ahash/
fallback_hash.rs

1use crate::convert::*;
2use crate::operations::folded_multiply;
3use crate::operations::read_small;
4use crate::operations::MULTIPLE;
5use crate::random_state::PI;
6use crate::RandomState;
7use core::hash::Hasher;
8
9const ROT: u32 = 23; //17
10
11/// A `Hasher` for hashing an arbitrary stream of bytes.
12///
13/// Instances of [`AHasher`] represent state that is updated while hashing data.
14///
15/// Each method updates the internal state based on the new data provided. Once
16/// all of the data has been provided, the resulting hash can be obtained by calling
17/// `finish()`
18///
19/// [Clone] is also provided in case you wish to calculate hashes for two different items that
20/// start with the same data.
21///
22#[derive(Debug, Clone)]
23pub struct AHasher {
24    buffer: u64,
25    pad: u64,
26    extra_keys: [u64; 2],
27}
28
29impl AHasher {
30    /// Creates a new hasher keyed to the provided key.
31    #[inline]
32    #[allow(dead_code)] // Is not called if non-fallback hash is used.
33    pub(crate) fn new_with_keys(key1: u128, key2: u128) -> AHasher {
34        let pi: [u128; 2] = PI.convert();
35        let key1: [u64; 2] = (key1 ^ pi[0]).convert();
36        let key2: [u64; 2] = (key2 ^ pi[1]).convert();
37        AHasher {
38            buffer: key1[0],
39            pad: key1[1],
40            extra_keys: key2,
41        }
42    }
43
44    #[allow(unused)] // False positive
45    pub(crate) fn test_with_keys(key1: u128, key2: u128) -> Self {
46        let key1: [u64; 2] = key1.convert();
47        let key2: [u64; 2] = key2.convert();
48        Self {
49            buffer: key1[0],
50            pad: key1[1],
51            extra_keys: key2,
52        }
53    }
54
55    #[inline]
56    #[allow(dead_code)] // Is not called if non-fallback hash is used.
57    pub(crate) fn from_random_state(rand_state: &RandomState) -> AHasher {
58        AHasher {
59            buffer: rand_state.k1,
60            pad: rand_state.k0,
61            extra_keys: [rand_state.k2, rand_state.k3],
62        }
63    }
64
65    /// This update function has the goal of updating the buffer with a single multiply
66    /// FxHash does this but is vulnerable to attack. To avoid this input needs to be masked to with an
67    /// unpredictable value. Other hashes such as murmurhash have taken this approach but were found vulnerable
68    /// to attack. The attack was based on the idea of reversing the pre-mixing (Which is necessarily
69    /// reversible otherwise bits would be lost) then placing a difference in the highest bit before the
70    /// multiply used to mix the data. Because a multiply can never affect the bits to the right of it, a
71    /// subsequent update that also differed in this bit could result in a predictable collision.
72    ///
73    /// This version avoids this vulnerability while still only using a single multiply. It takes advantage
74    /// of the fact that when a 64 bit multiply is performed the upper 64 bits are usually computed and thrown
75    /// away. Instead it creates two 128 bit values where the upper 64 bits are zeros and multiplies them.
76    /// (The compiler is smart enough to turn this into a 64 bit multiplication in the assembly)
77    /// Then the upper bits are xored with the lower bits to produce a single 64 bit result.
78    ///
79    /// To understand why this is a good scrambling function it helps to understand multiply-with-carry PRNGs:
80    /// https://en.wikipedia.org/wiki/Multiply-with-carry_pseudorandom_number_generator
81    /// If the multiple is chosen well, this creates a long period, decent quality PRNG.
82    /// Notice that this function is equivalent to this except the `buffer`/`state` is being xored with each
83    /// new block of data. In the event that data is all zeros, it is exactly equivalent to a MWC PRNG.
84    ///
85    /// This is impervious to attack because every bit buffer at the end is dependent on every bit in
86    /// `new_data ^ buffer`. For example suppose two inputs differed in only the 5th bit. Then when the
87    /// multiplication is performed the `result` will differ in bits 5-69. More specifically it will differ by
88    /// 2^5 * MULTIPLE. However in the next step bits 65-128 are turned into a separate 64 bit value. So the
89    /// differing bits will be in the lower 6 bits of this value. The two intermediate values that differ in
90    /// bits 5-63 and in bits 0-5 respectively get added together. Producing an output that differs in every
91    /// bit. The addition carries in the multiplication and at the end additionally mean that the even if an
92    /// attacker somehow knew part of (but not all) the contents of the buffer before hand,
93    /// they would not be able to predict any of the bits in the buffer at the end.
94    #[inline(always)]
95    fn update(&mut self, new_data: u64) {
96        self.buffer = folded_multiply(new_data ^ self.buffer, MULTIPLE);
97    }
98
99    /// Similar to the above this function performs an update using a "folded multiply".
100    /// However it takes in 128 bits of data instead of 64. Both halves must be masked.
101    ///
102    /// This makes it impossible for an attacker to place a single bit difference between
103    /// two blocks so as to cancel each other.
104    ///
105    /// However this is not sufficient. to prevent (a,b) from hashing the same as (b,a) the buffer itself must
106    /// be updated between calls in a way that does not commute. To achieve this XOR and Rotate are used.
107    /// Add followed by xor is not the same as xor followed by add, and rotate ensures that the same out bits
108    /// can't be changed by the same set of input bits. To cancel this sequence with subsequent input would require
109    /// knowing the keys.
110    #[inline(always)]
111    fn large_update(&mut self, new_data: u128) {
112        let block: [u64; 2] = new_data.convert();
113        let combined = folded_multiply(block[0] ^ self.extra_keys[0], block[1] ^ self.extra_keys[1]);
114        self.buffer = (self.buffer.wrapping_add(self.pad) ^ combined).rotate_left(ROT);
115    }
116
117    #[inline]
118    #[cfg(feature = "specialize")]
119    fn short_finish(&self) -> u64 {
120        folded_multiply(self.buffer, self.pad)
121    }
122}
123
124/// Provides [Hasher] methods to hash all of the primitive types.
125///
126/// [Hasher]: core::hash::Hasher
127impl Hasher for AHasher {
128    #[inline]
129    fn write_u8(&mut self, i: u8) {
130        self.update(i as u64);
131    }
132
133    #[inline]
134    fn write_u16(&mut self, i: u16) {
135        self.update(i as u64);
136    }
137
138    #[inline]
139    fn write_u32(&mut self, i: u32) {
140        self.update(i as u64);
141    }
142
143    #[inline]
144    fn write_u64(&mut self, i: u64) {
145        self.update(i as u64);
146    }
147
148    #[inline]
149    fn write_u128(&mut self, i: u128) {
150        self.large_update(i);
151    }
152
153    #[inline]
154    #[cfg(any(
155        target_pointer_width = "64",
156        target_pointer_width = "32",
157        target_pointer_width = "16"
158    ))]
159    fn write_usize(&mut self, i: usize) {
160        self.write_u64(i as u64);
161    }
162
163    #[inline]
164    #[cfg(target_pointer_width = "128")]
165    fn write_usize(&mut self, i: usize) {
166        self.write_u128(i as u128);
167    }
168
169    #[inline]
170    #[allow(clippy::collapsible_if)]
171    fn write(&mut self, input: &[u8]) {
172        let mut data = input;
173        let length = data.len() as u64;
174        //Needs to be an add rather than an xor because otherwise it could be canceled with carefully formed input.
175        self.buffer = self.buffer.wrapping_add(length).wrapping_mul(MULTIPLE);
176        //A 'binary search' on sizes reduces the number of comparisons.
177        if data.len() > 8 {
178            if data.len() > 16 {
179                let tail = data.read_last_u128();
180                self.large_update(tail);
181                while data.len() > 16 {
182                    let (block, rest) = data.read_u128();
183                    self.large_update(block);
184                    data = rest;
185                }
186            } else {
187                self.large_update([data.read_u64().0, data.read_last_u64()].convert());
188            }
189        } else {
190            let value = read_small(data);
191            self.large_update(value.convert());
192        }
193    }
194
195    #[inline]
196    fn finish(&self) -> u64 {
197        let rot = (self.buffer & 63) as u32;
198        folded_multiply(self.buffer, self.pad).rotate_left(rot)
199    }
200}
201
202#[cfg(feature = "specialize")]
203pub(crate) struct AHasherU64 {
204    pub(crate) buffer: u64,
205    pub(crate) pad: u64,
206}
207
208/// A specialized hasher for only primitives under 64 bits.
209#[cfg(feature = "specialize")]
210impl Hasher for AHasherU64 {
211    #[inline]
212    fn finish(&self) -> u64 {
213        folded_multiply(self.buffer, self.pad)
214        //self.buffer
215    }
216
217    #[inline]
218    fn write(&mut self, _bytes: &[u8]) {
219        unreachable!("Specialized hasher was called with a different type of object")
220    }
221
222    #[inline]
223    fn write_u8(&mut self, i: u8) {
224        self.write_u64(i as u64);
225    }
226
227    #[inline]
228    fn write_u16(&mut self, i: u16) {
229        self.write_u64(i as u64);
230    }
231
232    #[inline]
233    fn write_u32(&mut self, i: u32) {
234        self.write_u64(i as u64);
235    }
236
237    #[inline]
238    fn write_u64(&mut self, i: u64) {
239        self.buffer = folded_multiply(i ^ self.buffer, MULTIPLE);
240    }
241
242    #[inline]
243    fn write_u128(&mut self, _i: u128) {
244        unreachable!("Specialized hasher was called with a different type of object")
245    }
246
247    #[inline]
248    fn write_usize(&mut self, _i: usize) {
249        unreachable!("Specialized hasher was called with a different type of object")
250    }
251}
252
253#[cfg(feature = "specialize")]
254pub(crate) struct AHasherFixed(pub AHasher);
255
256/// A specialized hasher for fixed size primitives larger than 64 bits.
257#[cfg(feature = "specialize")]
258impl Hasher for AHasherFixed {
259    #[inline]
260    fn finish(&self) -> u64 {
261        self.0.short_finish()
262    }
263
264    #[inline]
265    fn write(&mut self, bytes: &[u8]) {
266        self.0.write(bytes)
267    }
268
269    #[inline]
270    fn write_u8(&mut self, i: u8) {
271        self.write_u64(i as u64);
272    }
273
274    #[inline]
275    fn write_u16(&mut self, i: u16) {
276        self.write_u64(i as u64);
277    }
278
279    #[inline]
280    fn write_u32(&mut self, i: u32) {
281        self.write_u64(i as u64);
282    }
283
284    #[inline]
285    fn write_u64(&mut self, i: u64) {
286        self.0.write_u64(i);
287    }
288
289    #[inline]
290    fn write_u128(&mut self, i: u128) {
291        self.0.write_u128(i);
292    }
293
294    #[inline]
295    fn write_usize(&mut self, i: usize) {
296        self.0.write_usize(i);
297    }
298}
299
300#[cfg(feature = "specialize")]
301pub(crate) struct AHasherStr(pub AHasher);
302
303/// A specialized hasher for a single string
304/// Note that the other types don't panic because the hash impl for String tacks on an unneeded call. (As does vec)
305#[cfg(feature = "specialize")]
306impl Hasher for AHasherStr {
307    #[inline]
308    fn finish(&self) -> u64 {
309        self.0.finish()
310    }
311
312    #[inline]
313    fn write(&mut self, bytes: &[u8]) {
314        if bytes.len() > 8 {
315            self.0.write(bytes)
316        } else {
317            let value = read_small(bytes);
318            self.0.buffer = folded_multiply(value[0] ^ self.0.buffer, value[1] ^ self.0.extra_keys[1]);
319            self.0.pad = self.0.pad.wrapping_add(bytes.len() as u64);
320        }
321    }
322
323    #[inline]
324    fn write_u8(&mut self, _i: u8) {}
325
326    #[inline]
327    fn write_u16(&mut self, _i: u16) {}
328
329    #[inline]
330    fn write_u32(&mut self, _i: u32) {}
331
332    #[inline]
333    fn write_u64(&mut self, _i: u64) {}
334
335    #[inline]
336    fn write_u128(&mut self, _i: u128) {}
337
338    #[inline]
339    fn write_usize(&mut self, _i: usize) {}
340}
341
342#[cfg(test)]
343mod tests {
344    use crate::fallback_hash::*;
345
346    #[test]
347    fn test_hash() {
348        let mut hasher = AHasher::new_with_keys(0, 0);
349        let value: u64 = 1 << 32;
350        hasher.update(value);
351        let result = hasher.buffer;
352        let mut hasher = AHasher::new_with_keys(0, 0);
353        let value2: u64 = 1;
354        hasher.update(value2);
355        let result2 = hasher.buffer;
356        let result: [u8; 8] = result.convert();
357        let result2: [u8; 8] = result2.convert();
358        assert_ne!(hex::encode(result), hex::encode(result2));
359    }
360
361    #[test]
362    fn test_conversion() {
363        let input: &[u8] = "dddddddd".as_bytes();
364        let bytes: u64 = as_array!(input, 8).convert();
365        assert_eq!(bytes, 0x6464646464646464);
366    }
367}