlush::None`]. |
| 172 | /// |
| 173 | /// # Errors |
| 174 | /// |
| 175 | /// Returns [`MZError::Buf`] if the size of the `output` slice is empty or no progress was made due |
| 176 | /// to lack of expected input data, or if called with [`MZFlush::Finish`] and input wasn't all |
| 177 | /// consumed. |
| 178 | /// |
| 179 | /// Returns [`MZError::Data`] if this or a a previous call failed with an error return from |
| 180 | /// [`TINFLStatus`]; probably indicates corrupted data. |
| 181 | /// |
| 182 | /// Returns [`MZError::Stream`] when called with [`MZFlush::Full`] (meaningless on |
| 183 | /// decompression), or when called without [`MZFlush::Finish`] after an earlier call with |
| 184 | /// [`MZFlush::Finish`] has been made. |
| 185 | pub fn inflate( |
| 186 | state: &mut InflateState, |
| 187 | input: &[u8], |
| 188 | output: &mut [u8], |
| 189 | flush: MZFlush, |
| 190 | ) -> StreamResult { |
| 191 | let mut bytes_consumed = 0; |
| 192 | let mut bytes_written = 0; |
| 193 | let mut next_in = input; |
| 194 | let mut next_out = output; |
| 195 | |
| 196 | if flush == MZFlush::Full { |
| 197 | return StreamResult::error(MZError::Stream); |
| 198 | } |
| 199 | |
| 200 | let mut decomp_flags = if state.data_format == DataFormat::Zlib { |
| 201 | inflate_flags::TINFL_FLAG_COMPUTE_ADLER32 |
| 202 | } else { |
| 203 | inflate_flags::TINFL_FLAG_IGNORE_ADLER32 |
| 204 | }; |
| 205 | |
| 206 | if (state.data_format == DataFormat::Zlib) |
| 207 | | (state.data_format == DataFormat::ZLibIgnoreChecksum) |
| 208 | { |
| 209 | decomp_flags |= inflate_flags::TINFL_FLAG_PARSE_ZLIB_HEADER; |
| 210 | } |
| 211 | |
| 212 | let first_call = state.first_call; |
| 213 | state.first_call = false; |
| 214 | if state.last_status == TINFLStatus::FailedCannotMakeProgress { |
| 215 | return StreamResult::error(MZError::Buf); |
| 216 | } |
| 217 | if (state.last_status as i32) < 0 { |
| 218 | return StreamResult::error(MZError::Data); |
| 219 | } |
| 220 | |
| 221 | if state.has_flushed && (flush != MZFlush::Finish) { |
| 222 | return StreamResult::error(MZError::Stream); |
| 223 | } |
| 224 | state.has_flushed |= flush == MZFlush::Finish; |
| 225 | |
| 226 | if (flush == MZFlush::Finish) && first_call { |
| 227 | decomp_flags |= inflate_flags::TINFL_FLAG_USING_NON_WRAPPING_OUTPUT_BUF; |
| 228 | |
| 229 | let status = decompress(&mut state.decomp, next_in, next_out, 0, decomp_flags); |
| 230 | let in_bytes = status.1; |
| 231 | let out_bytes = status.2; |
| 232 | let status = status.0; |
| 233 | |
| 234 | state.last_status = status; |
| 235 | |
| 236 | bytes_consumed += in_bytes; |
| 237 | bytes_written += out_bytes; |
| 238 | |
| 239 | let ret_status = { |
| 240 | if status == TINFLStatus::FailedCannotMakeProgress { |
| 241 | Err(MZError::Buf) |
| 242 | } else if (status as i32) < 0 { |
| 243 | Err(MZError::Data) |
| 244 | } else if status != TINFLStatus::Done { |
| 245 | state.last_status = TINFLStatus::Failed; |
| 246 | Err(MZError::Buf) |
| 247 | } else { |
| 248 | Ok(MZStatus::StreamEnd) |
| 249 | } |
| 250 | }; |
| 251 | return StreamResult { |
| 252 | bytes_consumed, |
| 253 | bytes_written, |
| 254 | status: ret_status, |
| 255 | }; |
| 256 | } |
| 257 | |
| 258 | if flush != MZFlush::Finish { |
| 259 | decomp_flags |= inflate_flags::TINFL_FLAG_HAS_MORE_INPUT; |
| 260 | } |
| 261 | |
| 262 | if state.dict_avail != 0 { |
| 263 | bytes_written += push_dict_out(state, &mut next_out); |
| 264 | return StreamResult { |
| 265 | bytes_consumed, |
| 266 | bytes_written, |
| 267 | status: Ok( |
| 268 | if (state.last_status == TINFLStatus::Done) && (state.dict_avail == 0) { |
| 269 | MZStatus::StreamEnd |
| 270 | } else { |
| 271 | MZStatus::Ok |
| 272 | }, |
| 273 | ), |
| 274 | }; |
| 275 | } |
| 276 | |
| 277 | let status = inflate_loop( |
| 278 | state, |
| 279 | &mut next_in, |
| 280 | &mut next_out, |
| 281 | &mut bytes_consumed, |
| 282 | &mut bytes_written, |
| 283 | decomp_flags, |
| 284 | flush, |
| 285 | ); |
| 286 | StreamResult { |
| 287 | bytes_consumed, |
| 288 | bytes_written, |
| 289 | status, |
| 290 | } |
| 291 | } |
| 292 | |
| 293 | fn inflate_loop( |
| 294 | state: &mut InflateState, |
| 295 | next_in: &mut &[u8], |
| 296 | next_out: &mut &mut [u8], |
| 297 | total_in: &mut usize, |
| 298 | total_out: &mut usize, |
| 299 | decomp_flags: u32, |
| 300 | flush: MZFlush, |
| 301 | ) -> MZResult { |
| 302 | let orig_in_len = next_in.len(); |
| 303 | loop { |
| 304 | let status = decompress( |
| 305 | &mut state.decomp, |
| 306 | next_in, |
| 307 | &mut state.dict, |
| 308 | state.dict_ofs, |
| 309 | decomp_flags, |
| 310 | ); |
| 311 | |
| 312 | let in_bytes = status.1; |
| 313 | let out_bytes = status.2; |
| 314 | let status = status.0; |
| 315 | |
| 316 | state.last_status = status; |
| 317 | |
| 318 | *next_in = &next_in[in_bytes..]; |
| 319 | *total_in += in_bytes; |
| 320 | |
| 321 | state.dict_avail = out_bytes; |
| 322 | *total_out += push_dict_out(state, next_out); |
| 323 | |
| 324 | // The stream was corrupted, and decompression failed. |
| 325 | if (status as i32) < 0 { |
| 326 | return Err(MZError::Data); |
| 327 | } |
| 328 | |
| 329 | // The decompressor has flushed all it's data and is waiting for more input, but |
| 330 | // there was no more input provided. |
| 331 | if (status == TINFLStatus::NeedsMoreInput) && orig_in_len == 0 { |
| 332 | return Err(MZError::Buf); |
| 333 | } |
| 334 | |
| 335 | if flush == MZFlush::Finish { |
| 336 | if status == TINFLStatus::Done { |
| 337 | // There is not enough space in the output buffer to flush the remaining |
| 338 | // decompressed data in the internal buffer. |
| 339 | return if state.dict_avail != 0 { |
| 340 | Err(MZError::Buf) |
| 341 | } else { |
| 342 | Ok(MZStatus::StreamEnd) |
| 343 | }; |
| 344 | // No more space in the output buffer, but we're not done. |
| 345 | } else if next_out.is_empty() { |
| 346 | return Err(MZError::Buf); |
| 347 | } |
| 348 | } else { |
| 349 | // We're not expected to finish, so it's fine if we can't flush everything yet. |
| 350 | let empty_buf = next_in.is_empty() || next_out.is_empty(); |
| 351 | if (status == TINFLStatus::Done) || empty_buf || (state.dict_avail != 0) { |
| 352 | return if (status == TINFLStatus::Done) && (state.dict_avail == 0) { |
| 353 | // No more data left, we're done. |
| 354 | Ok(MZStatus::StreamEnd) |
| 355 | } else { |
| 356 | // Ok for now, still waiting for more input data or output space. |
| 357 | Ok(MZStatus::Ok) |
| 358 | }; |
| 359 | } |
| 360 | } |
| 361 | } |
| 362 | } |
| 363 | |
| 364 | fn push_dict_out(state: &mut InflateState, next_out: &mut &mut [u8]) -> usize { |
| 365 | let n: usize = cmp::min(v1:state.dict_avail, v2:next_out.len()); |
| 366 | (next_out[..n]).copy_from_slice(&state.dict[state.dict_ofs..state.dict_ofs + n]); |
| 367 | *next_out = &mut mem::take(dest:next_out)[n..]; |
| 368 | state.dict_avail -= n; |
| 369 | state.dict_ofs = (state.dict_ofs + (n)) & (TINFL_LZ_DICT_SIZE - 1); |
| 370 | n |
| 371 | } |
| 372 | |
| 373 | #[cfg (all(test, feature = "with-alloc" ))] |
| 374 | mod test { |
| 375 | use super::{inflate, InflateState}; |
| 376 | use crate::{DataFormat, MZFlush, MZStatus}; |
| 377 | use alloc::vec; |
| 378 | |
| 379 | #[test ] |
| 380 | fn test_state() { |
| 381 | let encoded = [ |
| 382 | 120u8, 156, 243, 72, 205, 201, 201, 215, 81, 168, 202, 201, 76, 82, 4, 0, 27, 101, 4, |
| 383 | 19, |
| 384 | ]; |
| 385 | let mut out = vec![0; 50]; |
| 386 | let mut state = InflateState::new_boxed(DataFormat::Zlib); |
| 387 | let res = inflate(&mut state, &encoded, &mut out, MZFlush::Finish); |
| 388 | let status = res.status.expect("Failed to decompress!" ); |
| 389 | assert_eq!(status, MZStatus::StreamEnd); |
| 390 | assert_eq!(out[..res.bytes_written as usize], b"Hello, zlib!" [..]); |
| 391 | assert_eq!(res.bytes_consumed, encoded.len()); |
| 392 | |
| 393 | state.reset_as(super::ZeroReset); |
| 394 | out.iter_mut().map(|x| *x = 0).count(); |
| 395 | let res = inflate(&mut state, &encoded, &mut out, MZFlush::Finish); |
| 396 | let status = res.status.expect("Failed to decompress!" ); |
| 397 | assert_eq!(status, MZStatus::StreamEnd); |
| 398 | assert_eq!(out[..res.bytes_written as usize], b"Hello, zlib!" [..]); |
| 399 | assert_eq!(res.bytes_consumed, encoded.len()); |
| 400 | |
| 401 | state.reset_as(super::MinReset); |
| 402 | out.iter_mut().map(|x| *x = 0).count(); |
| 403 | let res = inflate(&mut state, &encoded, &mut out, MZFlush::Finish); |
| 404 | let status = res.status.expect("Failed to decompress!" ); |
| 405 | assert_eq!(status, MZStatus::StreamEnd); |
| 406 | assert_eq!(out[..res.bytes_written as usize], b"Hello, zlib!" [..]); |
| 407 | assert_eq!(res.bytes_consumed, encoded.len()); |
| 408 | assert_eq!(state.decompressor().adler32(), Some(459605011)); |
| 409 | |
| 410 | // Test state when not computing adler. |
| 411 | state = InflateState::new_boxed(DataFormat::ZLibIgnoreChecksum); |
| 412 | out.iter_mut().map(|x| *x = 0).count(); |
| 413 | let res = inflate(&mut state, &encoded, &mut out, MZFlush::Finish); |
| 414 | let status = res.status.expect("Failed to decompress!" ); |
| 415 | assert_eq!(status, MZStatus::StreamEnd); |
| 416 | assert_eq!(out[..res.bytes_written as usize], b"Hello, zlib!" [..]); |
| 417 | assert_eq!(res.bytes_consumed, encoded.len()); |
| 418 | // Not computed, so should be Some(1) |
| 419 | assert_eq!(state.decompressor().adler32(), Some(1)); |
| 420 | // Should still have the checksum read from the header file. |
| 421 | assert_eq!(state.decompressor().adler32_header(), Some(459605011)) |
| 422 | } |
| 423 | } |
| 424 | |