1 | // Copyright 2018 Developers of the Rand project. |
2 | // |
3 | // Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or |
4 | // https://www.apache.org/licenses/LICENSE-2.0> or the MIT license |
5 | // <LICENSE-MIT or https://opensource.org/licenses/MIT>, at your |
6 | // option. This file may not be copied, modified, or distributed |
7 | // except according to those terms. |
8 | |
9 | //! The `BlockRngCore` trait and implementation helpers |
10 | //! |
11 | //! The [`BlockRngCore`] trait exists to assist in the implementation of RNGs |
12 | //! which generate a block of data in a cache instead of returning generated |
13 | //! values directly. |
14 | //! |
15 | //! Usage of this trait is optional, but provides two advantages: |
16 | //! implementations only need to concern themselves with generation of the |
17 | //! block, not the various [`RngCore`] methods (especially [`fill_bytes`], where |
18 | //! the optimal implementations are not trivial), and this allows |
19 | //! `ReseedingRng` (see [`rand`](https://docs.rs/rand) crate) perform periodic |
20 | //! reseeding with very low overhead. |
21 | //! |
22 | //! # Example |
23 | //! |
24 | //! ```no_run |
25 | //! use rand_core::{RngCore, SeedableRng}; |
26 | //! use rand_core::block::{BlockRngCore, BlockRng}; |
27 | //! |
28 | //! struct MyRngCore; |
29 | //! |
30 | //! impl BlockRngCore for MyRngCore { |
31 | //! type Item = u32; |
32 | //! type Results = [u32; 16]; |
33 | //! |
34 | //! fn generate(&mut self, results: &mut Self::Results) { |
35 | //! unimplemented!() |
36 | //! } |
37 | //! } |
38 | //! |
39 | //! impl SeedableRng for MyRngCore { |
40 | //! type Seed = [u8; 32]; |
41 | //! fn from_seed(seed: Self::Seed) -> Self { |
42 | //! unimplemented!() |
43 | //! } |
44 | //! } |
45 | //! |
46 | //! // optionally, also implement CryptoRng for MyRngCore |
47 | //! |
48 | //! // Final RNG. |
49 | //! let mut rng = BlockRng::<MyRngCore>::seed_from_u64(0); |
50 | //! println!("First value: {}" , rng.next_u32()); |
51 | //! ``` |
52 | //! |
53 | //! [`BlockRngCore`]: crate::block::BlockRngCore |
54 | //! [`fill_bytes`]: RngCore::fill_bytes |
55 | |
56 | use crate::impls::{fill_via_u32_chunks, fill_via_u64_chunks}; |
57 | use crate::{CryptoRng, Error, RngCore, SeedableRng}; |
58 | use core::convert::AsRef; |
59 | use core::fmt; |
60 | #[cfg (feature = "serde1" )] |
61 | use serde::{Deserialize, Serialize}; |
62 | |
63 | /// A trait for RNGs which do not generate random numbers individually, but in |
64 | /// blocks (typically `[u32; N]`). This technique is commonly used by |
65 | /// cryptographic RNGs to improve performance. |
66 | /// |
67 | /// See the [module][crate::block] documentation for details. |
68 | pub trait BlockRngCore { |
69 | /// Results element type, e.g. `u32`. |
70 | type Item; |
71 | |
72 | /// Results type. This is the 'block' an RNG implementing `BlockRngCore` |
73 | /// generates, which will usually be an array like `[u32; 16]`. |
74 | type Results: AsRef<[Self::Item]> + AsMut<[Self::Item]> + Default; |
75 | |
76 | /// Generate a new block of results. |
77 | fn generate(&mut self, results: &mut Self::Results); |
78 | } |
79 | |
80 | /// A wrapper type implementing [`RngCore`] for some type implementing |
81 | /// [`BlockRngCore`] with `u32` array buffer; i.e. this can be used to implement |
82 | /// a full RNG from just a `generate` function. |
83 | /// |
84 | /// The `core` field may be accessed directly but the results buffer may not. |
85 | /// PRNG implementations can simply use a type alias |
86 | /// (`pub type MyRng = BlockRng<MyRngCore>;`) but might prefer to use a |
87 | /// wrapper type (`pub struct MyRng(BlockRng<MyRngCore>);`); the latter must |
88 | /// re-implement `RngCore` but hides the implementation details and allows |
89 | /// extra functionality to be defined on the RNG |
90 | /// (e.g. `impl MyRng { fn set_stream(...){...} }`). |
91 | /// |
92 | /// `BlockRng` has heavily optimized implementations of the [`RngCore`] methods |
93 | /// reading values from the results buffer, as well as |
94 | /// calling [`BlockRngCore::generate`] directly on the output array when |
95 | /// [`fill_bytes`] / [`try_fill_bytes`] is called on a large array. These methods |
96 | /// also handle the bookkeeping of when to generate a new batch of values. |
97 | /// |
98 | /// No whole generated `u32` values are thrown away and all values are consumed |
99 | /// in-order. [`next_u32`] simply takes the next available `u32` value. |
100 | /// [`next_u64`] is implemented by combining two `u32` values, least |
101 | /// significant first. [`fill_bytes`] and [`try_fill_bytes`] consume a whole |
102 | /// number of `u32` values, converting each `u32` to a byte slice in |
103 | /// little-endian order. If the requested byte length is not a multiple of 4, |
104 | /// some bytes will be discarded. |
105 | /// |
106 | /// See also [`BlockRng64`] which uses `u64` array buffers. Currently there is |
107 | /// no direct support for other buffer types. |
108 | /// |
109 | /// For easy initialization `BlockRng` also implements [`SeedableRng`]. |
110 | /// |
111 | /// [`next_u32`]: RngCore::next_u32 |
112 | /// [`next_u64`]: RngCore::next_u64 |
113 | /// [`fill_bytes`]: RngCore::fill_bytes |
114 | /// [`try_fill_bytes`]: RngCore::try_fill_bytes |
115 | #[derive (Clone)] |
116 | #[cfg_attr (feature = "serde1" , derive(Serialize, Deserialize))] |
117 | #[cfg_attr ( |
118 | feature = "serde1" , |
119 | serde( |
120 | bound = "for<'x> R: Serialize + Deserialize<'x> + Sized, for<'x> R::Results: Serialize + Deserialize<'x>" |
121 | ) |
122 | )] |
123 | pub struct BlockRng<R: BlockRngCore + ?Sized> { |
124 | results: R::Results, |
125 | index: usize, |
126 | /// The *core* part of the RNG, implementing the `generate` function. |
127 | pub core: R, |
128 | } |
129 | |
130 | // Custom Debug implementation that does not expose the contents of `results`. |
131 | impl<R: BlockRngCore + fmt::Debug> fmt::Debug for BlockRng<R> { |
132 | fn fmt(&self, fmt: &mut fmt::Formatter) -> fmt::Result { |
133 | fmt&mut DebugStruct<'_, '_>.debug_struct("BlockRng" ) |
134 | .field("core" , &self.core) |
135 | .field("result_len" , &self.results.as_ref().len()) |
136 | .field(name:"index" , &self.index) |
137 | .finish() |
138 | } |
139 | } |
140 | |
141 | impl<R: BlockRngCore> BlockRng<R> { |
142 | /// Create a new `BlockRng` from an existing RNG implementing |
143 | /// `BlockRngCore`. Results will be generated on first use. |
144 | #[inline ] |
145 | pub fn new(core: R) -> BlockRng<R> { |
146 | let results_empty = R::Results::default(); |
147 | BlockRng { |
148 | core, |
149 | index: results_empty.as_ref().len(), |
150 | results: results_empty, |
151 | } |
152 | } |
153 | |
154 | /// Get the index into the result buffer. |
155 | /// |
156 | /// If this is equal to or larger than the size of the result buffer then |
157 | /// the buffer is "empty" and `generate()` must be called to produce new |
158 | /// results. |
159 | #[inline (always)] |
160 | pub fn index(&self) -> usize { |
161 | self.index |
162 | } |
163 | |
164 | /// Reset the number of available results. |
165 | /// This will force a new set of results to be generated on next use. |
166 | #[inline ] |
167 | pub fn reset(&mut self) { |
168 | self.index = self.results.as_ref().len(); |
169 | } |
170 | |
171 | /// Generate a new set of results immediately, setting the index to the |
172 | /// given value. |
173 | #[inline ] |
174 | pub fn generate_and_set(&mut self, index: usize) { |
175 | assert!(index < self.results.as_ref().len()); |
176 | self.core.generate(&mut self.results); |
177 | self.index = index; |
178 | } |
179 | } |
180 | |
181 | impl<R: BlockRngCore<Item = u32>> RngCore for BlockRng<R> |
182 | where |
183 | <R as BlockRngCore>::Results: AsRef<[u32]> + AsMut<[u32]>, |
184 | { |
185 | #[inline ] |
186 | fn next_u32(&mut self) -> u32 { |
187 | if self.index >= self.results.as_ref().len() { |
188 | self.generate_and_set(0); |
189 | } |
190 | |
191 | let value = self.results.as_ref()[self.index]; |
192 | self.index += 1; |
193 | value |
194 | } |
195 | |
196 | #[inline ] |
197 | fn next_u64(&mut self) -> u64 { |
198 | let read_u64 = |results: &[u32], index| { |
199 | let data = &results[index..=index + 1]; |
200 | u64::from(data[1]) << 32 | u64::from(data[0]) |
201 | }; |
202 | |
203 | let len = self.results.as_ref().len(); |
204 | |
205 | let index = self.index; |
206 | if index < len - 1 { |
207 | self.index += 2; |
208 | // Read an u64 from the current index |
209 | read_u64(self.results.as_ref(), index) |
210 | } else if index >= len { |
211 | self.generate_and_set(2); |
212 | read_u64(self.results.as_ref(), 0) |
213 | } else { |
214 | let x = u64::from(self.results.as_ref()[len - 1]); |
215 | self.generate_and_set(1); |
216 | let y = u64::from(self.results.as_ref()[0]); |
217 | (y << 32) | x |
218 | } |
219 | } |
220 | |
221 | #[inline ] |
222 | fn fill_bytes(&mut self, dest: &mut [u8]) { |
223 | let mut read_len = 0; |
224 | while read_len < dest.len() { |
225 | if self.index >= self.results.as_ref().len() { |
226 | self.generate_and_set(0); |
227 | } |
228 | let (consumed_u32, filled_u8) = |
229 | fill_via_u32_chunks(&self.results.as_ref()[self.index..], &mut dest[read_len..]); |
230 | |
231 | self.index += consumed_u32; |
232 | read_len += filled_u8; |
233 | } |
234 | } |
235 | |
236 | #[inline (always)] |
237 | fn try_fill_bytes(&mut self, dest: &mut [u8]) -> Result<(), Error> { |
238 | self.fill_bytes(dest); |
239 | Ok(()) |
240 | } |
241 | } |
242 | |
243 | impl<R: BlockRngCore + SeedableRng> SeedableRng for BlockRng<R> { |
244 | type Seed = R::Seed; |
245 | |
246 | #[inline (always)] |
247 | fn from_seed(seed: Self::Seed) -> Self { |
248 | Self::new(R::from_seed(seed)) |
249 | } |
250 | |
251 | #[inline (always)] |
252 | fn seed_from_u64(seed: u64) -> Self { |
253 | Self::new(R::seed_from_u64(state:seed)) |
254 | } |
255 | |
256 | #[inline (always)] |
257 | fn from_rng<S: RngCore>(rng: S) -> Result<Self, Error> { |
258 | Ok(Self::new(R::from_rng(rng)?)) |
259 | } |
260 | } |
261 | |
262 | /// A wrapper type implementing [`RngCore`] for some type implementing |
263 | /// [`BlockRngCore`] with `u64` array buffer; i.e. this can be used to implement |
264 | /// a full RNG from just a `generate` function. |
265 | /// |
266 | /// This is similar to [`BlockRng`], but specialized for algorithms that operate |
267 | /// on `u64` values. |
268 | /// |
269 | /// No whole generated `u64` values are thrown away and all values are consumed |
270 | /// in-order. [`next_u64`] simply takes the next available `u64` value. |
271 | /// [`next_u32`] is however a bit special: half of a `u64` is consumed, leaving |
272 | /// the other half in the buffer. If the next function called is [`next_u32`] |
273 | /// then the other half is then consumed, however both [`next_u64`] and |
274 | /// [`fill_bytes`] discard the rest of any half-consumed `u64`s when called. |
275 | /// |
276 | /// [`fill_bytes`] and [`try_fill_bytes`] consume a whole number of `u64` |
277 | /// values. If the requested length is not a multiple of 8, some bytes will be |
278 | /// discarded. |
279 | /// |
280 | /// [`next_u32`]: RngCore::next_u32 |
281 | /// [`next_u64`]: RngCore::next_u64 |
282 | /// [`fill_bytes`]: RngCore::fill_bytes |
283 | /// [`try_fill_bytes`]: RngCore::try_fill_bytes |
284 | #[derive (Clone)] |
285 | #[cfg_attr (feature = "serde1" , derive(Serialize, Deserialize))] |
286 | pub struct BlockRng64<R: BlockRngCore + ?Sized> { |
287 | results: R::Results, |
288 | index: usize, |
289 | half_used: bool, // true if only half of the previous result is used |
290 | /// The *core* part of the RNG, implementing the `generate` function. |
291 | pub core: R, |
292 | } |
293 | |
294 | // Custom Debug implementation that does not expose the contents of `results`. |
295 | impl<R: BlockRngCore + fmt::Debug> fmt::Debug for BlockRng64<R> { |
296 | fn fmt(&self, fmt: &mut fmt::Formatter) -> fmt::Result { |
297 | fmt&mut DebugStruct<'_, '_>.debug_struct("BlockRng64" ) |
298 | .field("core" , &self.core) |
299 | .field("result_len" , &self.results.as_ref().len()) |
300 | .field("index" , &self.index) |
301 | .field(name:"half_used" , &self.half_used) |
302 | .finish() |
303 | } |
304 | } |
305 | |
306 | impl<R: BlockRngCore> BlockRng64<R> { |
307 | /// Create a new `BlockRng` from an existing RNG implementing |
308 | /// `BlockRngCore`. Results will be generated on first use. |
309 | #[inline ] |
310 | pub fn new(core: R) -> BlockRng64<R> { |
311 | let results_empty = R::Results::default(); |
312 | BlockRng64 { |
313 | core, |
314 | index: results_empty.as_ref().len(), |
315 | half_used: false, |
316 | results: results_empty, |
317 | } |
318 | } |
319 | |
320 | /// Get the index into the result buffer. |
321 | /// |
322 | /// If this is equal to or larger than the size of the result buffer then |
323 | /// the buffer is "empty" and `generate()` must be called to produce new |
324 | /// results. |
325 | #[inline (always)] |
326 | pub fn index(&self) -> usize { |
327 | self.index |
328 | } |
329 | |
330 | /// Reset the number of available results. |
331 | /// This will force a new set of results to be generated on next use. |
332 | #[inline ] |
333 | pub fn reset(&mut self) { |
334 | self.index = self.results.as_ref().len(); |
335 | self.half_used = false; |
336 | } |
337 | |
338 | /// Generate a new set of results immediately, setting the index to the |
339 | /// given value. |
340 | #[inline ] |
341 | pub fn generate_and_set(&mut self, index: usize) { |
342 | assert!(index < self.results.as_ref().len()); |
343 | self.core.generate(&mut self.results); |
344 | self.index = index; |
345 | self.half_used = false; |
346 | } |
347 | } |
348 | |
349 | impl<R: BlockRngCore<Item = u64>> RngCore for BlockRng64<R> |
350 | where |
351 | <R as BlockRngCore>::Results: AsRef<[u64]> + AsMut<[u64]>, |
352 | { |
353 | #[inline ] |
354 | fn next_u32(&mut self) -> u32 { |
355 | let mut index = self.index - self.half_used as usize; |
356 | if index >= self.results.as_ref().len() { |
357 | self.core.generate(&mut self.results); |
358 | self.index = 0; |
359 | index = 0; |
360 | // `self.half_used` is by definition `false` |
361 | self.half_used = false; |
362 | } |
363 | |
364 | let shift = 32 * (self.half_used as usize); |
365 | |
366 | self.half_used = !self.half_used; |
367 | self.index += self.half_used as usize; |
368 | |
369 | (self.results.as_ref()[index] >> shift) as u32 |
370 | } |
371 | |
372 | #[inline ] |
373 | fn next_u64(&mut self) -> u64 { |
374 | if self.index >= self.results.as_ref().len() { |
375 | self.core.generate(&mut self.results); |
376 | self.index = 0; |
377 | } |
378 | |
379 | let value = self.results.as_ref()[self.index]; |
380 | self.index += 1; |
381 | self.half_used = false; |
382 | value |
383 | } |
384 | |
385 | #[inline ] |
386 | fn fill_bytes(&mut self, dest: &mut [u8]) { |
387 | let mut read_len = 0; |
388 | self.half_used = false; |
389 | while read_len < dest.len() { |
390 | if self.index as usize >= self.results.as_ref().len() { |
391 | self.core.generate(&mut self.results); |
392 | self.index = 0; |
393 | } |
394 | |
395 | let (consumed_u64, filled_u8) = fill_via_u64_chunks( |
396 | &self.results.as_ref()[self.index as usize..], |
397 | &mut dest[read_len..], |
398 | ); |
399 | |
400 | self.index += consumed_u64; |
401 | read_len += filled_u8; |
402 | } |
403 | } |
404 | |
405 | #[inline (always)] |
406 | fn try_fill_bytes(&mut self, dest: &mut [u8]) -> Result<(), Error> { |
407 | self.fill_bytes(dest); |
408 | Ok(()) |
409 | } |
410 | } |
411 | |
412 | impl<R: BlockRngCore + SeedableRng> SeedableRng for BlockRng64<R> { |
413 | type Seed = R::Seed; |
414 | |
415 | #[inline (always)] |
416 | fn from_seed(seed: Self::Seed) -> Self { |
417 | Self::new(R::from_seed(seed)) |
418 | } |
419 | |
420 | #[inline (always)] |
421 | fn seed_from_u64(seed: u64) -> Self { |
422 | Self::new(R::seed_from_u64(state:seed)) |
423 | } |
424 | |
425 | #[inline (always)] |
426 | fn from_rng<S: RngCore>(rng: S) -> Result<Self, Error> { |
427 | Ok(Self::new(R::from_rng(rng)?)) |
428 | } |
429 | } |
430 | |
431 | impl<R: BlockRngCore + CryptoRng> CryptoRng for BlockRng<R> {} |
432 | |
433 | #[cfg (test)] |
434 | mod test { |
435 | use crate::{SeedableRng, RngCore}; |
436 | use crate::block::{BlockRng, BlockRng64, BlockRngCore}; |
437 | |
438 | #[derive (Debug, Clone)] |
439 | struct DummyRng { |
440 | counter: u32, |
441 | } |
442 | |
443 | impl BlockRngCore for DummyRng { |
444 | type Item = u32; |
445 | |
446 | type Results = [u32; 16]; |
447 | |
448 | fn generate(&mut self, results: &mut Self::Results) { |
449 | for r in results { |
450 | *r = self.counter; |
451 | self.counter = self.counter.wrapping_add(3511615421); |
452 | } |
453 | } |
454 | } |
455 | |
456 | impl SeedableRng for DummyRng { |
457 | type Seed = [u8; 4]; |
458 | |
459 | fn from_seed(seed: Self::Seed) -> Self { |
460 | DummyRng { counter: u32::from_le_bytes(seed) } |
461 | } |
462 | } |
463 | |
464 | #[test ] |
465 | fn blockrng_next_u32_vs_next_u64() { |
466 | let mut rng1 = BlockRng::<DummyRng>::from_seed([1, 2, 3, 4]); |
467 | let mut rng2 = rng1.clone(); |
468 | let mut rng3 = rng1.clone(); |
469 | |
470 | let mut a = [0; 16]; |
471 | (&mut a[..4]).copy_from_slice(&rng1.next_u32().to_le_bytes()); |
472 | (&mut a[4..12]).copy_from_slice(&rng1.next_u64().to_le_bytes()); |
473 | (&mut a[12..]).copy_from_slice(&rng1.next_u32().to_le_bytes()); |
474 | |
475 | let mut b = [0; 16]; |
476 | (&mut b[..4]).copy_from_slice(&rng2.next_u32().to_le_bytes()); |
477 | (&mut b[4..8]).copy_from_slice(&rng2.next_u32().to_le_bytes()); |
478 | (&mut b[8..]).copy_from_slice(&rng2.next_u64().to_le_bytes()); |
479 | assert_eq!(a, b); |
480 | |
481 | let mut c = [0; 16]; |
482 | (&mut c[..8]).copy_from_slice(&rng3.next_u64().to_le_bytes()); |
483 | (&mut c[8..12]).copy_from_slice(&rng3.next_u32().to_le_bytes()); |
484 | (&mut c[12..]).copy_from_slice(&rng3.next_u32().to_le_bytes()); |
485 | assert_eq!(a, c); |
486 | } |
487 | |
488 | #[derive (Debug, Clone)] |
489 | struct DummyRng64 { |
490 | counter: u64, |
491 | } |
492 | |
493 | impl BlockRngCore for DummyRng64 { |
494 | type Item = u64; |
495 | |
496 | type Results = [u64; 8]; |
497 | |
498 | fn generate(&mut self, results: &mut Self::Results) { |
499 | for r in results { |
500 | *r = self.counter; |
501 | self.counter = self.counter.wrapping_add(2781463553396133981); |
502 | } |
503 | } |
504 | } |
505 | |
506 | impl SeedableRng for DummyRng64 { |
507 | type Seed = [u8; 8]; |
508 | |
509 | fn from_seed(seed: Self::Seed) -> Self { |
510 | DummyRng64 { counter: u64::from_le_bytes(seed) } |
511 | } |
512 | } |
513 | |
514 | #[test ] |
515 | fn blockrng64_next_u32_vs_next_u64() { |
516 | let mut rng1 = BlockRng64::<DummyRng64>::from_seed([1, 2, 3, 4, 5, 6, 7, 8]); |
517 | let mut rng2 = rng1.clone(); |
518 | let mut rng3 = rng1.clone(); |
519 | |
520 | let mut a = [0; 16]; |
521 | (&mut a[..4]).copy_from_slice(&rng1.next_u32().to_le_bytes()); |
522 | (&mut a[4..12]).copy_from_slice(&rng1.next_u64().to_le_bytes()); |
523 | (&mut a[12..]).copy_from_slice(&rng1.next_u32().to_le_bytes()); |
524 | |
525 | let mut b = [0; 16]; |
526 | (&mut b[..4]).copy_from_slice(&rng2.next_u32().to_le_bytes()); |
527 | (&mut b[4..8]).copy_from_slice(&rng2.next_u32().to_le_bytes()); |
528 | (&mut b[8..]).copy_from_slice(&rng2.next_u64().to_le_bytes()); |
529 | assert_ne!(a, b); |
530 | assert_eq!(&a[..4], &b[..4]); |
531 | assert_eq!(&a[4..12], &b[8..]); |
532 | |
533 | let mut c = [0; 16]; |
534 | (&mut c[..8]).copy_from_slice(&rng3.next_u64().to_le_bytes()); |
535 | (&mut c[8..12]).copy_from_slice(&rng3.next_u32().to_le_bytes()); |
536 | (&mut c[12..]).copy_from_slice(&rng3.next_u32().to_le_bytes()); |
537 | assert_eq!(b, c); |
538 | } |
539 | } |
540 | |