1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Copyright (C) 2022-2024 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved. 4 */ 5 6 #include <tools/le_byteshift.h> 7 #include <sys/random.h> 8 #include <sys/auxv.h> 9 #include <string.h> 10 #include <stdint.h> 11 #include <stdbool.h> 12 #include "../kselftest.h" 13 14 #if defined(__aarch64__) 15 static bool cpu_has_capabilities(void) 16 { 17 return getauxval(AT_HWCAP) & HWCAP_ASIMD; 18 } 19 #elif defined(__s390x__) 20 static bool cpu_has_capabilities(void) 21 { 22 return getauxval(AT_HWCAP) & HWCAP_S390_VXRS; 23 } 24 #else 25 static bool cpu_has_capabilities(void) 26 { 27 return true; 28 } 29 #endif 30 31 static uint32_t rol32(uint32_t word, unsigned int shift) 32 { 33 return (word << (shift & 31)) | (word >> ((-shift) & 31)); 34 } 35 36 static void reference_chacha20_blocks(uint8_t *dst_bytes, const uint32_t *key, uint32_t *counter, size_t nblocks) 37 { 38 uint32_t s[16] = { 39 0x61707865U, 0x3320646eU, 0x79622d32U, 0x6b206574U, 40 key[0], key[1], key[2], key[3], key[4], key[5], key[6], key[7], 41 counter[0], counter[1], 0, 0 42 }; 43 44 while (nblocks--) { 45 uint32_t x[16]; 46 memcpy(x, s, sizeof(x)); 47 for (unsigned int r = 0; r < 20; r += 2) { 48 #define QR(a, b, c, d) ( \ 49 x[a] += x[b], \ 50 x[d] = rol32(x[d] ^ x[a], 16), \ 51 x[c] += x[d], \ 52 x[b] = rol32(x[b] ^ x[c], 12), \ 53 x[a] += x[b], \ 54 x[d] = rol32(x[d] ^ x[a], 8), \ 55 x[c] += x[d], \ 56 x[b] = rol32(x[b] ^ x[c], 7)) 57 58 QR(0, 4, 8, 12); 59 QR(1, 5, 9, 13); 60 QR(2, 6, 10, 14); 61 QR(3, 7, 11, 15); 62 QR(0, 5, 10, 15); 63 QR(1, 6, 11, 12); 64 QR(2, 7, 8, 13); 65 QR(3, 4, 9, 14); 66 } 67 for (unsigned int i = 0; i < 16; ++i, dst_bytes += sizeof(uint32_t)) 68 put_unaligned_le32(x[i] + s[i], dst_bytes); 69 if (!++s[12]) 70 ++s[13]; 71 } 72 counter[0] = s[12]; 73 counter[1] = s[13]; 74 } 75 76 typedef uint8_t u8; 77 typedef uint32_t u32; 78 typedef uint64_t u64; 79 #include <vdso/getrandom.h> 80 81 int main(int argc, char *argv[]) 82 { 83 enum { TRIALS = 1000, BLOCKS = 128, BLOCK_SIZE = 64 }; 84 uint32_t key[8], counter1[2], counter2[2]; 85 uint8_t output1[BLOCK_SIZE * BLOCKS], output2[BLOCK_SIZE * BLOCKS]; 86 87 ksft_print_header(); 88 if (!cpu_has_capabilities()) 89 ksft_exit_skip("Required CPU capabilities missing\n"); 90 ksft_set_plan(1); 91 92 for (unsigned int trial = 0; trial < TRIALS; ++trial) { 93 if (getrandom(key, sizeof(key), 0) != sizeof(key)) { 94 printf("getrandom() failed!\n"); 95 return KSFT_SKIP; 96 } 97 memset(counter1, 0, sizeof(counter1)); 98 reference_chacha20_blocks(output1, key, counter1, BLOCKS); 99 for (unsigned int split = 0; split < BLOCKS; ++split) { 100 memset(output2, 'X', sizeof(output2)); 101 memset(counter2, 0, sizeof(counter2)); 102 if (split) 103 __arch_chacha20_blocks_nostack(output2, key, counter2, split); 104 __arch_chacha20_blocks_nostack(output2 + split * BLOCK_SIZE, key, counter2, BLOCKS - split); 105 if (memcmp(output1, output2, sizeof(output1)) || memcmp(counter1, counter2, sizeof(counter1))) 106 return KSFT_FAIL; 107 } 108 } 109 memset(counter1, 0, sizeof(counter1)); 110 counter1[0] = (uint32_t)-BLOCKS + 2; 111 memset(counter2, 0, sizeof(counter2)); 112 counter2[0] = (uint32_t)-BLOCKS + 2; 113 114 reference_chacha20_blocks(output1, key, counter1, BLOCKS); 115 __arch_chacha20_blocks_nostack(output2, key, counter2, BLOCKS); 116 if (memcmp(output1, output2, sizeof(output1)) || memcmp(counter1, counter2, sizeof(counter1))) 117 return KSFT_FAIL; 118 119 reference_chacha20_blocks(output1, key, counter1, BLOCKS); 120 __arch_chacha20_blocks_nostack(output2, key, counter2, BLOCKS); 121 if (memcmp(output1, output2, sizeof(output1)) || memcmp(counter1, counter2, sizeof(counter1))) 122 return KSFT_FAIL; 123 124 ksft_test_result_pass("chacha: PASS\n"); 125 return KSFT_PASS; 126 } 127