sha256

Pure Haskell SHA-256, HMAC-SHA256 (docs.ppad.tech/sha256).
git clone git://git.ppad.tech/sha256.git
Log | Files | Refs | README | LICENSE

SHA256.hs (18178B)


      1 {-# OPTIONS_GHC -funbox-small-strict-fields #-}
      2 {-# LANGUAGE BangPatterns #-}
      3 {-# LANGUAGE RecordWildCards #-}
      4 {-# LANGUAGE ViewPatterns #-}
      5 
      6 -- |
      7 -- Module: Crypto.Hash.SHA256
      8 -- Copyright: (c) 2024 Jared Tobin
      9 -- License: MIT
     10 -- Maintainer: Jared Tobin <jared@ppad.tech>
     11 --
     12 -- Pure SHA-256 and HMAC-SHA256 implementations for
     13 -- strict and lazy ByteStrings, as specified by RFC's
     14 -- [6234](https://datatracker.ietf.org/doc/html/rfc6234) and
     15 -- [2104](https://datatracker.ietf.org/doc/html/rfc2104).
     16 
     17 module Crypto.Hash.SHA256 (
     18   -- * SHA-256 message digest functions
     19     hash
     20   , hash_lazy
     21 
     22   -- * SHA256-based MAC functions
     23   , hmac
     24   , hmac_lazy
     25   ) where
     26 
     27 import qualified Data.Bits as B
     28 import Data.Bits ((.|.), (.&.))
     29 import qualified Data.ByteString as BS
     30 import qualified Data.ByteString.Builder as BSB
     31 import qualified Data.ByteString.Builder.Extra as BE
     32 import qualified Data.ByteString.Internal as BI
     33 import qualified Data.ByteString.Lazy as BL
     34 import qualified Data.ByteString.Lazy.Internal as BLI
     35 import qualified Data.ByteString.Unsafe as BU
     36 import Data.Word (Word32, Word64)
     37 import Foreign.ForeignPtr (plusForeignPtr)
     38 
     39 -- preliminary utils
     40 
     41 -- keystroke saver
     42 fi :: (Integral a, Num b) => a -> b
     43 fi = fromIntegral
     44 {-# INLINE fi #-}
     45 
     46 -- parse strict ByteString in BE order to Word32 (verbatim from
     47 -- Data.Binary)
     48 --
     49 -- invariant:
     50 --   the input bytestring is at least 32 bits in length
     51 unsafe_word32be :: BS.ByteString -> Word32
     52 unsafe_word32be s =
     53   (fi (s `BU.unsafeIndex` 0) `B.unsafeShiftL` 24) .|.
     54   (fi (s `BU.unsafeIndex` 1) `B.unsafeShiftL` 16) .|.
     55   (fi (s `BU.unsafeIndex` 2) `B.unsafeShiftL`  8) .|.
     56   (fi (s `BU.unsafeIndex` 3))
     57 {-# INLINE unsafe_word32be #-}
     58 
     59 -- utility types for more efficient ByteString management
     60 
     61 data SSPair = SSPair
     62   {-# UNPACK #-} !BS.ByteString
     63   {-# UNPACK #-} !BS.ByteString
     64 
     65 data SLPair = SLPair {-# UNPACK #-} !BS.ByteString !BL.ByteString
     66 
     67 data WSPair = WSPair {-# UNPACK #-} !Word32 {-# UNPACK #-} !BS.ByteString
     68 
     69 -- unsafe version of splitAt that does no bounds checking
     70 --
     71 -- invariant:
     72 --   0 <= n <= l
     73 unsafe_splitAt :: Int -> BS.ByteString -> SSPair
     74 unsafe_splitAt n (BI.BS x l) =
     75   SSPair (BI.BS x n) (BI.BS (plusForeignPtr x n) (l - n))
     76 
     77 -- variant of Data.ByteString.Lazy.splitAt that returns the initial
     78 -- component as a strict, unboxed ByteString
     79 splitAt64 :: BL.ByteString -> SLPair
     80 splitAt64 = splitAt' (64 :: Int) where
     81   splitAt' _ BLI.Empty        = SLPair mempty BLI.Empty
     82   splitAt' n (BLI.Chunk c cs) =
     83     if    n < BS.length c
     84     then
     85       -- n < BS.length c, so unsafe_splitAt is safe
     86       let !(SSPair c0 c1) = unsafe_splitAt n c
     87       in  SLPair c0 (BLI.Chunk c1 cs)
     88     else
     89       let SLPair cs' cs'' = splitAt' (n - BS.length c) cs
     90       in  SLPair (c <> cs') cs''
     91 
     92 -- variant of Data.ByteString.splitAt that behaves like an incremental
     93 -- Word32 parser
     94 --
     95 -- invariant:
     96 --   the input bytestring is at least 32 bits in length
     97 unsafe_parseWsPair :: BS.ByteString -> WSPair
     98 unsafe_parseWsPair (BI.BS x l) =
     99   WSPair (unsafe_word32be (BI.BS x 4)) (BI.BS (plusForeignPtr x 4) (l - 4))
    100 {-# INLINE unsafe_parseWsPair #-}
    101 
    102 -- message padding and parsing
    103 -- https://datatracker.ietf.org/doc/html/rfc6234#section-4.1
    104 
    105 -- k such that (l + 1 + k) mod 64 = 56
    106 sol :: Word64 -> Word64
    107 sol l =
    108   let r = 56 - fi l `mod` 64 - 1 :: Integer -- fi prevents underflow
    109   in  fi (if r < 0 then r + 64 else r)
    110 
    111 -- RFC 6234 4.1 (strict)
    112 pad :: BS.ByteString -> BS.ByteString
    113 pad m = BL.toStrict . BSB.toLazyByteString $ padded where
    114   l = fi (BS.length m)
    115   padded = BSB.byteString m <> fill (sol l) (BSB.word8 0x80)
    116 
    117   fill j !acc
    118     | j == 0 = acc <> BSB.word64BE (l * 8)
    119     | otherwise = fill (pred j) (acc <> BSB.word8 0x00)
    120 
    121 -- RFC 6234 4.1 (lazy)
    122 pad_lazy :: BL.ByteString -> BL.ByteString
    123 pad_lazy (BL.toChunks -> m) = BL.fromChunks (walk 0 m) where
    124   walk !l bs = case bs of
    125     (c:cs) -> c : walk (l + fi (BS.length c)) cs
    126     [] -> padding l (sol l) (BSB.word8 0x80)
    127 
    128   padding l k bs
    129     | k == 0 =
    130           pure
    131         . BL.toStrict
    132           -- more efficient for small builder
    133         . BE.toLazyByteStringWith
    134             (BE.safeStrategy 128 BE.smallChunkSize) mempty
    135         $ bs <> BSB.word64BE (l * 8)
    136     | otherwise =
    137         let nacc = bs <> BSB.word8 0x00
    138         in  padding l (pred k) nacc
    139 
    140 -- functions and constants used
    141 -- https://datatracker.ietf.org/doc/html/rfc6234#section-5.1
    142 
    143 ch :: Word32 -> Word32 -> Word32 -> Word32
    144 ch x y z = (x .&. y) `B.xor` (B.complement x .&. z)
    145 {-# INLINE ch #-}
    146 
    147 -- credit to SHA authors for the following optimisation. their text:
    148 --
    149 -- > note:
    150 -- >   the original functions is (x & y) ^ (x & z) ^ (y & z)
    151 -- >   if you fire off truth tables, this is equivalent to
    152 -- >     (x & y) | (x & z) | (y & z)
    153 -- >   which you can the use distribution on:
    154 -- >     (x & (y | z)) | (y & z)
    155 -- >   which saves us one operation.
    156 maj :: Word32 -> Word32 -> Word32 -> Word32
    157 maj x y z = (x .&. (y .|. z)) .|. (y .&. z)
    158 {-# INLINE maj #-}
    159 
    160 bsig0 :: Word32 -> Word32
    161 bsig0 x = B.rotateR x 2 `B.xor` B.rotateR x 13 `B.xor` B.rotateR x 22
    162 {-# INLINE bsig0 #-}
    163 
    164 bsig1 :: Word32 -> Word32
    165 bsig1 x = B.rotateR x 6 `B.xor` B.rotateR x 11 `B.xor` B.rotateR x 25
    166 {-# INLINE bsig1 #-}
    167 
    168 ssig0 :: Word32 -> Word32
    169 ssig0 x = B.rotateR x 7 `B.xor` B.rotateR x 18 `B.xor` B.unsafeShiftR x 3
    170 {-# INLINE ssig0 #-}
    171 
    172 ssig1 :: Word32 -> Word32
    173 ssig1 x = B.rotateR x 17 `B.xor` B.rotateR x 19 `B.xor` B.unsafeShiftR x 10
    174 {-# INLINE ssig1 #-}
    175 
    176 data Schedule = Schedule {
    177     w00 :: !Word32, w01 :: !Word32, w02 :: !Word32, w03 :: !Word32
    178   , w04 :: !Word32, w05 :: !Word32, w06 :: !Word32, w07 :: !Word32
    179   , w08 :: !Word32, w09 :: !Word32, w10 :: !Word32, w11 :: !Word32
    180   , w12 :: !Word32, w13 :: !Word32, w14 :: !Word32, w15 :: !Word32
    181   , w16 :: !Word32, w17 :: !Word32, w18 :: !Word32, w19 :: !Word32
    182   , w20 :: !Word32, w21 :: !Word32, w22 :: !Word32, w23 :: !Word32
    183   , w24 :: !Word32, w25 :: !Word32, w26 :: !Word32, w27 :: !Word32
    184   , w28 :: !Word32, w29 :: !Word32, w30 :: !Word32, w31 :: !Word32
    185   , w32 :: !Word32, w33 :: !Word32, w34 :: !Word32, w35 :: !Word32
    186   , w36 :: !Word32, w37 :: !Word32, w38 :: !Word32, w39 :: !Word32
    187   , w40 :: !Word32, w41 :: !Word32, w42 :: !Word32, w43 :: !Word32
    188   , w44 :: !Word32, w45 :: !Word32, w46 :: !Word32, w47 :: !Word32
    189   , w48 :: !Word32, w49 :: !Word32, w50 :: !Word32, w51 :: !Word32
    190   , w52 :: !Word32, w53 :: !Word32, w54 :: !Word32, w55 :: !Word32
    191   , w56 :: !Word32, w57 :: !Word32, w58 :: !Word32, w59 :: !Word32
    192   , w60 :: !Word32, w61 :: !Word32, w62 :: !Word32, w63 :: !Word32
    193   }
    194 
    195 -- initialization
    196 -- https://datatracker.ietf.org/doc/html/rfc6234#section-6.1
    197 
    198 data Registers = Registers {
    199     h0 :: !Word32, h1 :: !Word32, h2 :: !Word32, h3 :: !Word32
    200   , h4 :: !Word32, h5 :: !Word32, h6 :: !Word32, h7 :: !Word32
    201   }
    202 
    203 -- first 32 bits of the fractional parts of the square roots of the
    204 -- first eight primes
    205 iv :: Registers
    206 iv = Registers
    207   0x6a09e667 0xbb67ae85 0x3c6ef372 0xa54ff53a
    208   0x510e527f 0x9b05688c 0x1f83d9ab 0x5be0cd19
    209 
    210 -- processing
    211 -- https://datatracker.ietf.org/doc/html/rfc6234#section-6.2
    212 
    213 data Block = Block {
    214     m00 :: !Word32, m01 :: !Word32, m02 :: !Word32, m03 :: !Word32
    215   , m04 :: !Word32, m05 :: !Word32, m06 :: !Word32, m07 :: !Word32
    216   , m08 :: !Word32, m09 :: !Word32, m10 :: !Word32, m11 :: !Word32
    217   , m12 :: !Word32, m13 :: !Word32, m14 :: !Word32, m15 :: !Word32
    218   }
    219 
    220 -- parse strict bytestring to block
    221 --
    222 -- invariant:
    223 --   the input bytestring is exactly 512 bits long
    224 unsafe_parse :: BS.ByteString -> Block
    225 unsafe_parse bs =
    226   let !(WSPair m00 t00) = unsafe_parseWsPair bs
    227       !(WSPair m01 t01) = unsafe_parseWsPair t00
    228       !(WSPair m02 t02) = unsafe_parseWsPair t01
    229       !(WSPair m03 t03) = unsafe_parseWsPair t02
    230       !(WSPair m04 t04) = unsafe_parseWsPair t03
    231       !(WSPair m05 t05) = unsafe_parseWsPair t04
    232       !(WSPair m06 t06) = unsafe_parseWsPair t05
    233       !(WSPair m07 t07) = unsafe_parseWsPair t06
    234       !(WSPair m08 t08) = unsafe_parseWsPair t07
    235       !(WSPair m09 t09) = unsafe_parseWsPair t08
    236       !(WSPair m10 t10) = unsafe_parseWsPair t09
    237       !(WSPair m11 t11) = unsafe_parseWsPair t10
    238       !(WSPair m12 t12) = unsafe_parseWsPair t11
    239       !(WSPair m13 t13) = unsafe_parseWsPair t12
    240       !(WSPair m14 t14) = unsafe_parseWsPair t13
    241       !(WSPair m15 t15) = unsafe_parseWsPair t14
    242   in  if   BS.null t15
    243       then Block {..}
    244       else error "ppad-sha256: internal error (bytes remaining)"
    245 
    246 -- RFC 6234 6.2 step 1
    247 prepare_schedule :: Block -> Schedule
    248 prepare_schedule Block {..} = Schedule {..} where
    249   w00 = m00; w01 = m01; w02 = m02; w03 = m03
    250   w04 = m04; w05 = m05; w06 = m06; w07 = m07
    251   w08 = m08; w09 = m09; w10 = m10; w11 = m11
    252   w12 = m12; w13 = m13; w14 = m14; w15 = m15
    253   w16 = ssig1 w14 + w09 + ssig0 w01 + w00
    254   w17 = ssig1 w15 + w10 + ssig0 w02 + w01
    255   w18 = ssig1 w16 + w11 + ssig0 w03 + w02
    256   w19 = ssig1 w17 + w12 + ssig0 w04 + w03
    257   w20 = ssig1 w18 + w13 + ssig0 w05 + w04
    258   w21 = ssig1 w19 + w14 + ssig0 w06 + w05
    259   w22 = ssig1 w20 + w15 + ssig0 w07 + w06
    260   w23 = ssig1 w21 + w16 + ssig0 w08 + w07
    261   w24 = ssig1 w22 + w17 + ssig0 w09 + w08
    262   w25 = ssig1 w23 + w18 + ssig0 w10 + w09
    263   w26 = ssig1 w24 + w19 + ssig0 w11 + w10
    264   w27 = ssig1 w25 + w20 + ssig0 w12 + w11
    265   w28 = ssig1 w26 + w21 + ssig0 w13 + w12
    266   w29 = ssig1 w27 + w22 + ssig0 w14 + w13
    267   w30 = ssig1 w28 + w23 + ssig0 w15 + w14
    268   w31 = ssig1 w29 + w24 + ssig0 w16 + w15
    269   w32 = ssig1 w30 + w25 + ssig0 w17 + w16
    270   w33 = ssig1 w31 + w26 + ssig0 w18 + w17
    271   w34 = ssig1 w32 + w27 + ssig0 w19 + w18
    272   w35 = ssig1 w33 + w28 + ssig0 w20 + w19
    273   w36 = ssig1 w34 + w29 + ssig0 w21 + w20
    274   w37 = ssig1 w35 + w30 + ssig0 w22 + w21
    275   w38 = ssig1 w36 + w31 + ssig0 w23 + w22
    276   w39 = ssig1 w37 + w32 + ssig0 w24 + w23
    277   w40 = ssig1 w38 + w33 + ssig0 w25 + w24
    278   w41 = ssig1 w39 + w34 + ssig0 w26 + w25
    279   w42 = ssig1 w40 + w35 + ssig0 w27 + w26
    280   w43 = ssig1 w41 + w36 + ssig0 w28 + w27
    281   w44 = ssig1 w42 + w37 + ssig0 w29 + w28
    282   w45 = ssig1 w43 + w38 + ssig0 w30 + w29
    283   w46 = ssig1 w44 + w39 + ssig0 w31 + w30
    284   w47 = ssig1 w45 + w40 + ssig0 w32 + w31
    285   w48 = ssig1 w46 + w41 + ssig0 w33 + w32
    286   w49 = ssig1 w47 + w42 + ssig0 w34 + w33
    287   w50 = ssig1 w48 + w43 + ssig0 w35 + w34
    288   w51 = ssig1 w49 + w44 + ssig0 w36 + w35
    289   w52 = ssig1 w50 + w45 + ssig0 w37 + w36
    290   w53 = ssig1 w51 + w46 + ssig0 w38 + w37
    291   w54 = ssig1 w52 + w47 + ssig0 w39 + w38
    292   w55 = ssig1 w53 + w48 + ssig0 w40 + w39
    293   w56 = ssig1 w54 + w49 + ssig0 w41 + w40
    294   w57 = ssig1 w55 + w50 + ssig0 w42 + w41
    295   w58 = ssig1 w56 + w51 + ssig0 w43 + w42
    296   w59 = ssig1 w57 + w52 + ssig0 w44 + w43
    297   w60 = ssig1 w58 + w53 + ssig0 w45 + w44
    298   w61 = ssig1 w59 + w54 + ssig0 w46 + w45
    299   w62 = ssig1 w60 + w55 + ssig0 w47 + w46
    300   w63 = ssig1 w61 + w56 + ssig0 w48 + w47
    301 
    302 -- RFC 6234 6.2 steps 2, 3, 4
    303 block_hash :: Registers -> Schedule -> Registers
    304 block_hash r00@Registers {..} Schedule {..} =
    305   -- constants are the first 32 bits of the fractional parts of the
    306   -- cube roots of the first sixty-four prime numbers
    307   let r01 = step r00 0x428a2f98 w00; r02 = step r01 0x71374491 w01
    308       r03 = step r02 0xb5c0fbcf w02; r04 = step r03 0xe9b5dba5 w03
    309       r05 = step r04 0x3956c25b w04; r06 = step r05 0x59f111f1 w05
    310       r07 = step r06 0x923f82a4 w06; r08 = step r07 0xab1c5ed5 w07
    311       r09 = step r08 0xd807aa98 w08; r10 = step r09 0x12835b01 w09
    312       r11 = step r10 0x243185be w10; r12 = step r11 0x550c7dc3 w11
    313       r13 = step r12 0x72be5d74 w12; r14 = step r13 0x80deb1fe w13
    314       r15 = step r14 0x9bdc06a7 w14; r16 = step r15 0xc19bf174 w15
    315       r17 = step r16 0xe49b69c1 w16; r18 = step r17 0xefbe4786 w17
    316       r19 = step r18 0x0fc19dc6 w18; r20 = step r19 0x240ca1cc w19
    317       r21 = step r20 0x2de92c6f w20; r22 = step r21 0x4a7484aa w21
    318       r23 = step r22 0x5cb0a9dc w22; r24 = step r23 0x76f988da w23
    319       r25 = step r24 0x983e5152 w24; r26 = step r25 0xa831c66d w25
    320       r27 = step r26 0xb00327c8 w26; r28 = step r27 0xbf597fc7 w27
    321       r29 = step r28 0xc6e00bf3 w28; r30 = step r29 0xd5a79147 w29
    322       r31 = step r30 0x06ca6351 w30; r32 = step r31 0x14292967 w31
    323       r33 = step r32 0x27b70a85 w32; r34 = step r33 0x2e1b2138 w33
    324       r35 = step r34 0x4d2c6dfc w34; r36 = step r35 0x53380d13 w35
    325       r37 = step r36 0x650a7354 w36; r38 = step r37 0x766a0abb w37
    326       r39 = step r38 0x81c2c92e w38; r40 = step r39 0x92722c85 w39
    327       r41 = step r40 0xa2bfe8a1 w40; r42 = step r41 0xa81a664b w41
    328       r43 = step r42 0xc24b8b70 w42; r44 = step r43 0xc76c51a3 w43
    329       r45 = step r44 0xd192e819 w44; r46 = step r45 0xd6990624 w45
    330       r47 = step r46 0xf40e3585 w46; r48 = step r47 0x106aa070 w47
    331       r49 = step r48 0x19a4c116 w48; r50 = step r49 0x1e376c08 w49
    332       r51 = step r50 0x2748774c w50; r52 = step r51 0x34b0bcb5 w51
    333       r53 = step r52 0x391c0cb3 w52; r54 = step r53 0x4ed8aa4a w53
    334       r55 = step r54 0x5b9cca4f w54; r56 = step r55 0x682e6ff3 w55
    335       r57 = step r56 0x748f82ee w56; r58 = step r57 0x78a5636f w57
    336       r59 = step r58 0x84c87814 w58; r60 = step r59 0x8cc70208 w59
    337       r61 = step r60 0x90befffa w60; r62 = step r61 0xa4506ceb w61
    338       r63 = step r62 0xbef9a3f7 w62; r64 = step r63 0xc67178f2 w63
    339       !(Registers a b c d e f g h) = r64
    340   in  Registers
    341         (a + h0) (b + h1) (c + h2) (d + h3)
    342         (e + h4) (f + h5) (g + h6) (h + h7)
    343 
    344 step :: Registers -> Word32 -> Word32 -> Registers
    345 step (Registers a b c d e f g h) k w =
    346   let t1 = h + bsig1 e + ch e f g + k + w
    347       t2 = bsig0 a + maj a b c
    348   in  Registers (t1 + t2) a b c (d + t1) e f g
    349 {-# INLINE step #-}
    350 
    351 -- RFC 6234 6.2 block pipeline
    352 --
    353 -- invariant:
    354 --   the input bytestring is exactly 512 bits in length
    355 unsafe_hash_alg :: Registers -> BS.ByteString -> Registers
    356 unsafe_hash_alg rs bs = block_hash rs (prepare_schedule (unsafe_parse bs))
    357 
    358 -- register concatenation
    359 cat :: Registers -> BS.ByteString
    360 cat Registers {..} =
    361     BL.toStrict
    362     -- more efficient for small builder
    363   . BE.toLazyByteStringWith (BE.safeStrategy 128 BE.smallChunkSize) mempty
    364   $ mconcat [
    365         BSB.word32BE h0, BSB.word32BE h1, BSB.word32BE h2, BSB.word32BE h3
    366       , BSB.word32BE h4, BSB.word32BE h5, BSB.word32BE h6, BSB.word32BE h7
    367       ]
    368 
    369 -- | Compute a condensed representation of a strict bytestring via
    370 --   SHA-256.
    371 --
    372 --   The 256-bit output digest is returned as a strict bytestring.
    373 --
    374 --   >>> hash "strict bytestring input"
    375 --   "<strict 256-bit message digest>"
    376 hash :: BS.ByteString -> BS.ByteString
    377 hash bs = cat (go iv (pad bs)) where
    378   -- proof that 'go' always terminates safely:
    379   --
    380   -- let b = pad bs
    381   -- then length(b) = n * 512 bits for some n >= 0                  (1)
    382   go :: Registers -> BS.ByteString -> Registers
    383   go !acc b
    384     -- if n == 0, then 'go' terminates safely                       (2)
    385     | BS.null b = acc
    386     -- if n > 0, then
    387     --
    388     -- let (c, r) = unsafe_splitAt 64 b
    389     -- then length(c) == 512 bits                                   by (1)
    390     --      length(r) == m * 512 bits for some m >= 0               by (1)
    391     --
    392     -- note 'unsafe_hash_alg' terminates safely for bytestring      (3)
    393     -- input of exactly 512 bits in length
    394     --
    395     -- length(c) == 512
    396     --   => 'unsafe_hash_alg' terminates safely                     by (3)
    397     --   => 'go' terminates safely                                  (4)
    398     -- length(r) == m * 512 bits for m >= 0
    399     --   => next invocation of 'go' terminates safely               by (2), (4)
    400     --
    401     -- then by induction, 'go' always terminates safely (QED)
    402     | otherwise = case unsafe_splitAt 64 b of
    403         SSPair c r -> go (unsafe_hash_alg acc c) r
    404 
    405 -- | Compute a condensed representation of a lazy bytestring via
    406 --   SHA-256.
    407 --
    408 --   The 256-bit output digest is returned as a strict bytestring.
    409 --
    410 --   >>> hash_lazy "lazy bytestring input"
    411 --   "<strict 256-bit message digest>"
    412 hash_lazy :: BL.ByteString -> BS.ByteString
    413 hash_lazy bl = cat (go iv (pad_lazy bl)) where
    414   -- proof of safety proceeds analogously
    415   go :: Registers -> BL.ByteString -> Registers
    416   go !acc bs
    417     | BL.null bs = acc
    418     | otherwise = case splitAt64 bs of
    419         SLPair c r -> go (unsafe_hash_alg acc c) r
    420 
    421 -- HMAC -----------------------------------------------------------------------
    422 -- https://datatracker.ietf.org/doc/html/rfc2104#section-2
    423 
    424 data KeyAndLen = KeyAndLen
    425   {-# UNPACK #-} !BS.ByteString
    426   {-# UNPACK #-} !Int
    427 
    428 -- | Produce a message authentication code for a strict bytestring,
    429 --   based on the provided (strict, bytestring) key, via SHA-256.
    430 --
    431 --   The 256-bit MAC is returned as a strict bytestring.
    432 --
    433 --   Per RFC 2104, the key /should/ be a minimum of 32 bytes long. Keys
    434 --   exceeding 64 bytes in length will first be hashed (via SHA-256).
    435 --
    436 --   >>> hmac "strict bytestring key" "strict bytestring input"
    437 --   "<strict 256-bit MAC>"
    438 hmac
    439   :: BS.ByteString -- ^ key
    440   -> BS.ByteString -- ^ text
    441   -> BS.ByteString
    442 hmac mk text =
    443     let step1 = k <> BS.replicate (64 - lk) 0x00
    444         step2 = BS.map (B.xor 0x36) step1
    445         step3 = step2 <> text
    446         step4 = hash step3
    447         step5 = BS.map (B.xor 0x5C) step1
    448         step6 = step5 <> step4
    449     in  hash step6
    450   where
    451     !(KeyAndLen k lk) =
    452       let l = BS.length mk
    453       in  if   l > 64
    454           then KeyAndLen (hash mk) 32
    455           else KeyAndLen mk l
    456 
    457 -- | Produce a message authentication code for a lazy bytestring, based
    458 --   on the provided (strict, bytestring) key, via SHA-256.
    459 --
    460 --   The 256-bit MAC is returned as a strict bytestring.
    461 --
    462 --   Per RFC 2104, the key /should/ be a minimum of 32 bytes long. Keys
    463 --   exceeding 64 bytes in length will first be hashed (via SHA-256).
    464 --
    465 --   >>> hmac_lazy "strict bytestring key" "lazy bytestring input"
    466 --   "<strict 256-bit MAC>"
    467 hmac_lazy
    468   :: BS.ByteString -- ^ key
    469   -> BL.ByteString -- ^ text
    470   -> BS.ByteString
    471 hmac_lazy mk text =
    472     let step1 = k <> BS.replicate (64 - lk) 0x00
    473         step2 = BS.map (B.xor 0x36) step1
    474         step3 = BL.fromStrict step2 <> text
    475         step4 = hash_lazy step3
    476         step5 = BS.map (B.xor 0x5C) step1
    477         step6 = step5 <> step4
    478     in  hash step6
    479   where
    480     !(KeyAndLen k lk) =
    481       let l = BS.length mk
    482       in  if   l > 64
    483           then KeyAndLen (hash mk) 32
    484           else KeyAndLen mk l
    485