diff --git a/CHANGELOG.md b/CHANGELOG.md --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -1,3 +1,13 @@ +## 0.27 + +* Optimise AES GCM and CCM +* Optimise P256R1 implementation +* Various AES-NI building improvements +* Add better ECDSA support +* Add XSalsa derive +* Implement square roots for ECC binary curve +* Various tests and benchmarks + ## 0.26 * Add Rabin cryptosystem (and variants) diff --git a/Crypto/Cipher/AES/Primitive.hs b/Crypto/Cipher/AES/Primitive.hs --- a/Crypto/Cipher/AES/Primitive.hs +++ b/Crypto/Cipher/AES/Primitive.hs @@ -37,6 +37,9 @@ , decryptCTR , decryptXTS + -- * CTR with 32-bit wrapping + , combineC32 + -- * Incremental GCM , gcmMode , gcmInit @@ -128,7 +131,7 @@ deriving (NFData) sizeGCM :: Int -sizeGCM = 80 +sizeGCM = 320 sizeOCB :: Int sizeOCB = 160 @@ -317,6 +320,21 @@ -> ba -- ^ output decrypted decryptXTS = doXTS c_aes_decrypt_xts +-- | encrypt/decrypt using Counter mode (32-bit wrapping used in AES-GCM-SIV) +{-# NOINLINE combineC32 #-} +combineC32 :: ByteArray ba + => AES -- ^ AES Context + -> IV AES -- ^ initial vector of AES block size (usually representing a 128 bit integer) + -> ba -- ^ plaintext input + -> ba -- ^ ciphertext output +combineC32 ctx iv input + | len <= 0 = B.empty + | B.length iv /= 16 = error $ "AES error: IV length must be block size (16). Its length is: " ++ show (B.length iv) + | otherwise = B.allocAndFreeze len doEncrypt + where doEncrypt o = withKeyAndIV ctx iv $ \k v -> withByteArray input $ \i -> + c_aes_encrypt_c32 (castPtr o) k v i (fromIntegral len) + len = B.length input + {-# INLINE doECB #-} doECB :: ByteArray ba => (Ptr b -> Ptr AES -> CString -> CUInt -> IO ()) @@ -577,6 +595,9 @@ foreign import ccall "cryptonite_aes.h cryptonite_aes_encrypt_ctr" c_aes_encrypt_ctr :: CString -> Ptr AES -> Ptr Word8 -> CString -> CUInt -> IO () + +foreign import ccall "cryptonite_aes.h cryptonite_aes_encrypt_c32" + c_aes_encrypt_c32 :: CString -> Ptr AES -> Ptr Word8 -> CString -> CUInt -> IO () foreign import ccall "cryptonite_aes.h cryptonite_aes_gcm_init" c_aes_gcm_init :: Ptr AESGCM -> Ptr AES -> Ptr Word8 -> CUInt -> IO () diff --git a/Crypto/Cipher/AESGCMSIV.hs b/Crypto/Cipher/AESGCMSIV.hs new file mode 100644 --- /dev/null +++ b/Crypto/Cipher/AESGCMSIV.hs @@ -0,0 +1,193 @@ +-- | +-- Module : Crypto.Cipher.AESGCMSIV +-- License : BSD-style +-- Maintainer : Olivier Chéron +-- Stability : experimental +-- Portability : unknown +-- +-- Implementation of AES-GCM-SIV, an AEAD scheme with nonce misuse resistance +-- defined in . +-- +-- To achieve the nonce misuse-resistance property, encryption requires two +-- passes on the plaintext, hence no streaming API is provided. This AEAD +-- operates on complete inputs held in memory. For simplicity, the +-- implementation of decryption uses a similar pattern, with performance +-- penalty compared to an implementation which is able to merge both passes. +-- +-- The specification allows inputs up to 2^36 bytes but this implementation +-- requires AAD and plaintext/ciphertext to be both smaller than 2^32 bytes. +{-# LANGUAGE ForeignFunctionInterface #-} +{-# LANGUAGE GeneralizedNewtypeDeriving #-} +module Crypto.Cipher.AESGCMSIV + ( Nonce + , nonce + , generateNonce + , encrypt + , decrypt + ) where + +import Data.Bits +import Data.Word + +import Foreign.C.Types +import Foreign.C.String +import Foreign.Ptr (Ptr, plusPtr) +import Foreign.Storable (peekElemOff, poke, pokeElemOff) + +import Data.ByteArray +import qualified Data.ByteArray as B +import Data.Memory.Endian (toLE) +import Data.Memory.PtrMethods (memXor) + +import Crypto.Cipher.AES.Primitive +import Crypto.Cipher.Types +import Crypto.Error +import Crypto.Internal.Compat (unsafeDoIO) +import Crypto.Random + + +-- 12-byte nonces + +-- | Nonce value for AES-GCM-SIV, always 12 bytes. +newtype Nonce = Nonce Bytes deriving (Show, Eq, ByteArrayAccess) + +-- | Nonce smart constructor. Accepts only 12-byte inputs. +nonce :: ByteArrayAccess iv => iv -> CryptoFailable Nonce +nonce iv + | B.length iv == 12 = CryptoPassed (Nonce $ B.convert iv) + | otherwise = CryptoFailed CryptoError_IvSizeInvalid + +-- | Generate a random nonce for use with AES-GCM-SIV. +generateNonce :: MonadRandom m => m Nonce +generateNonce = Nonce <$> getRandomBytes 12 + + +-- POLYVAL (mutable context) + +newtype Polyval = Polyval Bytes + +polyvalInit :: ScrubbedBytes -> IO Polyval +polyvalInit h = Polyval <$> doInit + where doInit = B.alloc 272 $ \pctx -> B.withByteArray h $ \ph -> + c_aes_polyval_init pctx ph + +polyvalUpdate :: ByteArrayAccess ba => Polyval -> ba -> IO () +polyvalUpdate (Polyval ctx) bs = B.withByteArray ctx $ \pctx -> + B.withByteArray bs $ \pbs -> c_aes_polyval_update pctx pbs sz + where sz = fromIntegral (B.length bs) + +polyvalFinalize :: Polyval -> IO ScrubbedBytes +polyvalFinalize (Polyval ctx) = B.alloc 16 $ \dst -> + B.withByteArray ctx $ \pctx -> c_aes_polyval_finalize pctx dst + +foreign import ccall unsafe "cryptonite_aes.h cryptonite_aes_polyval_init" + c_aes_polyval_init :: Ptr Polyval -> CString -> IO () + +foreign import ccall "cryptonite_aes.h cryptonite_aes_polyval_update" + c_aes_polyval_update :: Ptr Polyval -> CString -> CUInt -> IO () + +foreign import ccall unsafe "cryptonite_aes.h cryptonite_aes_polyval_finalize" + c_aes_polyval_finalize :: Ptr Polyval -> CString -> IO () + + +-- Key Generation + +le32iv :: Word32 -> Nonce -> Bytes +le32iv n (Nonce iv) = B.allocAndFreeze 16 $ \ptr -> do + poke ptr (toLE n) + copyByteArrayToPtr iv (ptr `plusPtr` 4) + +deriveKeys :: BlockCipher128 aes => aes -> Nonce -> (ScrubbedBytes, AES) +deriveKeys aes iv = + case cipherKeySize aes of + KeySizeFixed sz | sz `mod` 8 == 0 -> + let mak = buildKey [0 .. 1] + key = buildKey [2 .. fromIntegral (sz `div` 8) + 1] + mek = throwCryptoError (cipherInit key) + in (mak, mek) + _ -> error "AESGCMSIV: invalid cipher" + where + idx n = ecbEncrypt aes (le32iv n iv) `takeView` 8 + buildKey = B.concat . map idx + + +-- Encryption and decryption + +lengthInvalid :: ByteArrayAccess ba => ba -> Bool +lengthInvalid bs + | finiteBitSize len > 32 = len >= 1 `unsafeShiftL` 32 + | otherwise = False + where len = B.length bs + +-- | AEAD encryption with the specified key and nonce. The key must be given +-- as an initialized 'Crypto.Cipher.AES.AES128' or 'Crypto.Cipher.AES.AES256' +-- cipher. +-- +-- Lengths of additional data and plaintext must be less than 2^32 bytes, +-- otherwise an exception is thrown. +encrypt :: (BlockCipher128 aes, ByteArrayAccess aad, ByteArray ba) + => aes -> Nonce -> aad -> ba -> (AuthTag, ba) +encrypt aes iv aad plaintext + | lengthInvalid aad = error "AESGCMSIV: aad is too large" + | lengthInvalid plaintext = error "AESGCMSIV: plaintext is too large" + | otherwise = (AuthTag tag, ciphertext) + where + (mak, mek) = deriveKeys aes iv + ss = getSs mak aad plaintext + tag = buildTag mek ss iv + ciphertext = combineC32 mek (transformTag tag) plaintext + +-- | AEAD decryption with the specified key and nonce. The key must be given +-- as an initialized 'Crypto.Cipher.AES.AES128' or 'Crypto.Cipher.AES.AES256' +-- cipher. +-- +-- Lengths of additional data and ciphertext must be less than 2^32 bytes, +-- otherwise an exception is thrown. +decrypt :: (BlockCipher128 aes, ByteArrayAccess aad, ByteArray ba) + => aes -> Nonce -> aad -> ba -> AuthTag -> Maybe ba +decrypt aes iv aad ciphertext (AuthTag tag) + | lengthInvalid aad = error "AESGCMSIV: aad is too large" + | lengthInvalid ciphertext = error "AESGCMSIV: ciphertext is too large" + | tag `constEq` buildTag mek ss iv = Just plaintext + | otherwise = Nothing + where + (mak, mek) = deriveKeys aes iv + ss = getSs mak aad plaintext + plaintext = combineC32 mek (transformTag tag) ciphertext + +-- Calculate S_s = POLYVAL(mak, X_1, X_2, ...). +getSs :: (ByteArrayAccess aad, ByteArrayAccess ba) + => ScrubbedBytes -> aad -> ba -> ScrubbedBytes +getSs mak aad plaintext = unsafeDoIO $ do + ctx <- polyvalInit mak + polyvalUpdate ctx aad + polyvalUpdate ctx plaintext + polyvalUpdate ctx (lb :: Bytes) -- the "length block" + polyvalFinalize ctx + where + lb = B.allocAndFreeze 16 $ \ptr -> do + pokeElemOff ptr 0 (toLE64 $ B.length aad) + pokeElemOff ptr 1 (toLE64 $ B.length plaintext) + toLE64 x = toLE (fromIntegral x * 8 :: Word64) + +-- XOR the first 12 bytes of S_s with the nonce and clear the most significant +-- bit of the last byte. +tagInput :: ScrubbedBytes -> Nonce -> Bytes +tagInput ss (Nonce iv) = + B.copyAndFreeze ss $ \ptr -> + B.withByteArray iv $ \ivPtr -> do + memXor ptr ptr ivPtr 12 + b <- peekElemOff ptr 15 + pokeElemOff ptr 15 (b .&. (0x7f :: Word8)) + +-- Encrypt the result with AES using the message-encryption key to produce the +-- tag. +buildTag :: BlockCipher128 aes => aes -> ScrubbedBytes -> Nonce -> Bytes +buildTag mek ss iv = ecbEncrypt mek (tagInput ss iv) + +-- The initial counter block is the tag with the most significant bit of the +-- last byte set to one. +transformTag :: Bytes -> IV AES +transformTag tag = toIV $ B.copyAndFreeze tag $ \ptr -> + peekElemOff ptr 15 >>= pokeElemOff ptr 15 . (.|. (0x80 :: Word8)) + where toIV bs = let Just iv = makeIV (bs :: Bytes) in iv diff --git a/Crypto/Cipher/ChaCha.hs b/Crypto/Cipher/ChaCha.hs --- a/Crypto/Cipher/ChaCha.hs +++ b/Crypto/Cipher/ChaCha.hs @@ -41,9 +41,9 @@ -> nonce -- ^ the nonce (64 or 96 bits) -> State -- ^ the initial ChaCha state initialize nbRounds key nonce - | not (kLen `elem` [16,32]) = error "ChaCha: key length should be 128 or 256 bits" - | not (nonceLen `elem` [8,12]) = error "ChaCha: nonce length should be 64 or 96 bits" - | not (nbRounds `elem` [8,12,20]) = error "ChaCha: rounds should be 8, 12 or 20" + | kLen `notElem` [16,32] = error "ChaCha: key length should be 128 or 256 bits" + | nonceLen `notElem` [8,12] = error "ChaCha: nonce length should be 64 or 96 bits" + | nbRounds `notElem` [8,12,20] = error "ChaCha: rounds should be 8, 12 or 20" | otherwise = unsafeDoIO $ do stPtr <- B.alloc 132 $ \stPtr -> B.withByteArray nonce $ \noncePtr -> diff --git a/Crypto/Cipher/Salsa.hs b/Crypto/Cipher/Salsa.hs --- a/Crypto/Cipher/Salsa.hs +++ b/Crypto/Cipher/Salsa.hs @@ -33,9 +33,9 @@ -> nonce -- ^ the nonce (64 or 96 bits) -> State -- ^ the initial Salsa state initialize nbRounds key nonce - | not (kLen `elem` [16,32]) = error "Salsa: key length should be 128 or 256 bits" - | not (nonceLen `elem` [8,12]) = error "Salsa: nonce length should be 64 or 96 bits" - | not (nbRounds `elem` [8,12,20]) = error "Salsa: rounds should be 8, 12 or 20" + | kLen `notElem` [16,32] = error "Salsa: key length should be 128 or 256 bits" + | nonceLen `notElem` [8,12] = error "Salsa: nonce length should be 64 or 96 bits" + | nbRounds `notElem` [8,12,20] = error "Salsa: rounds should be 8, 12 or 20" | otherwise = unsafeDoIO $ do stPtr <- B.alloc 132 $ \stPtr -> B.withByteArray nonce $ \noncePtr -> diff --git a/Crypto/Cipher/XSalsa.hs b/Crypto/Cipher/XSalsa.hs --- a/Crypto/Cipher/XSalsa.hs +++ b/Crypto/Cipher/XSalsa.hs @@ -12,6 +12,7 @@ {-# LANGUAGE ForeignFunctionInterface #-} module Crypto.Cipher.XSalsa ( initialize + , derive , combine , generate , State @@ -34,7 +35,7 @@ initialize nbRounds key nonce | kLen /= 32 = error "XSalsa: key length should be 256 bits" | nonceLen /= 24 = error "XSalsa: nonce length should be 192 bits" - | not (nbRounds `elem` [8,12,20]) = error "XSalsa: rounds should be 8, 12 or 20" + | nbRounds `notElem` [8,12,20] = error "XSalsa: rounds should be 8, 12 or 20" | otherwise = unsafeDoIO $ do stPtr <- B.alloc 132 $ \stPtr -> B.withByteArray nonce $ \noncePtr -> @@ -44,5 +45,31 @@ where kLen = B.length key nonceLen = B.length nonce +-- | Use an already initialized context and new nonce material to derive another +-- XSalsa context. +-- +-- This allows a multi-level cascade where a first key @k1@ and nonce @n1@ is +-- used to get @HState(k1,n1)@, and this value is then used as key @k2@ to build +-- @XSalsa(k2,n2)@. Function 'initialize' is to be called with the first 192 +-- bits of @n1|n2@, and the call to @derive@ should add the remaining 128 bits. +-- +-- The output context always uses the same number of rounds as the input +-- context. +derive :: ByteArrayAccess nonce + => State -- ^ base XSalsa state + -> nonce -- ^ the remainder nonce (128 bits) + -> State -- ^ the new XSalsa state +derive (State stPtr') nonce + | nonceLen /= 16 = error "XSalsa: nonce length should be 128 bits" + | otherwise = unsafeDoIO $ do + stPtr <- B.copy stPtr' $ \stPtr -> + B.withByteArray nonce $ \noncePtr -> + ccryptonite_xsalsa_derive stPtr nonceLen noncePtr + return $ State stPtr + where nonceLen = B.length nonce + foreign import ccall "cryptonite_xsalsa_init" ccryptonite_xsalsa_init :: Ptr State -> Int -> Int -> Ptr Word8 -> Int -> Ptr Word8 -> IO () + +foreign import ccall "cryptonite_xsalsa_derive" + ccryptonite_xsalsa_derive :: Ptr State -> Int -> Ptr Word8 -> IO () diff --git a/Crypto/ConstructHash/MiyaguchiPreneel.hs b/Crypto/ConstructHash/MiyaguchiPreneel.hs --- a/Crypto/ConstructHash/MiyaguchiPreneel.hs +++ b/Crypto/ConstructHash/MiyaguchiPreneel.hs @@ -44,7 +44,7 @@ where (hd, tl) = B.splitAt bsz msg --- | Compute Miyaguchi-Preneel one way compress using the infered block cipher. +-- | Compute Miyaguchi-Preneel one way compress using the inferred block cipher. -- Only safe when KEY-SIZE equals to BLOCK-SIZE. -- -- Simple usage /mp' msg :: MiyaguchiPreneel AES128/ diff --git a/Crypto/Data/Padding.hs b/Crypto/Data/Padding.hs --- a/Crypto/Data/Padding.hs +++ b/Crypto/Data/Padding.hs @@ -6,7 +6,7 @@ -- Portability : unknown -- -- Various cryptographic padding commonly used for block ciphers --- or assymetric systems. +-- or asymmetric systems. -- module Crypto.Data.Padding ( Format(..) diff --git a/Crypto/ECC.hs b/Crypto/ECC.hs --- a/Crypto/ECC.hs +++ b/Crypto/ECC.hs @@ -8,6 +8,7 @@ -- Elliptic Curve Cryptography -- {-# LANGUAGE DeriveDataTypeable #-} +{-# LANGUAGE FlexibleContexts #-} {-# LANGUAGE GeneralizedNewtypeDeriving #-} {-# LANGUAGE TypeFamilies #-} {-# LANGUAGE ScopedTypeVariables #-} @@ -21,6 +22,7 @@ , EllipticCurve(..) , EllipticCurveDH(..) , EllipticCurveArith(..) + , EllipticCurveBasepointArith(..) , KeyPair(..) , SharedSecret(..) ) where @@ -34,7 +36,9 @@ import Crypto.Internal.Imports import Crypto.Internal.ByteArray (ByteArray, ByteArrayAccess, ScrubbedBytes) import qualified Crypto.Internal.ByteArray as B +import Crypto.Number.Basic (numBits) import Crypto.Number.Serialize (i2ospOf_, os2ip) +import qualified Crypto.Number.Serialize.LE as LE import qualified Crypto.PubKey.Curve25519 as X25519 import qualified Crypto.PubKey.Curve448 as X448 import Data.ByteArray (convert) @@ -98,7 +102,7 @@ -- value or an exception. ecdh :: proxy curve -> Scalar curve -> Point curve -> CryptoFailable SharedSecret -class EllipticCurve curve => EllipticCurveArith curve where +class (EllipticCurve curve, Eq (Point curve)) => EllipticCurveArith curve where -- | Add points on a curve pointAdd :: proxy curve -> Point curve -> Point curve -> Point curve @@ -111,6 +115,35 @@ -- -- | Scalar Inverse -- scalarInverse :: Scalar curve -> Scalar curve +class (EllipticCurveArith curve, Eq (Scalar curve)) => EllipticCurveBasepointArith curve where + -- | Get the curve order size in bits + curveOrderBits :: proxy curve -> Int + + -- | Multiply a scalar with the curve base point + pointBaseSmul :: proxy curve -> Scalar curve -> Point curve + + -- | Multiply the point @p@ with @s2@ and add a lifted to curve value @s1@ + pointsSmulVarTime :: proxy curve -> Scalar curve -> Scalar curve -> Point curve -> Point curve + pointsSmulVarTime prx s1 s2 p = pointAdd prx (pointBaseSmul prx s1) (pointSmul prx s2 p) + + -- | Encode an elliptic curve scalar into big-endian form + encodeScalar :: ByteArray bs => proxy curve -> Scalar curve -> bs + + -- | Try to decode the big-endian form of an elliptic curve scalar + decodeScalar :: ByteArray bs => proxy curve -> bs -> CryptoFailable (Scalar curve) + + -- | Convert an elliptic curve scalar to an integer + scalarToInteger :: proxy curve -> Scalar curve -> Integer + + -- | Try to create an elliptic curve scalar from an integer + scalarFromInteger :: proxy curve -> Integer -> CryptoFailable (Scalar curve) + + -- | Add two scalars and reduce modulo the curve order + scalarAdd :: proxy curve -> Scalar curve -> Scalar curve -> Scalar curve + + -- | Multiply two scalars and reduce modulo the curve order + scalarMul :: proxy curve -> Scalar curve -> Scalar curve -> Scalar curve + -- | P256 Curve -- -- also known as P256 @@ -133,11 +166,11 @@ uncompressed = B.singleton 4 xy = P256.pointToBinary p decodePoint _ mxy = case B.uncons mxy of - Nothing -> CryptoFailed $ CryptoError_PointSizeInvalid + Nothing -> CryptoFailed CryptoError_PointSizeInvalid Just (m,xy) -- uncompressed | m == 4 -> P256.pointFromBinary xy - | otherwise -> CryptoFailed $ CryptoError_PointFormatInvalid + | otherwise -> CryptoFailed CryptoError_PointFormatInvalid instance EllipticCurveArith Curve_P256R1 where pointAdd _ a b = P256.pointAdd a b @@ -148,6 +181,17 @@ ecdhRaw _ s p = SharedSecret $ P256.pointDh s p ecdh prx s p = checkNonZeroDH (ecdhRaw prx s p) +instance EllipticCurveBasepointArith Curve_P256R1 where + curveOrderBits _ = 256 + pointBaseSmul _ = P256.toPoint + pointsSmulVarTime _ = P256.pointsMulVarTime + encodeScalar _ = P256.scalarToBinary + decodeScalar _ = P256.scalarFromBinary + scalarToInteger _ = P256.scalarToInteger + scalarFromInteger _ = P256.scalarFromInteger + scalarAdd _ = P256.scalarAdd + scalarMul _ = P256.scalarMul + data Curve_P384R1 = Curve_P384R1 deriving (Show,Data) @@ -171,6 +215,17 @@ where prx = Proxy :: Proxy Simple.SEC_p384r1 +instance EllipticCurveBasepointArith Curve_P384R1 where + curveOrderBits _ = 384 + pointBaseSmul _ = Simple.pointBaseMul + pointsSmulVarTime _ = ecPointsMulVarTime + encodeScalar _ = ecScalarToBinary + decodeScalar _ = ecScalarFromBinary + scalarToInteger _ = ecScalarToInteger + scalarFromInteger _ = ecScalarFromInteger + scalarAdd _ = ecScalarAdd + scalarMul _ = ecScalarMul + data Curve_P521R1 = Curve_P521R1 deriving (Show,Data) @@ -194,6 +249,17 @@ where prx = Proxy :: Proxy Simple.SEC_p521r1 +instance EllipticCurveBasepointArith Curve_P521R1 where + curveOrderBits _ = 521 + pointBaseSmul _ = Simple.pointBaseMul + pointsSmulVarTime _ = ecPointsMulVarTime + encodeScalar _ = ecScalarToBinary + decodeScalar _ = ecScalarFromBinary + scalarToInteger _ = ecScalarToInteger + scalarFromInteger _ = ecScalarFromInteger + scalarAdd _ = ecScalarAdd + scalarMul _ = ecScalarMul + data Curve_X25519 = Curve_X25519 deriving (Show,Data) @@ -250,6 +316,22 @@ pointNegate _ p = Edwards25519.pointNegate p pointSmul _ s p = Edwards25519.pointMul s p +instance EllipticCurveBasepointArith Curve_Edwards25519 where + curveOrderBits _ = 253 + pointBaseSmul _ = Edwards25519.toPoint + pointsSmulVarTime _ = Edwards25519.pointsMulVarTime + encodeScalar _ = B.reverse . Edwards25519.scalarEncode + decodeScalar _ bs + | B.length bs == 32 = Edwards25519.scalarDecodeLong (B.reverse bs) + | otherwise = CryptoFailed CryptoError_SecretKeySizeInvalid + scalarToInteger _ s = LE.os2ip (Edwards25519.scalarEncode s :: B.Bytes) + scalarFromInteger _ i = + case LE.i2ospOf 32 i of + Nothing -> CryptoFailed CryptoError_SecretKeySizeInvalid + Just bs -> Edwards25519.scalarDecodeLong (bs :: B.Bytes) + scalarAdd _ = Edwards25519.scalarAdd + scalarMul _ = Edwards25519.scalarMul + checkNonZeroDH :: SharedSecret -> CryptoFailable SharedSecret checkNonZeroDH s@(SharedSecret b) | B.constAllZero b = CryptoFailed CryptoError_ScalarMultiplicationInvalid @@ -271,7 +353,7 @@ decodeECPoint :: (Simple.Curve curve, ByteArray bs) => bs -> CryptoFailable (Simple.Point curve) decodeECPoint mxy = case B.uncons mxy of - Nothing -> CryptoFailed $ CryptoError_PointSizeInvalid + Nothing -> CryptoFailed CryptoError_PointSizeInvalid Just (m,xy) -- uncompressed | m == 4 -> @@ -280,4 +362,47 @@ x = os2ip xb y = os2ip yb in Simple.pointFromIntegers (x,y) - | otherwise -> CryptoFailed $ CryptoError_PointFormatInvalid + | otherwise -> CryptoFailed CryptoError_PointFormatInvalid + +ecPointsMulVarTime :: forall curve . Simple.Curve curve + => Simple.Scalar curve + -> Simple.Scalar curve -> Simple.Point curve + -> Simple.Point curve +ecPointsMulVarTime n1 = Simple.pointAddTwoMuls n1 g + where g = Simple.curveEccG $ Simple.curveParameters (Proxy :: Proxy curve) + +ecScalarFromBinary :: forall curve bs . (Simple.Curve curve, ByteArrayAccess bs) + => bs -> CryptoFailable (Simple.Scalar curve) +ecScalarFromBinary ba + | B.length ba /= size = CryptoFailed CryptoError_SecretKeySizeInvalid + | otherwise = CryptoPassed (Simple.Scalar $ os2ip ba) + where size = ecCurveOrderBytes (Proxy :: Proxy curve) + +ecScalarToBinary :: forall curve bs . (Simple.Curve curve, ByteArray bs) + => Simple.Scalar curve -> bs +ecScalarToBinary (Simple.Scalar s) = i2ospOf_ size s + where size = ecCurveOrderBytes (Proxy :: Proxy curve) + +ecScalarFromInteger :: forall curve . Simple.Curve curve + => Integer -> CryptoFailable (Simple.Scalar curve) +ecScalarFromInteger s + | numBits s > nb = CryptoFailed CryptoError_SecretKeySizeInvalid + | otherwise = CryptoPassed (Simple.Scalar s) + where nb = 8 * ecCurveOrderBytes (Proxy :: Proxy curve) + +ecScalarToInteger :: Simple.Scalar curve -> Integer +ecScalarToInteger (Simple.Scalar s) = s + +ecCurveOrderBytes :: Simple.Curve c => proxy c -> Int +ecCurveOrderBytes prx = (numBits n + 7) `div` 8 + where n = Simple.curveEccN $ Simple.curveParameters prx + +ecScalarAdd :: forall curve . Simple.Curve curve + => Simple.Scalar curve -> Simple.Scalar curve -> Simple.Scalar curve +ecScalarAdd (Simple.Scalar a) (Simple.Scalar b) = Simple.Scalar ((a + b) `mod` n) + where n = Simple.curveEccN $ Simple.curveParameters (Proxy :: Proxy curve) + +ecScalarMul :: forall curve . Simple.Curve curve + => Simple.Scalar curve -> Simple.Scalar curve -> Simple.Scalar curve +ecScalarMul (Simple.Scalar a) (Simple.Scalar b) = Simple.Scalar ((a * b) `mod` n) + where n = Simple.curveEccN $ Simple.curveParameters (Proxy :: Proxy curve) diff --git a/Crypto/MAC/CMAC.hs b/Crypto/MAC/CMAC.hs --- a/Crypto/MAC/CMAC.hs +++ b/Crypto/MAC/CMAC.hs @@ -94,7 +94,7 @@ cipherIPT :: BlockCipher k => k -> [Word8] -cipherIPT = expandIPT . blockSize where +cipherIPT = expandIPT . blockSize -- Data type which represents the smallest irreducibule binary polynomial -- against specified degree. diff --git a/Crypto/Number/F2m.hs b/Crypto/Number/F2m.hs --- a/Crypto/Number/F2m.hs +++ b/Crypto/Number/F2m.hs @@ -16,7 +16,9 @@ , mulF2m , squareF2m' , squareF2m + , powF2m , modF2m + , sqrtF2m , invF2m , divF2m ) where @@ -66,8 +68,8 @@ mulF2m fx n1 n2 | fx < 0 || n1 < 0 - || n2 < 0 = error "mulF2m: negative number represent no binary binary polynomial" - | fx == 0 = error "modF2m: cannot multiply modulo zero polynomial" + || n2 < 0 = error "mulF2m: negative number represent no binary polynomial" + | fx == 0 = error "mulF2m: cannot multiply modulo zero polynomial" | otherwise = modF2m fx $ go (if n2 `mod` 2 == 1 then n1 else 0) (log2 n2) where go n s | s == 0 = n @@ -96,9 +98,36 @@ squareF2m' :: Integer -> Integer squareF2m' n - | n < 0 = error "mulF2m: negative number represent no binary binary polynomial" + | n < 0 = error "mulF2m: negative number represent no binary polynomial" | otherwise = foldl' (\acc s -> if testBit n s then setBit acc (2 * s) else acc) 0 [0 .. log2 n] {-# INLINE squareF2m' #-} + +-- | Exponentiation in F₂m by computing @a^b mod fx@. +-- +-- This implements an exponentiation by squaring based solution. It inherits the +-- same restrictions as 'squareF2m'. Negative exponents are disallowed. +powF2m :: BinaryPolynomial -- ^Modulus + -> Integer -- ^a + -> Integer -- ^b + -> Integer +powF2m fx a b + | b < 0 = error "powF2m: negative exponents disallowed" + | b == 0 = if fx > 1 then 1 else 0 + | even b = squareF2m fx x + | otherwise = mulF2m fx a (squareF2m' x) + where x = powF2m fx a (b `div` 2) + +-- | Square rooot in F₂m. +-- +-- We exploit the fact that @a^(2^m) = a@, or in particular, @a^(2^m - 1) = 1@ +-- from a classical result by Lagrange. Thus the square root is simply @a^(2^(m +-- - 1))@. +sqrtF2m :: BinaryPolynomial -- ^Modulus + -> Integer -- ^a + -> Integer +sqrtF2m fx a = go (log2 fx - 1) a + where go 0 x = x + go n x = go (n - 1) (squareF2m fx x) -- | Extended GCD algorithm for polynomials. For @a@ and @b@ returns @(g, u, v)@ such that @a * u + b * v == g@. -- diff --git a/Crypto/Number/ModArithmetic.hs b/Crypto/Number/ModArithmetic.hs --- a/Crypto/Number/ModArithmetic.hs +++ b/Crypto/Number/ModArithmetic.hs @@ -1,5 +1,4 @@ {-# LANGUAGE BangPatterns #-} -{-# LANGUAGE DeriveDataTypeable #-} -- | -- Module : Crypto.Number.ModArithmetic -- License : BSD-style @@ -15,7 +14,10 @@ -- * Inverse computing , inverse , inverseCoprimes + , inverseFermat + -- * Squares , jacobi + , squareRoot ) where import Control.Exception (throw, Exception) @@ -56,7 +58,7 @@ -- hiding parameters. -- -- Use this function when all the parameters are public, --- otherwise 'expSafe' should be prefered. +-- otherwise 'expSafe' should be preferred. expFast :: Integer -- ^ base -> Integer -- ^ exponent -> Integer -- ^ modulo @@ -70,7 +72,7 @@ | b == 1 = b | e == 0 = 1 | e == 1 = b `mod` m - | even e = let p = (exponentiation b (e `div` 2) m) `mod` m + | even e = let p = exponentiation b (e `div` 2) m `mod` m in (p^(2::Integer)) `mod` m | otherwise = (b * exponentiation b (e-1) m) `mod` m @@ -97,17 +99,17 @@ -- | Computes the Jacobi symbol (a/n). -- 0 ≤ a < n; n ≥ 3 and odd. --- +-- -- The Legendre and Jacobi symbols are indistinguishable exactly when the -- lower argument is an odd prime, in which case they have the same value. --- +-- -- See algorithm 2.149 in "Handbook of Applied Cryptography" by Alfred J. Menezes et al. jacobi :: Integer -> Integer -> Maybe Integer jacobi a n | n < 3 || even n = Nothing | a == 0 || a == 1 = Just a - | n <= a = jacobi (a `mod` n) n - | a < 0 = + | n <= a = jacobi (a `mod` n) n + | a < 0 = let b = if n `mod` 4 == 1 then 1 else -1 in fmap (*b) (jacobi (-a) n) | otherwise = @@ -120,3 +122,96 @@ n1 = n `mod` a1 in if a1 == 1 then Just s else fmap (*s) (jacobi n1 a1) + +-- | Modular inverse using Fermat's little theorem. This works only when +-- the modulus is prime but avoids side channels like in 'expSafe'. +inverseFermat :: Integer -> Integer -> Integer +inverseFermat g p = expSafe g (p - 2) p + +-- | Raised when the assumption about the modulus is invalid. +data ModulusAssertionError = ModulusAssertionError + deriving (Show) + +instance Exception ModulusAssertionError + +-- | Modular square root of @g@ modulo a prime @p@. +-- +-- If the modulus is found not to be prime, the function will raise a +-- 'ModulusAssertionError'. +-- +-- This implementation is variable time and should be used with public +-- parameters only. +squareRoot :: Integer -> Integer -> Maybe Integer +squareRoot p + | p < 2 = throw ModulusAssertionError + | otherwise = + case p `divMod` 8 of + (v, 3) -> method1 (2 * v + 1) + (v, 7) -> method1 (2 * v + 2) + (u, 5) -> method2 u + (_, 1) -> tonelliShanks p + (0, 2) -> \a -> Just (if even a then 0 else 1) + _ -> throw ModulusAssertionError + + where + x `eqMod` y = (x - y) `mod` p == 0 + + validate g y | (y * y) `eqMod` g = Just y + | otherwise = Nothing + + -- p == 4u + 3 and u' == u + 1 + method1 u' g = + let y = expFast g u' p + in validate g y + + -- p == 8u + 5 + method2 u g = + let gamma = expFast (2 * g) u p + g_gamma = g * gamma + i = (2 * g_gamma * gamma) `mod` p + y = (g_gamma * (i - 1)) `mod` p + in validate g y + +tonelliShanks :: Integer -> Integer -> Maybe Integer +tonelliShanks p a + | aa == 0 = Just 0 + | otherwise = + case expFast aa p2 p of + b | b == p1 -> Nothing + | b == 1 -> Just $ go (expFast aa ((s + 1) `div` 2) p) + (expFast aa s p) + (expFast n s p) + e + | otherwise -> throw ModulusAssertionError + where + aa = a `mod` p + p1 = p - 1 + p2 = p1 `div` 2 + n = findN 2 + + x `mul` y = (x * y) `mod` p + + pow2m 0 x = x + pow2m i x = pow2m (i - 1) (x `mul` x) + + (e, s) = asPowerOf2AndOdd p1 + + -- find a quadratic non-residue + findN i + | expFast i p2 p == p1 = i + | otherwise = findN (i + 1) + + -- find m such that b^(2^m) == 1 (mod p) + findM b i + | b == 1 = i + | otherwise = findM (b `mul` b) (i + 1) + + go !x b g !r + | b == 1 = x + | otherwise = + let r' = findM b 0 + z = pow2m (r - r' - 1) g + x' = x `mul` z + b' = b `mul` g' + g' = z `mul` z + in go x' b' g' r' diff --git a/Crypto/Number/Prime.hs b/Crypto/Number/Prime.hs --- a/Crypto/Number/Prime.hs +++ b/Crypto/Number/Prime.hs @@ -127,7 +127,7 @@ factorise :: Integer -> Integer -> (Integer, Integer) factorise !si !vi | vi `testBit` 0 = (si, vi) - | otherwise = factorise (si+1) (vi `shiftR` 1) -- probably faster to not shift v continously, but just once. + | otherwise = factorise (si+1) (vi `shiftR` 1) -- probably faster to not shift v continuously, but just once. expmod = expSafe -- when iteration reach zero, we have a probable prime diff --git a/Crypto/PubKey/ECC/P256.hs b/Crypto/PubKey/ECC/P256.hs --- a/Crypto/PubKey/ECC/P256.hs +++ b/Crypto/PubKey/ECC/P256.hs @@ -8,7 +8,6 @@ -- P256 support -- {-# LANGUAGE GeneralizedNewtypeDeriving #-} -{-# LANGUAGE BangPatterns #-} {-# LANGUAGE EmptyDataDecls #-} {-# OPTIONS_GHC -fno-warn-unused-binds #-} module Crypto.PubKey.ECC.P256 @@ -22,7 +21,9 @@ , pointDh , pointsMulVarTime , pointIsValid + , pointIsAtInfinity , toPoint + , pointX , pointToIntegers , pointFromIntegers , pointToBinary @@ -31,10 +32,13 @@ -- * Scalar arithmetic , scalarGenerate , scalarZero + , scalarN , scalarIsZero , scalarAdd , scalarSub + , scalarMul , scalarInv + , scalarInvSafe , scalarCmp , scalarFromBinary , scalarToBinary @@ -76,6 +80,9 @@ data P256Y data P256X +order :: Integer +order = 0xffffffff00000000ffffffffffffffffbce6faada7179e84f3b9cac2fc632551 + ------------------------------------------------------------------------ -- Point methods ------------------------------------------------------------------------ @@ -109,7 +116,7 @@ -- | Negate a point pointNegate :: Point -> Point pointNegate a = withNewPoint $ \dx dy -> - withPoint a $ \ax ay -> do + withPoint a $ \ax ay -> ccryptonite_p256e_point_negate ax ay dx dy -- | Multiply a point by a scalar @@ -117,16 +124,16 @@ -- warning: variable time pointMul :: Scalar -> Point -> Point pointMul scalar p = withNewPoint $ \dx dy -> - withScalar scalar $ \n -> withPoint p $ \px py -> withScalarZero $ \nzero -> - ccryptonite_p256_points_mul_vartime nzero n px py dx dy + withScalar scalar $ \n -> withPoint p $ \px py -> + ccryptonite_p256e_point_mul n px py dx dy -- | Similar to 'pointMul', serializing the x coordinate as binary. -- When scalar is multiple of point order the result is all zero. pointDh :: ByteArray binary => Scalar -> Point -> binary pointDh scalar p = B.unsafeCreate scalarSize $ \dst -> withTempPoint $ \dx dy -> do - withScalar scalar $ \n -> withPoint p $ \px py -> withScalarZero $ \nzero -> - ccryptonite_p256_points_mul_vartime nzero n px py dx dy + withScalar scalar $ \n -> withPoint p $ \px py -> + ccryptonite_p256e_point_mul n px py dx dy ccryptonite_p256_to_bin (castPtr dx) dst -- | multiply the point @p with @n2 and add a lifted to curve value @n1 @@ -145,6 +152,19 @@ r <- ccryptonite_p256_is_valid_point px py return (r /= 0) +-- | Check if a 'Point' is the point at infinity +pointIsAtInfinity :: Point -> Bool +pointIsAtInfinity (Point b) = constAllZero b + +-- | Return the x coordinate as a 'Scalar' if the point is not at infinity +pointX :: Point -> Maybe Scalar +pointX p + | pointIsAtInfinity p = Nothing + | otherwise = Just $ + withNewScalarFreeze $ \d -> + withPoint p $ \px _ -> + ccryptonite_p256_mod ccryptonite_SECP256r1_n (castPtr px) (castPtr d) + -- | Convert a point to (x,y) Integers pointToIntegers :: Point -> (Integer, Integer) pointToIntegers p = unsafeDoIO $ withPoint p $ \px py -> @@ -187,12 +207,12 @@ validatePoint :: Point -> CryptoFailable Point validatePoint p | pointIsValid p = CryptoPassed p - | otherwise = CryptoFailed $ CryptoError_PointCoordinatesInvalid + | otherwise = CryptoFailed CryptoError_PointCoordinatesInvalid -- | Convert from binary to a point, possibly invalid unsafePointFromBinary :: ByteArrayAccess ba => ba -> CryptoFailable Point unsafePointFromBinary ba - | B.length ba /= pointSize = CryptoFailed $ CryptoError_PublicKeySizeInvalid + | B.length ba /= pointSize = CryptoFailed CryptoError_PublicKeySizeInvalid | otherwise = CryptoPassed $ withNewPoint $ \px py -> B.withByteArray ba $ \src -> do ccryptonite_p256_from_bin src (castPtr px) @@ -215,6 +235,10 @@ scalarZero :: Scalar scalarZero = withNewScalarFreeze $ \d -> ccryptonite_p256_init d +-- | The scalar representing the curve order +scalarN :: Scalar +scalarN = throwCryptoError (scalarFromInteger order) + -- | Check if the scalar is 0 scalarIsZero :: Scalar -> Bool scalarIsZero s = unsafeDoIO $ withScalar s $ \d -> do @@ -237,6 +261,14 @@ withNewScalarFreeze $ \d -> withScalar a $ \pa -> withScalar b $ \pb -> ccryptonite_p256e_modsub ccryptonite_SECP256r1_n pa pb d +-- | Perform multiplication between two scalars +-- +-- > a * b +scalarMul :: Scalar -> Scalar -> Scalar +scalarMul a b = + withNewScalarFreeze $ \d -> withScalar a $ \pa -> withScalar b $ \pb -> + ccryptonite_p256_modmul ccryptonite_SECP256r1_n pa 0 pb d + -- | Give the inverse of the scalar -- -- > 1 / a @@ -247,6 +279,14 @@ withNewScalarFreeze $ \b -> withScalar a $ \pa -> ccryptonite_p256_modinv_vartime ccryptonite_SECP256r1_n pa b +-- | Give the inverse of the scalar using safe exponentiation +-- +-- > 1 / a +scalarInvSafe :: Scalar -> Scalar +scalarInvSafe a = + withNewScalarFreeze $ \b -> withScalar a $ \pa -> + ccryptonite_p256e_scalar_invert pa b + -- | Compare 2 Scalar scalarCmp :: Scalar -> Scalar -> Ordering scalarCmp a b = unsafeDoIO $ @@ -257,7 +297,7 @@ -- | convert a scalar from binary scalarFromBinary :: ByteArrayAccess ba => ba -> CryptoFailable Scalar scalarFromBinary ba - | B.length ba /= scalarSize = CryptoFailed $ CryptoError_SecretKeySizeInvalid + | B.length ba /= scalarSize = CryptoFailed CryptoError_SecretKeySizeInvalid | otherwise = CryptoPassed $ withNewScalarFreeze $ \p -> B.withByteArray ba $ \b -> ccryptonite_p256_from_bin b p @@ -298,18 +338,9 @@ withTempPoint :: (Ptr P256X -> Ptr P256Y -> IO a) -> IO a withTempPoint f = allocTempScrubbed pointSize (\p -> let px = castPtr p in f px (pxToPy px)) -withTempScalar :: (Ptr P256Scalar -> IO a) -> IO a -withTempScalar f = allocTempScrubbed scalarSize (f . castPtr) - withScalar :: Scalar -> (Ptr P256Scalar -> IO a) -> IO a withScalar (Scalar d) f = B.withByteArray d f -withScalarZero :: (Ptr P256Scalar -> IO a) -> IO a -withScalarZero f = - withTempScalar $ \d -> do - ccryptonite_p256_init d - f d - allocTemp :: Int -> (Ptr Word8 -> IO a) -> IO a allocTemp n f = ignoreSnd <$> B.allocRet n f where @@ -350,6 +381,8 @@ ccryptonite_p256_mod :: Ptr P256Scalar -> Ptr P256Scalar -> Ptr P256Scalar -> IO () foreign import ccall "cryptonite_p256_modmul" ccryptonite_p256_modmul :: Ptr P256Scalar -> Ptr P256Scalar -> P256Digit -> Ptr P256Scalar -> Ptr P256Scalar -> IO () +foreign import ccall "cryptonite_p256e_scalar_invert" + ccryptonite_p256e_scalar_invert :: Ptr P256Scalar -> Ptr P256Scalar -> IO () --foreign import ccall "cryptonite_p256_modinv" -- ccryptonite_p256_modinv :: Ptr P256Scalar -> Ptr P256Scalar -> Ptr P256Scalar -> IO () foreign import ccall "cryptonite_p256_modinv_vartime" @@ -369,6 +402,13 @@ ccryptonite_p256e_point_negate :: Ptr P256X -> Ptr P256Y -> Ptr P256X -> Ptr P256Y -> IO () + +-- compute (out_x,out_y) = n * (in_x,in_y) +foreign import ccall "cryptonite_p256e_point_mul" + ccryptonite_p256e_point_mul :: Ptr P256Scalar -- n + -> Ptr P256X -> Ptr P256Y -- in_{x,y} + -> Ptr P256X -> Ptr P256Y -- out_{x,y} + -> IO () -- compute (out_x,out,y) = n1 * G + n2 * (in_x,in_y) foreign import ccall "cryptonite_p256_points_mul_vartime" diff --git a/Crypto/PubKey/ECDSA.hs b/Crypto/PubKey/ECDSA.hs new file mode 100644 --- /dev/null +++ b/Crypto/PubKey/ECDSA.hs @@ -0,0 +1,272 @@ +-- | +-- Module : Crypto.PubKey.ECDSA +-- License : BSD-style +-- Maintainer : Vincent Hanquez +-- Stability : experimental +-- Portability : unknown +-- +-- Elliptic Curve Digital Signature Algorithm, with the parameterized +-- curve implementations provided by module "Crypto.ECC". +-- +-- Public/private key pairs can be generated using +-- 'curveGenerateKeyPair' or decoded from binary. +-- +-- /WARNING:/ Only curve P-256 has constant-time implementation. +-- Signature operations with P-384 and P-521 may leak the private key. +-- +-- Signature verification should be safe for all curves. +{-# LANGUAGE BangPatterns #-} +{-# LANGUAGE FlexibleContexts #-} +{-# LANGUAGE ScopedTypeVariables #-} +{-# LANGUAGE StandaloneDeriving #-} +{-# LANGUAGE TypeFamilies #-} +{-# LANGUAGE UndecidableInstances #-} +module Crypto.PubKey.ECDSA + ( EllipticCurveECDSA (..) + -- * Public keys + , PublicKey + , encodePublic + , decodePublic + , toPublic + -- * Private keys + , PrivateKey + , encodePrivate + , decodePrivate + -- * Signatures + , Signature(..) + , signatureFromIntegers + , signatureToIntegers + -- * Generation and verification + , signWith + , signDigestWith + , sign + , signDigest + , verify + , verifyDigest + ) where + +import Control.Monad + +import Crypto.ECC +import qualified Crypto.ECC.Simple.Types as Simple +import Crypto.Error +import Crypto.Hash +import Crypto.Hash.Types +import Crypto.Internal.ByteArray (ByteArray, ByteArrayAccess) +import Crypto.Internal.Imports +import Crypto.Number.ModArithmetic (inverseFermat) +import qualified Crypto.PubKey.ECC.P256 as P256 +import Crypto.Random.Types + +import Data.Bits +import qualified Data.ByteArray as B +import Data.Data + +import Foreign.Ptr (Ptr) +import Foreign.Storable (peekByteOff, pokeByteOff) + +-- | Represent a ECDSA signature namely R and S. +data Signature curve = Signature + { sign_r :: Scalar curve -- ^ ECDSA r + , sign_s :: Scalar curve -- ^ ECDSA s + } + +deriving instance Eq (Scalar curve) => Eq (Signature curve) +deriving instance Show (Scalar curve) => Show (Signature curve) + +instance NFData (Scalar curve) => NFData (Signature curve) where + rnf (Signature r s) = rnf r `seq` rnf s `seq` () + +-- | ECDSA Public Key. +type PublicKey curve = Point curve + +-- | ECDSA Private Key. +type PrivateKey curve = Scalar curve + +-- | Elliptic curves with ECDSA capabilities. +class EllipticCurveBasepointArith curve => EllipticCurveECDSA curve where + -- | Is a scalar in the accepted range for ECDSA + scalarIsValid :: proxy curve -> Scalar curve -> Bool + + -- | Test whether the scalar is zero + scalarIsZero :: proxy curve -> Scalar curve -> Bool + scalarIsZero prx s = s == throwCryptoError (scalarFromInteger prx 0) + + -- | Scalar inversion modulo the curve order + scalarInv :: proxy curve -> Scalar curve -> Maybe (Scalar curve) + + -- | Return the point X coordinate as a scalar + pointX :: proxy curve -> Point curve -> Maybe (Scalar curve) + +instance EllipticCurveECDSA Curve_P256R1 where + scalarIsValid _ s = not (P256.scalarIsZero s) + && P256.scalarCmp s P256.scalarN == LT + + scalarIsZero _ = P256.scalarIsZero + + scalarInv _ s = let inv = P256.scalarInvSafe s + in if P256.scalarIsZero inv then Nothing else Just inv + + pointX _ = P256.pointX + +instance EllipticCurveECDSA Curve_P384R1 where + scalarIsValid _ = ecScalarIsValid (Proxy :: Proxy Simple.SEC_p384r1) + + scalarIsZero _ = ecScalarIsZero + + scalarInv _ = ecScalarInv (Proxy :: Proxy Simple.SEC_p384r1) + + pointX _ = ecPointX (Proxy :: Proxy Simple.SEC_p384r1) + +instance EllipticCurveECDSA Curve_P521R1 where + scalarIsValid _ = ecScalarIsValid (Proxy :: Proxy Simple.SEC_p521r1) + + scalarIsZero _ = ecScalarIsZero + + scalarInv _ = ecScalarInv (Proxy :: Proxy Simple.SEC_p521r1) + + pointX _ = ecPointX (Proxy :: Proxy Simple.SEC_p521r1) + + +-- | Create a signature from integers (R, S). +signatureFromIntegers :: EllipticCurveECDSA curve + => proxy curve -> (Integer, Integer) -> CryptoFailable (Signature curve) +signatureFromIntegers prx (r, s) = + liftA2 Signature (scalarFromInteger prx r) (scalarFromInteger prx s) + +-- | Get integers (R, S) from a signature. +-- +-- The values can then be used to encode the signature to binary with +-- ASN.1. +signatureToIntegers :: EllipticCurveECDSA curve + => proxy curve -> Signature curve -> (Integer, Integer) +signatureToIntegers prx sig = + (scalarToInteger prx $ sign_r sig, scalarToInteger prx $ sign_s sig) + +-- | Encode a public key into binary form, i.e. the uncompressed encoding +-- referenced from section 2.2. +encodePublic :: (EllipticCurve curve, ByteArray bs) + => proxy curve -> PublicKey curve -> bs +encodePublic = encodePoint + +-- | Try to decode the binary form of a public key. +decodePublic :: (EllipticCurve curve, ByteArray bs) + => proxy curve -> bs -> CryptoFailable (PublicKey curve) +decodePublic = decodePoint + +-- | Encode a private key into binary form, i.e. the @privateKey@ field +-- described in . +encodePrivate :: (EllipticCurveECDSA curve, ByteArray bs) + => proxy curve -> PrivateKey curve -> bs +encodePrivate = encodeScalar + +-- | Try to decode the binary form of a private key. +decodePrivate :: (EllipticCurveECDSA curve, ByteArray bs) + => proxy curve -> bs -> CryptoFailable (PrivateKey curve) +decodePrivate = decodeScalar + +-- | Create a public key from a private key. +toPublic :: EllipticCurveECDSA curve + => proxy curve -> PrivateKey curve -> PublicKey curve +toPublic = pointBaseSmul + +-- | Sign digest using the private key and an explicit k scalar. +signDigestWith :: (EllipticCurveECDSA curve, HashAlgorithm hash) + => proxy curve -> Scalar curve -> PrivateKey curve -> Digest hash -> Maybe (Signature curve) +signDigestWith prx k d digest = do + let z = tHashDigest prx digest + point = pointBaseSmul prx k + r <- pointX prx point + kInv <- scalarInv prx k + let s = scalarMul prx kInv (scalarAdd prx z (scalarMul prx r d)) + when (scalarIsZero prx r || scalarIsZero prx s) Nothing + return $ Signature r s + +-- | Sign message using the private key and an explicit k scalar. +signWith :: (EllipticCurveECDSA curve, ByteArrayAccess msg, HashAlgorithm hash) + => proxy curve -> Scalar curve -> PrivateKey curve -> hash -> msg -> Maybe (Signature curve) +signWith prx k d hashAlg msg = signDigestWith prx k d (hashWith hashAlg msg) + +-- | Sign a digest using hash and private key. +signDigest :: (EllipticCurveECDSA curve, MonadRandom m, HashAlgorithm hash) + => proxy curve -> PrivateKey curve -> Digest hash -> m (Signature curve) +signDigest prx pk digest = do + k <- curveGenerateScalar prx + case signDigestWith prx k pk digest of + Nothing -> signDigest prx pk digest + Just sig -> return sig + +-- | Sign a message using hash and private key. +sign :: (EllipticCurveECDSA curve, MonadRandom m, ByteArrayAccess msg, HashAlgorithm hash) + => proxy curve -> PrivateKey curve -> hash -> msg -> m (Signature curve) +sign prx pk hashAlg msg = signDigest prx pk (hashWith hashAlg msg) + +-- | Verify a digest using hash and public key. +verifyDigest :: (EllipticCurveECDSA curve, HashAlgorithm hash) + => proxy curve -> PublicKey curve -> Signature curve -> Digest hash -> Bool +verifyDigest prx q (Signature r s) digest + | not (scalarIsValid prx r) = False + | not (scalarIsValid prx s) = False + | otherwise = maybe False (r ==) $ do + w <- scalarInv prx s + let z = tHashDigest prx digest + u1 = scalarMul prx z w + u2 = scalarMul prx r w + x = pointsSmulVarTime prx u1 u2 q + pointX prx x + -- Note: precondition q /= PointO is not tested because we assume + -- point decoding never decodes point at infinity. + +-- | Verify a signature using hash and public key. +verify :: (EllipticCurveECDSA curve, ByteArrayAccess msg, HashAlgorithm hash) + => proxy curve -> hash -> PublicKey curve -> Signature curve -> msg -> Bool +verify prx hashAlg q sig msg = verifyDigest prx q sig (hashWith hashAlg msg) + +-- | Truncate a digest based on curve order size. +tHashDigest :: (EllipticCurveECDSA curve, HashAlgorithm hash) + => proxy curve -> Digest hash -> Scalar curve +tHashDigest prx (Digest digest) = throwCryptoError $ decodeScalar prx encoded + where m = curveOrderBits prx + d = m - B.length digest * 8 + (n, r) = m `divMod` 8 + n' = if r > 0 then succ n else n + + encoded + | d > 0 = B.zero (n' - B.length digest) `B.append` digest + | d == 0 = digest + | r == 0 = B.take n digest + | otherwise = shiftBytes digest + + shiftBytes bs = B.allocAndFreeze n' $ \dst -> + B.withByteArray bs $ \src -> go dst src 0 0 + + go :: Ptr Word8 -> Ptr Word8 -> Word8 -> Int -> IO () + go dst src !a i + | i >= n' = return () + | otherwise = do + b <- peekByteOff src i + pokeByteOff dst i (unsafeShiftR b (8 - r) .|. unsafeShiftL a r) + go dst src b (succ i) + + +ecScalarIsValid :: Simple.Curve c => proxy c -> Simple.Scalar c -> Bool +ecScalarIsValid prx (Simple.Scalar s) = s > 0 && s < n + where n = Simple.curveEccN $ Simple.curveParameters prx + +ecScalarIsZero :: forall curve . Simple.Curve curve + => Simple.Scalar curve -> Bool +ecScalarIsZero (Simple.Scalar a) = a == 0 + +ecScalarInv :: Simple.Curve c + => proxy c -> Simple.Scalar c -> Maybe (Simple.Scalar c) +ecScalarInv prx (Simple.Scalar s) + | i == 0 = Nothing + | otherwise = Just $ Simple.Scalar i + where n = Simple.curveEccN $ Simple.curveParameters prx + i = inverseFermat s n + +ecPointX :: Simple.Curve c + => proxy c -> Simple.Point c -> Maybe (Simple.Scalar c) +ecPointX _ Simple.PointO = Nothing +ecPointX prx (Simple.Point x _) = Just (Simple.Scalar $ x `mod` n) + where n = Simple.curveEccN $ Simple.curveParameters prx diff --git a/Crypto/Random.hs b/Crypto/Random.hs --- a/Crypto/Random.hs +++ b/Crypto/Random.hs @@ -80,6 +80,10 @@ -- -- It can also be used in other contexts provided the input -- has been properly randomly generated. +-- +-- Note that the @Arbitrary@ instance provided by QuickCheck for 'Word64' does +-- not have a uniform distribution. It is often better to use instead +-- @arbitraryBoundedRandom@. drgNewTest :: (Word64, Word64, Word64, Word64, Word64) -> ChaChaDRG drgNewTest = initializeWords diff --git a/Crypto/Random/Types.hs b/Crypto/Random/Types.hs --- a/Crypto/Random/Types.hs +++ b/Crypto/Random/Types.hs @@ -17,7 +17,7 @@ import Crypto.Internal.ByteArray -- | A monad constraint that allows to generate random bytes -class (Functor m, Monad m) => MonadRandom m where +class Monad m => MonadRandom m where getRandomBytes :: ByteArray byteArray => Int -> m byteArray -- | A Deterministic Random Generator (DRG) class diff --git a/Crypto/System/CPU.hs b/Crypto/System/CPU.hs new file mode 100644 --- /dev/null +++ b/Crypto/System/CPU.hs @@ -0,0 +1,64 @@ +-- | +-- Module : Crypto.System.CPU +-- License : BSD-style +-- Maintainer : Olivier Chéron +-- Stability : experimental +-- Portability : unknown +-- +-- Gives information about cryptonite runtime environment. +-- +{-# LANGUAGE CPP #-} +{-# LANGUAGE DeriveDataTypeable #-} +{-# LANGUAGE ForeignFunctionInterface #-} +module Crypto.System.CPU + ( ProcessorOption (..) + , processorOptions + ) where + +import Data.Data +import Data.List (findIndices) +#ifdef SUPPORT_RDRAND +import Data.Maybe (isJust) +#endif +import Data.Word (Word8) +import Foreign.Ptr +import Foreign.Storable + +import Crypto.Internal.Compat + +#ifdef SUPPORT_RDRAND +import Crypto.Random.Entropy.RDRand +import Crypto.Random.Entropy.Source +#endif + +-- | CPU options impacting cryptography implementation and library performance. +data ProcessorOption + = AESNI -- ^ Support for AES instructions, with flag @support_aesni@ + | PCLMUL -- ^ Support for CLMUL instructions, with flag @support_pclmuldq@ + | RDRAND -- ^ Support for RDRAND instruction, with flag @support_rdrand@ + deriving (Show,Eq,Enum,Data) + +-- | Options which have been enabled at compile time and are supported by the +-- current CPU. +processorOptions :: [ProcessorOption] +processorOptions = unsafeDoIO $ do + p <- cryptonite_aes_cpu_init + options <- traverse (getOption p) aesOptions + rdrand <- hasRDRand + return (decodeOptions options ++ [ RDRAND | rdrand ]) + where + aesOptions = [ AESNI .. PCLMUL ] + getOption p = peekElemOff p . fromEnum + decodeOptions = map toEnum . findIndices (> 0) +{-# NOINLINE processorOptions #-} + +hasRDRand :: IO Bool +#ifdef SUPPORT_RDRAND +hasRDRand = fmap isJust getRDRand + where getRDRand = entropyOpen :: IO (Maybe RDRand) +#else +hasRDRand = return False +#endif + +foreign import ccall unsafe "cryptonite_aes_cpu_init" + cryptonite_aes_cpu_init :: IO (Ptr Word8) diff --git a/Crypto/Tutorial.hs b/Crypto/Tutorial.hs --- a/Crypto/Tutorial.hs +++ b/Crypto/Tutorial.hs @@ -8,6 +8,9 @@ -- * Symmetric block ciphers -- $symmetric_block_ciphers + + -- * Combining primitives + -- $combining_primitives ) where -- $api_design @@ -147,3 +150,46 @@ -- > putStrLn $ "Original Message: " ++ show msg -- > putStrLn $ "Message after encryption: " ++ show eMsg -- > putStrLn $ "Message after decryption: " ++ show dMsg + +-- $combining_primitives +-- +-- This example shows how to use Curve25519, XSalsa and Poly1305 primitives to +-- emulate NaCl's @crypto_box@ construct. +-- +-- > import qualified Data.ByteArray as BA +-- > import Data.ByteString (ByteString) +-- > import qualified Data.ByteString as B +-- > +-- > import qualified Crypto.Cipher.XSalsa as XSalsa +-- > import qualified Crypto.MAC.Poly1305 as Poly1305 +-- > import qualified Crypto.PubKey.Curve25519 as X25519 +-- > +-- > -- | Build a @crypto_box@ packet encrypting the specified content with a +-- > -- 192-bit nonce, receiver public key and sender private key. +-- > crypto_box content nonce pk sk = BA.convert tag `B.append` c +-- > where +-- > zero = B.replicate 16 0 +-- > shared = X25519.dh pk sk +-- > (iv0, iv1) = B.splitAt 8 nonce +-- > state0 = XSalsa.initialize 20 shared (zero `B.append` iv0) +-- > state1 = XSalsa.derive state0 iv1 +-- > (rs, state2) = XSalsa.generate state1 32 +-- > (c, _) = XSalsa.combine state2 content +-- > tag = Poly1305.auth (rs :: ByteString) c +-- > +-- > -- | Try to open a @crypto_box@ packet and recover the content using the +-- > -- 192-bit nonce, sender public key and receiver private key. +-- > crypto_box_open packet nonce pk sk +-- > | B.length packet < 16 = Nothing +-- > | BA.constEq tag' tag = Just content +-- > | otherwise = Nothing +-- > where +-- > (tag', c) = B.splitAt 16 packet +-- > zero = B.replicate 16 0 +-- > shared = X25519.dh pk sk +-- > (iv0, iv1) = B.splitAt 8 nonce +-- > state0 = XSalsa.initialize 20 shared (zero `B.append` iv0) +-- > state1 = XSalsa.derive state0 iv1 +-- > (rs, state2) = XSalsa.generate state1 32 +-- > (content, _) = XSalsa.combine state2 c +-- > tag = Poly1305.auth (rs :: ByteString) c diff --git a/benchs/Bench.hs b/benchs/Bench.hs --- a/benchs/Bench.hs +++ b/benchs/Bench.hs @@ -6,6 +6,7 @@ import Gauge.Main import Crypto.Cipher.AES +import qualified Crypto.Cipher.AESGCMSIV as AESGCMSIV import Crypto.Cipher.Blowfish import Crypto.Cipher.CAST5 import qualified Crypto.Cipher.ChaChaPoly1305 as CP @@ -22,12 +23,15 @@ import qualified Crypto.PubKey.DH as DH import qualified Crypto.PubKey.ECC.Types as ECC import qualified Crypto.PubKey.ECC.Prim as ECC +import qualified Crypto.PubKey.ECDSA as ECDSA import Crypto.Random import Control.DeepSeq (NFData) import Data.ByteArray (ByteArray, Bytes) import qualified Data.ByteString as B +import qualified Crypto.PubKey.ECC.P256 as P256 + import Number.F2m data HashAlg = forall alg . HashAlgorithm alg => HashAlg alg @@ -123,7 +127,7 @@ [ bgroup "ECB" benchECB , bgroup "CBC" benchCBC ] - where + where benchECB = [ bench "DES-input=1024" $ nf (run (undefined :: DES) cipherInit key8) input1024 , bench "Blowfish128-input=1024" $ nf (run (undefined :: Blowfish128) cipherInit key16) input1024 @@ -165,6 +169,7 @@ [ bench "ChaChaPoly1305" $ nf (cp key32) (input64, input1024) , bench "AES-GCM" $ nf (gcm key32) (input64, input1024) , bench "AES-CCM" $ nf (ccm key32) (input64, input1024) + , bench "AES-GCM-SIV" $ nf (gcmsiv key32) (input64, input1024) ] where cp k (ini, plain) = let iniState = throwCryptoError $ CP.initialize k (throwCryptoError $ CP.nonce12 nonce12) @@ -184,6 +189,11 @@ state = throwCryptoError $ aeadInit mode ctx nonce12 in aeadSimpleEncrypt state ini plain 16 + gcmsiv k (ini, plain) = + let ctx = throwCryptoError (cipherInit k) :: AES256 + iv = throwCryptoError (AESGCMSIV.nonce nonce12) + in AESGCMSIV.encrypt ctx iv ini plain + input64 = B.replicate 64 0 input1024 = B.replicate 1024 0 @@ -195,20 +205,42 @@ benchECC = [ bench "pointAddTwoMuls-baseline" $ nf run_b (n1, p1, n2, p2) , bench "pointAddTwoMuls-optimized" $ nf run_o (n1, p1, n2, p2) + , bench "pointAdd-ECC" $ nf run_c (p1, p2) + , bench "pointMul-ECC" $ nf run_d (n1, p2) ] where run_b (n, p, k, q) = ECC.pointAdd c (ECC.pointMul c n p) (ECC.pointMul c k q) run_o (n, p, k, q) = ECC.pointAddTwoMuls c n p k q + run_c (p, q) = ECC.pointAdd c p q + run_d (n, p) = ECC.pointMul c n p c = ECC.getCurveByName ECC.SEC_p256r1 - r1 = 7 - r2 = 11 - p1 = ECC.pointBaseMul c r1 - p2 = ECC.pointBaseMul c r2 + p1 = ECC.pointBaseMul c n1 + p2 = ECC.pointBaseMul c n2 n1 = 0x2ba9daf2363b2819e69b34a39cf496c2458a9b2a21505ea9e7b7cbca42dc7435 n2 = 0xf054a7f60d10b8c2cf847ee90e9e029f8b0e971b09ca5f55c4d49921a11fadc1 +benchP256 = + [ bench "pointAddTwoMuls-P256" $ nf run_p (n1, p1, n2, p2) + , bench "pointAdd-P256" $ nf run_q (p1, p2) + , bench "pointMul-P256" $ nf run_t (n1, p1) + ] + where run_p (n, p, k, q) = P256.pointAdd (P256.pointMul n p) (P256.pointMul k q) + run_q (p, q) = P256.pointAdd p q + run_t (n, p) = P256.pointMul n p + + xS = 0xde2444bebc8d36e682edd27e0f271508617519b3221a8fa0b77cab3989da97c9 + yS = 0xc093ae7ff36e5380fc01a5aad1e66659702de80f53cec576b6350b243042a256 + xT = 0x55a8b00f8da1d44e62f6b3b25316212e39540dc861c89575bb8cf92e35e0986b + yT = 0x5421c3209c2d6c704835d82ac4c3dd90f61a8a52598b9e7ab656e9d8c8b24316 + p1 = P256.pointFromIntegers (xS, yS) + p2 = P256.pointFromIntegers (xT, yT) + n1 = throwCryptoError $ P256.scalarFromInteger 0x2ba9daf2363b2819e69b34a39cf496c2458a9b2a21505ea9e7b7cbca42dc7435 + n2 = throwCryptoError $ P256.scalarFromInteger 0xf054a7f60d10b8c2cf847ee90e9e029f8b0e971b09ca5f55c4d49921a11fadc1 + + + benchFFDH = map doFFDHBench primes where doFFDHBench (e, p) = @@ -255,6 +287,44 @@ , ("X448", CurveDH Curve_X448) ] +data CurveHashECDSA = + forall curve hashAlg . (ECDSA.EllipticCurveECDSA curve, + NFData (Scalar curve), + NFData (Point curve), + HashAlgorithm hashAlg) => CurveHashECDSA curve hashAlg + +benchECDSA = map doECDSABench curveHashes + where + doECDSABench (name, CurveHashECDSA c hashAlg) = + let proxy = Just c -- using Maybe as Proxy + in bgroup name + [ env (signGenerate proxy) $ bench "sign" . nfIO . signRun proxy hashAlg + , env (verifyGenerate proxy hashAlg) $ bench "verify" . nf (verifyRun proxy hashAlg) + ] + + signGenerate proxy = do + m <- tenKB + s <- curveGenerateScalar proxy + return (s, m) + + signRun proxy hashAlg (priv, msg) = ECDSA.sign proxy priv hashAlg msg + + verifyGenerate proxy hashAlg = do + m <- tenKB + KeyPair p s <- curveGenerateKeyPair proxy + sig <- ECDSA.sign proxy s hashAlg m + return (p, sig, m) + + verifyRun proxy hashAlg (pub, sig, msg) = ECDSA.verify proxy hashAlg pub sig msg + + tenKB :: IO Bytes + tenKB = getRandomBytes 10240 + + curveHashes = [ ("secp256r1_sha256", CurveHashECDSA Curve_P256R1 SHA256) + , ("secp384r1_sha384", CurveHashECDSA Curve_P384R1 SHA384) + , ("secp521r1_sha512", CurveHashECDSA Curve_P521R1 SHA512) + ] + main = defaultMain [ bgroup "hash" benchHash , bgroup "block-cipher" benchBlockCipher @@ -262,9 +332,11 @@ , bgroup "pbkdf2" benchPBKDF2 , bgroup "bcrypt" benchBCrypt , bgroup "ECC" benchECC + , bgroup "P256" benchP256 , bgroup "DH" [ bgroup "FFDH" benchFFDH , bgroup "ECDH" benchECDH ] + , bgroup "ECDSA" benchECDSA , bgroup "F2m" benchF2m ] diff --git a/cbits/aes/block128.h b/cbits/aes/block128.h --- a/cbits/aes/block128.h +++ b/cbits/aes/block128.h @@ -108,6 +108,13 @@ } } +static inline void block128_byte_reverse(block128 *a) +{ + uint64_t s0 = a->q[0], s1 = a->q[1]; + a->q[0] = bitfn_swap64(s1); + a->q[1] = bitfn_swap64(s0); +} + static inline void block128_inc_be(block128 *b) { uint64_t v = be64_to_cpu(b->q[1]); @@ -116,6 +123,16 @@ b->q[1] = 0; } else b->q[1] = cpu_to_be64(v); +} + +static inline void block128_inc32_be(block128 *b) +{ + b->d[3] = cpu_to_be32(be32_to_cpu(b->d[3]) + 1); +} + +static inline void block128_inc32_le(block128 *b) +{ + b->d[0] = cpu_to_le32(le32_to_cpu(b->d[0]) + 1); } #ifdef IMPL_DEBUG diff --git a/cbits/aes/gf.c b/cbits/aes/gf.c --- a/cbits/aes/gf.c +++ b/cbits/aes/gf.c @@ -34,39 +34,113 @@ #include #include -/* this is a really inefficient way to GF multiply. - * the alternative without hw accel is building small tables - * to speed up the multiplication. - * TODO: optimise with tables +/* inplace GFMUL for xts mode */ +void cryptonite_aes_generic_gf_mulx(block128 *a) +{ + const uint64_t gf_mask = cpu_to_le64(0x8000000000000000ULL); + uint64_t r = ((a->q[1] & gf_mask) ? cpu_to_le64(0x87) : 0); + a->q[1] = cpu_to_le64((le64_to_cpu(a->q[1]) << 1) | (a->q[0] & gf_mask ? 1 : 0)); + a->q[0] = cpu_to_le64(le64_to_cpu(a->q[0]) << 1) ^ r; +} + + +/* + * GF multiplication with Shoup's method and 4-bit table. + * + * We precompute the products of H with all 4-bit polynomials and store them in + * a 'table_4bit' array. To avoid unnecessary byte swapping, the 16 blocks are + * written to the table with qwords already converted to CPU order. Table + * indices use the reflected bit ordering, i.e. polynomials X^0, X^1, X^2, X^3 + * map to bit positions 3, 2, 1, 0 respectively. + * + * To multiply an arbitrary block with H, the input block is decomposed in 4-bit + * segments. We get the final result after 32 table lookups and additions, one + * for each segment, interleaving multiplication by P(X)=X^4. */ -void cryptonite_aes_generic_gf_mul(block128 *a, block128 *b) + +/* convert block128 qwords between BE and CPU order */ +static inline void block128_cpu_swap_be(block128 *a, const block128 *b) { - uint64_t a0, a1, v0, v1; + a->q[1] = cpu_to_be64(b->q[1]); + a->q[0] = cpu_to_be64(b->q[0]); +} + +/* multiplication by P(X)=X, assuming qwords already in CPU order */ +static inline void cpu_gf_mulx(block128 *a, const block128 *b) +{ + uint64_t v0 = b->q[0]; + uint64_t v1 = b->q[1]; + a->q[1] = v1 >> 1 | v0 << 63; + a->q[0] = v0 >> 1 ^ ((0-(v1 & 1)) & 0xe100000000000000ULL); +} + +static const uint64_t r4_0[] = + { 0x0000000000000000ULL, 0x1c20000000000000ULL + , 0x3840000000000000ULL, 0x2460000000000000ULL + , 0x7080000000000000ULL, 0x6ca0000000000000ULL + , 0x48c0000000000000ULL, 0x54e0000000000000ULL + , 0xe100000000000000ULL, 0xfd20000000000000ULL + , 0xd940000000000000ULL, 0xc560000000000000ULL + , 0x9180000000000000ULL, 0x8da0000000000000ULL + , 0xa9c0000000000000ULL, 0xb5e0000000000000ULL + }; + +/* multiplication by P(X)=X^4, assuming qwords already in CPU order */ +static inline void cpu_gf_mulx4(block128 *a, const block128 *b) +{ + uint64_t v0 = b->q[0]; + uint64_t v1 = b->q[1]; + a->q[1] = v1 >> 4 | v0 << 60; + a->q[0] = v0 >> 4 ^ r4_0[v1 & 0xf]; +} + +/* initialize the 4-bit table given H */ +void cryptonite_aes_generic_hinit(table_4bit htable, const block128 *h) +{ + block128 v, *p; int i, j; - a0 = a1 = 0; - v0 = cpu_to_be64(a->q[0]); - v1 = cpu_to_be64(a->q[1]); + /* multiplication by 0 is 0 */ + block128_zero(&htable[0]); - for (i = 0; i < 16; i++) - for (j = 0x80; j != 0; j >>= 1) { - uint8_t x = b->b[i] & j; - a0 ^= x ? v0 : 0; - a1 ^= x ? v1 : 0; - x = (uint8_t) v1 & 1; - v1 = (v1 >> 1) | (v0 << 63); - v0 = (v0 >> 1) ^ (x ? (0xe1ULL << 56) : 0); + /* at index 8=2^3 we have H.X^0 = H */ + i = 8; + block128_cpu_swap_be(&htable[i], h); /* in CPU order */ + p = &htable[i]; + + /* for other powers of 2, repeat multiplication by P(X)=X */ + for (i = 4; i > 0; i >>= 1) + { + cpu_gf_mulx(&htable[i], p); + p = &htable[i]; + } + + /* remaining elements are linear combinations */ + for (i = 2; i < 16; i <<= 1) { + p = &htable[i]; + v = *p; + for (j = 1; j < i; j++) { + p[j] = v; + block128_xor_aligned(&p[j], &htable[j]); } - a->q[0] = cpu_to_be64(a0); - a->q[1] = cpu_to_be64(a1); + } } -/* inplace GFMUL for xts mode */ -void cryptonite_aes_generic_gf_mulx(block128 *a) +/* multiply a block with H */ +void cryptonite_aes_generic_gf_mul(block128 *a, const table_4bit htable) { - const uint64_t gf_mask = cpu_to_le64(0x8000000000000000ULL); - uint64_t r = ((a->q[1] & gf_mask) ? cpu_to_le64(0x87) : 0); - a->q[1] = cpu_to_le64((le64_to_cpu(a->q[1]) << 1) | (a->q[0] & gf_mask ? 1 : 0)); - a->q[0] = cpu_to_le64(le64_to_cpu(a->q[0]) << 1) ^ r; + block128 b; + int i; + block128_zero(&b); + for (i = 15; i >= 0; i--) + { + uint8_t v = a->b[i]; + block128_xor_aligned(&b, &htable[v & 0xf]); /* high bits (reflected) */ + cpu_gf_mulx4(&b, &b); + block128_xor_aligned(&b, &htable[v >> 4]); /* low bits (reflected) */ + if (i > 0) + cpu_gf_mulx4(&b, &b); + else + block128_cpu_swap_be(a, &b); /* restore BE order when done */ + } } - diff --git a/cbits/aes/gf.h b/cbits/aes/gf.h --- a/cbits/aes/gf.h +++ b/cbits/aes/gf.h @@ -32,7 +32,11 @@ #include "aes/block128.h" -void cryptonite_aes_generic_gf_mul(block128 *a, block128 *b); +typedef block128 table_4bit[16]; + void cryptonite_aes_generic_gf_mulx(block128 *a); + +void cryptonite_aes_generic_hinit(table_4bit htable, const block128 *h); +void cryptonite_aes_generic_gf_mul(block128 *a, const table_4bit htable); #endif diff --git a/cbits/aes/x86ni.c b/cbits/aes/x86ni.c --- a/cbits/aes/x86ni.c +++ b/cbits/aes/x86ni.c @@ -46,6 +46,7 @@ /* old GCC version doesn't cope with the shuffle parameters, that can take 2 values (0xff and 0xaa) * in our case, passed as argument despite being a immediate 8 bits constant anyway. * un-factorise aes_128_key_expansion into 2 version that have the shuffle parameter explicitly set */ +TARGET_AESNI static __m128i aes_128_key_expansion_ff(__m128i key, __m128i keygened) { keygened = _mm_shuffle_epi32(keygened, 0xff); @@ -55,6 +56,7 @@ return _mm_xor_si128(key, keygened); } +TARGET_AESNI static __m128i aes_128_key_expansion_aa(__m128i key, __m128i keygened) { keygened = _mm_shuffle_epi32(keygened, 0xaa); @@ -64,6 +66,7 @@ return _mm_xor_si128(key, keygened); } +TARGET_AESNI void cryptonite_aesni_init(aes_key *key, uint8_t *ikey, uint8_t size) { __m128i k[28]; @@ -145,6 +148,7 @@ /* TO OPTIMISE: use pcmulqdq... or some faster code. * this is the lamest way of doing it, but i'm out of time. * this is basically a copy of gf_mulx in gf.c */ +TARGET_AESNI static __m128i gfmulx(__m128i v) { uint64_t v_[2] ALIGNMENT(16); @@ -158,33 +162,34 @@ return v; } -static __m128i gfmul_generic(__m128i tag, __m128i h) +TARGET_AESNI +static __m128i gfmul_generic(__m128i tag, const table_4bit htable) { - aes_block _t, _h; + aes_block _t; _mm_store_si128((__m128i *) &_t, tag); - _mm_store_si128((__m128i *) &_h, h); - cryptonite_aes_generic_gf_mul(&_t, &_h); + cryptonite_aes_generic_gf_mul(&_t, htable); tag = _mm_load_si128((__m128i *) &_t); return tag; } #ifdef WITH_PCLMUL -__m128i (*gfmul_branch_ptr)(__m128i a, __m128i b) = gfmul_generic; -#define gfmul(a,b) ((*gfmul_branch_ptr)(a,b)) +__m128i (*gfmul_branch_ptr)(__m128i a, const table_4bit t) = gfmul_generic; +#define gfmul(a,t) ((*gfmul_branch_ptr)(a,t)) /* See Intel carry-less-multiplication-instruction-in-gcm-mode-paper.pdf * * Adapted from figure 5, with additional byte swapping so that interface * is simimar to cryptonite_aes_generic_gf_mul. */ -static __m128i gfmul_pclmuldq(__m128i a, __m128i b) +TARGET_AESNI_PCLMUL +static __m128i gfmul_pclmuldq(__m128i a, const table_4bit htable) { - __m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp8, tmp9; + __m128i b, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp8, tmp9; __m128i bswap_mask = _mm_set_epi8(0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15); a = _mm_shuffle_epi8(a, bswap_mask); - b = _mm_shuffle_epi8(b, bswap_mask); + b = _mm_loadu_si128((__m128i *) htable); tmp3 = _mm_clmulepi64_si128(a, b, 0x00); tmp4 = _mm_clmulepi64_si128(a, b, 0x10); @@ -231,28 +236,39 @@ return _mm_shuffle_epi8(tmp6, bswap_mask); } -void cryptonite_aesni_gf_mul(block128 *a, block128 *b) +void cryptonite_aesni_hinit_pclmul(table_4bit htable, const block128 *h) { - __m128i _a, _b, _c; + /* When pclmul is active we don't need to fill the table. Instead we just + * store H at index 0. It is written in reverse order, so function + * gfmul_pclmuldq will not byte-swap this value. + */ + htable->q[0] = bitfn_swap64(h->q[1]); + htable->q[1] = bitfn_swap64(h->q[0]); +} + +TARGET_AESNI_PCLMUL +void cryptonite_aesni_gf_mul_pclmul(block128 *a, const table_4bit htable) +{ + __m128i _a, _b; _a = _mm_loadu_si128((__m128i *) a); - _b = _mm_loadu_si128((__m128i *) b); - _c = gfmul_pclmuldq(_a, _b); - _mm_storeu_si128((__m128i *) a, _c); + _b = gfmul_pclmuldq(_a, htable); + _mm_storeu_si128((__m128i *) a, _b); } -void cryptonite_aesni_init_pclmul() +void cryptonite_aesni_init_pclmul(void) { gfmul_branch_ptr = gfmul_pclmuldq; } #else -#define gfmul(a,b) (gfmul_generic(a,b)) +#define gfmul(a,t) (gfmul_generic(a,t)) #endif -static inline __m128i ghash_add(__m128i tag, __m128i h, __m128i m) +TARGET_AESNI +static inline __m128i ghash_add(__m128i tag, const table_4bit htable, __m128i m) { tag = _mm_xor_si128(tag, m); - return gfmul(tag, h); + return gfmul(tag, htable); } #define PRELOAD_ENC_KEYS128(k) \ diff --git a/cbits/aes/x86ni.h b/cbits/aes/x86ni.h --- a/cbits/aes/x86ni.h +++ b/cbits/aes/x86ni.h @@ -40,7 +40,16 @@ #include #include +#ifdef WITH_TARGET_ATTRIBUTES +#define TARGET_AESNI __attribute__((target("ssse3,aes"))) +#define TARGET_AESNI_PCLMUL __attribute__((target("sse4.1,aes,pclmul"))) +#else +#define TARGET_AESNI +#define TARGET_AESNI_PCLMUL +#endif + #ifdef IMPL_DEBUG +TARGET_AESNI static void block128_sse_print(__m128i m) { block128 b; @@ -64,6 +73,8 @@ void cryptonite_aesni_decrypt_cbc256(aes_block *out, aes_key *key, aes_block *_iv, aes_block *in, uint32_t blocks); void cryptonite_aesni_encrypt_ctr128(uint8_t *out, aes_key *key, aes_block *_iv, uint8_t *in, uint32_t length); void cryptonite_aesni_encrypt_ctr256(uint8_t *out, aes_key *key, aes_block *_iv, uint8_t *in, uint32_t length); +void cryptonite_aesni_encrypt_c32_128(uint8_t *out, aes_key *key, aes_block *_iv, uint8_t *in, uint32_t length); +void cryptonite_aesni_encrypt_c32_256(uint8_t *out, aes_key *key, aes_block *_iv, uint8_t *in, uint32_t length); void cryptonite_aesni_encrypt_xts128(aes_block *out, aes_key *key1, aes_key *key2, aes_block *_tweak, uint32_t spoint, aes_block *in, uint32_t blocks); void cryptonite_aesni_encrypt_xts256(aes_block *out, aes_key *key1, aes_key *key2, @@ -73,8 +84,9 @@ void cryptonite_aesni_gcm_encrypt256(uint8_t *out, aes_gcm *gcm, aes_key *key, uint8_t *in, uint32_t length); #ifdef WITH_PCLMUL -void cryptonite_aesni_init_pclmul(); -void cryptonite_aesni_gf_mul(block128 *a, block128 *b); +void cryptonite_aesni_init_pclmul(void); +void cryptonite_aesni_hinit_pclmul(table_4bit htable, const block128 *h); +void cryptonite_aesni_gf_mul_pclmul(block128 *a, const table_4bit htable); #endif #endif diff --git a/cbits/aes/x86ni_impl.c b/cbits/aes/x86ni_impl.c --- a/cbits/aes/x86ni_impl.c +++ b/cbits/aes/x86ni_impl.c @@ -28,6 +28,7 @@ * SUCH DAMAGE. */ +TARGET_AESNI void SIZED(cryptonite_aesni_encrypt_block)(aes_block *out, aes_key *key, aes_block *in) { __m128i *k = (__m128i *) key->data; @@ -37,6 +38,7 @@ _mm_storeu_si128((__m128i *) out, m); } +TARGET_AESNI void SIZED(cryptonite_aesni_decrypt_block)(aes_block *out, aes_key *key, aes_block *in) { __m128i *k = (__m128i *) key->data; @@ -46,6 +48,7 @@ _mm_storeu_si128((__m128i *) out, m); } +TARGET_AESNI void SIZED(cryptonite_aesni_encrypt_ecb)(aes_block *out, aes_key *key, aes_block *in, uint32_t blocks) { __m128i *k = (__m128i *) key->data; @@ -58,6 +61,7 @@ } } +TARGET_AESNI void SIZED(cryptonite_aesni_decrypt_ecb)(aes_block *out, aes_key *key, aes_block *in, uint32_t blocks) { __m128i *k = (__m128i *) key->data; @@ -71,6 +75,7 @@ } } +TARGET_AESNI void SIZED(cryptonite_aesni_encrypt_cbc)(aes_block *out, aes_key *key, aes_block *_iv, aes_block *in, uint32_t blocks) { __m128i *k = (__m128i *) key->data; @@ -87,6 +92,7 @@ } } +TARGET_AESNI void SIZED(cryptonite_aesni_decrypt_cbc)(aes_block *out, aes_key *key, aes_block *_iv, aes_block *in, uint32_t blocks) { __m128i *k = (__m128i *) key->data; @@ -106,6 +112,7 @@ } } +TARGET_AESNI void SIZED(cryptonite_aesni_encrypt_ctr)(uint8_t *output, aes_key *key, aes_block *_iv, uint8_t *input, uint32_t len) { __m128i *k = (__m128i *) key->data; @@ -151,6 +158,49 @@ return ; } +TARGET_AESNI +void SIZED(cryptonite_aesni_encrypt_c32_)(uint8_t *output, aes_key *key, aes_block *_iv, uint8_t *input, uint32_t len) +{ + __m128i *k = (__m128i *) key->data; + __m128i one = _mm_set_epi32(0,0,0,1); + uint32_t nb_blocks = len / 16; + uint32_t part_block_len = len % 16; + + /* get the IV */ + __m128i iv = _mm_loadu_si128((__m128i *) _iv); + + PRELOAD_ENC(k); + + for (; nb_blocks-- > 0; output += 16, input += 16) { + /* encrypt the iv and and xor it the input block */ + __m128i tmp = iv; + DO_ENC_BLOCK(tmp); + __m128i m = _mm_loadu_si128((__m128i *) input); + m = _mm_xor_si128(m, tmp); + + _mm_storeu_si128((__m128i *) output, m); + /* iv += 1 */ + iv = _mm_add_epi32(iv, one); + } + + if (part_block_len != 0) { + aes_block block; + memset(&block.b, 0, 16); + memcpy(&block.b, input, part_block_len); + + __m128i m = _mm_loadu_si128((__m128i *) &block); + __m128i tmp = iv; + + DO_ENC_BLOCK(tmp); + m = _mm_xor_si128(m, tmp); + _mm_storeu_si128((__m128i *) &block.b, m); + memcpy(output, &block.b, part_block_len); + } + + return ; +} + +TARGET_AESNI void SIZED(cryptonite_aesni_encrypt_xts)(aes_block *out, aes_key *key1, aes_key *key2, aes_block *_tweak, uint32_t spoint, aes_block *in, uint32_t blocks) { @@ -181,6 +231,7 @@ } while (0); } +TARGET_AESNI void SIZED(cryptonite_aesni_gcm_encrypt)(uint8_t *output, aes_gcm *gcm, aes_key *key, uint8_t *input, uint32_t length) { __m128i *k = (__m128i *) key->data; @@ -191,7 +242,6 @@ gcm->length_input += length; - __m128i h = _mm_loadu_si128((__m128i *) &gcm->h); __m128i tag = _mm_loadu_si128((__m128i *) &gcm->tag); __m128i iv = _mm_loadu_si128((__m128i *) &gcm->civ); iv = _mm_shuffle_epi8(iv, bswap_mask); @@ -200,7 +250,7 @@ for (; nb_blocks-- > 0; output += 16, input += 16) { /* iv += 1 */ - iv = _mm_add_epi64(iv, one); + iv = _mm_add_epi32(iv, one); /* put back iv in big endian, encrypt it, * and xor it to input */ @@ -209,7 +259,7 @@ __m128i m = _mm_loadu_si128((__m128i *) input); m = _mm_xor_si128(m, tmp); - tag = ghash_add(tag, h, m); + tag = ghash_add(tag, gcm->htable, m); /* store it out */ _mm_storeu_si128((__m128i *) output, m); @@ -240,7 +290,7 @@ block128_copy_bytes(&block, input, part_block_len); /* iv += 1 */ - iv = _mm_add_epi64(iv, one); + iv = _mm_add_epi32(iv, one); /* put back iv in big endian mode, encrypt it and xor it with input */ __m128i tmp = _mm_shuffle_epi8(iv, bswap_mask); @@ -250,7 +300,7 @@ m = _mm_xor_si128(m, tmp); m = _mm_shuffle_epi8(m, mask); - tag = ghash_add(tag, h, m); + tag = ghash_add(tag, gcm->htable, m); /* make output */ _mm_storeu_si128((__m128i *) &block.b, m); diff --git a/cbits/cryptonite_aes.c b/cbits/cryptonite_aes.c --- a/cbits/cryptonite_aes.c +++ b/cbits/cryptonite_aes.c @@ -44,6 +44,7 @@ void cryptonite_aes_generic_encrypt_cbc(aes_block *output, aes_key *key, aes_block *iv, aes_block *input, uint32_t nb_blocks); void cryptonite_aes_generic_decrypt_cbc(aes_block *output, aes_key *key, aes_block *iv, aes_block *input, uint32_t nb_blocks); void cryptonite_aes_generic_encrypt_ctr(uint8_t *output, aes_key *key, aes_block *iv, uint8_t *input, uint32_t length); +void cryptonite_aes_generic_encrypt_c32(uint8_t *output, aes_key *key, aes_block *iv, uint8_t *input, uint32_t length); void cryptonite_aes_generic_encrypt_xts(aes_block *output, aes_key *k1, aes_key *k2, aes_block *dataunit, uint32_t spoint, aes_block *input, uint32_t nb_blocks); void cryptonite_aes_generic_decrypt_xts(aes_block *output, aes_key *k1, aes_key *k2, aes_block *dataunit, @@ -69,6 +70,8 @@ DECRYPT_CBC_128, DECRYPT_CBC_192, DECRYPT_CBC_256, /* ctr */ ENCRYPT_CTR_128, ENCRYPT_CTR_192, ENCRYPT_CTR_256, + /* ctr with 32-bit wrapping */ + ENCRYPT_C32_128, ENCRYPT_C32_192, ENCRYPT_C32_256, /* xts */ ENCRYPT_XTS_128, ENCRYPT_XTS_192, ENCRYPT_XTS_256, DECRYPT_XTS_128, DECRYPT_XTS_192, DECRYPT_XTS_256, @@ -82,7 +85,7 @@ ENCRYPT_CCM_128, ENCRYPT_CCM_192, ENCRYPT_CCM_256, DECRYPT_CCM_128, DECRYPT_CCM_192, DECRYPT_CCM_256, /* ghash */ - GHASH_GF_MUL, + GHASH_HINIT, GHASH_GF_MUL, }; void *cryptonite_aes_branch_table[] = { @@ -115,6 +118,10 @@ [ENCRYPT_CTR_128] = cryptonite_aes_generic_encrypt_ctr, [ENCRYPT_CTR_192] = cryptonite_aes_generic_encrypt_ctr, [ENCRYPT_CTR_256] = cryptonite_aes_generic_encrypt_ctr, + /* CTR with 32-bit wrapping */ + [ENCRYPT_C32_128] = cryptonite_aes_generic_encrypt_c32, + [ENCRYPT_C32_192] = cryptonite_aes_generic_encrypt_c32, + [ENCRYPT_C32_256] = cryptonite_aes_generic_encrypt_c32, /* XTS */ [ENCRYPT_XTS_128] = cryptonite_aes_generic_encrypt_xts, [ENCRYPT_XTS_192] = cryptonite_aes_generic_encrypt_xts, @@ -144,6 +151,7 @@ [DECRYPT_CCM_192] = cryptonite_aes_generic_ccm_decrypt, [DECRYPT_CCM_256] = cryptonite_aes_generic_ccm_decrypt, /* GHASH */ + [GHASH_HINIT] = cryptonite_aes_generic_hinit, [GHASH_GF_MUL] = cryptonite_aes_generic_gf_mul, }; @@ -156,7 +164,8 @@ typedef void (*ocb_crypt_f)(uint8_t *output, aes_ocb *ocb, aes_key *key, uint8_t *input, uint32_t length); typedef void (*ccm_crypt_f)(uint8_t *output, aes_ccm *ccm, aes_key *key, uint8_t *input, uint32_t length); typedef void (*block_f)(aes_block *output, aes_key *key, aes_block *input); -typedef void (*gf_mul_f)(aes_block *a, aes_block *b); +typedef void (*hinit_f)(table_4bit htable, const block128 *h); +typedef void (*gf_mul_f)(block128 *a, const table_4bit htable); #ifdef WITH_AESNI #define GET_INIT(strength) \ @@ -171,6 +180,8 @@ ((cbc_f) (cryptonite_aes_branch_table[DECRYPT_CBC_128 + strength])) #define GET_CTR_ENCRYPT(strength) \ ((ctr_f) (cryptonite_aes_branch_table[ENCRYPT_CTR_128 + strength])) +#define GET_C32_ENCRYPT(strength) \ + ((ctr_f) (cryptonite_aes_branch_table[ENCRYPT_C32_128 + strength])) #define GET_XTS_ENCRYPT(strength) \ ((xts_f) (cryptonite_aes_branch_table[ENCRYPT_XTS_128 + strength])) #define GET_XTS_DECRYPT(strength) \ @@ -191,8 +202,10 @@ (((block_f) (cryptonite_aes_branch_table[ENCRYPT_BLOCK_128 + k->strength]))(o,k,i)) #define cryptonite_aes_decrypt_block(o,k,i) \ (((block_f) (cryptonite_aes_branch_table[DECRYPT_BLOCK_128 + k->strength]))(o,k,i)) -#define cryptonite_gf_mul(a,b) \ - (((gf_mul_f) (cryptonite_aes_branch_table[GHASH_GF_MUL]))(a,b)) +#define cryptonite_hinit(t,h) \ + (((hinit_f) (cryptonite_aes_branch_table[GHASH_HINIT]))(t,h)) +#define cryptonite_gf_mul(a,t) \ + (((gf_mul_f) (cryptonite_aes_branch_table[GHASH_GF_MUL]))(a,t)) #else #define GET_INIT(strenght) cryptonite_aes_generic_init #define GET_ECB_ENCRYPT(strength) cryptonite_aes_generic_encrypt_ecb @@ -200,6 +213,7 @@ #define GET_CBC_ENCRYPT(strength) cryptonite_aes_generic_encrypt_cbc #define GET_CBC_DECRYPT(strength) cryptonite_aes_generic_decrypt_cbc #define GET_CTR_ENCRYPT(strength) cryptonite_aes_generic_encrypt_ctr +#define GET_C32_ENCRYPT(strength) cryptonite_aes_generic_encrypt_c32 #define GET_XTS_ENCRYPT(strength) cryptonite_aes_generic_encrypt_xts #define GET_XTS_DECRYPT(strength) cryptonite_aes_generic_decrypt_xts #define GET_GCM_ENCRYPT(strength) cryptonite_aes_generic_gcm_encrypt @@ -210,14 +224,23 @@ #define GET_CCM_DECRYPT(strength) cryptonite_aes_generic_ccm_decrypt #define cryptonite_aes_encrypt_block(o,k,i) cryptonite_aes_generic_encrypt_block(o,k,i) #define cryptonite_aes_decrypt_block(o,k,i) cryptonite_aes_generic_decrypt_block(o,k,i) -#define cryptonite_gf_mul(a,b) cryptonite_aes_generic_gf_mul(a,b) +#define cryptonite_hinit(t,h) cryptonite_aes_generic_hinit(t,h) +#define cryptonite_gf_mul(a,t) cryptonite_aes_generic_gf_mul(a,t) #endif +#define CPU_AESNI 0 +#define CPU_PCLMUL 1 +#define CPU_OPTION_COUNT 2 + +static uint8_t cryptonite_aes_cpu_options[CPU_OPTION_COUNT] = {}; + #if defined(ARCH_X86) && defined(WITH_AESNI) static void initialize_table_ni(int aesni, int pclmul) { if (!aesni) return; + cryptonite_aes_cpu_options[CPU_AESNI] = 1; + cryptonite_aes_branch_table[INIT_128] = cryptonite_aesni_init; cryptonite_aes_branch_table[INIT_256] = cryptonite_aesni_init; @@ -238,6 +261,9 @@ /* CTR */ cryptonite_aes_branch_table[ENCRYPT_CTR_128] = cryptonite_aesni_encrypt_ctr128; cryptonite_aes_branch_table[ENCRYPT_CTR_256] = cryptonite_aesni_encrypt_ctr256; + /* CTR with 32-bit wrapping */ + cryptonite_aes_branch_table[ENCRYPT_C32_128] = cryptonite_aesni_encrypt_c32_128; + cryptonite_aes_branch_table[ENCRYPT_C32_256] = cryptonite_aesni_encrypt_c32_256; /* XTS */ cryptonite_aes_branch_table[ENCRYPT_XTS_128] = cryptonite_aesni_encrypt_xts128; cryptonite_aes_branch_table[ENCRYPT_XTS_256] = cryptonite_aesni_encrypt_xts256; @@ -252,13 +278,24 @@ #ifdef WITH_PCLMUL if (!pclmul) return; + cryptonite_aes_cpu_options[CPU_PCLMUL] = 1; + /* GHASH */ - cryptonite_aes_branch_table[GHASH_GF_MUL] = cryptonite_aesni_gf_mul; + cryptonite_aes_branch_table[GHASH_HINIT] = cryptonite_aesni_hinit_pclmul, + cryptonite_aes_branch_table[GHASH_GF_MUL] = cryptonite_aesni_gf_mul_pclmul, cryptonite_aesni_init_pclmul(); #endif } #endif +uint8_t *cryptonite_aes_cpu_init(void) +{ +#if defined(ARCH_X86) && defined(WITH_AESNI) + cryptonite_aesni_initialize_hw(initialize_table_ni); +#endif + return cryptonite_aes_cpu_options; +} + void cryptonite_aes_initkey(aes_key *key, uint8_t *origkey, uint8_t size) { switch (size) { @@ -266,9 +303,7 @@ case 24: key->nbr = 12; key->strength = 1; break; case 32: key->nbr = 14; key->strength = 2; break; } -#if defined(ARCH_X86) && defined(WITH_AESNI) - cryptonite_aesni_initialize_hw(initialize_table_ni); -#endif + cryptonite_aes_cpu_init(); init_f _init = GET_INIT(key->strength); _init(key, origkey, size); } @@ -330,6 +365,12 @@ e(output, key, iv, input, len); } +void cryptonite_aes_encrypt_c32(uint8_t *output, aes_key *key, aes_block *iv, uint8_t *input, uint32_t len) +{ + ctr_f e = GET_C32_ENCRYPT(key->strength); + e(output, key, iv, input, len); +} + void cryptonite_aes_encrypt_xts(aes_block *output, aes_key *k1, aes_key *k2, aes_block *dataunit, uint32_t spoint, aes_block *input, uint32_t nb_blocks) { @@ -382,20 +423,22 @@ static void gcm_ghash_add(aes_gcm *gcm, block128 *b) { block128_xor(&gcm->tag, b); - cryptonite_gf_mul(&gcm->tag, &gcm->h); + cryptonite_gf_mul(&gcm->tag, gcm->htable); } void cryptonite_aes_gcm_init(aes_gcm *gcm, aes_key *key, uint8_t *iv, uint32_t len) { + block128 h; gcm->length_aad = 0; gcm->length_input = 0; - block128_zero(&gcm->h); + block128_zero(&h); block128_zero(&gcm->tag); block128_zero(&gcm->iv); /* prepare H : encrypt_K(0^128) */ - cryptonite_aes_encrypt_block(&gcm->h, key, &gcm->h); + cryptonite_aes_encrypt_block(&h, key, &h); + cryptonite_hinit(gcm->htable, &h); if (len == 12) { block128_copy_bytes(&gcm->iv, iv, 12); @@ -405,15 +448,15 @@ int i; for (; len >= 16; len -= 16, iv += 16) { block128_xor(&gcm->iv, (block128 *) iv); - cryptonite_gf_mul(&gcm->iv, &gcm->h); + cryptonite_gf_mul(&gcm->iv, gcm->htable); } if (len > 0) { block128_xor_bytes(&gcm->iv, iv, len); - cryptonite_gf_mul(&gcm->iv, &gcm->h); + cryptonite_gf_mul(&gcm->iv, gcm->htable); } for (i = 15; origlen; --i, origlen >>= 8) gcm->iv.b[i] ^= (uint8_t) origlen; - cryptonite_gf_mul(&gcm->iv, &gcm->h); + cryptonite_gf_mul(&gcm->iv, gcm->htable); } block128_copy_aligned(&gcm->civ, &gcm->iv); @@ -507,7 +550,7 @@ static void ccm_cbcmac_add(aes_ccm* ccm, aes_key* key, block128* bi) { block128_xor_aligned(&ccm->xi, bi); - cryptonite_aes_generic_encrypt_block(&ccm->xi, key, &ccm->xi); + cryptonite_aes_encrypt_block(&ccm->xi, key, &ccm->xi); } /* even though it is possible to support message size as large as 2^64, we support up to 2^32 only */ @@ -765,6 +808,30 @@ } } +void cryptonite_aes_generic_encrypt_c32(uint8_t *output, aes_key *key, aes_block *iv, uint8_t *input, uint32_t len) +{ + aes_block block, o; + uint32_t nb_blocks = len / 16; + int i; + + /* preload IV in block */ + block128_copy(&block, iv); + + for ( ; nb_blocks-- > 0; block128_inc32_le(&block), output += 16, input += 16) { + cryptonite_aes_encrypt_block(&o, key, &block); + block128_vxor((block128 *) output, &o, (block128 *) input); + } + + if ((len % 16) != 0) { + cryptonite_aes_encrypt_block(&o, key, &block); + for (i = 0; i < (len % 16); i++) { + *output = ((uint8_t *) &o)[i] ^ *input; + output++; + input++; + } + } +} + void cryptonite_aes_generic_encrypt_xts(aes_block *output, aes_key *k1, aes_key *k2, aes_block *dataunit, uint32_t spoint, aes_block *input, uint32_t nb_blocks) { @@ -811,7 +878,7 @@ gcm->length_input += length; for (; length >= 16; input += 16, output += 16, length -= 16) { - block128_inc_be(&gcm->civ); + block128_inc32_be(&gcm->civ); cryptonite_aes_encrypt_block(&out, key, &gcm->civ); block128_xor(&out, (block128 *) input); @@ -822,7 +889,7 @@ aes_block tmp; int i; - block128_inc_be(&gcm->civ); + block128_inc32_be(&gcm->civ); /* create e(civ) in out */ cryptonite_aes_encrypt_block(&out, key, &gcm->civ); /* initialize a tmp as input and xor it to e(civ) */ @@ -844,7 +911,7 @@ gcm->length_input += length; for (; length >= 16; input += 16, output += 16, length -= 16) { - block128_inc_be(&gcm->civ); + block128_inc32_be(&gcm->civ); cryptonite_aes_encrypt_block(&out, key, &gcm->civ); gcm_ghash_add(gcm, (block128 *) input); @@ -855,7 +922,7 @@ aes_block tmp; int i; - block128_inc_be(&gcm->civ); + block128_inc32_be(&gcm->civ); block128_zero(&tmp); block128_copy_bytes(&tmp, input, length); @@ -987,4 +1054,56 @@ void cryptonite_aes_generic_ocb_decrypt(uint8_t *output, aes_ocb *ocb, aes_key *key, uint8_t *input, uint32_t length) { ocb_generic_crypt(output, ocb, key, input, length, 0); +} + +static inline void gf_mulx_rev(block128 *a, const block128 *h) +{ + uint64_t v1 = cpu_to_le64(h->q[0]); + uint64_t v0 = cpu_to_le64(h->q[1]); + a->q[1] = cpu_to_be64(v1 >> 1 | v0 << 63); + a->q[0] = cpu_to_be64(v0 >> 1 ^ ((0-(v1 & 1)) & 0xe100000000000000ULL)); +} + +void cryptonite_aes_polyval_init(aes_polyval *ctx, const aes_block *h) +{ + aes_block r; + + /* ByteReverse(S_0) = 0 */ + block128_zero(&ctx->s); + + /* ByteReverse(H) * x */ + gf_mulx_rev(&r, h); + cryptonite_hinit(ctx->htable, &r); +} + +void cryptonite_aes_polyval_update(aes_polyval *ctx, const uint8_t *input, uint32_t length) +{ + aes_block r; + const uint8_t *p; + uint32_t sz; + + /* This automatically pads with zeros if input is not a multiple of the + block size. */ + for (p = input; length > 0; p += 16, length -= sz) + { + sz = length < 16 ? length : 16; + + /* ByteReverse(X_j) */ + block128_zero(&r); + memcpy(&r, p, sz); + block128_byte_reverse(&r); + + /* ByteReverse(S_{j-1}) + ByteReverse(X_j) */ + block128_xor_aligned(&ctx->s, &r); + + /* ByteReverse(S_j) */ + cryptonite_gf_mul(&ctx->s, ctx->htable); + } +} + +void cryptonite_aes_polyval_finalize(aes_polyval *ctx, aes_block *dst) +{ + /* S_s */ + block128_copy_aligned(dst, &ctx->s); + block128_byte_reverse(dst); } diff --git a/cbits/cryptonite_aes.h b/cbits/cryptonite_aes.h --- a/cbits/cryptonite_aes.h +++ b/cbits/cryptonite_aes.h @@ -45,10 +45,10 @@ uint8_t data[16*14*2]; } aes_key; -/* size = 4*16+2*8= 80 */ +/* size = 19*16+2*8= 320 */ typedef struct { aes_block tag; - aes_block h; + aes_block htable[16]; aes_block iv; aes_block civ; uint64_t length_aad; @@ -77,6 +77,12 @@ block128 li[4]; } aes_ocb; +/* size = 17*16= 272 */ +typedef struct { + aes_block htable[16]; + aes_block s; +} aes_polyval; + /* in bytes: either 16,24,32 */ void cryptonite_aes_initkey(aes_key *ctx, uint8_t *key, uint8_t size); @@ -114,5 +120,11 @@ void cryptonite_aes_ccm_encrypt(uint8_t *output, aes_ccm *ccm, aes_key *key, uint8_t *input, uint32_t length); void cryptonite_aes_ccm_decrypt(uint8_t *output, aes_ccm *ccm, aes_key *key, uint8_t *input, uint32_t length); void cryptonite_aes_ccm_finish(uint8_t *tag, aes_ccm *ccm, aes_key *key); + +uint8_t *cryptonite_aes_cpu_init(void); + +void cryptonite_aes_polyval_init(aes_polyval *ctx, const aes_block *h); +void cryptonite_aes_polyval_update(aes_polyval *ctx, const uint8_t *input, uint32_t length); +void cryptonite_aes_polyval_finalize(aes_polyval *ctx, aes_block *dst); #endif diff --git a/cbits/cryptonite_chacha.c b/cbits/cryptonite_chacha.c --- a/cbits/cryptonite_chacha.c +++ b/cbits/cryptonite_chacha.c @@ -98,7 +98,6 @@ uint32_t ivlen, const uint8_t *iv) { const uint8_t *constants = (keylen == 32) ? sigma : tau; - int i; ASSERT_ALIGNMENT(constants, 4); diff --git a/cbits/cryptonite_rdrand.c b/cbits/cryptonite_rdrand.c --- a/cbits/cryptonite_rdrand.c +++ b/cbits/cryptonite_rdrand.c @@ -91,7 +91,7 @@ } #endif -/* Returns the number of bytes succesfully generated */ +/* Returns the number of bytes successfully generated */ int cryptonite_get_rand_bytes(uint8_t *buffer, size_t len) { RDRAND_T tmp; diff --git a/cbits/cryptonite_salsa.c b/cbits/cryptonite_salsa.c --- a/cbits/cryptonite_salsa.c +++ b/cbits/cryptonite_salsa.c @@ -120,7 +120,6 @@ uint32_t ivlen, const uint8_t *iv) { const uint8_t *constants = (keylen == 32) ? sigma : tau; - int i; st->d[0] = load_le32_aligned(constants + 0); st->d[5] = load_le32_aligned(constants + 4); diff --git a/cbits/cryptonite_skein256.c b/cbits/cryptonite_skein256.c --- a/cbits/cryptonite_skein256.c +++ b/cbits/cryptonite_skein256.c @@ -167,7 +167,6 @@ void cryptonite_skein256_finalize(struct skein256_ctx *ctx, uint32_t hashlen, uint8_t *out) { uint32_t outsize; - uint64_t *p = (uint64_t *) out; uint64_t x[4]; int i, j, n; diff --git a/cbits/cryptonite_skein512.c b/cbits/cryptonite_skein512.c --- a/cbits/cryptonite_skein512.c +++ b/cbits/cryptonite_skein512.c @@ -185,7 +185,6 @@ void cryptonite_skein512_finalize(struct skein512_ctx *ctx, uint32_t hashlen, uint8_t *out) { uint32_t outsize; - uint64_t *p = (uint64_t *) out; uint64_t x[8]; int i, j, n; diff --git a/cbits/cryptonite_whirlpool.c b/cbits/cryptonite_whirlpool.c --- a/cbits/cryptonite_whirlpool.c +++ b/cbits/cryptonite_whirlpool.c @@ -777,7 +777,6 @@ uint64_t K[8]; /* the round key */ uint64_t block[8]; /* mu(buffer) */ uint64_t state[8]; /* the cipher state */ - uint64_t L[8]; uint8_t *buffer = ctx->buffer; /* diff --git a/cbits/cryptonite_xsalsa.c b/cbits/cryptonite_xsalsa.c --- a/cbits/cryptonite_xsalsa.c +++ b/cbits/cryptonite_xsalsa.c @@ -47,13 +47,27 @@ (x6, x7, x8, x9) is the first 128 bits of a 192-bit nonce */ cryptonite_salsa_init_core(&ctx->st, keylen, key, 8, iv); - ctx->st.d[ 8] = load_le32(iv + 8); - ctx->st.d[ 9] = load_le32(iv + 12); + /* Continue initialization in a separate function that may also + be called independently */ + cryptonite_xsalsa_derive(ctx, ivlen - 8, iv + 8); +} + +void cryptonite_xsalsa_derive(cryptonite_salsa_context *ctx, + uint32_t ivlen, const uint8_t *iv) +{ + /* Finish creating initial 512-bit input block: + (x6, x7, x8, x9) is the first 128 bits of a 192-bit nonce + + Except iv has been shifted by 64 bits so there are now only 128 bits ahead. + */ + ctx->st.d[ 8] += load_le32(iv + 0); + ctx->st.d[ 9] += load_le32(iv + 4); + /* Compute (z0, z1, . . . , z15) = doubleround ^(r/2) (x0, x1, . . . , x15) */ block hSalsa; memset(&hSalsa, 0, sizeof(block)); - cryptonite_salsa_core_xor(nb_rounds, &hSalsa, &ctx->st); + cryptonite_salsa_core_xor(ctx->nb_rounds, &hSalsa, &ctx->st); /* Build a new 512-bit input block (x′0, x′1, . . . , x′15): (x′0, x′5, x′10, x′15) is the Salsa20 constant @@ -69,8 +83,8 @@ ctx->st.d[12] = hSalsa.d[ 7] - ctx->st.d[ 7]; ctx->st.d[13] = hSalsa.d[ 8] - ctx->st.d[ 8]; ctx->st.d[14] = hSalsa.d[ 9] - ctx->st.d[ 9]; - ctx->st.d[ 6] = load_le32(iv + 16); - ctx->st.d[ 7] = load_le32(iv + 20); + ctx->st.d[ 6] = load_le32(iv + 8); + ctx->st.d[ 7] = load_le32(iv + 12); ctx->st.d[ 8] = 0; ctx->st.d[ 9] = 0; } diff --git a/cbits/cryptonite_xsalsa.h b/cbits/cryptonite_xsalsa.h --- a/cbits/cryptonite_xsalsa.h +++ b/cbits/cryptonite_xsalsa.h @@ -33,5 +33,6 @@ #include "cryptonite_salsa.h" void cryptonite_xsalsa_init(cryptonite_salsa_context *ctx, uint8_t nb_rounds, uint32_t keylen, const uint8_t *key, uint32_t ivlen, const uint8_t *iv); +void cryptonite_xsalsa_derive(cryptonite_salsa_context *ctx, uint32_t ivlen, const uint8_t *iv); #endif diff --git a/cbits/include32/p256/p256.h b/cbits/include32/p256/p256.h new file mode 100644 --- /dev/null +++ b/cbits/include32/p256/p256.h @@ -0,0 +1,167 @@ +/* + * Copyright 2013 The Android Open Source Project + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Google Inc. nor the names of its contributors may + * be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY Google Inc. ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO + * EVENT SHALL Google Inc. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; + * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, + * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR + * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF + * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#ifndef SYSTEM_CORE_INCLUDE_MINCRYPT_LITE_P256_H_ +#define SYSTEM_CORE_INCLUDE_MINCRYPT_LITE_P256_H_ + +// Collection of routines manipulating 256 bit unsigned integers. +// Just enough to implement ecdsa-p256 and related algorithms. + +#include + +#ifdef __cplusplus +extern "C" { +#endif + +#define P256_BITSPERDIGIT 32 +#define P256_NDIGITS 8 +#define P256_NBYTES 32 + +// n' such as n * n' = -1 mod (2^32) +#define P256_MONTGOMERY_FACTOR 0xEE00BC4F + +#define P256_LITERAL(lo,hi) (lo), (hi) + +typedef int cryptonite_p256_err; +typedef uint32_t cryptonite_p256_digit; +typedef int32_t cryptonite_p256_sdigit; +typedef uint64_t cryptonite_p256_ddigit; +typedef int64_t cryptonite_p256_sddigit; + +// Defining cryptonite_p256_int as struct to leverage struct assigment. +typedef struct { + cryptonite_p256_digit a[P256_NDIGITS]; +} cryptonite_p256_int; + +extern const cryptonite_p256_int cryptonite_SECP256r1_n; // Curve order +extern const cryptonite_p256_int cryptonite_SECP256r1_p; // Curve prime +extern const cryptonite_p256_int cryptonite_SECP256r1_b; // Curve param + +// Initialize a cryptonite_p256_int to zero. +void cryptonite_p256_init(cryptonite_p256_int* a); + +// Clear a cryptonite_p256_int to zero. +void cryptonite_p256_clear(cryptonite_p256_int* a); + +// Return bit. Index 0 is least significant. +int cryptonite_p256_get_bit(const cryptonite_p256_int* a, int index); + +// b := a % MOD +void cryptonite_p256_mod( + const cryptonite_p256_int* MOD, + const cryptonite_p256_int* a, + cryptonite_p256_int* b); + +// c := a * (top_b | b) % MOD +void cryptonite_p256_modmul( + const cryptonite_p256_int* MOD, + const cryptonite_p256_int* a, + const cryptonite_p256_digit top_b, + const cryptonite_p256_int* b, + cryptonite_p256_int* c); + +// b := 1 / a % MOD +// MOD best be SECP256r1_n +void cryptonite_p256_modinv( + const cryptonite_p256_int* MOD, + const cryptonite_p256_int* a, + cryptonite_p256_int* b); + +// b := 1 / a % MOD +// MOD best be SECP256r1_n +// Faster than cryptonite_p256_modinv() +void cryptonite_p256_modinv_vartime( + const cryptonite_p256_int* MOD, + const cryptonite_p256_int* a, + cryptonite_p256_int* b); + +// b := a << (n % P256_BITSPERDIGIT) +// Returns the bits shifted out of most significant digit. +cryptonite_p256_digit cryptonite_p256_shl(const cryptonite_p256_int* a, int n, cryptonite_p256_int* b); + +// b := a >> (n % P256_BITSPERDIGIT) +void cryptonite_p256_shr(const cryptonite_p256_int* a, int n, cryptonite_p256_int* b); + +int cryptonite_p256_is_zero(const cryptonite_p256_int* a); +int cryptonite_p256_is_odd(const cryptonite_p256_int* a); +int cryptonite_p256_is_even(const cryptonite_p256_int* a); + +// Returns -1, 0 or 1. +int cryptonite_p256_cmp(const cryptonite_p256_int* a, const cryptonite_p256_int *b); + +// c: = a - b +// Returns -1 on borrow. +int cryptonite_p256_sub(const cryptonite_p256_int* a, const cryptonite_p256_int* b, cryptonite_p256_int* c); + +// c := a + b +// Returns 1 on carry. +int cryptonite_p256_add(const cryptonite_p256_int* a, const cryptonite_p256_int* b, cryptonite_p256_int* c); + +// c := a + (single digit)b +// Returns carry 1 on carry. +int cryptonite_p256_add_d(const cryptonite_p256_int* a, cryptonite_p256_digit b, cryptonite_p256_int* c); + +// ec routines. + +// {out_x,out_y} := nG +void cryptonite_p256_base_point_mul(const cryptonite_p256_int *n, + cryptonite_p256_int *out_x, + cryptonite_p256_int *out_y); + +// {out_x,out_y} := n{in_x,in_y} +void cryptonite_p256_point_mul(const cryptonite_p256_int *n, + const cryptonite_p256_int *in_x, + const cryptonite_p256_int *in_y, + cryptonite_p256_int *out_x, + cryptonite_p256_int *out_y); + +// {out_x,out_y} := n1G + n2{in_x,in_y} +void cryptonite_p256_points_mul_vartime( + const cryptonite_p256_int *n1, const cryptonite_p256_int *n2, + const cryptonite_p256_int *in_x, const cryptonite_p256_int *in_y, + cryptonite_p256_int *out_x, cryptonite_p256_int *out_y); + +// Return whether point {x,y} is on curve. +int cryptonite_p256_is_valid_point(const cryptonite_p256_int* x, const cryptonite_p256_int* y); + +// Outputs big-endian binary form. No leading zero skips. +void cryptonite_p256_to_bin(const cryptonite_p256_int* src, uint8_t dst[P256_NBYTES]); + +// Reads from big-endian binary form, +// thus pre-pad with leading zeros if short. +void cryptonite_p256_from_bin(const uint8_t src[P256_NBYTES], cryptonite_p256_int* dst); + +#define P256_DIGITS(x) ((x)->a) +#define P256_DIGIT(x,y) ((x)->a[y]) + +#define P256_ZERO {{0}} +#define P256_ONE {{1}} + +#ifdef __cplusplus +} +#endif + +#endif // SYSTEM_CORE_INCLUDE_MINCRYPT_LITE_P256_H_ diff --git a/cbits/include32/p256/p256_gf.h b/cbits/include32/p256/p256_gf.h new file mode 100644 --- /dev/null +++ b/cbits/include32/p256/p256_gf.h @@ -0,0 +1,779 @@ +/* + * Copyright 2013 The Android Open Source Project + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Google Inc. nor the names of its contributors may + * be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY Google Inc. ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO + * EVENT SHALL Google Inc. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; + * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, + * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR + * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF + * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +// This is an implementation of the P256 finite field. It's written to be +// portable and still constant-time. +// +// WARNING: Implementing these functions in a constant-time manner is far from +// obvious. Be careful when touching this code. +// +// See http://www.imperialviolet.org/2010/12/04/ecc.html ([1]) for background. + +#include +#include + +#include +#include + +#include "p256/p256.h" + +typedef uint8_t u8; +typedef uint32_t u32; +typedef int32_t s32; +typedef uint64_t u64; + +/* Our field elements are represented as nine 32-bit limbs. + * + * The value of an felem (field element) is: + * x[0] + (x[1] * 2**29) + (x[2] * 2**57) + ... + (x[8] * 2**228) + * + * That is, each limb is alternately 29 or 28-bits wide in little-endian + * order. + * + * This means that an felem hits 2**257, rather than 2**256 as we would like. A + * 28, 29, ... pattern would cause us to hit 2**256, but that causes problems + * when multiplying as terms end up one bit short of a limb which would require + * much bit-shifting to correct. + * + * Finally, the values stored in an felem are in Montgomery form. So the value + * |y| is stored as (y*R) mod p, where p is the P-256 prime and R is 2**257. + */ +typedef u32 limb; +#define NLIMBS 9 +typedef limb felem[NLIMBS]; + +static const limb kBottom28Bits = 0xfffffff; +static const limb kBottom29Bits = 0x1fffffff; + +/* kOne is the number 1 as an felem. It's 2**257 mod p split up into 29 and + * 28-bit words. */ +static const felem kOne = { + 2, 0, 0, 0xffff800, + 0x1fffffff, 0xfffffff, 0x1fbfffff, 0x1ffffff, + 0 +}; +static const felem kZero = {0}; +static const felem kP = { + 0x1fffffff, 0xfffffff, 0x1fffffff, 0x3ff, + 0, 0, 0x200000, 0xf000000, + 0xfffffff +}; +static const felem k2P = { + 0x1ffffffe, 0xfffffff, 0x1fffffff, 0x7ff, + 0, 0, 0x400000, 0xe000000, + 0x1fffffff +}; +/* kPrecomputed contains precomputed values to aid the calculation of scalar + * multiples of the base point, G. It's actually two, equal length, tables + * concatenated. + * + * The first table contains (x,y) felem pairs for 16 multiples of the base + * point, G. + * + * Index | Index (binary) | Value + * 0 | 0000 | 0G (all zeros, omitted) + * 1 | 0001 | G + * 2 | 0010 | 2**64G + * 3 | 0011 | 2**64G + G + * 4 | 0100 | 2**128G + * 5 | 0101 | 2**128G + G + * 6 | 0110 | 2**128G + 2**64G + * 7 | 0111 | 2**128G + 2**64G + G + * 8 | 1000 | 2**192G + * 9 | 1001 | 2**192G + G + * 10 | 1010 | 2**192G + 2**64G + * 11 | 1011 | 2**192G + 2**64G + G + * 12 | 1100 | 2**192G + 2**128G + * 13 | 1101 | 2**192G + 2**128G + G + * 14 | 1110 | 2**192G + 2**128G + 2**64G + * 15 | 1111 | 2**192G + 2**128G + 2**64G + G + * + * The second table follows the same style, but the terms are 2**32G, + * 2**96G, 2**160G, 2**224G. + * + * This is ~2KB of data. */ +static const limb kPrecomputed[NLIMBS * 2 * 15 * 2] = { + 0x11522878, 0xe730d41, 0xdb60179, 0x4afe2ff, 0x12883add, 0xcaddd88, 0x119e7edc, 0xd4a6eab, 0x3120bee, + 0x1d2aac15, 0xf25357c, 0x19e45cdd, 0x5c721d0, 0x1992c5a5, 0xa237487, 0x154ba21, 0x14b10bb, 0xae3fe3, + 0xd41a576, 0x922fc51, 0x234994f, 0x60b60d3, 0x164586ae, 0xce95f18, 0x1fe49073, 0x3fa36cc, 0x5ebcd2c, + 0xb402f2f, 0x15c70bf, 0x1561925c, 0x5a26704, 0xda91e90, 0xcdc1c7f, 0x1ea12446, 0xe1ade1e, 0xec91f22, + 0x26f7778, 0x566847e, 0xa0bec9e, 0x234f453, 0x1a31f21a, 0xd85e75c, 0x56c7109, 0xa267a00, 0xb57c050, + 0x98fb57, 0xaa837cc, 0x60c0792, 0xcfa5e19, 0x61bab9e, 0x589e39b, 0xa324c5, 0x7d6dee7, 0x2976e4b, + 0x1fc4124a, 0xa8c244b, 0x1ce86762, 0xcd61c7e, 0x1831c8e0, 0x75774e1, 0x1d96a5a9, 0x843a649, 0xc3ab0fa, + 0x6e2e7d5, 0x7673a2a, 0x178b65e8, 0x4003e9b, 0x1a1f11c2, 0x7816ea, 0xf643e11, 0x58c43df, 0xf423fc2, + 0x19633ffa, 0x891f2b2, 0x123c231c, 0x46add8c, 0x54700dd, 0x59e2b17, 0x172db40f, 0x83e277d, 0xb0dd609, + 0xfd1da12, 0x35c6e52, 0x19ede20c, 0xd19e0c0, 0x97d0f40, 0xb015b19, 0x449e3f5, 0xe10c9e, 0x33ab581, + 0x56a67ab, 0x577734d, 0x1dddc062, 0xc57b10d, 0x149b39d, 0x26a9e7b, 0xc35df9f, 0x48764cd, 0x76dbcca, + 0xca4b366, 0xe9303ab, 0x1a7480e7, 0x57e9e81, 0x1e13eb50, 0xf466cf3, 0x6f16b20, 0x4ba3173, 0xc168c33, + 0x15cb5439, 0x6a38e11, 0x73658bd, 0xb29564f, 0x3f6dc5b, 0x53b97e, 0x1322c4c0, 0x65dd7ff, 0x3a1e4f6, + 0x14e614aa, 0x9246317, 0x1bc83aca, 0xad97eed, 0xd38ce4a, 0xf82b006, 0x341f077, 0xa6add89, 0x4894acd, + 0x9f162d5, 0xf8410ef, 0x1b266a56, 0xd7f223, 0x3e0cb92, 0xe39b672, 0x6a2901a, 0x69a8556, 0x7e7c0, + 0x9b7d8d3, 0x309a80, 0x1ad05f7f, 0xc2fb5dd, 0xcbfd41d, 0x9ceb638, 0x1051825c, 0xda0cf5b, 0x812e881, + 0x6f35669, 0x6a56f2c, 0x1df8d184, 0x345820, 0x1477d477, 0x1645db1, 0xbe80c51, 0xc22be3e, 0xe35e65a, + 0x1aeb7aa0, 0xc375315, 0xf67bc99, 0x7fdd7b9, 0x191fc1be, 0x61235d, 0x2c184e9, 0x1c5a839, 0x47a1e26, + 0xb7cb456, 0x93e225d, 0x14f3c6ed, 0xccc1ac9, 0x17fe37f3, 0x4988989, 0x1a90c502, 0x2f32042, 0xa17769b, + 0xafd8c7c, 0x8191c6e, 0x1dcdb237, 0x16200c0, 0x107b32a1, 0x66c08db, 0x10d06a02, 0x3fc93, 0x5620023, + 0x16722b27, 0x68b5c59, 0x270fcfc, 0xfad0ecc, 0xe5de1c2, 0xeab466b, 0x2fc513c, 0x407f75c, 0xbaab133, + 0x9705fe9, 0xb88b8e7, 0x734c993, 0x1e1ff8f, 0x19156970, 0xabd0f00, 0x10469ea7, 0x3293ac0, 0xcdc98aa, + 0x1d843fd, 0xe14bfe8, 0x15be825f, 0x8b5212, 0xeb3fb67, 0x81cbd29, 0xbc62f16, 0x2b6fcc7, 0xf5a4e29, + 0x13560b66, 0xc0b6ac2, 0x51ae690, 0xd41e271, 0xf3e9bd4, 0x1d70aab, 0x1029f72, 0x73e1c35, 0xee70fbc, + 0xad81baf, 0x9ecc49a, 0x86c741e, 0xfe6be30, 0x176752e7, 0x23d416, 0x1f83de85, 0x27de188, 0x66f70b8, + 0x181cd51f, 0x96b6e4c, 0x188f2335, 0xa5df759, 0x17a77eb6, 0xfeb0e73, 0x154ae914, 0x2f3ec51, 0x3826b59, + 0xb91f17d, 0x1c72949, 0x1362bf0a, 0xe23fddf, 0xa5614b0, 0xf7d8f, 0x79061, 0x823d9d2, 0x8213f39, + 0x1128ae0b, 0xd095d05, 0xb85c0c2, 0x1ecb2ef, 0x24ddc84, 0xe35e901, 0x18411a4a, 0xf5ddc3d, 0x3786689, + 0x52260e8, 0x5ae3564, 0x542b10d, 0x8d93a45, 0x19952aa4, 0x996cc41, 0x1051a729, 0x4be3499, 0x52b23aa, + 0x109f307e, 0x6f5b6bb, 0x1f84e1e7, 0x77a0cfa, 0x10c4df3f, 0x25a02ea, 0xb048035, 0xe31de66, 0xc6ecaa3, + 0x28ea335, 0x2886024, 0x1372f020, 0xf55d35, 0x15e4684c, 0xf2a9e17, 0x1a4a7529, 0xcb7beb1, 0xb2a78a1, + 0x1ab21f1f, 0x6361ccf, 0x6c9179d, 0xb135627, 0x1267b974, 0x4408bad, 0x1cbff658, 0xe3d6511, 0xc7d76f, + 0x1cc7a69, 0xe7ee31b, 0x54fab4f, 0x2b914f, 0x1ad27a30, 0xcd3579e, 0xc50124c, 0x50daa90, 0xb13f72, + 0xb06aa75, 0x70f5cc6, 0x1649e5aa, 0x84a5312, 0x329043c, 0x41c4011, 0x13d32411, 0xb04a838, 0xd760d2d, + 0x1713b532, 0xbaa0c03, 0x84022ab, 0x6bcf5c1, 0x2f45379, 0x18ae070, 0x18c9e11e, 0x20bca9a, 0x66f496b, + 0x3eef294, 0x67500d2, 0xd7f613c, 0x2dbbeb, 0xb741038, 0xe04133f, 0x1582968d, 0xbe985f7, 0x1acbc1a, + 0x1a6a939f, 0x33e50f6, 0xd665ed4, 0xb4b7bd6, 0x1e5a3799, 0x6b33847, 0x17fa56ff, 0x65ef930, 0x21dc4a, + 0x2b37659, 0x450fe17, 0xb357b65, 0xdf5efac, 0x15397bef, 0x9d35a7f, 0x112ac15f, 0x624e62e, 0xa90ae2f, + 0x107eecd2, 0x1f69bbe, 0x77d6bce, 0x5741394, 0x13c684fc, 0x950c910, 0x725522b, 0xdc78583, 0x40eeabb, + 0x1fde328a, 0xbd61d96, 0xd28c387, 0x9e77d89, 0x12550c40, 0x759cb7d, 0x367ef34, 0xae2a960, 0x91b8bdc, + 0x93462a9, 0xf469ef, 0xb2e9aef, 0xd2ca771, 0x54e1f42, 0x7aaa49, 0x6316abb, 0x2413c8e, 0x5425bf9, + 0x1bed3e3a, 0xf272274, 0x1f5e7326, 0x6416517, 0xea27072, 0x9cedea7, 0x6e7633, 0x7c91952, 0xd806dce, + 0x8e2a7e1, 0xe421e1a, 0x418c9e1, 0x1dbc890, 0x1b395c36, 0xa1dc175, 0x1dc4ef73, 0x8956f34, 0xe4b5cf2, + 0x1b0d3a18, 0x3194a36, 0x6c2641f, 0xe44124c, 0xa2f4eaa, 0xa8c25ba, 0xf927ed7, 0x627b614, 0x7371cca, + 0xba16694, 0x417bc03, 0x7c0a7e3, 0x9c35c19, 0x1168a205, 0x8b6b00d, 0x10e3edc9, 0x9c19bf2, 0x5882229, + 0x1b2b4162, 0xa5cef1a, 0x1543622b, 0x9bd433e, 0x364e04d, 0x7480792, 0x5c9b5b3, 0xe85ff25, 0x408ef57, + 0x1814cfa4, 0x121b41b, 0xd248a0f, 0x3b05222, 0x39bb16a, 0xc75966d, 0xa038113, 0xa4a1769, 0x11fbc6c, + 0x917e50e, 0xeec3da8, 0x169d6eac, 0x10c1699, 0xa416153, 0xf724912, 0x15cd60b7, 0x4acbad9, 0x5efc5fa, + 0xf150ed7, 0x122b51, 0x1104b40a, 0xcb7f442, 0xfbb28ff, 0x6ac53ca, 0x196142cc, 0x7bf0fa9, 0x957651, + 0x4e0f215, 0xed439f8, 0x3f46bd5, 0x5ace82f, 0x110916b6, 0x6db078, 0xffd7d57, 0xf2ecaac, 0xca86dec, + 0x15d6b2da, 0x965ecc9, 0x1c92b4c2, 0x1f3811, 0x1cb080f5, 0x2d8b804, 0x19d1c12d, 0xf20bd46, 0x1951fa7, + 0xa3656c3, 0x523a425, 0xfcd0692, 0xd44ddc8, 0x131f0f5b, 0xaf80e4a, 0xcd9fc74, 0x99bb618, 0x2db944c, + 0xa673090, 0x1c210e1, 0x178c8d23, 0x1474383, 0x10b8743d, 0x985a55b, 0x2e74779, 0x576138, 0x9587927, + 0x133130fa, 0xbe05516, 0x9f4d619, 0xbb62570, 0x99ec591, 0xd9468fe, 0x1d07782d, 0xfc72e0b, 0x701b298, + 0x1863863b, 0x85954b8, 0x121a0c36, 0x9e7fedf, 0xf64b429, 0x9b9d71e, 0x14e2f5d8, 0xf858d3a, 0x942eea8, + 0xda5b765, 0x6edafff, 0xa9d18cc, 0xc65e4ba, 0x1c747e86, 0xe4ea915, 0x1981d7a1, 0x8395659, 0x52ed4e2, + 0x87d43b7, 0x37ab11b, 0x19d292ce, 0xf8d4692, 0x18c3053f, 0x8863e13, 0x4c146c0, 0x6bdf55a, 0x4e4457d, + 0x16152289, 0xac78ec2, 0x1a59c5a2, 0x2028b97, 0x71c2d01, 0x295851f, 0x404747b, 0x878558d, 0x7d29aa4, + 0x13d8341f, 0x8daefd7, 0x139c972d, 0x6b7ea75, 0xd4a9dde, 0xff163d8, 0x81d55d7, 0xa5bef68, 0xb7b30d8, + 0xbe73d6f, 0xaa88141, 0xd976c81, 0x7e7a9cc, 0x18beb771, 0xd773cbd, 0x13f51951, 0x9d0c177, 0x1c49a78, +}; + + +/* Field element operations: */ + +/* NON_ZERO_TO_ALL_ONES returns: + * 0xffffffff for 0 < x <= 2**31 + * 0 for x == 0 or x > 2**31. + * + * x must be a u32 or an equivalent type such as limb. */ +#define NON_ZERO_TO_ALL_ONES(x) ((((u32)(x) - 1) >> 31) - 1) + +/* felem_reduce_carry adds a multiple of p in order to cancel |carry|, + * which is a term at 2**257. + * + * On entry: carry < 2**3, inout[0,2,...] < 2**29, inout[1,3,...] < 2**28. + * On exit: inout[0,2,..] < 2**30, inout[1,3,...] < 2**29. */ +static void felem_reduce_carry(felem inout, limb carry) { + const u32 carry_mask = NON_ZERO_TO_ALL_ONES(carry); + + inout[0] += carry << 1; + inout[3] += 0x10000000 & carry_mask; + /* carry < 2**3 thus (carry << 11) < 2**14 and we added 2**28 in the + * previous line therefore this doesn't underflow. */ + inout[3] -= carry << 11; + inout[4] += (0x20000000 - 1) & carry_mask; + inout[5] += (0x10000000 - 1) & carry_mask; + inout[6] += (0x20000000 - 1) & carry_mask; + inout[6] -= carry << 22; + /* This may underflow if carry is non-zero but, if so, we'll fix it in the + * next line. */ + inout[7] -= 1 & carry_mask; + inout[7] += carry << 25; +} + +/* felem_sum sets out = in+in2. + * + * On entry, in[i]+in2[i] must not overflow a 32-bit word. + * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29 */ +static void felem_sum(felem out, const felem in, const felem in2) { + limb carry = 0; + unsigned i; + + for (i = 0;; i++) { + out[i] = in[i] + in2[i]; + out[i] += carry; + carry = out[i] >> 29; + out[i] &= kBottom29Bits; + + i++; + if (i == NLIMBS) + break; + + out[i] = in[i] + in2[i]; + out[i] += carry; + carry = out[i] >> 28; + out[i] &= kBottom28Bits; + } + + felem_reduce_carry(out, carry); +} + +#define two31m3 (((limb)1) << 31) - (((limb)1) << 3) +#define two30m2 (((limb)1) << 30) - (((limb)1) << 2) +#define two30p13m2 (((limb)1) << 30) + (((limb)1) << 13) - (((limb)1) << 2) +#define two31m2 (((limb)1) << 31) - (((limb)1) << 2) +#define two31p24m2 (((limb)1) << 31) + (((limb)1) << 24) - (((limb)1) << 2) +#define two30m27m2 (((limb)1) << 30) - (((limb)1) << 27) - (((limb)1) << 2) + +/* zero31 is 0 mod p. */ +static const felem zero31 = { two31m3, two30m2, two31m2, two30p13m2, two31m2, two30m2, two31p24m2, two30m27m2, two31m2 }; + +/* felem_diff sets out = in-in2. + * + * On entry: in[0,2,...] < 2**30, in[1,3,...] < 2**29 and + * in2[0,2,...] < 2**30, in2[1,3,...] < 2**29. + * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29. */ +static void felem_diff(felem out, const felem in, const felem in2) { + limb carry = 0; + unsigned i; + + for (i = 0;; i++) { + out[i] = in[i] - in2[i]; + out[i] += zero31[i]; + out[i] += carry; + carry = out[i] >> 29; + out[i] &= kBottom29Bits; + + i++; + if (i == NLIMBS) + break; + + out[i] = in[i] - in2[i]; + out[i] += zero31[i]; + out[i] += carry; + carry = out[i] >> 28; + out[i] &= kBottom28Bits; + } + + felem_reduce_carry(out, carry); +} + +/* felem_reduce_degree sets out = tmp/R mod p where tmp contains 64-bit words + * with the same 29,28,... bit positions as an felem. + * + * The values in felems are in Montgomery form: x*R mod p where R = 2**257. + * Since we just multiplied two Montgomery values together, the result is + * x*y*R*R mod p. We wish to divide by R in order for the result also to be + * in Montgomery form. + * + * On entry: tmp[i] < 2**64 + * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29 */ +static void felem_reduce_degree(felem out, u64 tmp[17]) { + /* The following table may be helpful when reading this code: + * + * Limb number: 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10... + * Width (bits): 29| 28| 29| 28| 29| 28| 29| 28| 29| 28| 29 + * Start bit: 0 | 29| 57| 86|114|143|171|200|228|257|285 + * (odd phase): 0 | 28| 57| 85|114|142|171|199|228|256|285 */ + limb tmp2[18], carry, x, xMask; + unsigned i; + + /* tmp contains 64-bit words with the same 29,28,29-bit positions as an + * felem. So the top of an element of tmp might overlap with another + * element two positions down. The following loop eliminates this + * overlap. */ + tmp2[0] = (limb)(tmp[0] & kBottom29Bits); + + /* In the following we use "(limb) tmp[x]" and "(limb) (tmp[x]>>32)" to try + * and hint to the compiler that it can do a single-word shift by selecting + * the right register rather than doing a double-word shift and truncating + * afterwards. */ + tmp2[1] = ((limb) tmp[0]) >> 29; + tmp2[1] |= (((limb)(tmp[0] >> 32)) << 3) & kBottom28Bits; + tmp2[1] += ((limb) tmp[1]) & kBottom28Bits; + carry = tmp2[1] >> 28; + tmp2[1] &= kBottom28Bits; + + for (i = 2; i < 17; i++) { + tmp2[i] = ((limb)(tmp[i - 2] >> 32)) >> 25; + tmp2[i] += ((limb)(tmp[i - 1])) >> 28; + tmp2[i] += (((limb)(tmp[i - 1] >> 32)) << 4) & kBottom29Bits; + tmp2[i] += ((limb) tmp[i]) & kBottom29Bits; + tmp2[i] += carry; + carry = tmp2[i] >> 29; + tmp2[i] &= kBottom29Bits; + + i++; + if (i == 17) + break; + tmp2[i] = ((limb)(tmp[i - 2] >> 32)) >> 25; + tmp2[i] += ((limb)(tmp[i - 1])) >> 29; + tmp2[i] += (((limb)(tmp[i - 1] >> 32)) << 3) & kBottom28Bits; + tmp2[i] += ((limb) tmp[i]) & kBottom28Bits; + tmp2[i] += carry; + carry = tmp2[i] >> 28; + tmp2[i] &= kBottom28Bits; + } + + tmp2[17] = ((limb)(tmp[15] >> 32)) >> 25; + tmp2[17] += ((limb)(tmp[16])) >> 29; + tmp2[17] += (((limb)(tmp[16] >> 32)) << 3); + tmp2[17] += carry; + + /* Montgomery elimination of terms. + * + * Since R is 2**257, we can divide by R with a bitwise shift if we can + * ensure that the right-most 257 bits are all zero. We can make that true by + * adding multiplies of p without affecting the value. + * + * So we eliminate limbs from right to left. Since the bottom 29 bits of p + * are all ones, then by adding tmp2[0]*p to tmp2 we'll make tmp2[0] == 0. + * We can do that for 8 further limbs and then right shift to eliminate the + * extra factor of R. */ + for (i = 0;; i += 2) { + tmp2[i + 1] += tmp2[i] >> 29; + x = tmp2[i] & kBottom29Bits; + xMask = NON_ZERO_TO_ALL_ONES(x); + tmp2[i] = 0; + + /* The bounds calculations for this loop are tricky. Each iteration of + * the loop eliminates two words by adding values to words to their + * right. + * + * The following table contains the amounts added to each word (as an + * offset from the value of i at the top of the loop). The amounts are + * accounted for from the first and second half of the loop separately + * and are written as, for example, 28 to mean a value <2**28. + * + * Word: 3 4 5 6 7 8 9 10 + * Added in top half: 28 11 29 21 29 28 + * 28 29 + * 29 + * Added in bottom half: 29 10 28 21 28 28 + * 29 + * + * The value that is currently offset 7 will be offset 5 for the next + * iteration and then offset 3 for the iteration after that. Therefore + * the total value added will be the values added at 7, 5 and 3. + * + * The following table accumulates these values. The sums at the bottom + * are written as, for example, 29+28, to mean a value < 2**29+2**28. + * + * Word: 3 4 5 6 7 8 9 10 11 12 13 + * 28 11 10 29 21 29 28 28 28 28 28 + * 29 28 11 28 29 28 29 28 29 28 + * 29 28 21 21 29 21 29 21 + * 10 29 28 21 28 21 28 + * 28 29 28 29 28 29 28 + * 11 10 29 10 29 10 + * 29 28 11 28 11 + * 29 29 + * -------------------------------------------- + * 30+ 31+ 30+ 31+ 30+ + * 28+ 29+ 28+ 29+ 21+ + * 21+ 28+ 21+ 28+ 10 + * 10 21+ 10 21+ + * 11 11 + * + * So the greatest amount is added to tmp2[10] and tmp2[12]. If + * tmp2[10/12] has an initial value of <2**29, then the maximum value + * will be < 2**31 + 2**30 + 2**28 + 2**21 + 2**11, which is < 2**32, + * as required. */ + tmp2[i + 3] += (x << 10) & kBottom28Bits; + tmp2[i + 4] += (x >> 18); + + tmp2[i + 6] += (x << 21) & kBottom29Bits; + tmp2[i + 7] += x >> 8; + + /* At position 200, which is the starting bit position for word 7, we + * have a factor of 0xf000000 = 2**28 - 2**24. */ + tmp2[i + 7] += 0x10000000 & xMask; + /* Word 7 is 28 bits wide, so the 2**28 term exactly hits word 8. */ + tmp2[i + 8] += (x - 1) & xMask; + tmp2[i + 7] -= (x << 24) & kBottom28Bits; + tmp2[i + 8] -= x >> 4; + + tmp2[i + 8] += 0x20000000 & xMask; + tmp2[i + 8] -= x; + tmp2[i + 8] += (x << 28) & kBottom29Bits; + tmp2[i + 9] += ((x >> 1) - 1) & xMask; + + if (i+1 == NLIMBS) + break; + tmp2[i + 2] += tmp2[i + 1] >> 28; + x = tmp2[i + 1] & kBottom28Bits; + xMask = NON_ZERO_TO_ALL_ONES(x); + tmp2[i + 1] = 0; + + tmp2[i + 4] += (x << 11) & kBottom29Bits; + tmp2[i + 5] += (x >> 18); + + tmp2[i + 7] += (x << 21) & kBottom28Bits; + tmp2[i + 8] += x >> 7; + + /* At position 199, which is the starting bit of the 8th word when + * dealing with a context starting on an odd word, we have a factor of + * 0x1e000000 = 2**29 - 2**25. Since we have not updated i, the 8th + * word from i+1 is i+8. */ + tmp2[i + 8] += 0x20000000 & xMask; + tmp2[i + 9] += (x - 1) & xMask; + tmp2[i + 8] -= (x << 25) & kBottom29Bits; + tmp2[i + 9] -= x >> 4; + + tmp2[i + 9] += 0x10000000 & xMask; + tmp2[i + 9] -= x; + tmp2[i + 10] += (x - 1) & xMask; + } + + /* We merge the right shift with a carry chain. The words above 2**257 have + * widths of 28,29,... which we need to correct when copying them down. */ + carry = 0; + for (i = 0; i < 8; i++) { + /* The maximum value of tmp2[i + 9] occurs on the first iteration and + * is < 2**30+2**29+2**28. Adding 2**29 (from tmp2[i + 10]) is + * therefore safe. */ + out[i] = tmp2[i + 9]; + out[i] += carry; + out[i] += (tmp2[i + 10] << 28) & kBottom29Bits; + carry = out[i] >> 29; + out[i] &= kBottom29Bits; + + i++; + out[i] = tmp2[i + 9] >> 1; + out[i] += carry; + carry = out[i] >> 28; + out[i] &= kBottom28Bits; + } + + out[8] = tmp2[17]; + out[8] += carry; + carry = out[8] >> 29; + out[8] &= kBottom29Bits; + + felem_reduce_carry(out, carry); +} + +/* felem_square sets out=in*in. + * + * On entry: in[0,2,...] < 2**30, in[1,3,...] < 2**29. + * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29. */ +static void felem_square(felem out, const felem in) { + u64 tmp[17]; + + tmp[0] = ((u64) in[0]) * in[0]; + tmp[1] = ((u64) in[0]) * (in[1] << 1); + tmp[2] = ((u64) in[0]) * (in[2] << 1) + + ((u64) in[1]) * (in[1] << 1); + tmp[3] = ((u64) in[0]) * (in[3] << 1) + + ((u64) in[1]) * (in[2] << 1); + tmp[4] = ((u64) in[0]) * (in[4] << 1) + + ((u64) in[1]) * (in[3] << 2) + ((u64) in[2]) * in[2]; + tmp[5] = ((u64) in[0]) * (in[5] << 1) + ((u64) in[1]) * + (in[4] << 1) + ((u64) in[2]) * (in[3] << 1); + tmp[6] = ((u64) in[0]) * (in[6] << 1) + ((u64) in[1]) * + (in[5] << 2) + ((u64) in[2]) * (in[4] << 1) + + ((u64) in[3]) * (in[3] << 1); + tmp[7] = ((u64) in[0]) * (in[7] << 1) + ((u64) in[1]) * + (in[6] << 1) + ((u64) in[2]) * (in[5] << 1) + + ((u64) in[3]) * (in[4] << 1); + /* tmp[8] has the greatest value of 2**61 + 2**60 + 2**61 + 2**60 + 2**60, + * which is < 2**64 as required. */ + tmp[8] = ((u64) in[0]) * (in[8] << 1) + ((u64) in[1]) * + (in[7] << 2) + ((u64) in[2]) * (in[6] << 1) + + ((u64) in[3]) * (in[5] << 2) + ((u64) in[4]) * in[4]; + tmp[9] = ((u64) in[1]) * (in[8] << 1) + ((u64) in[2]) * + (in[7] << 1) + ((u64) in[3]) * (in[6] << 1) + + ((u64) in[4]) * (in[5] << 1); + tmp[10] = ((u64) in[2]) * (in[8] << 1) + ((u64) in[3]) * + (in[7] << 2) + ((u64) in[4]) * (in[6] << 1) + + ((u64) in[5]) * (in[5] << 1); + tmp[11] = ((u64) in[3]) * (in[8] << 1) + ((u64) in[4]) * + (in[7] << 1) + ((u64) in[5]) * (in[6] << 1); + tmp[12] = ((u64) in[4]) * (in[8] << 1) + + ((u64) in[5]) * (in[7] << 2) + ((u64) in[6]) * in[6]; + tmp[13] = ((u64) in[5]) * (in[8] << 1) + + ((u64) in[6]) * (in[7] << 1); + tmp[14] = ((u64) in[6]) * (in[8] << 1) + + ((u64) in[7]) * (in[7] << 1); + tmp[15] = ((u64) in[7]) * (in[8] << 1); + tmp[16] = ((u64) in[8]) * in[8]; + + felem_reduce_degree(out, tmp); +} + +/* felem_mul sets out=in*in2. + * + * On entry: in[0,2,...] < 2**30, in[1,3,...] < 2**29 and + * in2[0,2,...] < 2**30, in2[1,3,...] < 2**29. + * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29. */ +static void felem_mul(felem out, const felem in, const felem in2) { + u64 tmp[17]; + + tmp[0] = ((u64) in[0]) * in2[0]; + tmp[1] = ((u64) in[0]) * (in2[1] << 0) + + ((u64) in[1]) * (in2[0] << 0); + tmp[2] = ((u64) in[0]) * (in2[2] << 0) + ((u64) in[1]) * + (in2[1] << 1) + ((u64) in[2]) * (in2[0] << 0); + tmp[3] = ((u64) in[0]) * (in2[3] << 0) + ((u64) in[1]) * + (in2[2] << 0) + ((u64) in[2]) * (in2[1] << 0) + + ((u64) in[3]) * (in2[0] << 0); + tmp[4] = ((u64) in[0]) * (in2[4] << 0) + ((u64) in[1]) * + (in2[3] << 1) + ((u64) in[2]) * (in2[2] << 0) + + ((u64) in[3]) * (in2[1] << 1) + + ((u64) in[4]) * (in2[0] << 0); + tmp[5] = ((u64) in[0]) * (in2[5] << 0) + ((u64) in[1]) * + (in2[4] << 0) + ((u64) in[2]) * (in2[3] << 0) + + ((u64) in[3]) * (in2[2] << 0) + ((u64) in[4]) * + (in2[1] << 0) + ((u64) in[5]) * (in2[0] << 0); + tmp[6] = ((u64) in[0]) * (in2[6] << 0) + ((u64) in[1]) * + (in2[5] << 1) + ((u64) in[2]) * (in2[4] << 0) + + ((u64) in[3]) * (in2[3] << 1) + ((u64) in[4]) * + (in2[2] << 0) + ((u64) in[5]) * (in2[1] << 1) + + ((u64) in[6]) * (in2[0] << 0); + tmp[7] = ((u64) in[0]) * (in2[7] << 0) + ((u64) in[1]) * + (in2[6] << 0) + ((u64) in[2]) * (in2[5] << 0) + + ((u64) in[3]) * (in2[4] << 0) + ((u64) in[4]) * + (in2[3] << 0) + ((u64) in[5]) * (in2[2] << 0) + + ((u64) in[6]) * (in2[1] << 0) + + ((u64) in[7]) * (in2[0] << 0); + /* tmp[8] has the greatest value but doesn't overflow. See logic in + * felem_square. */ + tmp[8] = ((u64) in[0]) * (in2[8] << 0) + ((u64) in[1]) * + (in2[7] << 1) + ((u64) in[2]) * (in2[6] << 0) + + ((u64) in[3]) * (in2[5] << 1) + ((u64) in[4]) * + (in2[4] << 0) + ((u64) in[5]) * (in2[3] << 1) + + ((u64) in[6]) * (in2[2] << 0) + ((u64) in[7]) * + (in2[1] << 1) + ((u64) in[8]) * (in2[0] << 0); + tmp[9] = ((u64) in[1]) * (in2[8] << 0) + ((u64) in[2]) * + (in2[7] << 0) + ((u64) in[3]) * (in2[6] << 0) + + ((u64) in[4]) * (in2[5] << 0) + ((u64) in[5]) * + (in2[4] << 0) + ((u64) in[6]) * (in2[3] << 0) + + ((u64) in[7]) * (in2[2] << 0) + + ((u64) in[8]) * (in2[1] << 0); + tmp[10] = ((u64) in[2]) * (in2[8] << 0) + ((u64) in[3]) * + (in2[7] << 1) + ((u64) in[4]) * (in2[6] << 0) + + ((u64) in[5]) * (in2[5] << 1) + ((u64) in[6]) * + (in2[4] << 0) + ((u64) in[7]) * (in2[3] << 1) + + ((u64) in[8]) * (in2[2] << 0); + tmp[11] = ((u64) in[3]) * (in2[8] << 0) + ((u64) in[4]) * + (in2[7] << 0) + ((u64) in[5]) * (in2[6] << 0) + + ((u64) in[6]) * (in2[5] << 0) + ((u64) in[7]) * + (in2[4] << 0) + ((u64) in[8]) * (in2[3] << 0); + tmp[12] = ((u64) in[4]) * (in2[8] << 0) + ((u64) in[5]) * + (in2[7] << 1) + ((u64) in[6]) * (in2[6] << 0) + + ((u64) in[7]) * (in2[5] << 1) + + ((u64) in[8]) * (in2[4] << 0); + tmp[13] = ((u64) in[5]) * (in2[8] << 0) + ((u64) in[6]) * + (in2[7] << 0) + ((u64) in[7]) * (in2[6] << 0) + + ((u64) in[8]) * (in2[5] << 0); + tmp[14] = ((u64) in[6]) * (in2[8] << 0) + ((u64) in[7]) * + (in2[7] << 1) + ((u64) in[8]) * (in2[6] << 0); + tmp[15] = ((u64) in[7]) * (in2[8] << 0) + + ((u64) in[8]) * (in2[7] << 0); + tmp[16] = ((u64) in[8]) * (in2[8] << 0); + + felem_reduce_degree(out, tmp); +} + +static void felem_assign(felem out, const felem in) { + memcpy(out, in, sizeof(felem)); +} + +/* felem_scalar_3 sets out=3*out. + * + * On entry: out[0,2,...] < 2**30, out[1,3,...] < 2**29. + * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29. */ +static void felem_scalar_3(felem out) { + limb carry = 0; + unsigned i; + + for (i = 0;; i++) { + out[i] *= 3; + out[i] += carry; + carry = out[i] >> 29; + out[i] &= kBottom29Bits; + + i++; + if (i == NLIMBS) + break; + + out[i] *= 3; + out[i] += carry; + carry = out[i] >> 28; + out[i] &= kBottom28Bits; + } + + felem_reduce_carry(out, carry); +} + +/* felem_scalar_4 sets out=4*out. + * + * On entry: out[0,2,...] < 2**30, out[1,3,...] < 2**29. + * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29. */ +static void felem_scalar_4(felem out) { + limb carry = 0, next_carry; + unsigned i; + + for (i = 0;; i++) { + next_carry = out[i] >> 27; + out[i] <<= 2; + out[i] &= kBottom29Bits; + out[i] += carry; + carry = next_carry + (out[i] >> 29); + out[i] &= kBottom29Bits; + + i++; + if (i == NLIMBS) + break; + + next_carry = out[i] >> 26; + out[i] <<= 2; + out[i] &= kBottom28Bits; + out[i] += carry; + carry = next_carry + (out[i] >> 28); + out[i] &= kBottom28Bits; + } + + felem_reduce_carry(out, carry); +} + +/* felem_scalar_8 sets out=8*out. + * + * On entry: out[0,2,...] < 2**30, out[1,3,...] < 2**29. + * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29. */ +static void felem_scalar_8(felem out) { + limb carry = 0, next_carry; + unsigned i; + + for (i = 0;; i++) { + next_carry = out[i] >> 26; + out[i] <<= 3; + out[i] &= kBottom29Bits; + out[i] += carry; + carry = next_carry + (out[i] >> 29); + out[i] &= kBottom29Bits; + + i++; + if (i == NLIMBS) + break; + + next_carry = out[i] >> 25; + out[i] <<= 3; + out[i] &= kBottom28Bits; + out[i] += carry; + carry = next_carry + (out[i] >> 28); + out[i] &= kBottom28Bits; + } + + felem_reduce_carry(out, carry); +} + +/* felem_is_zero_vartime returns 1 iff |in| == 0. It takes a variable amount of + * time depending on the value of |in|. */ +static char felem_is_zero_vartime(const felem in) { + limb carry; + int i; + limb tmp[NLIMBS]; + + felem_assign(tmp, in); + + /* First, reduce tmp to a minimal form. */ + do { + carry = 0; + for (i = 0;; i++) { + tmp[i] += carry; + carry = tmp[i] >> 29; + tmp[i] &= kBottom29Bits; + + i++; + if (i == NLIMBS) + break; + + tmp[i] += carry; + carry = tmp[i] >> 28; + tmp[i] &= kBottom28Bits; + } + + felem_reduce_carry(tmp, carry); + } while (carry); + + /* tmp < 2**257, so the only possible zero values are 0, p and 2p. */ + return memcmp(tmp, kZero, sizeof(tmp)) == 0 || + memcmp(tmp, kP, sizeof(tmp)) == 0 || + memcmp(tmp, k2P, sizeof(tmp)) == 0; +} + + +/* Montgomery operations: */ + +#define kRDigits {2, 0, 0, 0xfffffffe, 0xffffffff, 0xffffffff, 0xfffffffd, 1} // 2^257 mod p256.p + +#define kRInvDigits {0x80000000, 1, 0xffffffff, 0, 0x80000001, 0xfffffffe, 1, 0x7fffffff} // 1 / 2^257 mod p256.p + +static const cryptonite_p256_int kR = { kRDigits }; +static const cryptonite_p256_int kRInv = { kRInvDigits }; + +/* to_montgomery sets out = R*in. */ +static void to_montgomery(felem out, const cryptonite_p256_int* in) { + cryptonite_p256_int in_shifted; + int i; + + cryptonite_p256_init(&in_shifted); + cryptonite_p256_modmul(&cryptonite_SECP256r1_p, in, 0, &kR, &in_shifted); + + for (i = 0; i < NLIMBS; i++) { + if ((i & 1) == 0) { + out[i] = P256_DIGIT(&in_shifted, 0) & kBottom29Bits; + cryptonite_p256_shr(&in_shifted, 29, &in_shifted); + } else { + out[i] = P256_DIGIT(&in_shifted, 0) & kBottom28Bits; + cryptonite_p256_shr(&in_shifted, 28, &in_shifted); + } + } + + cryptonite_p256_clear(&in_shifted); +} + +/* from_montgomery sets out=in/R. */ +static void from_montgomery(cryptonite_p256_int* out, const felem in) { + cryptonite_p256_int result, tmp; + int i, top; + + cryptonite_p256_init(&result); + cryptonite_p256_init(&tmp); + + cryptonite_p256_add_d(&tmp, in[NLIMBS - 1], &result); + for (i = NLIMBS - 2; i >= 0; i--) { + if ((i & 1) == 0) { + top = cryptonite_p256_shl(&result, 29, &tmp); + } else { + top = cryptonite_p256_shl(&result, 28, &tmp); + } + top |= cryptonite_p256_add_d(&tmp, in[i], &result); + } + + cryptonite_p256_modmul(&cryptonite_SECP256r1_p, &kRInv, top, &result, out); + + cryptonite_p256_clear(&result); + cryptonite_p256_clear(&tmp); +} diff --git a/cbits/include64/p256/p256.h b/cbits/include64/p256/p256.h new file mode 100644 --- /dev/null +++ b/cbits/include64/p256/p256.h @@ -0,0 +1,167 @@ +/* + * Copyright 2013 The Android Open Source Project + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Google Inc. nor the names of its contributors may + * be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY Google Inc. ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO + * EVENT SHALL Google Inc. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; + * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, + * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR + * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF + * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#ifndef SYSTEM_CORE_INCLUDE_MINCRYPT_LITE_P256_H_ +#define SYSTEM_CORE_INCLUDE_MINCRYPT_LITE_P256_H_ + +// Collection of routines manipulating 256 bit unsigned integers. +// Just enough to implement ecdsa-p256 and related algorithms. + +#include + +#ifdef __cplusplus +extern "C" { +#endif + +#define P256_BITSPERDIGIT 64 +#define P256_NDIGITS 4 +#define P256_NBYTES 32 + +// n' such as n * n' = -1 mod (2^64) +#define P256_MONTGOMERY_FACTOR 0xCCD1C8AAEE00BC4F + +#define P256_LITERAL(lo,hi) (((uint32_t) (lo)) + (((uint64_t) (hi)) << 32)) + +typedef int cryptonite_p256_err; +typedef uint64_t cryptonite_p256_digit; +typedef int64_t cryptonite_p256_sdigit; +typedef __uint128_t cryptonite_p256_ddigit; +typedef __int128_t cryptonite_p256_sddigit; + +// Defining cryptonite_p256_int as struct to leverage struct assigment. +typedef struct { + cryptonite_p256_digit a[P256_NDIGITS]; +} cryptonite_p256_int; + +extern const cryptonite_p256_int cryptonite_SECP256r1_n; // Curve order +extern const cryptonite_p256_int cryptonite_SECP256r1_p; // Curve prime +extern const cryptonite_p256_int cryptonite_SECP256r1_b; // Curve param + +// Initialize a cryptonite_p256_int to zero. +void cryptonite_p256_init(cryptonite_p256_int* a); + +// Clear a cryptonite_p256_int to zero. +void cryptonite_p256_clear(cryptonite_p256_int* a); + +// Return bit. Index 0 is least significant. +int cryptonite_p256_get_bit(const cryptonite_p256_int* a, int index); + +// b := a % MOD +void cryptonite_p256_mod( + const cryptonite_p256_int* MOD, + const cryptonite_p256_int* a, + cryptonite_p256_int* b); + +// c := a * (top_b | b) % MOD +void cryptonite_p256_modmul( + const cryptonite_p256_int* MOD, + const cryptonite_p256_int* a, + const cryptonite_p256_digit top_b, + const cryptonite_p256_int* b, + cryptonite_p256_int* c); + +// b := 1 / a % MOD +// MOD best be SECP256r1_n +void cryptonite_p256_modinv( + const cryptonite_p256_int* MOD, + const cryptonite_p256_int* a, + cryptonite_p256_int* b); + +// b := 1 / a % MOD +// MOD best be SECP256r1_n +// Faster than cryptonite_p256_modinv() +void cryptonite_p256_modinv_vartime( + const cryptonite_p256_int* MOD, + const cryptonite_p256_int* a, + cryptonite_p256_int* b); + +// b := a << (n % P256_BITSPERDIGIT) +// Returns the bits shifted out of most significant digit. +cryptonite_p256_digit cryptonite_p256_shl(const cryptonite_p256_int* a, int n, cryptonite_p256_int* b); + +// b := a >> (n % P256_BITSPERDIGIT) +void cryptonite_p256_shr(const cryptonite_p256_int* a, int n, cryptonite_p256_int* b); + +int cryptonite_p256_is_zero(const cryptonite_p256_int* a); +int cryptonite_p256_is_odd(const cryptonite_p256_int* a); +int cryptonite_p256_is_even(const cryptonite_p256_int* a); + +// Returns -1, 0 or 1. +int cryptonite_p256_cmp(const cryptonite_p256_int* a, const cryptonite_p256_int *b); + +// c: = a - b +// Returns -1 on borrow. +int cryptonite_p256_sub(const cryptonite_p256_int* a, const cryptonite_p256_int* b, cryptonite_p256_int* c); + +// c := a + b +// Returns 1 on carry. +int cryptonite_p256_add(const cryptonite_p256_int* a, const cryptonite_p256_int* b, cryptonite_p256_int* c); + +// c := a + (single digit)b +// Returns carry 1 on carry. +int cryptonite_p256_add_d(const cryptonite_p256_int* a, cryptonite_p256_digit b, cryptonite_p256_int* c); + +// ec routines. + +// {out_x,out_y} := nG +void cryptonite_p256_base_point_mul(const cryptonite_p256_int *n, + cryptonite_p256_int *out_x, + cryptonite_p256_int *out_y); + +// {out_x,out_y} := n{in_x,in_y} +void cryptonite_p256_point_mul(const cryptonite_p256_int *n, + const cryptonite_p256_int *in_x, + const cryptonite_p256_int *in_y, + cryptonite_p256_int *out_x, + cryptonite_p256_int *out_y); + +// {out_x,out_y} := n1G + n2{in_x,in_y} +void cryptonite_p256_points_mul_vartime( + const cryptonite_p256_int *n1, const cryptonite_p256_int *n2, + const cryptonite_p256_int *in_x, const cryptonite_p256_int *in_y, + cryptonite_p256_int *out_x, cryptonite_p256_int *out_y); + +// Return whether point {x,y} is on curve. +int cryptonite_p256_is_valid_point(const cryptonite_p256_int* x, const cryptonite_p256_int* y); + +// Outputs big-endian binary form. No leading zero skips. +void cryptonite_p256_to_bin(const cryptonite_p256_int* src, uint8_t dst[P256_NBYTES]); + +// Reads from big-endian binary form, +// thus pre-pad with leading zeros if short. +void cryptonite_p256_from_bin(const uint8_t src[P256_NBYTES], cryptonite_p256_int* dst); + +#define P256_DIGITS(x) ((x)->a) +#define P256_DIGIT(x,y) ((x)->a[y]) + +#define P256_ZERO {{0}} +#define P256_ONE {{1}} + +#ifdef __cplusplus +} +#endif + +#endif // SYSTEM_CORE_INCLUDE_MINCRYPT_LITE_P256_H_ diff --git a/cbits/include64/p256/p256_gf.h b/cbits/include64/p256/p256_gf.h new file mode 100644 --- /dev/null +++ b/cbits/include64/p256/p256_gf.h @@ -0,0 +1,713 @@ +/* + * Copyright 2013 The Android Open Source Project + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Google Inc. nor the names of its contributors may + * be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY Google Inc. ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO + * EVENT SHALL Google Inc. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; + * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, + * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR + * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF + * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +// This is an implementation of the P256 finite field. It's written to be +// portable and still constant-time. +// +// WARNING: Implementing these functions in a constant-time manner is far from +// obvious. Be careful when touching this code. +// +// See http://www.imperialviolet.org/2010/12/04/ecc.html ([1]) for background. + +#include +#include + +#include +#include + +#include "p256/p256.h" + +typedef uint8_t u8; +typedef uint32_t u32; +typedef uint64_t u64; +typedef int64_t s64; +typedef __uint128_t u128; + +/* Our field elements are represented as five 64-bit limbs. + * + * The value of an felem (field element) is: + * x[0] + (x[1] * 2**51) + (x[2] * 2**103) + ... + (x[4] * 2**206) + * + * That is, each limb is alternately 51 or 52-bits wide in little-endian + * order. + * + * This means that an felem hits 2**257, rather than 2**256 as we would like. + * + * Finally, the values stored in an felem are in Montgomery form. So the value + * |y| is stored as (y*R) mod p, where p is the P-256 prime and R is 2**257. + */ +typedef u64 limb; +#define NLIMBS 5 +typedef limb felem[NLIMBS]; + +static const limb kBottom51Bits = 0x7ffffffffffff; +static const limb kBottom52Bits = 0xfffffffffffff; + +/* kOne is the number 1 as an felem. It's 2**257 mod p split up into 51 and + * 52-bit words. */ +static const felem kOne = { + 2, 0xfc00000000000, 0x7ffffffffffff, 0xfff7fffffffff, 0x7ffff +}; +static const felem kZero = {0}; +static const felem kP = { + 0x7ffffffffffff, 0x1fffffffffff, 0, 0x4000000000, 0x3fffffffc0000 +}; +static const felem k2P = { + 0x7fffffffffffe, 0x3fffffffffff, 0, 0x8000000000, 0x7fffffff80000 +}; +/* kPrecomputed contains precomputed values to aid the calculation of scalar + * multiples of the base point, G. It's actually two, equal length, tables + * concatenated. + * + * The first table contains (x,y) felem pairs for 16 multiples of the base + * point, G. + * + * Index | Index (binary) | Value + * 0 | 0000 | 0G (all zeros, omitted) + * 1 | 0001 | G + * 2 | 0010 | 2**64G + * 3 | 0011 | 2**64G + G + * 4 | 0100 | 2**128G + * 5 | 0101 | 2**128G + G + * 6 | 0110 | 2**128G + 2**64G + * 7 | 0111 | 2**128G + 2**64G + G + * 8 | 1000 | 2**192G + * 9 | 1001 | 2**192G + G + * 10 | 1010 | 2**192G + 2**64G + * 11 | 1011 | 2**192G + 2**64G + G + * 12 | 1100 | 2**192G + 2**128G + * 13 | 1101 | 2**192G + 2**128G + G + * 14 | 1110 | 2**192G + 2**128G + 2**64G + * 15 | 1111 | 2**192G + 2**128G + 2**64G + G + * + * The second table follows the same style, but the terms are 2**32G, + * 2**96G, 2**160G, 2**224G. + * + * This is ~2KB of data. */ +static const limb kPrecomputed[NLIMBS * 2 * 15 * 2] = { + 0x661a831522878, 0xf17fb6d805e79, 0x5889441d6ea57, 0xae33cfdb995bb, 0xc482fbb529ba, + 0x4a6af9d2aac15, 0x90e867917377c, 0x487cc962d2ae3, 0xec2a97443446e, 0x2b8ff8c52c42, + 0x45f8a2d41a576, 0xb06988d2653e4, 0x718b22c357305, 0x33fc920e79d2b, 0x17af34b0fe8db, + 0x38e17eb402f2f, 0x3382558649705, 0x47f6d48f482d1, 0x7bd42488d9b83, 0x3b247c8b86b78, + 0x4d08fc26f7778, 0x7a29a82fb2795, 0x75cd18f90d11a, 0xad8e213b0bc, 0x2d5f0142899e8, + 0x506f98098fb57, 0x2f0c98301e4aa, 0x39b30dd5cf67d, 0x9c146498ab13c, 0xa5db92df5b7b, + 0x184897fc4124a, 0xe3f73a19d8aa, 0x4e1c18e47066b, 0x27b2d4b52eaee, 0x30eac3ea10e99, + 0x4e74546e2e7d5, 0x1f4dde2d97a1d, 0x6ead0f88e1200, 0x7dec87c220f02, 0x3d08ff096310f, + 0x23e5659633ffa, 0x6ec648f08c722, 0x3172a3806ea35, 0xf6e5b681eb3c5, 0x2c3758260f89d, + 0x38dca4fd1da12, 0xf06067b78830d, 0x3194be87a068c, 0x78893c7eb602b, 0xcead60438432, + 0x6ee69a56a67ab, 0xd886f77701895, 0x67b0a4d9cee2b, 0x3586bbf3e4d53, 0x1db6f32921d93, + 0x260756ca4b366, 0x4f40e9d2039fa, 0x4f3f09f5a82bf, 0xccde2d641e8cd, 0x305a30cd2e8c5, + 0x471c235cb5439, 0xab279cd962f5a, 0x17e1fb6e2dd94, 0xfe64589800a77, 0xe8793d99775f, + 0x48c62f4e614aa, 0xbf76ef20eb2a4, 0x669c672556c, 0x24683e0eff056, 0x12252b369ab76, + 0x821de9f162d5, 0xf911ec99a95be, 0x6721f065c906b, 0x58d452035c736, 0x1f9f01a6a15, + 0x6135009b7d8d3, 0xdaeeeb417dfc0, 0x63865fea0ee17, 0x6e0a304b939d6, 0x204ba2076833d, + 0x4ade586f35669, 0x2c1077e34611a, 0x5b1a3bea3b81a, 0xf97d018a22c8b, 0x38d7996b08af8, + 0x6ea62baeb7aa0, 0xebdcbd9ef2670, 0x35dc8fe0df3fe, 0xe458309d20c24, 0x11e87898716a0, + 0x7c44bab7cb456, 0xd64d3cf1bb64, 0x189bff1bf9e66, 0xb5218a049311, 0x285dda6cbcc81, + 0x3238dcafd8c7c, 0x607736c8de0, 0xdb83d99508b1, 0x4e1a0d404cd81, 0x1588008c00ff2, + 0x16b8b36722b27, 0x876609c3f3f1a, 0x66b72ef0e17d6, 0x705f8a279d568, 0x2eaac4cd01fdd, + 0x1171ce9705fe9, 0xffc79cd3264ee, 0x700c8ab4b80f0, 0x208d3d4f57a1, 0x337262a8ca4eb, + 0x297fd01d843fd, 0xa90956fa097f8, 0x529759fdb3845, 0x1d78c5e2d0397, 0x3d6938a4adbf3, + 0x16d5853560b66, 0xf138946b9a430, 0x2ab79f4dea6a0, 0xd42053ee43ae1, 0x3b9c3ef1cf870, + 0x598934ad81baf, 0x5f1821b1d07a7, 0x416bb3a973ff3, 0x23f07bd0a047a, 0x19bdc2e09f786, + 0x56dc9981cd51f, 0xfbace23c8cd65, 0x673bd3bf5b52e, 0x46a95d229fd61, 0xe09ad64bcfb1, + 0xe5292b91f17d, 0xfeefcd8afc287, 0x58f52b0a58711, 0x4800f20c201ef, 0x2084fce608f67, + 0x12ba0b128ae0b, 0x5977ae17030b4, 0x101126ee420f6, 0xf70823495c6bd, 0xde19a27d7770, + 0x5c6ac852260e8, 0x9d22950ac4356, 0x441cca955246c, 0x660a34e5332d9, 0x14ac8ea92f8d2, + 0x6b6d7709f307e, 0x67d7e13879db, 0x2ea8626f9fbbd, 0x99609006a4b40, 0x31bb2a8f8c779, + 0x10c04828ea335, 0xae9acdcbc080a, 0x617af2342607a, 0xc7494ea53e553, 0x2ca9e2872defa, + 0x6c399fab21f1f, 0xab139b245e758, 0x3ad933dcba589, 0x4797fecb08811, 0x31f5dbf8f594, + 0x7dc6361cc7a69, 0xc8a7953ead3f9, 0x79ed693d18015, 0x418a024999a6a, 0x2c4fdc9436aa, + 0x1eb98cb06aa75, 0x2989592796a9c, 0x11194821e425, 0xe27a648228388, 0x35d834b6c12a0, + 0x541807713b532, 0x7ae0a1008aaee, 0x7017a29bcb5e, 0x6b193c23c315c, 0x19bd25ac82f2a, + 0x6a01a43eef294, 0xddf5b5fd84f19, 0x33f5ba081c016, 0xdeb052d1bc082, 0x6b2f06afa617, + 0x7ca1eda6a939f, 0xbdeb35997b50c, 0x47f2d1bccda5, 0xc2ff4adfed667, 0x87712997be4, + 0x21fc2e2b37659, 0xf7d62cd5ed951, 0x27fa9cbdf7efa, 0xba25582bf3a6b, 0x2a42b8bd89398, + 0x6d377d07eecd2, 0x9ca1df5af387, 0x1109e3427e2ba, 0xce4aa4572a19, 0x103baaef71e16, + 0x2c3b2dfde328a, 0xbec4b4a30e1ef, 0x37d92a86204f3, 0x806cfde68eb39, 0x246e2f72b8aa5, + 0x68d3de93462a9, 0x53b8acba6bbc3, 0x2492a70fa1696, 0x38c62d5760f55, 0x15096fe4904f2, + 0x4e44e9bed3e3a, 0xb28bfd79cc9bc, 0x6a77513839320, 0x480dcec6739db, 0x3601b739f2465, + 0x43c348e2a7e1, 0xe448106327879, 0x175d9cae1b0ed, 0xd3b89dee743b8, 0x392d73ca255bc, + 0x32946db0d3a18, 0x9261b09907cc, 0x5ba517a755722, 0x51f24fdaf5184, 0x1cdc732989ed8, + 0x2f7806ba16694, 0xae0c9f029f8d0, 0xd8b45102ce1, 0xca1c7db9316d6, 0x162088a67066f, + 0x39de35b2b4162, 0xa19f550d88ae9, 0x7921b27026cde, 0x94b936b66e900, 0x1023bd5fa17fc, + 0x436837814cfa4, 0x29113492283c4, 0x66d1cdd8b51d8, 0xa540702278eb2, 0x47ef1b29285d, + 0x587b50917e50e, 0xb4cda75bab3b, 0x112520b0a9886, 0x66b9ac16fee49, 0x17bf17e92b2eb, + 0x2456a2f150ed7, 0xfa214412d0280, 0x3ca7dd947fe5b, 0xa72c28598d58a, 0x255d945efc3e, + 0x2873f04e0f215, 0x74178fd1af57b, 0x788848b5b2d6, 0xb1ffafaae0db6, 0x32a1b7b3cbb2a, + 0x4bd9935d6b2da, 0x9c08f24ad30a5, 0x4e58407a80f, 0x1b3a3825a5b17, 0x6547e9fc82f5, + 0x47484aa3656c3, 0x6ee43f341a494, 0x64a98f87adea2, 0x619b3f8e95f01, 0xb6e513266ed8, + 0x421c2a673090, 0xa1c1de32348c7, 0x55b85c3a1e8a3, 0xe05ce8ef330b4, 0x2561e49c15d84, + 0x40aa2d33130fa, 0x12b827d35866f, 0xfe4cf62c8ddb, 0x2fa0ef05bb28d, 0x1c06ca63f1cb8, + 0x32a971863863b, 0xff6fc86830da1, 0x71e7b25a14cf3, 0xea9c5ebb1373a, 0x250bbaa3e1634, + 0x5b5ffeda5b765, 0xf25d2a746331b, 0x115e3a3f43632, 0x67303af43c9d5, 0x14bb538a0e559, + 0x75623687d43b7, 0xa349674a4b38d, 0x613c61829ffc6, 0x689828d8110c7, 0x139115f5af7d5, + 0xf1d856152289, 0x45cbe967168ab, 0x51f38e1680901, 0x34808e8f652b0, 0x1f4a6a921e156, + 0x35dfaf3d8341f, 0xf53ace725cb63, 0x3d86a54eef35b, 0xa103aabaffe2c, 0x2decc36296fbd, + 0x510282be73d6f, 0xd4e6365db206a, 0x4bdc5f5bb8bf3, 0xde7ea32a3aee7, 0x71269e274305, +}; + + +/* Field element operations: */ + +/* NON_ZERO_TO_ALL_ONES returns: + * 0xffffffffffffffff for 0 < x <= 2**63 + * 0 for x == 0 or x > 2**63. + * + * x must be a u64 or an equivalent type such as limb. */ +#define NON_ZERO_TO_ALL_ONES(x) ((((u64)(x) - 1) >> 63) - 1) + +/* felem_reduce_carry adds a multiple of p in order to cancel |carry|, + * which is a term at 2**257. + * + * On entry: carry < 2**6, inout[0,2,...] < 2**51, inout[1,3,...] < 2**52. + * On exit: inout[0,2,..] < 2**52, inout[1,3,...] < 2**53. */ +static void felem_reduce_carry(felem inout, limb carry) { + const u64 carry_mask = NON_ZERO_TO_ALL_ONES(carry); + + inout[0] += carry << 1; + inout[1] += 0x10000000000000 & carry_mask; + /* carry < 2**6 thus (carry << 46) < 2**52 and we added 2**52 in the + * previous line therefore this doesn't underflow. */ + inout[1] -= carry << 46; + inout[2] += (0x8000000000000 - 1) & carry_mask; + inout[3] += (0x10000000000000 - 1) & carry_mask; + inout[3] -= carry << 39; + /* This may underflow if carry is non-zero but, if so, we'll fix it in the + * next line. */ + inout[4] -= 1 & carry_mask; + inout[4] += carry << 19; +} + +/* felem_sum sets out = in+in2. + * + * On entry, in[i]+in2[i] must not overflow a 64-bit word. + * On exit: out[0,2,...] < 2**52, out[1,3,...] < 2**53 */ +static void felem_sum(felem out, const felem in, const felem in2) { + limb carry = 0; + unsigned i; + + for (i = 0;; i++) { + out[i] = in[i] + in2[i]; + out[i] += carry; + carry = out[i] >> 51; + out[i] &= kBottom51Bits; + + i++; + if (i == NLIMBS) + break; + + out[i] = in[i] + in2[i]; + out[i] += carry; + carry = out[i] >> 52; + out[i] &= kBottom52Bits; + } + + felem_reduce_carry(out, carry); +} + +#define two53m3 (((limb)1) << 53) - (((limb)1) << 3) +#define two54m52p48m2 (((limb)1) << 54) - (((limb)1) << 52) + (((limb)1) << 48) - (((limb)1) << 2) +#define two53m2p0 (((limb)1) << 53) - (((limb)1) << 2) + (((limb)1) << 0) +#define two54m52p41m2 (((limb)1) << 54) - (((limb)1) << 52) + (((limb)1) << 41) - (((limb)1) << 2) +#define two53m21m2p0 (((limb)1) << 53) - (((limb)1) << 21) - (((limb)1) << 2) + (((limb)1) << 0) + +/* zero53 is 0 mod p. */ +static const felem zero53 = { two53m3, two54m52p48m2, two53m2p0, two54m52p41m2, two53m21m2p0 }; + +/* felem_diff sets out = in-in2. + * + * On entry: in[0,2,...] < 2**52, in[1,3,...] < 2**53 and + * in2[0,2,...] < 2**52, in2[1,3,...] < 2**53. + * On exit: out[0,2,...] < 2**52, out[1,3,...] < 2**53. */ +static void felem_diff(felem out, const felem in, const felem in2) { + limb carry = 0; + unsigned i; + + for (i = 0;; i++) { + out[i] = in[i] - in2[i]; + out[i] += zero53[i]; + out[i] += carry; + carry = out[i] >> 51; + out[i] &= kBottom51Bits; + + i++; + if (i == NLIMBS) + break; + + out[i] = in[i] - in2[i]; + out[i] += zero53[i]; + out[i] += carry; + carry = out[i] >> 52; + out[i] &= kBottom52Bits; + } + + felem_reduce_carry(out, carry); +} + +/* felem_reduce_degree sets out = tmp/R mod p where tmp contains 64-bit words + * with the same 51,52,... bit positions as an felem. + * + * The values in felems are in Montgomery form: x*R mod p where R = 2**257. + * Since we just multiplied two Montgomery values together, the result is + * x*y*R*R mod p. We wish to divide by R in order for the result also to be + * in Montgomery form. + * + * On entry: tmp[i] < 2**128 + * On exit: out[0,2,...] < 2**52, out[1,3,...] < 2**53 */ +static void felem_reduce_degree(felem out, u128 tmp[9]) { + /* The following table may be helpful when reading this code: + * + * Limb number: 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 + * Width (bits): 51| 52| 51| 52| 51| 52| 51| 52| 51| 52| 51 + * Start bit: 0 | 51|103|154|206|257|309|360|412|463|515 + * (odd phase): 0 | 52|103|155|206|258|309|361|412|464|515 */ + limb tmp2[10], carry, x, xShiftedMask; + unsigned i; + + /* tmp contains 128-bit words with the same 51,52,51-bit positions as an + * felem. So the top of an element of tmp might overlap with another + * element two positions down. The following loop eliminates this + * overlap. */ + tmp2[0] = (limb)(tmp[0] & kBottom51Bits); + + /* In the following we use "(limb) tmp[x]" and "(limb) (tmp[x]>>64)" to try + * and hint to the compiler that it can do a single-word shift by selecting + * the right register rather than doing a double-word shift and truncating + * afterwards. */ + tmp2[1] = ((limb) tmp[0]) >> 51; + tmp2[1] |= (((limb)(tmp[0] >> 64)) << 13) & kBottom52Bits; + tmp2[1] += ((limb) tmp[1]) & kBottom52Bits; + carry = tmp2[1] >> 52; + tmp2[1] &= kBottom52Bits; + + for (i = 2; i < 9; i++) { + tmp2[i] = ((limb)(tmp[i - 2] >> 64)) >> 39; + tmp2[i] += ((limb)(tmp[i - 1])) >> 52; + tmp2[i] += (((limb)(tmp[i - 1] >> 64)) << 12) & kBottom51Bits; + tmp2[i] += ((limb) tmp[i]) & kBottom51Bits; + tmp2[i] += carry; + carry = tmp2[i] >> 51; + tmp2[i] &= kBottom51Bits; + + i++; + if (i == 9) + break; + tmp2[i] = ((limb)(tmp[i - 2] >> 64)) >> 39; + tmp2[i] += ((limb)(tmp[i - 1])) >> 51; + tmp2[i] += (((limb)(tmp[i - 1] >> 64)) << 13) & kBottom52Bits; + tmp2[i] += ((limb) tmp[i]) & kBottom52Bits; + tmp2[i] += carry; + carry = tmp2[i] >> 52; + tmp2[i] &= kBottom52Bits; + } + + tmp2[9] = ((limb)(tmp[7] >> 64)) >> 39; + tmp2[9] += ((limb)(tmp[8])) >> 51; + tmp2[9] += (((limb)(tmp[8] >> 64)) << 13); + tmp2[9] += carry; + + /* Montgomery elimination of terms. + * + * Since R is 2**257, we can divide by R with a bitwise shift if we can + * ensure that the right-most 257 bits are all zero. We can make that true by + * adding multiplies of p without affecting the value. + * + * So we eliminate limbs from right to left. Since the bottom 51 bits of p + * are all ones, then by adding tmp2[0]*p to tmp2 we'll make tmp2[0] == 0. + * We can do that for 8 further limbs and then right shift to eliminate the + * extra factor of R. */ + for (i = 0;; i += 2) { + tmp2[i + 1] += tmp2[i] >> 51; + x = tmp2[i] & kBottom51Bits; + xShiftedMask = NON_ZERO_TO_ALL_ONES(x >> 1); + tmp2[i] = 0; + + /* The bounds calculations for this loop are tricky. Each iteration of + * the loop eliminates two words by adding values to words to their + * right. + * + * The following table contains the amounts added to each word (as an + * offset from the value of i at the top of the loop). The amounts are + * accounted for from the first and second half of the loop separately + * and are written as, for example, 51 to mean a value <2**51. + * + * Word: 1 2 3 4 5 6 + * Added in top half: 52 44 52 37 50 + * 51 + * 51 + * Added in bottom half: 51 45 51 38 50 + * 52 + * 52 + * + * The value that is currently offset 5 will be offset 3 for the next + * iteration and then offset 1 for the iteration after that. Therefore + * the total value added will be the values added at 5, 3 and 1. + * + * The following table accumulates these values. The sums at the bottom + * are written as, for example, 53+45, to mean a value < 2**53+2**45. + * + * Word: 1 2 3 4 5 6 7 8 9 + * 52 44 52 37 50 50 50 50 50 + * 51 45 51 38 37 38 37 + * 52 51 52 51 52 51 + * 51 52 51 52 51 + * 44 52 51 52 + * 51 45 44 + * 52 + * ------------------------------------ + * 53+ 53+ 54+ 52+ 53+ 52+ + * 45 44+ 50+ 51+ 52+ 50+ + * 37 45+ 50+ 50+ 37 + * 38 44+ 38 + * 37 + * + * So the greatest amount is added to tmp2[5]. If tmp2[5] has an initial + * value of <2**52, then the maximum value will be < 2**54 + 2**52 + 2**50 + + * 2**45 + 2**38, which is < 2**64, as required. */ + tmp2[i + 1] += (x << 45) & kBottom52Bits; + tmp2[i + 2] += x >> 7; + + tmp2[i + 3] += (x << 38) & kBottom52Bits; + tmp2[i + 4] += x >> 14; + + /* On tmp2[i + 4], when x < 2**1, the subtraction with (x << 18) will not + * underflow because it is balanced with the (x << 50) term. On the next + * word tmp2[i + 5], terms with (x >> 1) and (x >> 33) are both zero and + * there is no underflow either. + * + * When x >= 2**1, we add 2**51 to tmp2[i + 4] to avoid an underflow. + * Removing 1 from tmp2[i + 5] is safe because (x >> 1) - (x >> 33) is + * strictly positive. + */ + tmp2[i + 4] += 0x8000000000000 & xShiftedMask; + tmp2[i + 5] -= 1 & xShiftedMask; + + tmp2[i + 4] -= (x << 18) & kBottom51Bits; + tmp2[i + 4] += (x << 50) & kBottom51Bits; + tmp2[i + 5] += (x >> 1) - (x >> 33); + + if (i+1 == NLIMBS) + break; + tmp2[i + 2] += tmp2[i + 1] >> 52; + x = tmp2[i + 1] & kBottom52Bits; + xShiftedMask = NON_ZERO_TO_ALL_ONES(x >> 2); + tmp2[i + 1] = 0; + + tmp2[i + 2] += (x << 44) & kBottom51Bits; + tmp2[i + 3] += x >> 7; + + tmp2[i + 4] += (x << 37) & kBottom51Bits; + tmp2[i + 5] += x >> 14; + + /* On tmp2[i + 5], when x < 2**2, the subtraction with (x << 18) will not + * underflow because it is balanced with the (x << 50) term. On the next + * word tmp2[i + 6], terms with (x >> 2) and (x >> 34) are both zero and + * there is no underflow either. + * + * When x >= 2**2, we add 2**52 to tmp2[i + 5] to avoid an underflow. + * Removing 1 from tmp2[i + 6] is safe because (x >> 2) - (x >> 34) is + * stricly positive. + */ + tmp2[i + 5] += 0x10000000000000 & xShiftedMask; + tmp2[i + 6] -= 1 & xShiftedMask; + + tmp2[i + 5] -= (x << 18) & kBottom52Bits; + tmp2[i + 5] += (x << 50) & kBottom52Bits; + tmp2[i + 6] += (x >> 2) - (x >> 34); + } + + /* We merge the right shift with a carry chain. The words above 2**257 have + * widths of 52,51,... which we need to correct when copying them down. */ + carry = 0; + for (i = 0; i < 4; i++) { + out[i] = tmp2[i + 5]; + out[i] += carry; + carry = out[i] >> 51; + out[i] &= kBottom51Bits; + + i++; + out[i] = tmp2[i + 5] << 1; + out[i] += carry; + carry = out[i] >> 52; + out[i] &= kBottom52Bits; + } + + out[4] = tmp2[9]; + out[4] += carry; + carry = out[4] >> 51; + out[4] &= kBottom51Bits; + + felem_reduce_carry(out, carry); +} + +/* felem_square sets out=in*in. + * + * On entry: in[0,2,...] < 2**52, in[1,3,...] < 2**53. + * On exit: out[0,2,...] < 2**52, out[1,3,...] < 2**53. */ +static void felem_square(felem out, const felem in) { + u128 tmp[9], x1x1, x3x3; + + x1x1 = ((u128) in[1]) * in[1]; + x3x3 = ((u128) in[3]) * in[3]; + + tmp[0] = ((u128) in[0]) * (in[0] << 0); + tmp[1] = ((u128) in[0]) * (in[1] << 1) + ((x1x1 & 1) << 51); + tmp[2] = ((u128) in[0]) * (in[2] << 1) + (x1x1 >> 1); + tmp[3] = ((u128) in[0]) * (in[3] << 1) + + ((u128) in[1]) * (in[2] << 1); + tmp[4] = ((u128) in[0]) * (in[4] << 1) + + ((u128) in[1]) * (in[3] << 0) + + ((u128) in[2]) * (in[2] << 0); + tmp[5] = ((u128) in[1]) * (in[4] << 1) + + ((u128) in[2]) * (in[3] << 1) + ((x3x3 & 1) << 51); + tmp[6] = ((u128) in[2]) * (in[4] << 1) + (x3x3 >> 1); + tmp[7] = ((u128) in[3]) * (in[4] << 1); + tmp[8] = ((u128) in[4]) * (in[4] << 0); + + felem_reduce_degree(out, tmp); +} + +/* felem_mul sets out=in*in2. + * + * On entry: in[0,2,...] < 2**52, in[1,3,...] < 2**53 and + * in2[0,2,...] < 2**52, in2[1,3,...] < 2**53. + * On exit: out[0,2,...] < 2**52, out[1,3,...] < 2**53. */ +static void felem_mul(felem out, const felem in, const felem in2) { + u128 tmp[9], x1y1, x1y3, x3y1, x3y3; + + x1y1 = ((u128) in[1]) * in2[1]; + x1y3 = ((u128) in[1]) * in2[3]; + x3y1 = ((u128) in[3]) * in2[1]; + x3y3 = ((u128) in[3]) * in2[3]; + + tmp[0] = ((u128) in[0]) * in2[0]; + tmp[1] = ((u128) in[0]) * in2[1] + + ((u128) in[1]) * in2[0] + ((x1y1 & 1) << 51); + tmp[2] = ((u128) in[0]) * in2[2] + (x1y1 >> 1) + + ((u128) in[2]) * in2[0]; + tmp[3] = ((u128) in[0]) * in2[3] + + ((u128) in[1]) * in2[2] + + ((u128) in[2]) * in2[1] + ((x1y3 & 1) << 51) + + ((u128) in[3]) * in2[0] + ((x3y1 & 1) << 51); + tmp[4] = ((u128) in[0]) * in2[4] + (x1y3 >> 1) + + ((u128) in[2]) * in2[2] + (x3y1 >> 1) + + ((u128) in[4]) * in2[0]; + tmp[5] = ((u128) in[1]) * in2[4] + + ((u128) in[2]) * in2[3] + + ((u128) in[3]) * in2[2] + + ((u128) in[4]) * in2[1] + ((x3y3 & 1) << 51); + tmp[6] = ((u128) in[2]) * in2[4] + (x3y3 >> 1) + + ((u128) in[4]) * in2[2]; + tmp[7] = ((u128) in[3]) * in2[4] + + ((u128) in[4]) * in2[3]; + tmp[8] = ((u128) in[4]) * in2[4]; + + felem_reduce_degree(out, tmp); +} + +static void felem_assign(felem out, const felem in) { + memcpy(out, in, sizeof(felem)); +} + +/* felem_scalar_3 sets out=3*out. + * + * On entry: out[0,2,...] < 2**52, out[1,3,...] < 2**53. + * On exit: out[0,2,...] < 2**52, out[1,3,...] < 2**53. */ +static void felem_scalar_3(felem out) { + limb carry = 0; + unsigned i; + + for (i = 0;; i++) { + out[i] *= 3; + out[i] += carry; + carry = out[i] >> 51; + out[i] &= kBottom51Bits; + + i++; + if (i == NLIMBS) + break; + + out[i] *= 3; + out[i] += carry; + carry = out[i] >> 52; + out[i] &= kBottom52Bits; + } + + felem_reduce_carry(out, carry); +} + +/* felem_scalar_4 sets out=4*out. + * + * On entry: out[0,2,...] < 2**52, out[1,3,...] < 2**53. + * On exit: out[0,2,...] < 2**52, out[1,3,...] < 2**53. */ +static void felem_scalar_4(felem out) { + limb carry = 0, next_carry; + unsigned i; + + for (i = 0;; i++) { + next_carry = out[i] >> 49; + out[i] <<= 2; + out[i] &= kBottom51Bits; + out[i] += carry; + carry = next_carry + (out[i] >> 51); + out[i] &= kBottom51Bits; + + i++; + if (i == NLIMBS) + break; + + next_carry = out[i] >> 50; + out[i] <<= 2; + out[i] &= kBottom52Bits; + out[i] += carry; + carry = next_carry + (out[i] >> 52); + out[i] &= kBottom52Bits; + } + + felem_reduce_carry(out, carry); +} + +/* felem_scalar_8 sets out=8*out. + * + * On entry: out[0,2,...] < 2**52, out[1,3,...] < 2**53. + * On exit: out[0,2,...] < 2**52, out[1,3,...] < 2**53. */ +static void felem_scalar_8(felem out) { + limb carry = 0, next_carry; + unsigned i; + + for (i = 0;; i++) { + next_carry = out[i] >> 48; + out[i] <<= 3; + out[i] &= kBottom51Bits; + out[i] += carry; + carry = next_carry + (out[i] >> 51); + out[i] &= kBottom51Bits; + + i++; + if (i == NLIMBS) + break; + + next_carry = out[i] >> 49; + out[i] <<= 3; + out[i] &= kBottom52Bits; + out[i] += carry; + carry = next_carry + (out[i] >> 52); + out[i] &= kBottom52Bits; + } + + felem_reduce_carry(out, carry); +} + +/* felem_is_zero_vartime returns 1 iff |in| == 0. It takes a variable amount of + * time depending on the value of |in|. */ +static char felem_is_zero_vartime(const felem in) { + limb carry; + int i; + limb tmp[NLIMBS]; + + felem_assign(tmp, in); + + /* First, reduce tmp to a minimal form. */ + do { + carry = 0; + for (i = 0;; i++) { + tmp[i] += carry; + carry = tmp[i] >> 51; + tmp[i] &= kBottom51Bits; + + i++; + if (i == NLIMBS) + break; + + tmp[i] += carry; + carry = tmp[i] >> 52; + tmp[i] &= kBottom52Bits; + } + + felem_reduce_carry(tmp, carry); + } while (carry); + + /* tmp < 2**257, so the only possible zero values are 0, p and 2p. */ + return memcmp(tmp, kZero, sizeof(tmp)) == 0 || + memcmp(tmp, kP, sizeof(tmp)) == 0 || + memcmp(tmp, k2P, sizeof(tmp)) == 0; +} + + +/* Montgomery operations: */ + +#define kRDigits {2, 0xfffffffe00000000, 0xffffffffffffffff, 0x1fffffffd} // 2^257 mod p256.p + +#define kRInvDigits {0x180000000, 0xffffffff, 0xfffffffe80000001, 0x7fffffff00000001} // 1 / 2^257 mod p256.p + +static const cryptonite_p256_int kR = { kRDigits }; +static const cryptonite_p256_int kRInv = { kRInvDigits }; + +/* to_montgomery sets out = R*in. */ +static void to_montgomery(felem out, const cryptonite_p256_int* in) { + cryptonite_p256_int in_shifted; + int i; + + cryptonite_p256_init(&in_shifted); + cryptonite_p256_modmul(&cryptonite_SECP256r1_p, in, 0, &kR, &in_shifted); + + for (i = 0; i < NLIMBS; i++) { + if ((i & 1) == 0) { + out[i] = P256_DIGIT(&in_shifted, 0) & kBottom51Bits; + cryptonite_p256_shr(&in_shifted, 51, &in_shifted); + } else { + out[i] = P256_DIGIT(&in_shifted, 0) & kBottom52Bits; + cryptonite_p256_shr(&in_shifted, 52, &in_shifted); + } + } + + cryptonite_p256_clear(&in_shifted); +} + +/* from_montgomery sets out=in/R. */ +static void from_montgomery(cryptonite_p256_int* out, const felem in) { + cryptonite_p256_int result, tmp; + int i, top; + + cryptonite_p256_init(&result); + cryptonite_p256_init(&tmp); + + cryptonite_p256_add_d(&tmp, in[NLIMBS - 1], &result); + for (i = NLIMBS - 2; i >= 0; i--) { + if ((i & 1) == 0) { + top = cryptonite_p256_shl(&result, 51, &tmp); + } else { + top = cryptonite_p256_shl(&result, 52, &tmp); + } + top += cryptonite_p256_add_d(&tmp, in[i], &result); + } + + cryptonite_p256_modmul(&cryptonite_SECP256r1_p, &kRInv, top, &result, out); + + cryptonite_p256_clear(&result); + cryptonite_p256_clear(&tmp); +} diff --git a/cbits/p256/p256.c b/cbits/p256/p256.c --- a/cbits/p256/p256.c +++ b/cbits/p256/p256.c @@ -25,7 +25,7 @@ */ // This is an implementation of the P256 elliptic curve group. It's written to -// be portable 32-bit, although it's still constant-time. +// be portable and still constant-time. // // WARNING: Implementing these functions in a constant-time manner is far from // obvious. Be careful when touching this code. @@ -40,14 +40,16 @@ #include "p256/p256.h" const cryptonite_p256_int cryptonite_SECP256r1_n = // curve order - {{0xfc632551, 0xf3b9cac2, 0xa7179e84, 0xbce6faad, -1, -1, 0, -1}}; + {{P256_LITERAL(0xfc632551, 0xf3b9cac2), P256_LITERAL(0xa7179e84, 0xbce6faad), + P256_LITERAL(-1, -1), P256_LITERAL(0, -1)}}; const cryptonite_p256_int cryptonite_SECP256r1_p = // curve field size - {{-1, -1, -1, 0, 0, 0, 1, -1 }}; + {{P256_LITERAL(-1, -1), P256_LITERAL(-1, 0), + P256_LITERAL(0, 0), P256_LITERAL(1, -1) }}; const cryptonite_p256_int cryptonite_SECP256r1_b = // curve b - {{0x27d2604b, 0x3bce3c3e, 0xcc53b0f6, 0x651d06b0, - 0x769886bc, 0xb3ebbd55, 0xaa3a93e7, 0x5ac635d8}}; + {{P256_LITERAL(0x27d2604b, 0x3bce3c3e), P256_LITERAL(0xcc53b0f6, 0x651d06b0), + P256_LITERAL(0x769886bc, 0xb3ebbd55), P256_LITERAL(0xaa3a93e7, 0x5ac635d8)}}; void cryptonite_p256_init(cryptonite_p256_int* a) { memset(a, 0, sizeof(*a)); @@ -61,9 +63,10 @@ } int cryptonite_p256_is_zero(const cryptonite_p256_int* a) { - int i, result = 0; + cryptonite_p256_digit result = 0; + int i = 0; for (i = 0; i < P256_NDIGITS; ++i) result |= P256_DIGIT(a, i); - return !result; + return result == 0; } // top, c[] += a[] * b @@ -167,6 +170,10 @@ // top can be any value at this point. // Guestimate reducer as top * MOD, since msw of MOD is -1. top_reducer = mulAdd(MOD, top, 0, reducer); +#if P256_BITSPERDIGIT > 32 + // Correction when msw of MOD has only high 32 bits set + top_reducer += mulAdd(MOD, top >> 32, 0, reducer); +#endif // Subtract reducer from top | tmp. top = subTop(top_reducer, reducer, top, tmp + i); @@ -229,7 +236,7 @@ P256_DIGIT(b, i) = accu; } P256_DIGIT(b, i) = (P256_DIGIT(a, i) >> 1) | - (highbit << (P256_BITSPERDIGIT - 1)); + (((cryptonite_p256_sdigit) highbit) << (P256_BITSPERDIGIT - 1)); } // Return -1, 0, 1 for a < b, a == b or a > b respectively. @@ -359,31 +366,32 @@ } void cryptonite_p256_from_bin(const uint8_t src[P256_NBYTES], cryptonite_p256_int* dst) { - int i; + int i, n; const uint8_t* p = &src[0]; for (i = P256_NDIGITS - 1; i >= 0; --i) { - P256_DIGIT(dst, i) = - (p[0] << 24) | - (p[1] << 16) | - (p[2] << 8) | - p[3]; - p += 4; + cryptonite_p256_digit dig = 0; + n = P256_BITSPERDIGIT; + while (n > 0) { + n -= 8; + dig |= ((cryptonite_p256_digit) *(p++)) << n; + } + P256_DIGIT(dst, i) = dig; } } void cryptonite_p256_to_bin(const cryptonite_p256_int* src, uint8_t dst[P256_NBYTES]) { - int i; + int i, n; uint8_t* p = &dst[0]; for (i = P256_NDIGITS -1; i >= 0; --i) { const cryptonite_p256_digit dig = P256_DIGIT(src, i); - p[0] = dig >> 24; - p[1] = dig >> 16; - p[2] = dig >> 8; - p[3] = dig; - p += 4; + n = P256_BITSPERDIGIT; + while (n > 0) { + n -= 8; + *(p++) = dig >> n; + } } } @@ -395,6 +403,7 @@ // c = a + b mod MOD void cryptonite_p256e_modadd(const cryptonite_p256_int* MOD, const cryptonite_p256_int* a, const cryptonite_p256_int* b, cryptonite_p256_int* c) { + assert(c); /* avoid repeated checks inside inlined cryptonite_p256_add */ cryptonite_p256_digit top = cryptonite_p256_add(a, b, c); top = subM(MOD, top, P256_DIGITS(c), -1); top = subM(MOD, top, P256_DIGITS(c), MSB_COMPLEMENT(top)); @@ -403,8 +412,117 @@ // c = a - b mod MOD void cryptonite_p256e_modsub(const cryptonite_p256_int* MOD, const cryptonite_p256_int* a, const cryptonite_p256_int* b, cryptonite_p256_int* c) { + assert(c); /* avoid repeated checks inside inlined cryptonite_p256_sub */ cryptonite_p256_digit top = cryptonite_p256_sub(a, b, c); top = addM(MOD, top, P256_DIGITS(c), ~MSB_COMPLEMENT(top)); top = subM(MOD, top, P256_DIGITS(c), MSB_COMPLEMENT(top)); addM(MOD, 0, P256_DIGITS(c), top); +} + +#define NTH_DOUBLE_THEN_ADD(i, a, nth, b, out) \ + cryptonite_p256e_montmul(a, a, out); \ + for (i = 1; i < nth; i++) \ + cryptonite_p256e_montmul(out, out, out); \ + cryptonite_p256e_montmul(out, b, out); + +const cryptonite_p256_int cryptonite_SECP256r1_r2 = // r^2 mod n + {{P256_LITERAL(0xBE79EEA2, 0x83244C95), P256_LITERAL(0x49BD6FA6, 0x4699799C), + P256_LITERAL(0x2B6BEC59, 0x2845B239), P256_LITERAL(0xF3D95620, 0x66E12D94)}}; + +const cryptonite_p256_int cryptonite_SECP256r1_one = {{1}}; + +// Montgomery multiplication, i.e. c = ab/r mod n with r = 2^256. +// Implementation is adapted from 'sc_montmul' in libdecaf. +static void cryptonite_p256e_montmul(const cryptonite_p256_int* a, const cryptonite_p256_int* b, cryptonite_p256_int* c) { + int i, j, borrow; + cryptonite_p256_digit accum[P256_NDIGITS+1] = {0}; + cryptonite_p256_digit hi_carry = 0; + + for (i=0; i>= P256_BITSPERDIGIT; + } + accum[j] = chain; + + mand = accum[0] * P256_MONTGOMERY_FACTOR; + chain = 0; + mier = P256_DIGITS(&cryptonite_SECP256r1_n); + for (j=0; j>= P256_BITSPERDIGIT; + } + chain += accum[j]; + chain += hi_carry; + accum[j-1] = chain; + hi_carry = chain >> P256_BITSPERDIGIT; + } + + memcpy(P256_DIGITS(c), accum, sizeof(*c)); + borrow = cryptonite_p256_sub(c, &cryptonite_SECP256r1_n, c); + addM(&cryptonite_SECP256r1_n, 0, P256_DIGITS(c), borrow + hi_carry); +} + +// b = 1/a mod n, using Fermat's little theorem. +void cryptonite_p256e_scalar_invert(const cryptonite_p256_int* a, cryptonite_p256_int* b) { + cryptonite_p256_int _1, _10, _11, _101, _111, _1010, _1111; + cryptonite_p256_int _10101, _101010, _101111, x6, x8, x16, x32; + int i; + + // Montgomerize + cryptonite_p256e_montmul(a, &cryptonite_SECP256r1_r2, &_1); + + // P-256 (secp256r1) Scalar Inversion + // + cryptonite_p256e_montmul(&_1 , &_1 , &_10); + cryptonite_p256e_montmul(&_10 , &_1 , &_11); + cryptonite_p256e_montmul(&_10 , &_11 , &_101); + cryptonite_p256e_montmul(&_10 , &_101 , &_111); + cryptonite_p256e_montmul(&_101 , &_101 , &_1010); + cryptonite_p256e_montmul(&_101 , &_1010 , &_1111); + NTH_DOUBLE_THEN_ADD(i, &_1010, 1 , &_1 , &_10101); + cryptonite_p256e_montmul(&_10101 , &_10101 , &_101010); + cryptonite_p256e_montmul(&_101 , &_101010, &_101111); + cryptonite_p256e_montmul(&_10101 , &_101010, &x6); + NTH_DOUBLE_THEN_ADD(i, &x6 , 2 , &_11 , &x8); + NTH_DOUBLE_THEN_ADD(i, &x8 , 8 , &x8 , &x16); + NTH_DOUBLE_THEN_ADD(i, &x16 , 16 , &x16 , &x32); + + NTH_DOUBLE_THEN_ADD(i, &x32 , 32+32, &x32 , b); + NTH_DOUBLE_THEN_ADD(i, b , 32, &x32 , b); + NTH_DOUBLE_THEN_ADD(i, b , 6, &_101111, b); + NTH_DOUBLE_THEN_ADD(i, b , 2 + 3, &_111 , b); + NTH_DOUBLE_THEN_ADD(i, b , 2 + 2, &_11 , b); + NTH_DOUBLE_THEN_ADD(i, b , 1 + 4, &_1111 , b); + NTH_DOUBLE_THEN_ADD(i, b , 5, &_10101 , b); + NTH_DOUBLE_THEN_ADD(i, b , 1 + 3, &_101 , b); + NTH_DOUBLE_THEN_ADD(i, b , 3, &_101 , b); + NTH_DOUBLE_THEN_ADD(i, b , 3, &_101 , b); + NTH_DOUBLE_THEN_ADD(i, b , 2 + 3, &_111 , b); + NTH_DOUBLE_THEN_ADD(i, b , 3 + 6, &_101111, b); + NTH_DOUBLE_THEN_ADD(i, b , 2 + 4, &_1111 , b); + NTH_DOUBLE_THEN_ADD(i, b , 1 + 1, &_1 , b); + NTH_DOUBLE_THEN_ADD(i, b , 4 + 1, &_1 , b); + NTH_DOUBLE_THEN_ADD(i, b , 2 + 4, &_1111 , b); + NTH_DOUBLE_THEN_ADD(i, b , 2 + 3, &_111 , b); + NTH_DOUBLE_THEN_ADD(i, b , 1 + 3, &_111 , b); + NTH_DOUBLE_THEN_ADD(i, b , 2 + 3, &_111 , b); + NTH_DOUBLE_THEN_ADD(i, b , 2 + 3, &_101 , b); + NTH_DOUBLE_THEN_ADD(i, b , 1 + 2, &_11 , b); + NTH_DOUBLE_THEN_ADD(i, b , 4 + 6, &_101111, b); + NTH_DOUBLE_THEN_ADD(i, b , 2, &_11 , b); + NTH_DOUBLE_THEN_ADD(i, b , 3 + 2, &_11 , b); + NTH_DOUBLE_THEN_ADD(i, b , 3 + 2, &_11 , b); + NTH_DOUBLE_THEN_ADD(i, b , 2 + 1, &_1 , b); + NTH_DOUBLE_THEN_ADD(i, b , 2 + 5, &_10101 , b); + NTH_DOUBLE_THEN_ADD(i, b , 2 + 4, &_1111 , b); + + // Demontgomerize + cryptonite_p256e_montmul(b, &cryptonite_SECP256r1_one, b); } diff --git a/cbits/p256/p256.h b/cbits/p256/p256.h deleted file mode 100644 --- a/cbits/p256/p256.h +++ /dev/null @@ -1,162 +0,0 @@ -/* - * Copyright 2013 The Android Open Source Project - * - * Redistribution and use in source and binary forms, with or without - * modification, are permitted provided that the following conditions are met: - * * Redistributions of source code must retain the above copyright - * notice, this list of conditions and the following disclaimer. - * * Redistributions in binary form must reproduce the above copyright - * notice, this list of conditions and the following disclaimer in the - * documentation and/or other materials provided with the distribution. - * * Neither the name of Google Inc. nor the names of its contributors may - * be used to endorse or promote products derived from this software - * without specific prior written permission. - * - * THIS SOFTWARE IS PROVIDED BY Google Inc. ``AS IS'' AND ANY EXPRESS OR - * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF - * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO - * EVENT SHALL Google Inc. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, - * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, - * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; - * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, - * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR - * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF - * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - */ - -#ifndef SYSTEM_CORE_INCLUDE_MINCRYPT_LITE_P256_H_ -#define SYSTEM_CORE_INCLUDE_MINCRYPT_LITE_P256_H_ - -// Collection of routines manipulating 256 bit unsigned integers. -// Just enough to implement ecdsa-p256 and related algorithms. - -#include - -#ifdef __cplusplus -extern "C" { -#endif - -#define P256_BITSPERDIGIT 32 -#define P256_NDIGITS 8 -#define P256_NBYTES 32 - -typedef int cryptonite_p256_err; -typedef uint32_t cryptonite_p256_digit; -typedef int32_t cryptonite_p256_sdigit; -typedef uint64_t cryptonite_p256_ddigit; -typedef int64_t cryptonite_p256_sddigit; - -// Defining cryptonite_p256_int as struct to leverage struct assigment. -typedef struct { - cryptonite_p256_digit a[P256_NDIGITS]; -} cryptonite_p256_int; - -extern const cryptonite_p256_int cryptonite_SECP256r1_n; // Curve order -extern const cryptonite_p256_int cryptonite_SECP256r1_p; // Curve prime -extern const cryptonite_p256_int cryptonite_SECP256r1_b; // Curve param - -// Initialize a cryptonite_p256_int to zero. -void cryptonite_p256_init(cryptonite_p256_int* a); - -// Clear a cryptonite_p256_int to zero. -void cryptonite_p256_clear(cryptonite_p256_int* a); - -// Return bit. Index 0 is least significant. -int cryptonite_p256_get_bit(const cryptonite_p256_int* a, int index); - -// b := a % MOD -void cryptonite_p256_mod( - const cryptonite_p256_int* MOD, - const cryptonite_p256_int* a, - cryptonite_p256_int* b); - -// c := a * (top_b | b) % MOD -void cryptonite_p256_modmul( - const cryptonite_p256_int* MOD, - const cryptonite_p256_int* a, - const cryptonite_p256_digit top_b, - const cryptonite_p256_int* b, - cryptonite_p256_int* c); - -// b := 1 / a % MOD -// MOD best be SECP256r1_n -void cryptonite_p256_modinv( - const cryptonite_p256_int* MOD, - const cryptonite_p256_int* a, - cryptonite_p256_int* b); - -// b := 1 / a % MOD -// MOD best be SECP256r1_n -// Faster than cryptonite_p256_modinv() -void cryptonite_p256_modinv_vartime( - const cryptonite_p256_int* MOD, - const cryptonite_p256_int* a, - cryptonite_p256_int* b); - -// b := a << (n % P256_BITSPERDIGIT) -// Returns the bits shifted out of most significant digit. -cryptonite_p256_digit cryptonite_p256_shl(const cryptonite_p256_int* a, int n, cryptonite_p256_int* b); - -// b := a >> (n % P256_BITSPERDIGIT) -void cryptonite_p256_shr(const cryptonite_p256_int* a, int n, cryptonite_p256_int* b); - -int cryptonite_p256_is_zero(const cryptonite_p256_int* a); -int cryptonite_p256_is_odd(const cryptonite_p256_int* a); -int cryptonite_p256_is_even(const cryptonite_p256_int* a); - -// Returns -1, 0 or 1. -int cryptonite_p256_cmp(const cryptonite_p256_int* a, const cryptonite_p256_int *b); - -// c: = a - b -// Returns -1 on borrow. -int cryptonite_p256_sub(const cryptonite_p256_int* a, const cryptonite_p256_int* b, cryptonite_p256_int* c); - -// c := a + b -// Returns 1 on carry. -int cryptonite_p256_add(const cryptonite_p256_int* a, const cryptonite_p256_int* b, cryptonite_p256_int* c); - -// c := a + (single digit)b -// Returns carry 1 on carry. -int cryptonite_p256_add_d(const cryptonite_p256_int* a, cryptonite_p256_digit b, cryptonite_p256_int* c); - -// ec routines. - -// {out_x,out_y} := nG -void cryptonite_p256_base_point_mul(const cryptonite_p256_int *n, - cryptonite_p256_int *out_x, - cryptonite_p256_int *out_y); - -// {out_x,out_y} := n{in_x,in_y} -void cryptonite_p256_point_mul(const cryptonite_p256_int *n, - const cryptonite_p256_int *in_x, - const cryptonite_p256_int *in_y, - cryptonite_p256_int *out_x, - cryptonite_p256_int *out_y); - -// {out_x,out_y} := n1G + n2{in_x,in_y} -void cryptonite_p256_points_mul_vartime( - const cryptonite_p256_int *n1, const cryptonite_p256_int *n2, - const cryptonite_p256_int *in_x, const cryptonite_p256_int *in_y, - cryptonite_p256_int *out_x, cryptonite_p256_int *out_y); - -// Return whether point {x,y} is on curve. -int cryptonite_p256_is_valid_point(const cryptonite_p256_int* x, const cryptonite_p256_int* y); - -// Outputs big-endian binary form. No leading zero skips. -void cryptonite_p256_to_bin(const cryptonite_p256_int* src, uint8_t dst[P256_NBYTES]); - -// Reads from big-endian binary form, -// thus pre-pad with leading zeros if short. -void cryptonite_p256_from_bin(const uint8_t src[P256_NBYTES], cryptonite_p256_int* dst); - -#define P256_DIGITS(x) ((x)->a) -#define P256_DIGIT(x,y) ((x)->a[y]) - -#define P256_ZERO {{0}} -#define P256_ONE {{1}} - -#ifdef __cplusplus -} -#endif - -#endif // SYSTEM_CORE_INCLUDE_MINCRYPT_LITE_P256_H_ diff --git a/cbits/p256/p256_ec.c b/cbits/p256/p256_ec.c --- a/cbits/p256/p256_ec.c +++ b/cbits/p256/p256_ec.c @@ -25,580 +25,18 @@ */ // This is an implementation of the P256 elliptic curve group. It's written to -// be portable 32-bit, although it's still constant-time. +// be portable and still constant-time. // // WARNING: Implementing these functions in a constant-time manner is far from // obvious. Be careful when touching this code. // // See http://www.imperialviolet.org/2010/12/04/ecc.html ([1]) for background. -#include -#include - -#include -#include - -#include "p256/p256.h" - -typedef uint8_t u8; -typedef uint32_t u32; -typedef int32_t s32; -typedef uint64_t u64; - -/* Our field elements are represented as nine 32-bit limbs. - * - * The value of an felem (field element) is: - * x[0] + (x[1] * 2**29) + (x[2] * 2**57) + ... + (x[8] * 2**228) - * - * That is, each limb is alternately 29 or 28-bits wide in little-endian - * order. - * - * This means that an felem hits 2**257, rather than 2**256 as we would like. A - * 28, 29, ... pattern would cause us to hit 2**256, but that causes problems - * when multiplying as terms end up one bit short of a limb which would require - * much bit-shifting to correct. - * - * Finally, the values stored in an felem are in Montgomery form. So the value - * |y| is stored as (y*R) mod p, where p is the P-256 prime and R is 2**257. - */ -typedef u32 limb; -#define NLIMBS 9 -typedef limb felem[NLIMBS]; - -static const limb kBottom28Bits = 0xfffffff; -static const limb kBottom29Bits = 0x1fffffff; - -/* kOne is the number 1 as an felem. It's 2**257 mod p split up into 29 and - * 28-bit words. */ -static const felem kOne = { - 2, 0, 0, 0xffff800, - 0x1fffffff, 0xfffffff, 0x1fbfffff, 0x1ffffff, - 0 -}; -static const felem kZero = {0}; -static const felem kP = { - 0x1fffffff, 0xfffffff, 0x1fffffff, 0x3ff, - 0, 0, 0x200000, 0xf000000, - 0xfffffff -}; -static const felem k2P = { - 0x1ffffffe, 0xfffffff, 0x1fffffff, 0x7ff, - 0, 0, 0x400000, 0xe000000, - 0x1fffffff -}; -/* kPrecomputed contains precomputed values to aid the calculation of scalar - * multiples of the base point, G. It's actually two, equal length, tables - * concatenated. - * - * The first table contains (x,y) felem pairs for 16 multiples of the base - * point, G. - * - * Index | Index (binary) | Value - * 0 | 0000 | 0G (all zeros, omitted) - * 1 | 0001 | G - * 2 | 0010 | 2**64G - * 3 | 0011 | 2**64G + G - * 4 | 0100 | 2**128G - * 5 | 0101 | 2**128G + G - * 6 | 0110 | 2**128G + 2**64G - * 7 | 0111 | 2**128G + 2**64G + G - * 8 | 1000 | 2**192G - * 9 | 1001 | 2**192G + G - * 10 | 1010 | 2**192G + 2**64G - * 11 | 1011 | 2**192G + 2**64G + G - * 12 | 1100 | 2**192G + 2**128G - * 13 | 1101 | 2**192G + 2**128G + G - * 14 | 1110 | 2**192G + 2**128G + 2**64G - * 15 | 1111 | 2**192G + 2**128G + 2**64G + G - * - * The second table follows the same style, but the terms are 2**32G, - * 2**96G, 2**160G, 2**224G. - * - * This is ~2KB of data. */ -static const limb kPrecomputed[NLIMBS * 2 * 15 * 2] = { - 0x11522878, 0xe730d41, 0xdb60179, 0x4afe2ff, 0x12883add, 0xcaddd88, 0x119e7edc, 0xd4a6eab, 0x3120bee, - 0x1d2aac15, 0xf25357c, 0x19e45cdd, 0x5c721d0, 0x1992c5a5, 0xa237487, 0x154ba21, 0x14b10bb, 0xae3fe3, - 0xd41a576, 0x922fc51, 0x234994f, 0x60b60d3, 0x164586ae, 0xce95f18, 0x1fe49073, 0x3fa36cc, 0x5ebcd2c, - 0xb402f2f, 0x15c70bf, 0x1561925c, 0x5a26704, 0xda91e90, 0xcdc1c7f, 0x1ea12446, 0xe1ade1e, 0xec91f22, - 0x26f7778, 0x566847e, 0xa0bec9e, 0x234f453, 0x1a31f21a, 0xd85e75c, 0x56c7109, 0xa267a00, 0xb57c050, - 0x98fb57, 0xaa837cc, 0x60c0792, 0xcfa5e19, 0x61bab9e, 0x589e39b, 0xa324c5, 0x7d6dee7, 0x2976e4b, - 0x1fc4124a, 0xa8c244b, 0x1ce86762, 0xcd61c7e, 0x1831c8e0, 0x75774e1, 0x1d96a5a9, 0x843a649, 0xc3ab0fa, - 0x6e2e7d5, 0x7673a2a, 0x178b65e8, 0x4003e9b, 0x1a1f11c2, 0x7816ea, 0xf643e11, 0x58c43df, 0xf423fc2, - 0x19633ffa, 0x891f2b2, 0x123c231c, 0x46add8c, 0x54700dd, 0x59e2b17, 0x172db40f, 0x83e277d, 0xb0dd609, - 0xfd1da12, 0x35c6e52, 0x19ede20c, 0xd19e0c0, 0x97d0f40, 0xb015b19, 0x449e3f5, 0xe10c9e, 0x33ab581, - 0x56a67ab, 0x577734d, 0x1dddc062, 0xc57b10d, 0x149b39d, 0x26a9e7b, 0xc35df9f, 0x48764cd, 0x76dbcca, - 0xca4b366, 0xe9303ab, 0x1a7480e7, 0x57e9e81, 0x1e13eb50, 0xf466cf3, 0x6f16b20, 0x4ba3173, 0xc168c33, - 0x15cb5439, 0x6a38e11, 0x73658bd, 0xb29564f, 0x3f6dc5b, 0x53b97e, 0x1322c4c0, 0x65dd7ff, 0x3a1e4f6, - 0x14e614aa, 0x9246317, 0x1bc83aca, 0xad97eed, 0xd38ce4a, 0xf82b006, 0x341f077, 0xa6add89, 0x4894acd, - 0x9f162d5, 0xf8410ef, 0x1b266a56, 0xd7f223, 0x3e0cb92, 0xe39b672, 0x6a2901a, 0x69a8556, 0x7e7c0, - 0x9b7d8d3, 0x309a80, 0x1ad05f7f, 0xc2fb5dd, 0xcbfd41d, 0x9ceb638, 0x1051825c, 0xda0cf5b, 0x812e881, - 0x6f35669, 0x6a56f2c, 0x1df8d184, 0x345820, 0x1477d477, 0x1645db1, 0xbe80c51, 0xc22be3e, 0xe35e65a, - 0x1aeb7aa0, 0xc375315, 0xf67bc99, 0x7fdd7b9, 0x191fc1be, 0x61235d, 0x2c184e9, 0x1c5a839, 0x47a1e26, - 0xb7cb456, 0x93e225d, 0x14f3c6ed, 0xccc1ac9, 0x17fe37f3, 0x4988989, 0x1a90c502, 0x2f32042, 0xa17769b, - 0xafd8c7c, 0x8191c6e, 0x1dcdb237, 0x16200c0, 0x107b32a1, 0x66c08db, 0x10d06a02, 0x3fc93, 0x5620023, - 0x16722b27, 0x68b5c59, 0x270fcfc, 0xfad0ecc, 0xe5de1c2, 0xeab466b, 0x2fc513c, 0x407f75c, 0xbaab133, - 0x9705fe9, 0xb88b8e7, 0x734c993, 0x1e1ff8f, 0x19156970, 0xabd0f00, 0x10469ea7, 0x3293ac0, 0xcdc98aa, - 0x1d843fd, 0xe14bfe8, 0x15be825f, 0x8b5212, 0xeb3fb67, 0x81cbd29, 0xbc62f16, 0x2b6fcc7, 0xf5a4e29, - 0x13560b66, 0xc0b6ac2, 0x51ae690, 0xd41e271, 0xf3e9bd4, 0x1d70aab, 0x1029f72, 0x73e1c35, 0xee70fbc, - 0xad81baf, 0x9ecc49a, 0x86c741e, 0xfe6be30, 0x176752e7, 0x23d416, 0x1f83de85, 0x27de188, 0x66f70b8, - 0x181cd51f, 0x96b6e4c, 0x188f2335, 0xa5df759, 0x17a77eb6, 0xfeb0e73, 0x154ae914, 0x2f3ec51, 0x3826b59, - 0xb91f17d, 0x1c72949, 0x1362bf0a, 0xe23fddf, 0xa5614b0, 0xf7d8f, 0x79061, 0x823d9d2, 0x8213f39, - 0x1128ae0b, 0xd095d05, 0xb85c0c2, 0x1ecb2ef, 0x24ddc84, 0xe35e901, 0x18411a4a, 0xf5ddc3d, 0x3786689, - 0x52260e8, 0x5ae3564, 0x542b10d, 0x8d93a45, 0x19952aa4, 0x996cc41, 0x1051a729, 0x4be3499, 0x52b23aa, - 0x109f307e, 0x6f5b6bb, 0x1f84e1e7, 0x77a0cfa, 0x10c4df3f, 0x25a02ea, 0xb048035, 0xe31de66, 0xc6ecaa3, - 0x28ea335, 0x2886024, 0x1372f020, 0xf55d35, 0x15e4684c, 0xf2a9e17, 0x1a4a7529, 0xcb7beb1, 0xb2a78a1, - 0x1ab21f1f, 0x6361ccf, 0x6c9179d, 0xb135627, 0x1267b974, 0x4408bad, 0x1cbff658, 0xe3d6511, 0xc7d76f, - 0x1cc7a69, 0xe7ee31b, 0x54fab4f, 0x2b914f, 0x1ad27a30, 0xcd3579e, 0xc50124c, 0x50daa90, 0xb13f72, - 0xb06aa75, 0x70f5cc6, 0x1649e5aa, 0x84a5312, 0x329043c, 0x41c4011, 0x13d32411, 0xb04a838, 0xd760d2d, - 0x1713b532, 0xbaa0c03, 0x84022ab, 0x6bcf5c1, 0x2f45379, 0x18ae070, 0x18c9e11e, 0x20bca9a, 0x66f496b, - 0x3eef294, 0x67500d2, 0xd7f613c, 0x2dbbeb, 0xb741038, 0xe04133f, 0x1582968d, 0xbe985f7, 0x1acbc1a, - 0x1a6a939f, 0x33e50f6, 0xd665ed4, 0xb4b7bd6, 0x1e5a3799, 0x6b33847, 0x17fa56ff, 0x65ef930, 0x21dc4a, - 0x2b37659, 0x450fe17, 0xb357b65, 0xdf5efac, 0x15397bef, 0x9d35a7f, 0x112ac15f, 0x624e62e, 0xa90ae2f, - 0x107eecd2, 0x1f69bbe, 0x77d6bce, 0x5741394, 0x13c684fc, 0x950c910, 0x725522b, 0xdc78583, 0x40eeabb, - 0x1fde328a, 0xbd61d96, 0xd28c387, 0x9e77d89, 0x12550c40, 0x759cb7d, 0x367ef34, 0xae2a960, 0x91b8bdc, - 0x93462a9, 0xf469ef, 0xb2e9aef, 0xd2ca771, 0x54e1f42, 0x7aaa49, 0x6316abb, 0x2413c8e, 0x5425bf9, - 0x1bed3e3a, 0xf272274, 0x1f5e7326, 0x6416517, 0xea27072, 0x9cedea7, 0x6e7633, 0x7c91952, 0xd806dce, - 0x8e2a7e1, 0xe421e1a, 0x418c9e1, 0x1dbc890, 0x1b395c36, 0xa1dc175, 0x1dc4ef73, 0x8956f34, 0xe4b5cf2, - 0x1b0d3a18, 0x3194a36, 0x6c2641f, 0xe44124c, 0xa2f4eaa, 0xa8c25ba, 0xf927ed7, 0x627b614, 0x7371cca, - 0xba16694, 0x417bc03, 0x7c0a7e3, 0x9c35c19, 0x1168a205, 0x8b6b00d, 0x10e3edc9, 0x9c19bf2, 0x5882229, - 0x1b2b4162, 0xa5cef1a, 0x1543622b, 0x9bd433e, 0x364e04d, 0x7480792, 0x5c9b5b3, 0xe85ff25, 0x408ef57, - 0x1814cfa4, 0x121b41b, 0xd248a0f, 0x3b05222, 0x39bb16a, 0xc75966d, 0xa038113, 0xa4a1769, 0x11fbc6c, - 0x917e50e, 0xeec3da8, 0x169d6eac, 0x10c1699, 0xa416153, 0xf724912, 0x15cd60b7, 0x4acbad9, 0x5efc5fa, - 0xf150ed7, 0x122b51, 0x1104b40a, 0xcb7f442, 0xfbb28ff, 0x6ac53ca, 0x196142cc, 0x7bf0fa9, 0x957651, - 0x4e0f215, 0xed439f8, 0x3f46bd5, 0x5ace82f, 0x110916b6, 0x6db078, 0xffd7d57, 0xf2ecaac, 0xca86dec, - 0x15d6b2da, 0x965ecc9, 0x1c92b4c2, 0x1f3811, 0x1cb080f5, 0x2d8b804, 0x19d1c12d, 0xf20bd46, 0x1951fa7, - 0xa3656c3, 0x523a425, 0xfcd0692, 0xd44ddc8, 0x131f0f5b, 0xaf80e4a, 0xcd9fc74, 0x99bb618, 0x2db944c, - 0xa673090, 0x1c210e1, 0x178c8d23, 0x1474383, 0x10b8743d, 0x985a55b, 0x2e74779, 0x576138, 0x9587927, - 0x133130fa, 0xbe05516, 0x9f4d619, 0xbb62570, 0x99ec591, 0xd9468fe, 0x1d07782d, 0xfc72e0b, 0x701b298, - 0x1863863b, 0x85954b8, 0x121a0c36, 0x9e7fedf, 0xf64b429, 0x9b9d71e, 0x14e2f5d8, 0xf858d3a, 0x942eea8, - 0xda5b765, 0x6edafff, 0xa9d18cc, 0xc65e4ba, 0x1c747e86, 0xe4ea915, 0x1981d7a1, 0x8395659, 0x52ed4e2, - 0x87d43b7, 0x37ab11b, 0x19d292ce, 0xf8d4692, 0x18c3053f, 0x8863e13, 0x4c146c0, 0x6bdf55a, 0x4e4457d, - 0x16152289, 0xac78ec2, 0x1a59c5a2, 0x2028b97, 0x71c2d01, 0x295851f, 0x404747b, 0x878558d, 0x7d29aa4, - 0x13d8341f, 0x8daefd7, 0x139c972d, 0x6b7ea75, 0xd4a9dde, 0xff163d8, 0x81d55d7, 0xa5bef68, 0xb7b30d8, - 0xbe73d6f, 0xaa88141, 0xd976c81, 0x7e7a9cc, 0x18beb771, 0xd773cbd, 0x13f51951, 0x9d0c177, 0x1c49a78, -}; +#include "p256/p256_gf.h" /* Field element operations: */ -/* NON_ZERO_TO_ALL_ONES returns: - * 0xffffffff for 0 < x <= 2**31 - * 0 for x == 0 or x > 2**31. - * - * x must be a u32 or an equivalent type such as limb. */ -#define NON_ZERO_TO_ALL_ONES(x) ((((u32)(x) - 1) >> 31) - 1) - -/* felem_reduce_carry adds a multiple of p in order to cancel |carry|, - * which is a term at 2**257. - * - * On entry: carry < 2**3, inout[0,2,...] < 2**29, inout[1,3,...] < 2**28. - * On exit: inout[0,2,..] < 2**30, inout[1,3,...] < 2**29. */ -static void felem_reduce_carry(felem inout, limb carry) { - const u32 carry_mask = NON_ZERO_TO_ALL_ONES(carry); - - inout[0] += carry << 1; - inout[3] += 0x10000000 & carry_mask; - /* carry < 2**3 thus (carry << 11) < 2**14 and we added 2**28 in the - * previous line therefore this doesn't underflow. */ - inout[3] -= carry << 11; - inout[4] += (0x20000000 - 1) & carry_mask; - inout[5] += (0x10000000 - 1) & carry_mask; - inout[6] += (0x20000000 - 1) & carry_mask; - inout[6] -= carry << 22; - /* This may underflow if carry is non-zero but, if so, we'll fix it in the - * next line. */ - inout[7] -= 1 & carry_mask; - inout[7] += carry << 25; -} - -/* felem_sum sets out = in+in2. - * - * On entry, in[i]+in2[i] must not overflow a 32-bit word. - * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29 */ -static void felem_sum(felem out, const felem in, const felem in2) { - limb carry = 0; - unsigned i; - - for (i = 0;; i++) { - out[i] = in[i] + in2[i]; - out[i] += carry; - carry = out[i] >> 29; - out[i] &= kBottom29Bits; - - i++; - if (i == NLIMBS) - break; - - out[i] = in[i] + in2[i]; - out[i] += carry; - carry = out[i] >> 28; - out[i] &= kBottom28Bits; - } - - felem_reduce_carry(out, carry); -} - -#define two31m3 (((limb)1) << 31) - (((limb)1) << 3) -#define two30m2 (((limb)1) << 30) - (((limb)1) << 2) -#define two30p13m2 (((limb)1) << 30) + (((limb)1) << 13) - (((limb)1) << 2) -#define two31m2 (((limb)1) << 31) - (((limb)1) << 2) -#define two31p24m2 (((limb)1) << 31) + (((limb)1) << 24) - (((limb)1) << 2) -#define two30m27m2 (((limb)1) << 30) - (((limb)1) << 27) - (((limb)1) << 2) - -/* zero31 is 0 mod p. */ -static const felem zero31 = { two31m3, two30m2, two31m2, two30p13m2, two31m2, two30m2, two31p24m2, two30m27m2, two31m2 }; - -/* felem_diff sets out = in-in2. - * - * On entry: in[0,2,...] < 2**30, in[1,3,...] < 2**29 and - * in2[0,2,...] < 2**30, in2[1,3,...] < 2**29. - * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29. */ -static void felem_diff(felem out, const felem in, const felem in2) { - limb carry = 0; - unsigned i; - - for (i = 0;; i++) { - out[i] = in[i] - in2[i]; - out[i] += zero31[i]; - out[i] += carry; - carry = out[i] >> 29; - out[i] &= kBottom29Bits; - - i++; - if (i == NLIMBS) - break; - - out[i] = in[i] - in2[i]; - out[i] += zero31[i]; - out[i] += carry; - carry = out[i] >> 28; - out[i] &= kBottom28Bits; - } - - felem_reduce_carry(out, carry); -} - -/* felem_reduce_degree sets out = tmp/R mod p where tmp contains 64-bit words - * with the same 29,28,... bit positions as an felem. - * - * The values in felems are in Montgomery form: x*R mod p where R = 2**257. - * Since we just multiplied two Montgomery values together, the result is - * x*y*R*R mod p. We wish to divide by R in order for the result also to be - * in Montgomery form. - * - * On entry: tmp[i] < 2**64 - * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29 */ -static void felem_reduce_degree(felem out, u64 tmp[17]) { - /* The following table may be helpful when reading this code: - * - * Limb number: 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10... - * Width (bits): 29| 28| 29| 28| 29| 28| 29| 28| 29| 28| 29 - * Start bit: 0 | 29| 57| 86|114|143|171|200|228|257|285 - * (odd phase): 0 | 28| 57| 85|114|142|171|199|228|256|285 */ - limb tmp2[18], carry, x, xMask; - unsigned i; - - /* tmp contains 64-bit words with the same 29,28,29-bit positions as an - * felem. So the top of an element of tmp might overlap with another - * element two positions down. The following loop eliminates this - * overlap. */ - tmp2[0] = (limb)(tmp[0] & kBottom29Bits); - - /* In the following we use "(limb) tmp[x]" and "(limb) (tmp[x]>>32)" to try - * and hint to the compiler that it can do a single-word shift by selecting - * the right register rather than doing a double-word shift and truncating - * afterwards. */ - tmp2[1] = ((limb) tmp[0]) >> 29; - tmp2[1] |= (((limb)(tmp[0] >> 32)) << 3) & kBottom28Bits; - tmp2[1] += ((limb) tmp[1]) & kBottom28Bits; - carry = tmp2[1] >> 28; - tmp2[1] &= kBottom28Bits; - - for (i = 2; i < 17; i++) { - tmp2[i] = ((limb)(tmp[i - 2] >> 32)) >> 25; - tmp2[i] += ((limb)(tmp[i - 1])) >> 28; - tmp2[i] += (((limb)(tmp[i - 1] >> 32)) << 4) & kBottom29Bits; - tmp2[i] += ((limb) tmp[i]) & kBottom29Bits; - tmp2[i] += carry; - carry = tmp2[i] >> 29; - tmp2[i] &= kBottom29Bits; - - i++; - if (i == 17) - break; - tmp2[i] = ((limb)(tmp[i - 2] >> 32)) >> 25; - tmp2[i] += ((limb)(tmp[i - 1])) >> 29; - tmp2[i] += (((limb)(tmp[i - 1] >> 32)) << 3) & kBottom28Bits; - tmp2[i] += ((limb) tmp[i]) & kBottom28Bits; - tmp2[i] += carry; - carry = tmp2[i] >> 28; - tmp2[i] &= kBottom28Bits; - } - - tmp2[17] = ((limb)(tmp[15] >> 32)) >> 25; - tmp2[17] += ((limb)(tmp[16])) >> 29; - tmp2[17] += (((limb)(tmp[16] >> 32)) << 3); - tmp2[17] += carry; - - /* Montgomery elimination of terms. - * - * Since R is 2**257, we can divide by R with a bitwise shift if we can - * ensure that the right-most 257 bits are all zero. We can make that true by - * adding multiplies of p without affecting the value. - * - * So we eliminate limbs from right to left. Since the bottom 29 bits of p - * are all ones, then by adding tmp2[0]*p to tmp2 we'll make tmp2[0] == 0. - * We can do that for 8 further limbs and then right shift to eliminate the - * extra factor of R. */ - for (i = 0;; i += 2) { - tmp2[i + 1] += tmp2[i] >> 29; - x = tmp2[i] & kBottom29Bits; - xMask = NON_ZERO_TO_ALL_ONES(x); - tmp2[i] = 0; - - /* The bounds calculations for this loop are tricky. Each iteration of - * the loop eliminates two words by adding values to words to their - * right. - * - * The following table contains the amounts added to each word (as an - * offset from the value of i at the top of the loop). The amounts are - * accounted for from the first and second half of the loop separately - * and are written as, for example, 28 to mean a value <2**28. - * - * Word: 3 4 5 6 7 8 9 10 - * Added in top half: 28 11 29 21 29 28 - * 28 29 - * 29 - * Added in bottom half: 29 10 28 21 28 28 - * 29 - * - * The value that is currently offset 7 will be offset 5 for the next - * iteration and then offset 3 for the iteration after that. Therefore - * the total value added will be the values added at 7, 5 and 3. - * - * The following table accumulates these values. The sums at the bottom - * are written as, for example, 29+28, to mean a value < 2**29+2**28. - * - * Word: 3 4 5 6 7 8 9 10 11 12 13 - * 28 11 10 29 21 29 28 28 28 28 28 - * 29 28 11 28 29 28 29 28 29 28 - * 29 28 21 21 29 21 29 21 - * 10 29 28 21 28 21 28 - * 28 29 28 29 28 29 28 - * 11 10 29 10 29 10 - * 29 28 11 28 11 - * 29 29 - * -------------------------------------------- - * 30+ 31+ 30+ 31+ 30+ - * 28+ 29+ 28+ 29+ 21+ - * 21+ 28+ 21+ 28+ 10 - * 10 21+ 10 21+ - * 11 11 - * - * So the greatest amount is added to tmp2[10] and tmp2[12]. If - * tmp2[10/12] has an initial value of <2**29, then the maximum value - * will be < 2**31 + 2**30 + 2**28 + 2**21 + 2**11, which is < 2**32, - * as required. */ - tmp2[i + 3] += (x << 10) & kBottom28Bits; - tmp2[i + 4] += (x >> 18); - - tmp2[i + 6] += (x << 21) & kBottom29Bits; - tmp2[i + 7] += x >> 8; - - /* At position 200, which is the starting bit position for word 7, we - * have a factor of 0xf000000 = 2**28 - 2**24. */ - tmp2[i + 7] += 0x10000000 & xMask; - /* Word 7 is 28 bits wide, so the 2**28 term exactly hits word 8. */ - tmp2[i + 8] += (x - 1) & xMask; - tmp2[i + 7] -= (x << 24) & kBottom28Bits; - tmp2[i + 8] -= x >> 4; - - tmp2[i + 8] += 0x20000000 & xMask; - tmp2[i + 8] -= x; - tmp2[i + 8] += (x << 28) & kBottom29Bits; - tmp2[i + 9] += ((x >> 1) - 1) & xMask; - - if (i+1 == NLIMBS) - break; - tmp2[i + 2] += tmp2[i + 1] >> 28; - x = tmp2[i + 1] & kBottom28Bits; - xMask = NON_ZERO_TO_ALL_ONES(x); - tmp2[i + 1] = 0; - - tmp2[i + 4] += (x << 11) & kBottom29Bits; - tmp2[i + 5] += (x >> 18); - - tmp2[i + 7] += (x << 21) & kBottom28Bits; - tmp2[i + 8] += x >> 7; - - /* At position 199, which is the starting bit of the 8th word when - * dealing with a context starting on an odd word, we have a factor of - * 0x1e000000 = 2**29 - 2**25. Since we have not updated i, the 8th - * word from i+1 is i+8. */ - tmp2[i + 8] += 0x20000000 & xMask; - tmp2[i + 9] += (x - 1) & xMask; - tmp2[i + 8] -= (x << 25) & kBottom29Bits; - tmp2[i + 9] -= x >> 4; - - tmp2[i + 9] += 0x10000000 & xMask; - tmp2[i + 9] -= x; - tmp2[i + 10] += (x - 1) & xMask; - } - - /* We merge the right shift with a carry chain. The words above 2**257 have - * widths of 28,29,... which we need to correct when copying them down. */ - carry = 0; - for (i = 0; i < 8; i++) { - /* The maximum value of tmp2[i + 9] occurs on the first iteration and - * is < 2**30+2**29+2**28. Adding 2**29 (from tmp2[i + 10]) is - * therefore safe. */ - out[i] = tmp2[i + 9]; - out[i] += carry; - out[i] += (tmp2[i + 10] << 28) & kBottom29Bits; - carry = out[i] >> 29; - out[i] &= kBottom29Bits; - - i++; - out[i] = tmp2[i + 9] >> 1; - out[i] += carry; - carry = out[i] >> 28; - out[i] &= kBottom28Bits; - } - - out[8] = tmp2[17]; - out[8] += carry; - carry = out[8] >> 29; - out[8] &= kBottom29Bits; - - felem_reduce_carry(out, carry); -} - -/* felem_square sets out=in*in. - * - * On entry: in[0,2,...] < 2**30, in[1,3,...] < 2**29. - * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29. */ -static void felem_square(felem out, const felem in) { - u64 tmp[17]; - - tmp[0] = ((u64) in[0]) * in[0]; - tmp[1] = ((u64) in[0]) * (in[1] << 1); - tmp[2] = ((u64) in[0]) * (in[2] << 1) + - ((u64) in[1]) * (in[1] << 1); - tmp[3] = ((u64) in[0]) * (in[3] << 1) + - ((u64) in[1]) * (in[2] << 1); - tmp[4] = ((u64) in[0]) * (in[4] << 1) + - ((u64) in[1]) * (in[3] << 2) + ((u64) in[2]) * in[2]; - tmp[5] = ((u64) in[0]) * (in[5] << 1) + ((u64) in[1]) * - (in[4] << 1) + ((u64) in[2]) * (in[3] << 1); - tmp[6] = ((u64) in[0]) * (in[6] << 1) + ((u64) in[1]) * - (in[5] << 2) + ((u64) in[2]) * (in[4] << 1) + - ((u64) in[3]) * (in[3] << 1); - tmp[7] = ((u64) in[0]) * (in[7] << 1) + ((u64) in[1]) * - (in[6] << 1) + ((u64) in[2]) * (in[5] << 1) + - ((u64) in[3]) * (in[4] << 1); - /* tmp[8] has the greatest value of 2**61 + 2**60 + 2**61 + 2**60 + 2**60, - * which is < 2**64 as required. */ - tmp[8] = ((u64) in[0]) * (in[8] << 1) + ((u64) in[1]) * - (in[7] << 2) + ((u64) in[2]) * (in[6] << 1) + - ((u64) in[3]) * (in[5] << 2) + ((u64) in[4]) * in[4]; - tmp[9] = ((u64) in[1]) * (in[8] << 1) + ((u64) in[2]) * - (in[7] << 1) + ((u64) in[3]) * (in[6] << 1) + - ((u64) in[4]) * (in[5] << 1); - tmp[10] = ((u64) in[2]) * (in[8] << 1) + ((u64) in[3]) * - (in[7] << 2) + ((u64) in[4]) * (in[6] << 1) + - ((u64) in[5]) * (in[5] << 1); - tmp[11] = ((u64) in[3]) * (in[8] << 1) + ((u64) in[4]) * - (in[7] << 1) + ((u64) in[5]) * (in[6] << 1); - tmp[12] = ((u64) in[4]) * (in[8] << 1) + - ((u64) in[5]) * (in[7] << 2) + ((u64) in[6]) * in[6]; - tmp[13] = ((u64) in[5]) * (in[8] << 1) + - ((u64) in[6]) * (in[7] << 1); - tmp[14] = ((u64) in[6]) * (in[8] << 1) + - ((u64) in[7]) * (in[7] << 1); - tmp[15] = ((u64) in[7]) * (in[8] << 1); - tmp[16] = ((u64) in[8]) * in[8]; - - felem_reduce_degree(out, tmp); -} - -/* felem_mul sets out=in*in2. - * - * On entry: in[0,2,...] < 2**30, in[1,3,...] < 2**29 and - * in2[0,2,...] < 2**30, in2[1,3,...] < 2**29. - * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29. */ -static void felem_mul(felem out, const felem in, const felem in2) { - u64 tmp[17]; - - tmp[0] = ((u64) in[0]) * in2[0]; - tmp[1] = ((u64) in[0]) * (in2[1] << 0) + - ((u64) in[1]) * (in2[0] << 0); - tmp[2] = ((u64) in[0]) * (in2[2] << 0) + ((u64) in[1]) * - (in2[1] << 1) + ((u64) in[2]) * (in2[0] << 0); - tmp[3] = ((u64) in[0]) * (in2[3] << 0) + ((u64) in[1]) * - (in2[2] << 0) + ((u64) in[2]) * (in2[1] << 0) + - ((u64) in[3]) * (in2[0] << 0); - tmp[4] = ((u64) in[0]) * (in2[4] << 0) + ((u64) in[1]) * - (in2[3] << 1) + ((u64) in[2]) * (in2[2] << 0) + - ((u64) in[3]) * (in2[1] << 1) + - ((u64) in[4]) * (in2[0] << 0); - tmp[5] = ((u64) in[0]) * (in2[5] << 0) + ((u64) in[1]) * - (in2[4] << 0) + ((u64) in[2]) * (in2[3] << 0) + - ((u64) in[3]) * (in2[2] << 0) + ((u64) in[4]) * - (in2[1] << 0) + ((u64) in[5]) * (in2[0] << 0); - tmp[6] = ((u64) in[0]) * (in2[6] << 0) + ((u64) in[1]) * - (in2[5] << 1) + ((u64) in[2]) * (in2[4] << 0) + - ((u64) in[3]) * (in2[3] << 1) + ((u64) in[4]) * - (in2[2] << 0) + ((u64) in[5]) * (in2[1] << 1) + - ((u64) in[6]) * (in2[0] << 0); - tmp[7] = ((u64) in[0]) * (in2[7] << 0) + ((u64) in[1]) * - (in2[6] << 0) + ((u64) in[2]) * (in2[5] << 0) + - ((u64) in[3]) * (in2[4] << 0) + ((u64) in[4]) * - (in2[3] << 0) + ((u64) in[5]) * (in2[2] << 0) + - ((u64) in[6]) * (in2[1] << 0) + - ((u64) in[7]) * (in2[0] << 0); - /* tmp[8] has the greatest value but doesn't overflow. See logic in - * felem_square. */ - tmp[8] = ((u64) in[0]) * (in2[8] << 0) + ((u64) in[1]) * - (in2[7] << 1) + ((u64) in[2]) * (in2[6] << 0) + - ((u64) in[3]) * (in2[5] << 1) + ((u64) in[4]) * - (in2[4] << 0) + ((u64) in[5]) * (in2[3] << 1) + - ((u64) in[6]) * (in2[2] << 0) + ((u64) in[7]) * - (in2[1] << 1) + ((u64) in[8]) * (in2[0] << 0); - tmp[9] = ((u64) in[1]) * (in2[8] << 0) + ((u64) in[2]) * - (in2[7] << 0) + ((u64) in[3]) * (in2[6] << 0) + - ((u64) in[4]) * (in2[5] << 0) + ((u64) in[5]) * - (in2[4] << 0) + ((u64) in[6]) * (in2[3] << 0) + - ((u64) in[7]) * (in2[2] << 0) + - ((u64) in[8]) * (in2[1] << 0); - tmp[10] = ((u64) in[2]) * (in2[8] << 0) + ((u64) in[3]) * - (in2[7] << 1) + ((u64) in[4]) * (in2[6] << 0) + - ((u64) in[5]) * (in2[5] << 1) + ((u64) in[6]) * - (in2[4] << 0) + ((u64) in[7]) * (in2[3] << 1) + - ((u64) in[8]) * (in2[2] << 0); - tmp[11] = ((u64) in[3]) * (in2[8] << 0) + ((u64) in[4]) * - (in2[7] << 0) + ((u64) in[5]) * (in2[6] << 0) + - ((u64) in[6]) * (in2[5] << 0) + ((u64) in[7]) * - (in2[4] << 0) + ((u64) in[8]) * (in2[3] << 0); - tmp[12] = ((u64) in[4]) * (in2[8] << 0) + ((u64) in[5]) * - (in2[7] << 1) + ((u64) in[6]) * (in2[6] << 0) + - ((u64) in[7]) * (in2[5] << 1) + - ((u64) in[8]) * (in2[4] << 0); - tmp[13] = ((u64) in[5]) * (in2[8] << 0) + ((u64) in[6]) * - (in2[7] << 0) + ((u64) in[7]) * (in2[6] << 0) + - ((u64) in[8]) * (in2[5] << 0); - tmp[14] = ((u64) in[6]) * (in2[8] << 0) + ((u64) in[7]) * - (in2[7] << 1) + ((u64) in[8]) * (in2[6] << 0); - tmp[15] = ((u64) in[7]) * (in2[8] << 0) + - ((u64) in[8]) * (in2[7] << 0); - tmp[16] = ((u64) in[8]) * (in2[8] << 0); - - felem_reduce_degree(out, tmp); -} - -static void felem_assign(felem out, const felem in) { - memcpy(out, in, sizeof(felem)); -} - /* felem_inv calculates |out| = |in|^{-1} * * Based on Fermat's Little Theorem: @@ -667,131 +105,7 @@ felem_mul(out, ftmp2, ftmp); /* 2^256 - 2^224 + 2^192 + 2^96 - 3 */ } -/* felem_scalar_3 sets out=3*out. - * - * On entry: out[0,2,...] < 2**30, out[1,3,...] < 2**29. - * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29. */ -static void felem_scalar_3(felem out) { - limb carry = 0; - unsigned i; - for (i = 0;; i++) { - out[i] *= 3; - out[i] += carry; - carry = out[i] >> 29; - out[i] &= kBottom29Bits; - - i++; - if (i == NLIMBS) - break; - - out[i] *= 3; - out[i] += carry; - carry = out[i] >> 28; - out[i] &= kBottom28Bits; - } - - felem_reduce_carry(out, carry); -} - -/* felem_scalar_4 sets out=4*out. - * - * On entry: out[0,2,...] < 2**30, out[1,3,...] < 2**29. - * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29. */ -static void felem_scalar_4(felem out) { - limb carry = 0, next_carry; - unsigned i; - - for (i = 0;; i++) { - next_carry = out[i] >> 27; - out[i] <<= 2; - out[i] &= kBottom29Bits; - out[i] += carry; - carry = next_carry + (out[i] >> 29); - out[i] &= kBottom29Bits; - - i++; - if (i == NLIMBS) - break; - - next_carry = out[i] >> 26; - out[i] <<= 2; - out[i] &= kBottom28Bits; - out[i] += carry; - carry = next_carry + (out[i] >> 28); - out[i] &= kBottom28Bits; - } - - felem_reduce_carry(out, carry); -} - -/* felem_scalar_8 sets out=8*out. - * - * On entry: out[0,2,...] < 2**30, out[1,3,...] < 2**29. - * On exit: out[0,2,...] < 2**30, out[1,3,...] < 2**29. */ -static void felem_scalar_8(felem out) { - limb carry = 0, next_carry; - unsigned i; - - for (i = 0;; i++) { - next_carry = out[i] >> 26; - out[i] <<= 3; - out[i] &= kBottom29Bits; - out[i] += carry; - carry = next_carry + (out[i] >> 29); - out[i] &= kBottom29Bits; - - i++; - if (i == NLIMBS) - break; - - next_carry = out[i] >> 25; - out[i] <<= 3; - out[i] &= kBottom28Bits; - out[i] += carry; - carry = next_carry + (out[i] >> 28); - out[i] &= kBottom28Bits; - } - - felem_reduce_carry(out, carry); -} - -/* felem_is_zero_vartime returns 1 iff |in| == 0. It takes a variable amount of - * time depending on the value of |in|. */ -static char felem_is_zero_vartime(const felem in) { - limb carry; - int i; - limb tmp[NLIMBS]; - - felem_assign(tmp, in); - - /* First, reduce tmp to a minimal form. */ - do { - carry = 0; - for (i = 0;; i++) { - tmp[i] += carry; - carry = tmp[i] >> 29; - tmp[i] &= kBottom29Bits; - - i++; - if (i == NLIMBS) - break; - - tmp[i] += carry; - carry = tmp[i] >> 28; - tmp[i] &= kBottom28Bits; - } - - felem_reduce_carry(tmp, carry); - } while (carry); - - /* tmp < 2**257, so the only possible zero values are 0, p and 2p. */ - return memcmp(tmp, kZero, sizeof(tmp)) == 0 || - memcmp(tmp, kP, sizeof(tmp)) == 0 || - memcmp(tmp, k2P, sizeof(tmp)) == 0; -} - - /* Group operations: * * Elements of the elliptic curve group are represented in Jacobian @@ -971,9 +285,9 @@ felem_diff(y_out, y_out, tmp); } -/* copy_conditional sets out=in if mask = 0xffffffff in constant time. +/* copy_conditional sets out=in if mask = -1 in constant time. * - * On entry: mask is either 0 or 0xffffffff. */ + * On entry: mask is either 0 or -1. */ static void copy_conditional(felem out, const felem in, limb mask) { int i; @@ -1168,58 +482,6 @@ } } -#define kRDigits {2, 0, 0, 0xfffffffe, 0xffffffff, 0xffffffff, 0xfffffffd, 1} // 2^257 mod p256.p - -#define kRInvDigits {0x80000000, 1, 0xffffffff, 0, 0x80000001, 0xfffffffe, 1, 0x7fffffff} // 1 / 2^257 mod p256.p - -static const cryptonite_p256_int kR = { kRDigits }; -static const cryptonite_p256_int kRInv = { kRInvDigits }; - -/* to_montgomery sets out = R*in. */ -static void to_montgomery(felem out, const cryptonite_p256_int* in) { - cryptonite_p256_int in_shifted; - int i; - - cryptonite_p256_init(&in_shifted); - cryptonite_p256_modmul(&cryptonite_SECP256r1_p, in, 0, &kR, &in_shifted); - - for (i = 0; i < NLIMBS; i++) { - if ((i & 1) == 0) { - out[i] = P256_DIGIT(&in_shifted, 0) & kBottom29Bits; - cryptonite_p256_shr(&in_shifted, 29, &in_shifted); - } else { - out[i] = P256_DIGIT(&in_shifted, 0) & kBottom28Bits; - cryptonite_p256_shr(&in_shifted, 28, &in_shifted); - } - } - - cryptonite_p256_clear(&in_shifted); -} - -/* from_montgomery sets out=in/R. */ -static void from_montgomery(cryptonite_p256_int* out, const felem in) { - cryptonite_p256_int result, tmp; - int i, top; - - cryptonite_p256_init(&result); - cryptonite_p256_init(&tmp); - - cryptonite_p256_add_d(&tmp, in[NLIMBS - 1], &result); - for (i = NLIMBS - 2; i >= 0; i--) { - if ((i & 1) == 0) { - top = cryptonite_p256_shl(&result, 29, &tmp); - } else { - top = cryptonite_p256_shl(&result, 28, &tmp); - } - top |= cryptonite_p256_add_d(&tmp, in[i], &result); - } - - cryptonite_p256_modmul(&cryptonite_SECP256r1_p, &kRInv, top, &result, out); - - cryptonite_p256_clear(&result); - cryptonite_p256_clear(&tmp); -} - /* cryptonite_p256_base_point_mul sets {out_x,out_y} = nG, where n is < the * order of the group. */ void cryptonite_p256_base_point_mul(const cryptonite_p256_int* n, cryptonite_p256_int* out_x, cryptonite_p256_int* out_y) { @@ -1287,19 +549,16 @@ const cryptonite_p256_int *in_x2, const cryptonite_p256_int *in_y2, cryptonite_p256_int *out_x, cryptonite_p256_int *out_y) { - felem x1, y1, z1, x2, y2, z2, px1, py1, px2, py2; - const cryptonite_p256_int one = P256_ONE; + felem x, y, z, px1, py1, px2, py2; to_montgomery(px1, in_x1); to_montgomery(py1, in_y1); to_montgomery(px2, in_x2); to_montgomery(py2, in_y2); - scalar_mult(x1, y1, z1, px1, py1, &one); - scalar_mult(x2, y2, z2, px2, py2, &one); - point_add_or_double_vartime(x1, y1, z1, x1, y1, z1, x2, y2, z2); + point_add_or_double_vartime(x, y, z, px1, py1, kOne, px2, py2, kOne); - point_to_affine(px1, py1, x1, y1, z1); + point_to_affine(px1, py1, x, y, z); from_montgomery(out_x, px1); from_montgomery(out_y, py1); } @@ -1313,4 +572,21 @@ { memcpy(out_x, in_x, P256_NBYTES); cryptonite_p256_sub(&cryptonite_SECP256r1_p, in_y, out_y); +} + +/* this function is not part of the original source + cryptonite_p256e_point_mul sets {out_x,out_y} = n*{in_x,in_y}, where + n is < the order of the group. + */ +void cryptonite_p256e_point_mul(const cryptonite_p256_int* n, + const cryptonite_p256_int* in_x, const cryptonite_p256_int* in_y, + cryptonite_p256_int* out_x, cryptonite_p256_int* out_y) { + felem x, y, z, px, py; + + to_montgomery(px, in_x); + to_montgomery(py, in_y); + scalar_mult(x, y, z, px, py, n); + point_to_affine(px, py, x, y, z); + from_montgomery(out_x, px); + from_montgomery(out_y, py); } diff --git a/cryptonite.cabal b/cryptonite.cabal --- a/cryptonite.cabal +++ b/cryptonite.cabal @@ -1,5 +1,5 @@ Name: cryptonite -version: 0.26 +version: 0.27 Synopsis: Cryptography Primitives sink Description: A repository of cryptographic primitives. @@ -19,7 +19,7 @@ * Data related: Anti-Forensic Information Splitter (AFIS) . If anything cryptographic related is missing from here, submit - a pull request to have it added. This package strive to be a + a pull request to have it added. This package strives to be a cryptographic kitchen sink that provides cryptography for everyone. . Evaluate the security related to your requirements before using. @@ -36,7 +36,7 @@ Homepage: https://github.com/haskell-crypto/cryptonite Bug-reports: https://github.com/haskell-crypto/cryptonite/issues Cabal-Version: 1.18 -tested-with: GHC==8.6.5, GHC==8.4.4, GHC==8.2.2, GHC==8.0.2 +tested-with: GHC==8.8.2, GHC==8.6.5, GHC==8.4.4, GHC==8.2.2, GHC==8.0.2 extra-doc-files: README.md CHANGELOG.md extra-source-files: cbits/*.h cbits/aes/*.h @@ -50,7 +50,8 @@ cbits/decaf/p448/*.h cbits/decaf/ed448goldilocks/decaf_tables.c cbits/decaf/ed448goldilocks/decaf.c - cbits/p256/*.h + cbits/include32/p256/*.h + cbits/include64/p256/*.h cbits/blake2/ref/*.h cbits/blake2/sse/*.h cbits/argon2/*.h @@ -102,8 +103,14 @@ Default: False Manual: True +Flag use_target_attributes + Description: use GCC / clang function attributes instead of global target options. + Default: True + Manual: True + Library Exposed-modules: Crypto.Cipher.AES + Crypto.Cipher.AESGCMSIV Crypto.Cipher.Blowfish Crypto.Cipher.CAST5 Crypto.Cipher.Camellia @@ -158,6 +165,7 @@ Crypto.PubKey.ECC.ECDSA Crypto.PubKey.ECC.P256 Crypto.PubKey.ECC.Types + Crypto.PubKey.ECDSA Crypto.PubKey.ECIES Crypto.PubKey.Ed25519 Crypto.PubKey.Ed448 @@ -177,6 +185,7 @@ Crypto.Random.Entropy Crypto.Random.EntropyPool Crypto.Random.Entropy.Unsafe + Crypto.System.CPU Crypto.Tutorial Other-modules: Crypto.Cipher.AES.Primitive Crypto.Cipher.Blowfish.Box @@ -282,6 +291,11 @@ , cbits/decaf/p448 if arch(x86_64) || arch(aarch64) + include-dirs: cbits/include64 + else + include-dirs: cbits/include32 + + if arch(x86_64) || arch(aarch64) C-sources: cbits/decaf/p448/arch_ref64/f_impl.c , cbits/decaf/p448/f_generic.c , cbits/decaf/p448/f_arithmetic.c @@ -327,9 +341,13 @@ c-sources: cbits/cryptonite_rdrand.c if flag(support_aesni) && (os(linux) || os(freebsd) || os(osx)) && (arch(i386) || arch(x86_64)) - CC-options: -mssse3 -maes -DWITH_AESNI + CC-options: -DWITH_AESNI + if !flag(use_target_attributes) + CC-options: -mssse3 -maes if flag(support_pclmuldq) - CC-options: -msse4.1 -mpclmul -DWITH_PCLMUL + CC-options: -DWITH_PCLMUL + if !flag(use_target_attributes) + CC-options: -msse4.1 -mpclmul C-sources: cbits/aes/x86ni.c , cbits/aes/generic.c , cbits/aes/gf.c @@ -354,6 +372,8 @@ if arch(x86_64) || flag(support_sse) CPP-options: -DSUPPORT_SSE + if arch(i386) + CC-options: -msse2 C-sources: cbits/argon2/argon2.c include-dirs: cbits/argon2 @@ -374,6 +394,8 @@ Build-depends: deepseq if flag(check_alignment) cc-options: -DWITH_ASSERT_ALIGNMENT + if flag(use_target_attributes) + cc-options: -DWITH_TARGET_ATTRIBUTES Test-Suite test-cryptonite type: exitcode-stdio-1.0 @@ -385,6 +407,7 @@ BCryptPBKDF ECC ECC.Edwards25519 + ECDSA Hash Imports KAT_AES.KATCBC @@ -394,6 +417,7 @@ KAT_AES.KATOCB3 KAT_AES.KATXTS KAT_AES + KAT_AESGCMSIV KAT_AFIS KAT_Argon2 KAT_Blowfish diff --git a/tests/ECC.hs b/tests/ECC.hs --- a/tests/ECC.hs +++ b/tests/ECC.hs @@ -24,6 +24,19 @@ , Curve ECC.Curve_X448 ] +data CurveArith = forall curve. (ECC.EllipticCurveBasepointArith curve, Show curve) => CurveArith curve + +instance Show CurveArith where + showsPrec d (CurveArith curve) = showsPrec d curve + +instance Arbitrary CurveArith where + arbitrary = elements + [ CurveArith ECC.Curve_P256R1 + , CurveArith ECC.Curve_P384R1 + , CurveArith ECC.Curve_P521R1 + , CurveArith ECC.Curve_Edwards25519 + ] + data VectorPoint = VectorPoint { vpCurve :: Curve , vpHex :: ByteString @@ -262,13 +275,13 @@ cryptoError :: CryptoFailable a -> Maybe CryptoError cryptoError = onCryptoFailure Just (const Nothing) -doPointDecodeTest (i, vector) = +doPointDecodeTest i vector = case vpCurve vector of Curve curve -> let prx = Just curve -- using Maybe as Proxy in testCase (show i) (vpError vector @=? cryptoError (ECC.decodePoint prx $ vpEncodedPoint vector)) -doWeakPointECDHTest (i, vector) = +doWeakPointECDHTest i vector = case vpCurve vector of Curve curve -> testCase (show i) $ do let prx = Just curve -- using Maybe as Proxy @@ -277,10 +290,10 @@ vpError vector @=? cryptoError (ECC.ecdh prx (ECC.keypairGetPrivate keyPair) public) tests = testGroup "ECC" - [ testGroup "decodePoint" $ map doPointDecodeTest (zip [katZero..] vectorsPoint) - , testGroup "ECDH weak points" $ map doWeakPointECDHTest (zip [katZero..] vectorsWeakPoint) + [ testGroup "decodePoint" $ zipWith doPointDecodeTest [katZero..] vectorsPoint + , testGroup "ECDH weak points" $ zipWith doWeakPointECDHTest [katZero..] vectorsWeakPoint , testGroup "property" - [ testProperty "decodePoint.encodePoint==id" $ \testDRG (Curve curve) -> do + [ testProperty "decodePoint.encodePoint==id" $ \testDRG (Curve curve) -> let prx = Just curve -- using Maybe as Proxy keyPair = withTestDRG testDRG $ ECC.curveGenerateKeyPair prx p1 = ECC.keypairGetPublic keyPair @@ -298,5 +311,33 @@ bobShared' = ECC.ecdhRaw prx (ECC.keypairGetPrivate bob) (ECC.keypairGetPublic alice) in aliceShared == bobShared && aliceShared == CryptoPassed aliceShared' && bobShared == CryptoPassed bobShared' + , testProperty "decodeScalar.encodeScalar==id" $ \testDRG (CurveArith curve) -> + let prx = Just curve -- using Maybe as Proxy + s1 = withTestDRG testDRG $ ECC.curveGenerateScalar prx + bs = ECC.encodeScalar prx s1 :: ByteString + s2 = ECC.decodeScalar prx bs + in CryptoPassed s1 == s2 + , testProperty "scalarFromInteger.scalarToInteger==id" $ \testDRG (CurveArith curve) -> + let prx = Just curve -- using Maybe as Proxy + s1 = withTestDRG testDRG $ ECC.curveGenerateScalar prx + bs = ECC.scalarToInteger prx s1 + s2 = ECC.scalarFromInteger prx bs + in CryptoPassed s1 == s2 + , localOption (QuickCheckTests 20) $ testProperty "(a + b).P = a.P + b.P" $ \testDRG (CurveArith curve) -> + let prx = Just curve -- using Maybe as Proxy + (s, a, b) = withTestDRG testDRG $ + (,,) <$> ECC.curveGenerateScalar prx + <*> ECC.curveGenerateScalar prx + <*> ECC.curveGenerateScalar prx + p = ECC.pointBaseSmul prx s + in ECC.pointSmul prx (ECC.scalarAdd prx a b) p == ECC.pointAdd prx (ECC.pointSmul prx a p) (ECC.pointSmul prx b p) + , localOption (QuickCheckTests 20) $ testProperty "(a * b).P = a.(b.P)" $ \testDRG (CurveArith curve) -> + let prx = Just curve -- using Maybe as Proxy + (s, a, b) = withTestDRG testDRG $ + (,,) <$> ECC.curveGenerateScalar prx + <*> ECC.curveGenerateScalar prx + <*> ECC.curveGenerateScalar prx + p = ECC.pointBaseSmul prx s + in ECC.pointSmul prx (ECC.scalarMul prx a b) p == ECC.pointSmul prx a (ECC.pointSmul prx b p) ] ] diff --git a/tests/ECDSA.hs b/tests/ECDSA.hs new file mode 100644 --- /dev/null +++ b/tests/ECDSA.hs @@ -0,0 +1,61 @@ +{-# LANGUAGE ExistentialQuantification #-} +{-# LANGUAGE FlexibleContexts #-} +module ECDSA (tests) where + +import qualified Crypto.ECC as ECDSA +import qualified Crypto.PubKey.ECC.ECDSA as ECC +import qualified Crypto.PubKey.ECC.Types as ECC +import qualified Crypto.PubKey.ECDSA as ECDSA +import Crypto.Hash.Algorithms +import Crypto.Error +import qualified Data.ByteString as B + +import Imports + +data Curve = forall curve. (ECDSA.EllipticCurveECDSA curve, Show (ECDSA.Scalar curve)) => Curve curve ECC.Curve ECC.CurveName + +instance Show Curve where + showsPrec d (Curve _ _ name) = showsPrec d name + +instance Arbitrary Curve where + arbitrary = elements + [ makeCurve ECDSA.Curve_P256R1 ECC.SEC_p256r1 + , makeCurve ECDSA.Curve_P384R1 ECC.SEC_p384r1 + , makeCurve ECDSA.Curve_P521R1 ECC.SEC_p521r1 + ] + where + makeCurve c name = Curve c (ECC.getCurveByName name) name + +arbitraryScalar curve = choose (1, n - 1) + where n = ECC.ecc_n (ECC.common_curve curve) + +sigECCToECDSA :: ECDSA.EllipticCurveECDSA curve + => proxy curve -> ECC.Signature -> ECDSA.Signature curve +sigECCToECDSA prx (ECC.Signature r s) = + ECDSA.Signature (throwCryptoError $ ECDSA.scalarFromInteger prx r) + (throwCryptoError $ ECDSA.scalarFromInteger prx s) + +tests = localOption (QuickCheckTests 5) $ testGroup "ECDSA" + [ testProperty "SHA1" $ propertyECDSA SHA1 + , testProperty "SHA224" $ propertyECDSA SHA224 + , testProperty "SHA256" $ propertyECDSA SHA256 + , testProperty "SHA384" $ propertyECDSA SHA384 + , testProperty "SHA512" $ propertyECDSA SHA512 + ] + where + propertyECDSA hashAlg (Curve c curve _) (ArbitraryBS0_2901 msg) = do + d <- arbitraryScalar curve + kECC <- arbitraryScalar curve + let privECC = ECC.PrivateKey curve d + prx = Just c -- using Maybe as Proxy + kECDSA = throwCryptoError $ ECDSA.scalarFromInteger prx kECC + privECDSA = throwCryptoError $ ECDSA.scalarFromInteger prx d + pubECDSA = ECDSA.toPublic prx privECDSA + Just sigECC = ECC.signWith kECC privECC hashAlg msg + Just sigECDSA = ECDSA.signWith prx kECDSA privECDSA hashAlg msg + sigECDSA' = sigECCToECDSA prx sigECC + msg' = msg `B.append` B.singleton 42 + return $ propertyHold [ eqTest "signature" sigECDSA sigECDSA' + , eqTest "verification" True (ECDSA.verify prx hashAlg pubECDSA sigECDSA' msg) + , eqTest "alteration" False (ECDSA.verify prx hashAlg pubECDSA sigECDSA msg') + ] diff --git a/tests/KAT_AES/KATGCM.hs b/tests/KAT_AES/KATGCM.hs --- a/tests/KAT_AES/KATGCM.hs +++ b/tests/KAT_AES/KATGCM.hs @@ -56,6 +56,14 @@ , {-out = -}"\xe4\x42\xf8\xc4\xc6\x67\x84\x86\x4a\x5a\x6e\xc7\xe0\xca\x68\xac\x16\xbc\x5b\xbf\xf7\xd5\xf3\xfa\xf3\xb2\xcb\xb0\xa2\x14\xa1" , {-taglen = -}16 , {-tag = -}"\x94\xd1\x47\xc3\xa2\xca\x93\xe9\x66\x93\x1e\x3b\xb3\xbb\x67\x01") + -- vector 6 tests 32-bit counter wrapping + , ( {-key = -}"\x01\x02\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , {-iv = -}"\xe8\x38\x84\x1d\x75\xae\x33\xb5\x4b\x51\x57\x89\xc9\x5f\xbe\x65" + , {-aad = -}"\x54\x68\x65\x20\x66\x69\x76\x65\x20\x62\x6f\x78\x69\x6e\x67\x20\x77\x69\x7a\x61\x72\x64\x73\x20\x6a\x75\x6d\x70\x20\x71\x75\x69\x63\x6b\x6c\x79\x2e" + , {-input = -}"\x54\x68\x65\x20\x71\x75\x69\x63\x6b\x20\x62\x72\x6f\x77\x6e\x20\x66\x6f\x78\x20\x6a\x75\x6d\x70\x73\x20\x6f\x76\x65\x72\x20\x74\x68\x65\x20\x6c\x61\x7a\x79\x20\x64\x6f\x67" + , {-out = -}"\x82\x31\x9e\x5a\x6a\x7f\x43\xd0\x42\x8c\xf1\x01\xcf\x0c\x75\xf1\x5d\xda\x4f\xa1\x28\x95\xcd\xd7\x7b\xd5\x42\x68\x2f\xcd\x10\x1b\x0c\x75\x05\x54\xf4\x2f\x2b\xf6\x69\x96\x29" + , {-taglen = -}16 + , {-tag = -}"\x9a\xfa\xf4\xea\xae\x2e\x6f\x40\x00\xf4\x89\x77\xd0\x1e\xd5\x14") ] vectors_aes256_enc :: [KATGCM] diff --git a/tests/KAT_AESGCMSIV.hs b/tests/KAT_AESGCMSIV.hs new file mode 100644 --- /dev/null +++ b/tests/KAT_AESGCMSIV.hs @@ -0,0 +1,494 @@ +{-# LANGUAGE FlexibleInstances #-} +{-# LANGUAGE OverloadedStrings #-} +{-# LANGUAGE Rank2Types #-} +{-# LANGUAGE RecordWildCards #-} +module KAT_AESGCMSIV (tests) where + +import Imports + +import Data.Proxy +import qualified Data.ByteArray as B + +import Crypto.Cipher.AES +import Crypto.Cipher.AESGCMSIV +import Crypto.Cipher.Types +import Crypto.Error + +data Vector c = Vector + { vecPlaintext :: ByteString + , vecAAD :: ByteString + , vecKey :: ByteString + , vecNonce :: ByteString + , vecTag :: ByteString + , vecCiphertext :: ByteString + } + +vecCipher :: Cipher c => Vector c -> c +vecCipher = throwCryptoError . cipherInit . vecKey + +vectors128 :: [Vector AES128] +vectors128 = + [ Vector + { vecPlaintext = "" + , vecAAD = "" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\xdc\x20\xe2\xd8\x3f\x25\x70\x5b\xb4\x9e\x43\x9e\xca\x56\xde\x25" + , vecCiphertext = "" + } + , Vector + { vecPlaintext = "\x01\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x57\x87\x82\xff\xf6\x01\x3b\x81\x5b\x28\x7c\x22\x49\x3a\x36\x4c" + , vecCiphertext = "\xb5\xd8\x39\x33\x0a\xc7\xb7\x86" + } + , Vector + { vecPlaintext = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\xa4\x97\x8d\xb3\x57\x39\x1a\x0b\xc4\xfd\xec\x8b\x0d\x10\x66\x39" + , vecCiphertext = "\x73\x23\xea\x61\xd0\x59\x32\x26\x00\x47\xd9\x42" + } + , Vector + { vecPlaintext = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x30\x3a\xaf\x90\xf6\xfe\x21\x19\x9c\x60\x68\x57\x74\x37\xa0\xc4" + , vecCiphertext = "\x74\x3f\x7c\x80\x77\xab\x25\xf8\x62\x4e\x2e\x94\x85\x79\xcf\x77" + } + , Vector + { vecPlaintext = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x1a\x8e\x45\xdc\xd4\x57\x8c\x66\x7c\xd8\x68\x47\xbf\x61\x55\xff" + , vecCiphertext = "\x84\xe0\x7e\x62\xba\x83\xa6\x58\x54\x17\x24\x5d\x7e\xc4\x13\xa9\xfe\x42\x7d\x63\x15\xc0\x9b\x57\xce\x45\xf2\xe3\x93\x6a\x94\x45" + } + , Vector + { vecPlaintext = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x5e\x6e\x31\x1d\xbf\x39\x5d\x35\xb0\xfe\x39\xc2\x71\x43\x88\xf8" + , vecCiphertext = "\x3f\xd2\x4c\xe1\xf5\xa6\x7b\x75\xbf\x23\x51\xf1\x81\xa4\x75\xc7\xb8\x00\xa5\xb4\xd3\xdc\xf7\x01\x06\xb1\xee\xa8\x2f\xa1\xd6\x4d\xf4\x2b\xf7\x22\x61\x22\xfa\x92\xe1\x7a\x40\xee\xaa\xc1\x20\x1b" + } + , Vector + { vecPlaintext = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x04\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x8a\x26\x3d\xd3\x17\xaa\x88\xd5\x6b\xdf\x39\x36\xdb\xa7\x5b\xb8" + , vecCiphertext = "\x24\x33\x66\x8f\x10\x58\x19\x0f\x6d\x43\xe3\x60\xf4\xf3\x5c\xd8\xe4\x75\x12\x7c\xfc\xa7\x02\x8e\xa8\xab\x5c\x20\xf7\xab\x2a\xf0\x25\x16\xa2\xbd\xcb\xc0\x8d\x52\x1b\xe3\x7f\xf2\x8c\x15\x2b\xba\x36\x69\x7f\x25\xb4\xcd\x16\x9c\x65\x90\xd1\xdd\x39\x56\x6d\x3f" + } + , Vector + { vecPlaintext = "\x02\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "\x01" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x3b\x0a\x1a\x25\x60\x96\x9c\xdf\x79\x0d\x99\x75\x9a\xbd\x15\x08" + , vecCiphertext = "\x1e\x6d\xab\xa3\x56\x69\xf4\x27" + } + , Vector + { vecPlaintext = "\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "\x01" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x08\x29\x9c\x51\x02\x74\x5a\xaa\x3a\x0c\x46\x9f\xad\x9e\x07\x5a" + , vecCiphertext = "\x29\x6c\x78\x89\xfd\x99\xf4\x19\x17\xf4\x46\x20" + } + , Vector + { vecPlaintext = "\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "\x01" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x8f\x89\x36\xec\x03\x9e\x4e\x4b\xb9\x7e\xbd\x8c\x44\x57\x44\x1f" + , vecCiphertext = "\xe2\xb0\xc5\xda\x79\xa9\x01\xc1\x74\x5f\x70\x05\x25\xcb\x33\x5b" + } + , Vector + { vecPlaintext = "\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "\x01" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\xe6\xaf\x6a\x7f\x87\x28\x7d\xa0\x59\xa7\x16\x84\xed\x34\x98\xe1" + , vecCiphertext = "\x62\x00\x48\xef\x3c\x1e\x73\xe5\x7e\x02\xbb\x85\x62\xc4\x16\xa3\x19\xe7\x3e\x4c\xaa\xc8\xe9\x6a\x1e\xcb\x29\x33\x14\x5a\x1d\x71" + } + , Vector + { vecPlaintext = "\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x04\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "\x01" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x6a\x8c\xc3\x86\x5f\x76\x89\x7c\x2e\x4b\x24\x5c\xf3\x1c\x51\xf2" + , vecCiphertext = "\x50\xc8\x30\x3e\xa9\x39\x25\xd6\x40\x90\xd0\x7b\xd1\x09\xdf\xd9\x51\x5a\x5a\x33\x43\x10\x19\xc1\x7d\x93\x46\x59\x99\xa8\xb0\x05\x32\x01\xd7\x23\x12\x0a\x85\x62\xb8\x38\xcd\xff\x25\xbf\x9d\x1e" + } + , Vector + { vecPlaintext = "\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x04\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x05\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "\x01" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\xcd\xc4\x6a\xe4\x75\x56\x3d\xe0\x37\x00\x1e\xf8\x4a\xe2\x17\x44" + , vecCiphertext = "\x2f\x5c\x64\x05\x9d\xb5\x5e\xe0\xfb\x84\x7e\xd5\x13\x00\x37\x46\xac\xa4\xe6\x1c\x71\x1b\x5d\xe2\xe7\xa7\x7f\xfd\x02\xda\x42\xfe\xec\x60\x19\x10\xd3\x46\x7b\xb8\xb3\x6e\xbb\xae\xbc\xe5\xfb\xa3\x0d\x36\xc9\x5f\x48\xa3\xe7\x98\x0f\x0e\x7a\xc2\x99\x33\x2a\x80" + } + , Vector + { vecPlaintext = "\x02\x00\x00\x00" + , vecAAD = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x07\xeb\x1f\x84\xfb\x28\xf8\xcb\x73\xde\x8e\x99\xe2\xf4\x8a\x14" + , vecCiphertext = "\xa8\xfe\x3e\x87" + } + , Vector + { vecPlaintext = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x04\x00\x00\x00" + , vecAAD = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x24\xaf\xc9\x80\x5e\x97\x6f\x45\x1e\x6d\x87\xf6\xfe\x10\x65\x14" + , vecCiphertext = "\x6b\xb0\xfe\xcf\x5d\xed\x9b\x77\xf9\x02\xc7\xd5\xda\x23\x6a\x43\x91\xdd\x02\x97" + } + , Vector + { vecPlaintext = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x04\x00" + , vecAAD = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x00\x00" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\xbf\xf9\xb2\xef\x00\xfb\x47\x92\x0c\xc7\x2a\x0c\x0f\x13\xb9\xfd" + , vecCiphertext = "\x44\xd0\xaa\xf6\xfb\x2f\x1f\x34\xad\xd5\xe8\x06\x4e\x83\xe1\x2a\x2a\xda" + } + , Vector + { vecPlaintext = "" + , vecAAD = "" + , vecKey = "\xe6\x60\x21\xd5\xeb\x8e\x4f\x40\x66\xd4\xad\xb9\xc3\x35\x60\xe4" + , vecNonce = "\xf4\x6e\x44\xbb\x3d\xa0\x01\x5c\x94\xf7\x08\x87" + , vecTag = "\xa4\x19\x4b\x79\x07\x1b\x01\xa8\x7d\x65\xf7\x06\xe3\x94\x95\x78" + , vecCiphertext = "" + } + , Vector + { vecPlaintext = "\x7a\x80\x6c" + , vecAAD = "\x46\xbb\x91\xc3\xc5" + , vecKey = "\x36\x86\x42\x00\xe0\xea\xf5\x28\x4d\x88\x4a\x0e\x77\xd3\x16\x46" + , vecNonce = "\xba\xe8\xe3\x7f\xc8\x34\x41\xb1\x60\x34\x56\x6b" + , vecTag = "\x71\x1b\xd8\x5b\xc1\xe4\xd3\xe0\xa4\x62\xe0\x74\xee\xa4\x28\xa8" + , vecCiphertext = "\xaf\x60\xeb" + } + , Vector + { vecPlaintext = "\xbd\xc6\x6f\x14\x65\x45" + , vecAAD = "\xfc\x88\x0c\x94\xa9\x51\x98\x87\x42\x96" + , vecKey = "\xae\xdb\x64\xa6\xc5\x90\xbc\x84\xd1\xa5\xe2\x69\xe4\xb4\x78\x01" + , vecNonce = "\xaf\xc0\x57\x7e\x34\x69\x9b\x9e\x67\x1f\xdd\x4f" + , vecTag = "\xd6\xa9\xc4\x55\x45\xcf\xc1\x1f\x03\xad\x74\x3d\xba\x20\xf9\x66" + , vecCiphertext = "\xbb\x93\xa3\xe3\x4d\x3c" + } + , Vector + { vecPlaintext = "\x11\x77\x44\x1f\x19\x54\x95\x86\x0f" + , vecAAD = "\x04\x67\x87\xf3\xea\x22\xc1\x27\xaa\xf1\x95\xd1\x89\x47\x28" + , vecKey = "\xd5\xcc\x1f\xd1\x61\x32\x0b\x69\x20\xce\x07\x78\x7f\x86\x74\x3b" + , vecNonce = "\x27\x5d\x1a\xb3\x2f\x6d\x1f\x04\x34\xd8\x84\x8c" + , vecTag = "\x1d\x02\xfd\x0c\xd1\x74\xc8\x4f\xc5\xda\xe2\xf6\x0f\x52\xfd\x2b" + , vecCiphertext = "\x4f\x37\x28\x1f\x7a\xd1\x29\x49\xd0" + } + , Vector + { vecPlaintext = "\x9f\x57\x2c\x61\x4b\x47\x45\x91\x44\x74\xe7\xc7" + , vecAAD = "\xc9\x88\x2e\x53\x86\xfd\x9f\x92\xec\x48\x9c\x8f\xde\x2b\xe2\xcf\x97\xe7\x4e\x93" + , vecKey = "\xb3\xfe\xd1\x47\x3c\x52\x8b\x84\x26\xa5\x82\x99\x59\x29\xa1\x49" + , vecNonce = "\x9e\x9a\xd8\x78\x0c\x8d\x63\xd0\xab\x41\x49\xc0" + , vecTag = "\xc1\xdc\x2f\x87\x1f\xb7\x56\x1d\xa1\x28\x6e\x65\x5e\x24\xb7\xb0" + , vecCiphertext = "\xf5\x46\x73\xc5\xdd\xf7\x10\xc7\x45\x64\x1c\x8b" + } + , Vector + { vecPlaintext = "\x0d\x8c\x84\x51\x17\x80\x82\x35\x5c\x9e\x94\x0f\xea\x2f\x58" + , vecAAD = "\x29\x50\xa7\x0d\x5a\x1d\xb2\x31\x6f\xd5\x68\x37\x8d\xa1\x07\xb5\x2b\x0d\xa5\x52\x10\xcc\x1c\x1b\x0a" + , vecKey = "\x2d\x4e\xd8\x7d\xa4\x41\x02\x95\x2e\xf9\x4b\x02\xb8\x05\x24\x9b" + , vecNonce = "\xac\x80\xe6\xf6\x14\x55\xbf\xac\x83\x08\xa2\xd4" + , vecTag = "\x83\xb3\x44\x9b\x9f\x39\x55\x2d\xe9\x9d\xc2\x14\xa1\x19\x0b\x0b" + , vecCiphertext = "\xc9\xff\x54\x5e\x07\xb8\x8a\x01\x5f\x05\xb2\x74\x54\x0a\xa1" + } + , Vector + { vecPlaintext = "\x6b\x3d\xb4\xda\x3d\x57\xaa\x94\x84\x2b\x98\x03\xa9\x6e\x07\xfb\x6d\xe7" + , vecAAD = "\x18\x60\xf7\x62\xeb\xfb\xd0\x82\x84\xe4\x21\x70\x2d\xe0\xde\x18\xba\xa9\xc9\x59\x62\x91\xb0\x84\x66\xf3\x7d\xe2\x1c\x7f" + , vecKey = "\xbd\xe3\xb2\xf2\x04\xd1\xe9\xf8\xb0\x6b\xc4\x7f\x97\x45\xb3\xd1" + , vecNonce = "\xae\x06\x55\x6f\xb6\xaa\x78\x90\xbe\xbc\x18\xfe" + , vecTag = "\x3e\x37\x70\x94\xf0\x47\x09\xf6\x4d\x7b\x98\x53\x10\xa4\xdb\x84" + , vecCiphertext = "\x62\x98\xb2\x96\xe2\x4e\x8c\xc3\x5d\xce\x0b\xed\x48\x4b\x7f\x30\xd5\x80" + } + , Vector + { vecPlaintext = "\xe4\x2a\x3c\x02\xc2\x5b\x64\x86\x9e\x14\x6d\x7b\x23\x39\x87\xbd\xdf\xc2\x40\x87\x1d" + , vecAAD = "\x75\x76\xf7\x02\x8e\xc6\xeb\x5e\xa7\xe2\x98\x34\x2a\x94\xd4\xb2\x02\xb3\x70\xef\x97\x68\xec\x65\x61\xc4\xfe\x6b\x7e\x72\x96\xfa\x85\x9c\x21" + , vecKey = "\xf9\x01\xcf\xe8\xa6\x96\x15\xa9\x3f\xdf\x7a\x98\xca\xd4\x81\x79" + , vecNonce = "\x62\x45\x70\x9f\xb1\x88\x53\xf6\x8d\x83\x36\x40" + , vecTag = "\x2d\x15\x50\x6c\x84\xa9\xed\xd6\x5e\x13\xe9\xd2\x4a\x2a\x6e\x70" + , vecCiphertext = "\x39\x1c\xc3\x28\xd4\x84\xa4\xf4\x64\x06\x18\x1b\xcd\x62\xef\xd9\xb3\xee\x19\x7d\x05" + } + ] + +vectors256 :: [Vector AES256] +vectors256 = + [ Vector + { vecPlaintext = "" + , vecAAD = "" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x07\xf5\xf4\x16\x9b\xbf\x55\xa8\x40\x0c\xd4\x7e\xa6\xfd\x40\x0f" + , vecCiphertext = "" + } + , Vector + { vecPlaintext = "\x01\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x84\x31\x22\x13\x0f\x73\x64\xb7\x61\xe0\xb9\x74\x27\xe3\xdf\x28" + , vecCiphertext = "\xc2\xef\x32\x8e\x5c\x71\xc8\x3b" + } + , Vector + { vecPlaintext = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x8c\xa5\x0d\xa9\xae\x65\x59\xe4\x8f\xd1\x0f\x6e\x5c\x9c\xa1\x7e" + , vecCiphertext = "\x9a\xab\x2a\xeb\x3f\xaa\x0a\x34\xae\xa8\xe2\xb1" + } + , Vector + { vecPlaintext = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\xc9\xea\xc6\xfa\x70\x09\x42\x70\x2e\x90\x86\x23\x83\xc6\xc3\x66" + , vecCiphertext = "\x85\xa0\x1b\x63\x02\x5b\xa1\x9b\x7f\xd3\xdd\xfc\x03\x3b\x3e\x76" + } + , Vector + { vecPlaintext = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\xe8\x19\xe6\x3a\xbc\xd0\x20\xb0\x06\xa9\x76\x39\x76\x32\xeb\x5d" + , vecCiphertext = "\x4a\x6a\x9d\xb4\xc8\xc6\x54\x92\x01\xb9\xed\xb5\x30\x06\xcb\xa8\x21\xec\x9c\xf8\x50\x94\x8a\x7c\x86\xc6\x8a\xc7\x53\x9d\x02\x7f" + } + , Vector + { vecPlaintext = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x79\x0b\xc9\x68\x80\xa9\x9b\xa8\x04\xbd\x12\xc0\xe6\xa2\x2c\xc4" + , vecCiphertext = "\xc0\x0d\x12\x18\x93\xa9\xfa\x60\x3f\x48\xcc\xc1\xca\x3c\x57\xce\x74\x99\x24\x5e\xa0\x04\x6d\xb1\x6c\x53\xc7\xc6\x6f\xe7\x17\xe3\x9c\xf6\xc7\x48\x83\x7b\x61\xf6\xee\x3a\xdc\xee\x17\x53\x4e\xd5" + } + , Vector + { vecPlaintext = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x04\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x11\x28\x64\xc2\x69\xfc\x0d\x9d\x88\xc6\x1f\xa4\x7e\x39\xaa\x08" + , vecCiphertext = "\xc2\xd5\x16\x0a\x1f\x86\x83\x83\x49\x10\xac\xda\xfc\x41\xfb\xb1\x63\x2d\x4a\x35\x3e\x8b\x90\x5e\xc9\xa5\x49\x9a\xc3\x4f\x96\xc7\xe1\x04\x9e\xb0\x80\x88\x38\x91\xa4\xdb\x8c\xaa\xa1\xf9\x9d\xd0\x04\xd8\x04\x87\x54\x07\x35\x23\x4e\x37\x44\x51\x2c\x6f\x90\xce" + } + , Vector + { vecPlaintext = "\x02\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "\x01" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x91\x21\x3f\x26\x7e\x3b\x45\x2f\x02\xd0\x1a\xe3\x3e\x4e\xc8\x54" + , vecCiphertext = "\x1d\xe2\x29\x67\x23\x7a\x81\x32" + } + , Vector + { vecPlaintext = "\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "\x01" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\xc1\xa4\xa1\x9a\xe8\x00\x94\x1c\xcd\xc5\x7c\xc8\x41\x3c\x27\x7f" + , vecCiphertext = "\x16\x3d\x6f\x9c\xc1\xb3\x46\xcd\x45\x3a\x2e\x4c" + } + , Vector + { vecPlaintext = "\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "\x01" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\xb2\x92\xd2\x8f\xf6\x11\x89\xe8\xe4\x9f\x38\x75\xef\x91\xaf\xf7" + , vecCiphertext = "\xc9\x15\x45\x82\x3c\xc2\x4f\x17\xdb\xb0\xe9\xe8\x07\xd5\xec\x17" + } + , Vector + { vecPlaintext = "\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "\x01" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\xae\xa1\xba\xd1\x27\x02\xe1\x96\x56\x04\x37\x4a\xab\x96\xdb\xbc" + , vecCiphertext = "\x07\xda\xd3\x64\xbf\xc2\xb9\xda\x89\x11\x6d\x7b\xef\x6d\xaa\xaf\x6f\x25\x55\x10\xaa\x65\x4f\x92\x0a\xc8\x1b\x94\xe8\xba\xd3\x65" + } + , Vector + { vecPlaintext = "\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x04\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "\x01" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x03\x33\x27\x42\xb2\x28\xc6\x47\x17\x36\x16\xcf\xd4\x4c\x54\xeb" + , vecCiphertext = "\xc6\x7a\x1f\x0f\x56\x7a\x51\x98\xaa\x1f\xcc\x8e\x3f\x21\x31\x43\x36\xf7\xf5\x1c\xa8\xb1\xaf\x61\xfe\xac\x35\xa8\x64\x16\xfa\x47\xfb\xca\x3b\x5f\x74\x9c\xdf\x56\x45\x27\xf2\x31\x4f\x42\xfe\x25" + } + , Vector + { vecPlaintext = "\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x04\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x05\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "\x01" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x5b\xde\x02\x85\x03\x7c\x5d\xe8\x1e\x5b\x57\x0a\x04\x9b\x62\xa0" + , vecCiphertext = "\x67\xfd\x45\xe1\x26\xbf\xb9\xa7\x99\x30\xc4\x3a\xad\x2d\x36\x96\x7d\x3f\x0e\x4d\x21\x7c\x1e\x55\x1f\x59\x72\x78\x70\xbe\xef\xc9\x8c\xb9\x33\xa8\xfc\xe9\xde\x88\x7b\x1e\x40\x79\x99\x88\xdb\x1f\xc3\xf9\x18\x80\xed\x40\x5b\x2d\xd2\x98\x31\x88\x58\x46\x7c\x89" + } + , Vector + { vecPlaintext = "\x02\x00\x00\x00" + , vecAAD = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\x18\x35\xe5\x17\x74\x1d\xfd\xdc\xcf\xa0\x7f\xa4\x66\x1b\x74\xcf" + , vecCiphertext = "\x22\xb3\xf4\xcd" + } + , Vector + { vecPlaintext = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x04\x00\x00\x00" + , vecAAD = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\xb8\x79\xad\x97\x6d\x82\x42\xac\xc1\x88\xab\x59\xca\xbf\xe3\x07" + , vecCiphertext = "\x43\xdd\x01\x63\xcd\xb4\x8f\x9f\xe3\x21\x2b\xf6\x1b\x20\x19\x76\x06\x7f\x34\x2b" + } + , Vector + { vecPlaintext = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x04\x00" + , vecAAD = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x00\x00" + , vecKey = "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\xcf\xcd\xf5\x04\x21\x12\xaa\x29\x68\x5c\x91\x2f\xc2\x05\x65\x43" + , vecCiphertext = "\x46\x24\x01\x72\x4b\x5c\xe6\x58\x8d\x5a\x54\xaa\xe5\x37\x55\x13\xa0\x75" + } + , Vector + { vecPlaintext = "" + , vecAAD = "" + , vecKey = "\xe6\x60\x21\xd5\xeb\x8e\x4f\x40\x66\xd4\xad\xb9\xc3\x35\x60\xe4\xf4\x6e\x44\xbb\x3d\xa0\x01\x5c\x94\xf7\x08\x87\x36\x86\x42\x00" + , vecNonce = "\xe0\xea\xf5\x28\x4d\x88\x4a\x0e\x77\xd3\x16\x46" + , vecTag = "\x16\x9f\xbb\x2f\xbf\x38\x9a\x99\x5f\x63\x90\xaf\x22\x22\x8a\x62" + , vecCiphertext = "" + } + , Vector + { vecPlaintext = "\x67\x1f\xdd" + , vecAAD = "\x4f\xbd\xc6\x6f\x14" + , vecKey = "\xba\xe8\xe3\x7f\xc8\x34\x41\xb1\x60\x34\x56\x6b\x7a\x80\x6c\x46\xbb\x91\xc3\xc5\xae\xdb\x64\xa6\xc5\x90\xbc\x84\xd1\xa5\xe2\x69" + , vecNonce = "\xe4\xb4\x78\x01\xaf\xc0\x57\x7e\x34\x69\x9b\x9e" + , vecTag = "\x93\xda\x9b\xb8\x13\x33\xae\xe0\xc7\x85\xb2\x40\xd3\x19\x71\x9d" + , vecCiphertext = "\x0e\xac\xcb" + } + , Vector + { vecPlaintext = "\x19\x54\x95\x86\x0f\x04" + , vecAAD = "\x67\x87\xf3\xea\x22\xc1\x27\xaa\xf1\x95" + , vecKey = "\x65\x45\xfc\x88\x0c\x94\xa9\x51\x98\x87\x42\x96\xd5\xcc\x1f\xd1\x61\x32\x0b\x69\x20\xce\x07\x78\x7f\x86\x74\x3b\x27\x5d\x1a\xb3" + , vecNonce = "\x2f\x6d\x1f\x04\x34\xd8\x84\x8c\x11\x77\x44\x1f" + , vecTag = "\x6b\x62\xb8\x4d\xc4\x0c\x84\x63\x6a\x5e\xc1\x20\x20\xec\x8c\x2c" + , vecCiphertext = "\xa2\x54\xda\xd4\xf3\xf9" + } + , Vector + { vecPlaintext = "\xc9\x88\x2e\x53\x86\xfd\x9f\x92\xec" + , vecAAD = "\x48\x9c\x8f\xde\x2b\xe2\xcf\x97\xe7\x4e\x93\x2d\x4e\xd8\x7d" + , vecKey = "\xd1\x89\x47\x28\xb3\xfe\xd1\x47\x3c\x52\x8b\x84\x26\xa5\x82\x99\x59\x29\xa1\x49\x9e\x9a\xd8\x78\x0c\x8d\x63\xd0\xab\x41\x49\xc0" + , vecNonce = "\x9f\x57\x2c\x61\x4b\x47\x45\x91\x44\x74\xe7\xc7" + , vecTag = "\xc0\xfd\x3d\xc6\x62\x8d\xfe\x55\xeb\xb0\xb9\xfb\x22\x95\xc8\xc2" + , vecCiphertext = "\x0d\xf9\xe3\x08\x67\x82\x44\xc4\x4b" + } + , Vector + { vecPlaintext = "\x1d\xb2\x31\x6f\xd5\x68\x37\x8d\xa1\x07\xb5\x2b" + , vecAAD = "\x0d\xa5\x52\x10\xcc\x1c\x1b\x0a\xbd\xe3\xb2\xf2\x04\xd1\xe9\xf8\xb0\x6b\xc4\x7f" + , vecKey = "\xa4\x41\x02\x95\x2e\xf9\x4b\x02\xb8\x05\x24\x9b\xac\x80\xe6\xf6\x14\x55\xbf\xac\x83\x08\xa2\xd4\x0d\x8c\x84\x51\x17\x80\x82\x35" + , vecNonce = "\x5c\x9e\x94\x0f\xea\x2f\x58\x29\x50\xa7\x0d\x5a" + , vecTag = "\x40\x40\x99\xc2\x58\x7f\x64\x97\x9f\x21\x82\x67\x06\xd4\x97\xd5" + , vecCiphertext = "\x8d\xbe\xb9\xf7\x25\x5b\xf5\x76\x9d\xd5\x66\x92" + } + , Vector + { vecPlaintext = "\x21\x70\x2d\xe0\xde\x18\xba\xa9\xc9\x59\x62\x91\xb0\x84\x66" + , vecAAD = "\xf3\x7d\xe2\x1c\x7f\xf9\x01\xcf\xe8\xa6\x96\x15\xa9\x3f\xdf\x7a\x98\xca\xd4\x81\x79\x62\x45\x70\x9f" + , vecKey = "\x97\x45\xb3\xd1\xae\x06\x55\x6f\xb6\xaa\x78\x90\xbe\xbc\x18\xfe\x6b\x3d\xb4\xda\x3d\x57\xaa\x94\x84\x2b\x98\x03\xa9\x6e\x07\xfb" + , vecNonce = "\x6d\xe7\x18\x60\xf7\x62\xeb\xfb\xd0\x82\x84\xe4" + , vecTag = "\xb3\x08\x0d\x28\xf6\xeb\xb5\xd3\x64\x8c\xe9\x7b\xd5\xba\x67\xfd" + , vecCiphertext = "\x79\x35\x76\xdf\xa5\xc0\xf8\x87\x29\xa7\xed\x3c\x2f\x1b\xff" + } + , Vector + { vecPlaintext = "\xb2\x02\xb3\x70\xef\x97\x68\xec\x65\x61\xc4\xfe\x6b\x7e\x72\x96\xfa\x85" + , vecAAD = "\x9c\x21\x59\x05\x8b\x1f\x0f\xe9\x14\x33\xa5\xbd\xc2\x0e\x21\x4e\xab\x7f\xec\xef\x44\x54\xa1\x0e\xf0\x65\x7d\xf2\x1a\xc7" + , vecKey = "\xb1\x88\x53\xf6\x8d\x83\x36\x40\xe4\x2a\x3c\x02\xc2\x5b\x64\x86\x9e\x14\x6d\x7b\x23\x39\x87\xbd\xdf\xc2\x40\x87\x1d\x75\x76\xf7" + , vecNonce = "\x02\x8e\xc6\xeb\x5e\xa7\xe2\x98\x34\x2a\x94\xd4" + , vecTag = "\x45\x4f\xc2\xa1\x54\xfe\xa9\x1f\x83\x63\xa3\x9f\xec\x7d\x0a\x49" + , vecCiphertext = "\x85\x7e\x16\xa6\x49\x15\xa7\x87\x63\x76\x87\xdb\x4a\x95\x19\x63\x5c\xdd" + } + , Vector + { vecPlaintext = "\xce\xd5\x32\xce\x41\x59\xb0\x35\x27\x7d\x4d\xfb\xb7\xdb\x62\x96\x8b\x13\xcd\x4e\xec" + , vecAAD = "\x73\x43\x20\xcc\xc9\xd9\xbb\xbb\x19\xcb\x81\xb2\xaf\x4e\xcb\xc3\xe7\x28\x34\x32\x1f\x7a\xa0\xf7\x0b\x72\x82\xb4\xf3\x3d\xf2\x3f\x16\x75\x41" + , vecKey = "\x3c\x53\x5d\xe1\x92\xea\xed\x38\x22\xa2\xfb\xbe\x2c\xa9\xdf\xc8\x82\x55\xe1\x4a\x66\x1b\x8a\xa8\x2c\xc5\x42\x36\x09\x3b\xbc\x23" + , vecNonce = "\x68\x80\x89\xe5\x55\x40\xdb\x18\x72\x50\x4e\x1c" + , vecTag = "\x9d\x6c\x70\x29\x67\x5b\x89\xea\xf4\xba\x1d\xed\x1a\x28\x65\x94" + , vecCiphertext = "\x62\x66\x60\xc2\x6e\xa6\x61\x2f\xb1\x7a\xd9\x1e\x8e\x76\x76\x39\xed\xd6\xc9\xfa\xee" + } + ] + +vectorsWrap256 :: [Vector AES256] +vectorsWrap256 = + [ Vector + { vecPlaintext = "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x4d\xb9\x23\xdc\x79\x3e\xe6\x49\x7c\x76\xdc\xc0\x3a\x98\xe1\x08" + , vecAAD = "" + , vecKey = "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\xff\xff\xff\xff\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecCiphertext = "\xf3\xf8\x0f\x2c\xf0\xcb\x2d\xd9\xc5\x98\x4f\xcd\xa9\x08\x45\x6c\xc5\x37\x70\x3b\x5b\xa7\x03\x24\xa6\x79\x3a\x7b\xf2\x18\xd3\xea" + } + , Vector + { vecPlaintext = "\xeb\x36\x40\x27\x7c\x7f\xfd\x13\x03\xc7\xa5\x42\xd0\x2d\x3e\x4c\x00\x00\x00\x00\x00\x00\x00\x00" + , vecAAD = "" + , vecKey = "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecNonce = "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecTag = "\xff\xff\xff\xff\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + , vecCiphertext = "\x18\xce\x4f\x0b\x8c\xb4\xd0\xca\xc6\x5f\xea\x8f\x79\x25\x7b\x20\x88\x8e\x53\xe7\x22\x99\xe5\x6d" + } + ] + +makeEncryptionTest :: BlockCipher128 aes => Int -> Vector aes -> TestTree +makeEncryptionTest i vec@Vector{..} = + testCase (show i) $ + (t, vecCiphertext) @=? encrypt (vecCipher vec) n vecAAD vecPlaintext + where t = AuthTag (B.convert vecTag) + n = throwCryptoError (nonce vecNonce) + +makeDecryptionTest :: BlockCipher128 aes => Int -> Vector aes -> TestTree +makeDecryptionTest i vec@Vector{..} = + testCase (show i) $ + Just vecPlaintext @=? decrypt (vecCipher vec) n vecAAD vecCiphertext t + where t = AuthTag (B.convert vecTag) + n = throwCryptoError (nonce vecNonce) + +katTests :: TestName + -> (forall c . BlockCipher128 c => Int -> Vector c -> TestTree) + -> TestTree +katTests name makeTest = testGroup name + [ testGroup "AES128" $ zipWith makeTest [1..] vectors128 + , testGroup "AES256" $ zipWith makeTest [1..] vectors256 + , testGroup "CounterWrap" $ zipWith makeTest [1..] vectorsWrap256 + ] + +newtype Key c = Key ByteString + deriving (Show,Eq) + +instance Arbitrary (Key AES128) where + arbitrary = Key <$> arbitraryBS 16 + +instance Arbitrary (Key AES256) where + arbitrary = Key <$> arbitraryBS 32 + +instance Arbitrary Nonce where + arbitrary = throwCryptoError . nonce <$> arbitraryBS 12 + +encDecTest :: BlockCipher128 c + => Proxy c -> Key c -> Nonce + -> ArbitraryBS0_2901 -> ArbitraryBS0_2901 -> Property +encDecTest prx (Key key) iv (ArbitraryBS0_2901 aad) (ArbitraryBS0_2901 input) = + let c = throwCryptoError (cipherInit key) `asProxyTypeOf` prx + (tag, ciphertext) = encrypt c iv aad input + in decrypt c iv aad ciphertext tag === Just input + +tests :: TestTree +tests = testGroup "AES-GCM-SIV" + [ testGroup "KATs" + [ katTests "encrypt" makeEncryptionTest + , katTests "decrypt" makeDecryptionTest + ] + , testGroup "properties" + [ testProperty "AES128" $ encDecTest (Proxy :: Proxy AES128) + , testProperty "AES256" $ encDecTest (Proxy :: Proxy AES256) + ] + ] diff --git a/tests/KAT_AFIS.hs b/tests/KAT_AFIS.hs --- a/tests/KAT_AFIS.hs +++ b/tests/KAT_AFIS.hs @@ -23,8 +23,8 @@ ) ] -mergeKATs = map toProp $ zip mergeVec [(0 :: Int)..] - where toProp ((nbExpands, hashAlg, expected, dat), i) = +mergeKATs = zipWith toProp mergeVec [(0 :: Int)..] + where toProp (nbExpands, hashAlg, expected, dat) i = testCase ("merge " ++ show i) (expected @=? AFIS.merge hashAlg nbExpands dat) data AFISParams = AFISParams B.ByteString Int SHA1 ChaChaDRG diff --git a/tests/KAT_Argon2.hs b/tests/KAT_Argon2.hs --- a/tests/KAT_Argon2.hs +++ b/tests/KAT_Argon2.hs @@ -28,9 +28,9 @@ ] kdfTests :: [TestTree] -kdfTests = map toKDFTest $ zip is vectors +kdfTests = zipWith toKDFTest is vectors where - toKDFTest (i, v) = + toKDFTest i v = testCase (show i) (CryptoPassed (kdfResult v) @=? Argon2.hash (kdfOptions v) (kdfPass v) (kdfSalt v) (B.length $ kdfResult v)) diff --git a/tests/KAT_Ed25519.hs b/tests/KAT_Ed25519.hs --- a/tests/KAT_Ed25519.hs +++ b/tests/KAT_Ed25519.hs @@ -47,18 +47,18 @@ ] -doPublicKeyTest (i, vec) = testCase (show i) (pub @=? Ed25519.toPublic sec) +doPublicKeyTest i vec = testCase (show i) (pub @=? Ed25519.toPublic sec) where !pub = throwCryptoError $ Ed25519.publicKey (vecPub vec) !sec = throwCryptoError $ Ed25519.secretKey (vecSec vec) -doSignatureTest (i, vec) = testCase (show i) (sig @=? Ed25519.sign sec pub (vecMsg vec)) +doSignatureTest i vec = testCase (show i) (sig @=? Ed25519.sign sec pub (vecMsg vec)) where !sig = throwCryptoError $ Ed25519.signature (vecSig vec) !pub = throwCryptoError $ Ed25519.publicKey (vecPub vec) !sec = throwCryptoError $ Ed25519.secretKey (vecSec vec) -doVerifyTest (i, vec) = testCase (show i) (True @=? Ed25519.verify pub (vecMsg vec) sig) +doVerifyTest i vec = testCase (show i) (True @=? Ed25519.verify pub (vecMsg vec) sig) where !sig = throwCryptoError $ Ed25519.signature (vecSig vec) !pub = throwCryptoError $ Ed25519.publicKey (vecPub vec) @@ -66,7 +66,7 @@ tests = testGroup "Ed25519" [ testCase "gen secretkey" (Ed25519.generateSecretKey *> pure ()) - , testGroup "gen publickey" $ map doPublicKeyTest (zip [katZero..] vectors) - , testGroup "gen signature" $ map doSignatureTest (zip [katZero..] vectors) - , testGroup "verify sig" $ map doVerifyTest (zip [katZero..] vectors) + , testGroup "gen publickey" $ zipWith doPublicKeyTest [katZero..] vectors + , testGroup "gen signature" $ zipWith doSignatureTest [katZero..] vectors + , testGroup "verify sig" $ zipWith doVerifyTest [katZero..] vectors ] diff --git a/tests/KAT_Ed448.hs b/tests/KAT_Ed448.hs --- a/tests/KAT_Ed448.hs +++ b/tests/KAT_Ed448.hs @@ -65,18 +65,18 @@ ] -doPublicKeyTest (i, vec) = testCase (show i) (pub @=? Ed448.toPublic sec) +doPublicKeyTest i vec = testCase (show i) (pub @=? Ed448.toPublic sec) where !pub = throwCryptoError $ Ed448.publicKey (vecPub vec) !sec = throwCryptoError $ Ed448.secretKey (vecSec vec) -doSignatureTest (i, vec) = testCase (show i) (sig @=? Ed448.sign sec pub (vecMsg vec)) +doSignatureTest i vec = testCase (show i) (sig @=? Ed448.sign sec pub (vecMsg vec)) where !sig = throwCryptoError $ Ed448.signature (vecSig vec) !pub = throwCryptoError $ Ed448.publicKey (vecPub vec) !sec = throwCryptoError $ Ed448.secretKey (vecSec vec) -doVerifyTest (i, vec) = testCase (show i) (True @=? Ed448.verify pub (vecMsg vec) sig) +doVerifyTest i vec = testCase (show i) (True @=? Ed448.verify pub (vecMsg vec) sig) where !sig = throwCryptoError $ Ed448.signature (vecSig vec) !pub = throwCryptoError $ Ed448.publicKey (vecPub vec) @@ -84,7 +84,7 @@ tests = testGroup "Ed448" [ testCase "gen secretkey" (Ed448.generateSecretKey *> pure ()) - , testGroup "gen publickey" $ map doPublicKeyTest (zip [katZero..] vectors) - , testGroup "gen signature" $ map doSignatureTest (zip [katZero..] vectors) - , testGroup "verify sig" $ map doVerifyTest (zip [katZero..] vectors) + , testGroup "gen publickey" $ zipWith doPublicKeyTest [katZero..] vectors + , testGroup "gen signature" $ zipWith doSignatureTest [katZero..] vectors + , testGroup "verify sig" $ zipWith doVerifyTest [katZero..] vectors ] diff --git a/tests/KAT_PBKDF2.hs b/tests/KAT_PBKDF2.hs --- a/tests/KAT_PBKDF2.hs +++ b/tests/KAT_PBKDF2.hs @@ -67,21 +67,21 @@ , testGroup "KATs-HMAC-SHA512" (katTests (PBKDF2.prfHMAC SHA512) vectors_hmac_sha512) , testGroup "KATs-HMAC-SHA512 (fast)" (katTestFastPBKDF2_SHA512 vectors_hmac_sha512) ] - where katTests prf vects = map (toKatTest prf) $ zip is vects + where katTests prf = zipWith (toKatTest prf) is - toKatTest prf (i, ((pass, salt, iter, dkLen), output)) = + toKatTest prf i ((pass, salt, iter, dkLen), output) = testCase (show i) (output @=? PBKDF2.generate prf (PBKDF2.Parameters iter dkLen) pass salt) - katTestFastPBKDF2_SHA1 = map toKatTestFastPBKDF2_SHA1 . zip is - toKatTestFastPBKDF2_SHA1 (i, ((pass, salt, iter, dkLen), output)) = + katTestFastPBKDF2_SHA1 = zipWith toKatTestFastPBKDF2_SHA1 is + toKatTestFastPBKDF2_SHA1 i ((pass, salt, iter, dkLen), output) = testCase (show i) (output @=? PBKDF2.fastPBKDF2_SHA1 (PBKDF2.Parameters iter dkLen) pass salt) - katTestFastPBKDF2_SHA256 = map toKatTestFastPBKDF2_SHA256 . zip is - toKatTestFastPBKDF2_SHA256 (i, ((pass, salt, iter, dkLen), output)) = + katTestFastPBKDF2_SHA256 = zipWith toKatTestFastPBKDF2_SHA256 is + toKatTestFastPBKDF2_SHA256 i ((pass, salt, iter, dkLen), output) = testCase (show i) (output @=? PBKDF2.fastPBKDF2_SHA256 (PBKDF2.Parameters iter dkLen) pass salt) - katTestFastPBKDF2_SHA512 = map toKatTestFastPBKDF2_SHA512 . zip is - toKatTestFastPBKDF2_SHA512 (i, ((pass, salt, iter, dkLen), output)) = + katTestFastPBKDF2_SHA512 = zipWith toKatTestFastPBKDF2_SHA512 is + toKatTestFastPBKDF2_SHA512 i ((pass, salt, iter, dkLen), output) = testCase (show i) (output @=? PBKDF2.fastPBKDF2_SHA512 (PBKDF2.Parameters iter dkLen) pass salt) diff --git a/tests/KAT_PubKey.hs b/tests/KAT_PubKey.hs --- a/tests/KAT_PubKey.hs +++ b/tests/KAT_PubKey.hs @@ -25,7 +25,7 @@ , dbMask :: ByteString } -doMGFTest (i, vmgf) = testCase (show i) (dbMask vmgf @=? actual) +doMGFTest i vmgf = testCase (show i) (dbMask vmgf @=? actual) where actual = mgf1 SHA1 (seed vmgf) (B.length $ dbMask vmgf) vectorsMGF = @@ -36,7 +36,7 @@ ] tests = testGroup "PubKey" - [ testGroup "MGF1" $ map doMGFTest (zip [katZero..] vectorsMGF) + [ testGroup "MGF1" $ zipWith doMGFTest [katZero..] vectorsMGF , rsaTests , pssTests , oaepTests diff --git a/tests/KAT_PubKey/DSA.hs b/tests/KAT_PubKey/DSA.hs --- a/tests/KAT_PubKey/DSA.hs +++ b/tests/KAT_PubKey/DSA.hs @@ -331,32 +331,32 @@ , DSA.public_params = pgq vector } -doSignatureTest hashAlg (i, vector) = testCase (show i) (expected @=? actual) +doSignatureTest hashAlg i vector = testCase (show i) (expected @=? actual) where expected = Just $ DSA.Signature (r vector) (s vector) actual = DSA.signWith (k vector) (vectorToPrivate vector) hashAlg (msg vector) -doVerifyTest hashAlg (i, vector) = testCase (show i) (True @=? actual) +doVerifyTest hashAlg i vector = testCase (show i) (True @=? actual) where actual = DSA.verify hashAlg (vectorToPublic vector) (DSA.Signature (r vector) (s vector)) (msg vector) dsaTests = testGroup "DSA" [ testGroup "SHA1" - [ testGroup "signature" $ map (doSignatureTest SHA1) (zip [katZero..] vectorsSHA1) - , testGroup "verify" $ map (doVerifyTest SHA1) (zip [katZero..] vectorsSHA1) + [ testGroup "signature" $ zipWith (doSignatureTest SHA1) [katZero..] vectorsSHA1 + , testGroup "verify" $ zipWith (doVerifyTest SHA1) [katZero..] vectorsSHA1 ] , testGroup "SHA224" - [ testGroup "signature" $ map (doSignatureTest SHA224) (zip [katZero..] vectorsSHA224) - , testGroup "verify" $ map (doVerifyTest SHA224) (zip [katZero..] vectorsSHA224) + [ testGroup "signature" $ zipWith (doSignatureTest SHA224) [katZero..] vectorsSHA224 + , testGroup "verify" $ zipWith (doVerifyTest SHA224) [katZero..] vectorsSHA224 ] , testGroup "SHA256" - [ testGroup "signature" $ map (doSignatureTest SHA256) (zip [katZero..] vectorsSHA256) - , testGroup "verify" $ map (doVerifyTest SHA256) (zip [katZero..] vectorsSHA256) + [ testGroup "signature" $ zipWith (doSignatureTest SHA256) [katZero..] vectorsSHA256 + , testGroup "verify" $ zipWith (doVerifyTest SHA256) [katZero..] vectorsSHA256 ] , testGroup "SHA384" - [ testGroup "signature" $ map (doSignatureTest SHA384) (zip [katZero..] vectorsSHA384) - , testGroup "verify" $ map (doVerifyTest SHA384) (zip [katZero..] vectorsSHA384) + [ testGroup "signature" $ zipWith (doSignatureTest SHA384) [katZero..] vectorsSHA384 + , testGroup "verify" $ zipWith (doVerifyTest SHA384) [katZero..] vectorsSHA384 ] , testGroup "SHA512" - [ testGroup "signature" $ map (doSignatureTest SHA512) (zip [katZero..] vectorsSHA512) - , testGroup "verify" $ map (doVerifyTest SHA512) (zip [katZero..] vectorsSHA512) + [ testGroup "signature" $ zipWith (doSignatureTest SHA512) [katZero..] vectorsSHA512 + , testGroup "verify" $ zipWith (doVerifyTest SHA512) [katZero..] vectorsSHA512 ] ] diff --git a/tests/KAT_PubKey/ECC.hs b/tests/KAT_PubKey/ECC.hs --- a/tests/KAT_PubKey/ECC.hs +++ b/tests/KAT_PubKey/ECC.hs @@ -136,7 +136,7 @@ } ] -doPointValidTest (i, vector) = testCase (show i) (valid vector @=? ECC.isPointValid (curve vector) (ECC.Point (x vector) (y vector))) +doPointValidTest i vector = testCase (show i) (valid vector @=? ECC.isPointValid (curve vector) (ECC.Point (x vector) (y vector))) arbitraryPoint :: ECC.Curve -> Gen ECC.Point arbitraryPoint aCurve = @@ -146,7 +146,7 @@ pointGen = ECC.pointBaseMul aCurve <$> choose (1, n - 1) eccTests = testGroup "ECC" - [ testGroup "valid-point" $ map doPointValidTest (zip [katZero..] vectorsPoint) + [ testGroup "valid-point" $ zipWith doPointValidTest [katZero..] vectorsPoint , localOption (QuickCheckTests 20) $ testGroup "property" [ testProperty "point-add" $ \aCurve (QAInteger r1) (QAInteger r2) -> let curveN = ECC.ecc_n . ECC.common_curve $ aCurve diff --git a/tests/KAT_PubKey/ECDSA.hs b/tests/KAT_PubKey/ECDSA.hs --- a/tests/KAT_PubKey/ECDSA.hs +++ b/tests/KAT_PubKey/ECDSA.hs @@ -490,32 +490,32 @@ vectorToPublic :: VectorECDSA -> ECDSA.PublicKey vectorToPublic vector = ECDSA.PublicKey (curve vector) (q vector) -doSignatureTest hashAlg (i, vector) = testCase (show i) (expected @=? actual) +doSignatureTest hashAlg i vector = testCase (show i) (expected @=? actual) where expected = Just $ ECDSA.Signature (r vector) (s vector) actual = ECDSA.signWith (k vector) (vectorToPrivate vector) hashAlg (msg vector) -doVerifyTest hashAlg (i, vector) = testCase (show i) (True @=? actual) +doVerifyTest hashAlg i vector = testCase (show i) (True @=? actual) where actual = ECDSA.verify hashAlg (vectorToPublic vector) (ECDSA.Signature (r vector) (s vector)) (msg vector) ecdsaTests = testGroup "ECDSA" [ testGroup "SHA1" - [ testGroup "signature" $ map (doSignatureTest SHA1) (zip [katZero..] vectorsSHA1) - , testGroup "verify" $ map (doVerifyTest SHA1) (zip [katZero..] vectorsSHA1) + [ testGroup "signature" $ zipWith (doSignatureTest SHA1) [katZero..] vectorsSHA1 + , testGroup "verify" $ zipWith (doVerifyTest SHA1) [katZero..] vectorsSHA1 ] , testGroup "SHA224" - [ testGroup "signature" $ map (doSignatureTest SHA224) (zip [katZero..] rfc6979_vectorsSHA224) - , testGroup "verify" $ map (doVerifyTest SHA224) (zip [katZero..] rfc6979_vectorsSHA224) + [ testGroup "signature" $ zipWith (doSignatureTest SHA224) [katZero..] rfc6979_vectorsSHA224 + , testGroup "verify" $ zipWith (doVerifyTest SHA224) [katZero..] rfc6979_vectorsSHA224 ] , testGroup "SHA256" - [ testGroup "signature" $ map (doSignatureTest SHA256) (zip [katZero..] rfc6979_vectorsSHA256) - , testGroup "verify" $ map (doVerifyTest SHA256) (zip [katZero..] rfc6979_vectorsSHA256) + [ testGroup "signature" $ zipWith (doSignatureTest SHA256) [katZero..] rfc6979_vectorsSHA256 + , testGroup "verify" $ zipWith (doVerifyTest SHA256) [katZero..] rfc6979_vectorsSHA256 ] , testGroup "SHA384" - [ testGroup "signature" $ map (doSignatureTest SHA384) (zip [katZero..] rfc6979_vectorsSHA384) - , testGroup "verify" $ map (doVerifyTest SHA384) (zip [katZero..] rfc6979_vectorsSHA384) + [ testGroup "signature" $ zipWith (doSignatureTest SHA384) [katZero..] rfc6979_vectorsSHA384 + , testGroup "verify" $ zipWith (doVerifyTest SHA384) [katZero..] rfc6979_vectorsSHA384 ] , testGroup "SHA512" - [ testGroup "signature" $ map (doSignatureTest SHA512) (zip [katZero..] rfc6979_vectorsSHA512) - , testGroup "verify" $ map (doVerifyTest SHA512) (zip [katZero..] rfc6979_vectorsSHA512) + [ testGroup "signature" $ zipWith (doSignatureTest SHA512) [katZero..] rfc6979_vectorsSHA512 + , testGroup "verify" $ zipWith (doVerifyTest SHA512) [katZero..] rfc6979_vectorsSHA512 ] ] diff --git a/tests/KAT_PubKey/OAEP.hs b/tests/KAT_PubKey/OAEP.hs --- a/tests/KAT_PubKey/OAEP.hs +++ b/tests/KAT_PubKey/OAEP.hs @@ -81,17 +81,17 @@ } ] -doEncryptionTest key (i, vec) = testCase (show i) (Right (cipherText vec) @=? actual) - where actual = OAEP.encryptWithSeed (seed vec) (OAEP.defaultOAEPParams SHA1) key (message vec) +doEncryptionTest key i vec = testCase (show i) (Right (cipherText vec) @=? actual) + where actual = OAEP.encryptWithSeed (seed vec) (OAEP.defaultOAEPParams SHA1) key (message vec) -doDecryptionTest key (i, vec) = testCase (show i) (Right (message vec) @=? actual) +doDecryptionTest key i vec = testCase (show i) (Right (message vec) @=? actual) where actual = OAEP.decrypt Nothing (OAEP.defaultOAEPParams SHA1) key (cipherText vec) oaepTests = testGroup "RSA-OAEP" [ testGroup "internal" - [ doEncryptionTest (private_pub rsaKeyInt) (0 :: Int, vectorInt) - , doDecryptionTest rsaKeyInt (0 :: Int, vectorInt) + [ doEncryptionTest (private_pub rsaKeyInt) (0 :: Int) vectorInt + , doDecryptionTest rsaKeyInt (0 :: Int) vectorInt ] - , testGroup "encryption key 1024 bits" $ map (doEncryptionTest $ private_pub rsaKey1) (zip [katZero..] vectorsKey1) - , testGroup "decryption key 1024 bits" $ map (doDecryptionTest rsaKey1) (zip [katZero..] vectorsKey1) + , testGroup "encryption key 1024 bits" $ zipWith (doEncryptionTest $ private_pub rsaKey1) [katZero..] vectorsKey1 + , testGroup "decryption key 1024 bits" $ zipWith (doDecryptionTest rsaKey1) [katZero..] vectorsKey1 ] diff --git a/tests/KAT_PubKey/P256.hs b/tests/KAT_PubKey/P256.hs --- a/tests/KAT_PubKey/P256.hs +++ b/tests/KAT_PubKey/P256.hs @@ -54,6 +54,9 @@ unP256 :: P256Scalar -> Integer unP256 (P256Scalar r) = r +modP256Scalar :: P256Scalar -> P256Scalar +modP256Scalar (P256Scalar r) = P256Scalar (r `mod` curveN) + p256ScalarToInteger :: P256.Scalar -> Integer p256ScalarToInteger s = os2ip (P256.scalarToBinary s :: Bytes) @@ -92,10 +95,28 @@ let v = unP256 r `mod` curveN v' = P256.scalarSub (unP256Scalar r) P256.scalarZero in v `propertyEq` p256ScalarToInteger v' + , testProperty "mul" $ \r1 r2 -> + let r = (unP256 r1 * unP256 r2) `mod` curveN + r' = P256.scalarMul (unP256Scalar r1) (unP256Scalar r2) + in r `propertyEq` p256ScalarToInteger r' , testProperty "inv" $ \r' -> let inv = inverseCoprimes (unP256 r') curveN inv' = P256.scalarInv (unP256Scalar r') - in if unP256 r' == 0 then True else inv `propertyEq` p256ScalarToInteger inv' + in unP256 r' /= 0 ==> inv `propertyEq` p256ScalarToInteger inv' + , testProperty "inv-safe" $ \r' -> + let inv = P256.scalarInv (unP256Scalar r') + inv' = P256.scalarInvSafe (unP256Scalar r') + in unP256 r' /= 0 ==> inv `propertyEq` inv' + , testProperty "inv-safe-mul" $ \r' -> + let inv = P256.scalarInvSafe (unP256Scalar r') + res = P256.scalarMul (unP256Scalar r') inv + in unP256 r' /= 0 ==> 1 `propertyEq` p256ScalarToInteger res + , testProperty "inv-safe-zero" $ + let inv0 = P256.scalarInvSafe P256.scalarZero + invN = P256.scalarInvSafe P256.scalarN + in propertyHold [ eqTest "scalarZero" P256.scalarZero inv0 + , eqTest "scalarN" P256.scalarZero invN + ] ] , testGroup "point" [ testProperty "marshalling" $ \rx ry -> @@ -115,9 +136,16 @@ t = P256.pointFromIntegers (xT, yT) r = P256.pointFromIntegers (xR, yR) in r @=? P256.pointAdd s t - , testProperty "lift-to-curve" $ propertyLiftToCurve - , testProperty "point-add" $ propertyPointAdd - , testProperty "point-negate" $ propertyPointNegate + , testProperty "lift-to-curve" propertyLiftToCurve + , testProperty "point-add" propertyPointAdd + , testProperty "point-negate" propertyPointNegate + , testProperty "point-mul" propertyPointMul + , testProperty "infinity" $ + let gN = P256.toPoint P256.scalarN + g1 = P256.pointBase + in propertyHold [ eqTest "zero" True (P256.pointIsAtInfinity gN) + , eqTest "base" False (P256.pointIsAtInfinity g1) + ] ] ] where @@ -146,4 +174,15 @@ let p = P256.toPoint (unP256Scalar r) pe = ECC.pointMul curve (unP256 r) curveGen pR = P256.pointNegate p - in ECC.pointNegate curve pe `propertyEq` (pointP256ToECC pR) + in ECC.pointNegate curve pe `propertyEq` pointP256ToECC pR + + propertyPointMul s' r' = + let s = modP256Scalar s' + r = modP256Scalar r' + p = P256.toPoint (unP256Scalar r) + pe = ECC.pointMul curve (unP256 r) curveGen + pR = P256.toPoint (P256.scalarMul (unP256Scalar s) (unP256Scalar r)) + peR = ECC.pointMul curve (unP256 s) pe + in propertyHold [ eqTest "p256" pR (P256.pointMul (unP256Scalar s) p) + , eqTest "ecc" peR (pointP256ToECC pR) + ] diff --git a/tests/KAT_PubKey/PSS.hs b/tests/KAT_PubKey/PSS.hs --- a/tests/KAT_PubKey/PSS.hs +++ b/tests/KAT_PubKey/PSS.hs @@ -326,23 +326,23 @@ } ] -doSignTest key (i, vector) = testCase (show i) (Right (signature vector) @=? actual) +doSignTest key i vector = testCase (show i) (Right (signature vector) @=? actual) where actual = PSS.signWithSalt (salt vector) Nothing PSS.defaultPSSParamsSHA1 key (message vector) -doVerifyTest key (i, vector) = testCase (show i) (True @=? actual) +doVerifyTest key i vector = testCase (show i) (True @=? actual) where actual = PSS.verify PSS.defaultPSSParamsSHA1 (private_pub key) (message vector) (signature vector) pssTests = testGroup "RSA-PSS" [ testGroup "signature internal" - [ doSignTest rsaKeyInt (katZero, vectorInt) ] + [ doSignTest rsaKeyInt katZero vectorInt ] , testGroup "verify internal" - [ doVerifyTest rsaKeyInt (katZero, vectorInt) ] - , testGroup "signature key 1024" $ map (doSignTest rsaKey1) (zip [katZero..] vectorsKey1) - , testGroup "verify key 1024" $ map (doVerifyTest rsaKey1) (zip [katZero..] vectorsKey1) - , testGroup "signature key 1025" $ map (doSignTest rsaKey2) (zip [katZero..] vectorsKey2) - , testGroup "verify key 1025" $ map (doVerifyTest rsaKey2) (zip [katZero..] vectorsKey2) - , testGroup "signature key 1026" $ map (doSignTest rsaKey3) (zip [katZero..] vectorsKey3) - , testGroup "verify key 1026" $ map (doVerifyTest rsaKey3) (zip [katZero..] vectorsKey3) - , testGroup "signature key 1031" $ map (doSignTest rsaKey8) (zip [katZero..] vectorsKey8) - , testGroup "verify key 1031" $ map (doVerifyTest rsaKey8) (zip [katZero..] vectorsKey8) + [ doVerifyTest rsaKeyInt katZero vectorInt ] + , testGroup "signature key 1024" $ zipWith (doSignTest rsaKey1) [katZero..] vectorsKey1 + , testGroup "verify key 1024" $ zipWith (doVerifyTest rsaKey1) [katZero..] vectorsKey1 + , testGroup "signature key 1025" $ zipWith (doSignTest rsaKey2) [katZero..] vectorsKey2 + , testGroup "verify key 1025" $ zipWith (doVerifyTest rsaKey2) [katZero..] vectorsKey2 + , testGroup "signature key 1026" $ zipWith (doSignTest rsaKey3) [katZero..] vectorsKey3 + , testGroup "verify key 1026" $ zipWith (doVerifyTest rsaKey3) [katZero..] vectorsKey3 + , testGroup "signature key 1031" $ zipWith (doSignTest rsaKey8) [katZero..] vectorsKey8 + , testGroup "verify key 1031" $ zipWith (doVerifyTest rsaKey8) [katZero..] vectorsKey8 ] diff --git a/tests/KAT_PubKey/RSA.hs b/tests/KAT_PubKey/RSA.hs --- a/tests/KAT_PubKey/RSA.hs +++ b/tests/KAT_PubKey/RSA.hs @@ -86,17 +86,17 @@ vectorHasSignature :: VectorRSA -> Bool vectorHasSignature = isRight . sig -doSignatureTest (i, vector) = testCase (show i) (expected @=? actual) +doSignatureTest i vector = testCase (show i) (expected @=? actual) where expected = sig vector actual = RSA.sign Nothing (Just SHA1) (vectorToPrivate vector) (msg vector) -doVerifyTest (i, vector) = testCase (show i) (True @=? actual) +doVerifyTest i vector = testCase (show i) (True @=? actual) where actual = RSA.verify (Just SHA1) (vectorToPublic vector) (msg vector) bs Right bs = sig vector rsaTests = testGroup "RSA" [ testGroup "SHA1" - [ testGroup "signature" $ map doSignatureTest (zip [katZero..] vectorsSHA1) - , testGroup "verify" $ map doVerifyTest $ filter (vectorHasSignature . snd) (zip [katZero..] vectorsSHA1) + [ testGroup "signature" $ zipWith doSignatureTest [katZero..] vectorsSHA1 + , testGroup "verify" $ zipWith doVerifyTest [katZero..] $ filter vectorHasSignature vectorsSHA1 ] ] diff --git a/tests/KAT_PubKey/Rabin.hs b/tests/KAT_PubKey/Rabin.hs --- a/tests/KAT_PubKey/Rabin.hs +++ b/tests/KAT_PubKey/Rabin.hs @@ -95,51 +95,51 @@ } ] -doBasicRabinEncryptTest key (i, vector) = testCase (show i) (Right (cipherText vector) @=? actual) +doBasicRabinEncryptTest key i vector = testCase (show i) (Right (cipherText vector) @=? actual) where actual = BRabin.encryptWithSeed (seed vector) (OAEP.defaultOAEPParams SHA1) key (plainText vector) -doBasicRabinDecryptTest key (i, vector) = testCase (show i) (Just (plainText vector) @=? actual) +doBasicRabinDecryptTest key i vector = testCase (show i) (Just (plainText vector) @=? actual) where actual = BRabin.decrypt (OAEP.defaultOAEPParams SHA1) key (cipherText vector) -doBasicRabinSignTest key (i, vector) = testCase (show i) (Right (BRabin.Signature ((os2ip $ padding vector), (signature vector))) @=? actual) +doBasicRabinSignTest key i vector = testCase (show i) (Right (BRabin.Signature ((os2ip $ padding vector), (signature vector))) @=? actual) where actual = BRabin.signWith (padding vector) key SHA1 (message vector) -doBasicRabinVerifyTest key (i, vector) = testCase (show i) (True @=? actual) +doBasicRabinVerifyTest key i vector = testCase (show i) (True @=? actual) where actual = BRabin.verify key SHA1 (message vector) (BRabin.Signature ((os2ip $ padding vector), (signature vector))) -doModifiedRabinSignTest key (i, vector) = testCase (show i) (Right (signature vector) @=? actual) +doModifiedRabinSignTest key i vector = testCase (show i) (Right (signature vector) @=? actual) where actual = MRabin.sign key SHA1 (message vector) -doModifiedRabinVerifyTest key (i, vector) = testCase (show i) (True @=? actual) +doModifiedRabinVerifyTest key i vector = testCase (show i) (True @=? actual) where actual = MRabin.verify key SHA1 (message vector) (signature vector) -doRwEncryptTest key (i, vector) = testCase (show i) (Right (cipherText vector) @=? actual) - where actual = RW.encryptWithSeed (seed vector) (OAEP.defaultOAEPParams SHA1) key (plainText vector) +doRwEncryptTest key i vector = testCase (show i) (Right (cipherText vector) @=? actual) + where actual = RW.encryptWithSeed (seed vector) (OAEP.defaultOAEPParams SHA1) key (plainText vector) -doRwDecryptTest key (i, vector) = testCase (show i) (Just (plainText vector) @=? actual) +doRwDecryptTest key i vector = testCase (show i) (Just (plainText vector) @=? actual) where actual = RW.decrypt (OAEP.defaultOAEPParams SHA1) key (cipherText vector) -doRwSignTest key (i, vector) = testCase (show i) (Right (signature vector) @=? actual) +doRwSignTest key i vector = testCase (show i) (Right (signature vector) @=? actual) where actual = RW.sign key SHA1 (message vector) -doRwVerifyTest key (i, vector) = testCase (show i) (True @=? actual) +doRwVerifyTest key i vector = testCase (show i) (True @=? actual) where actual = RW.verify key SHA1 (message vector) (signature vector) rabinTests = testGroup "Rabin" [ testGroup "Basic" - [ testGroup "encrypt" $ map (doBasicRabinEncryptTest $ BRabin.private_pub basicRabinKey) (zip [katZero..] basicRabinEncryptionVectors) - , testGroup "decrypt" $ map (doBasicRabinDecryptTest $ basicRabinKey) (zip [katZero..] basicRabinEncryptionVectors) - , testGroup "sign" $ map (doBasicRabinSignTest $ basicRabinKey) (zip [katZero..] basicRabinSignatureVectors) - , testGroup "verify" $ map (doBasicRabinVerifyTest $ BRabin.private_pub basicRabinKey) (zip [katZero..] basicRabinSignatureVectors) + [ testGroup "encrypt" $ zipWith (doBasicRabinEncryptTest $ BRabin.private_pub basicRabinKey) [katZero..] basicRabinEncryptionVectors + , testGroup "decrypt" $ zipWith (doBasicRabinDecryptTest basicRabinKey) [katZero..] basicRabinEncryptionVectors + , testGroup "sign" $ zipWith (doBasicRabinSignTest basicRabinKey) [katZero..] basicRabinSignatureVectors + , testGroup "verify" $ zipWith (doBasicRabinVerifyTest $ BRabin.private_pub basicRabinKey) [katZero..] basicRabinSignatureVectors ] , testGroup "Modified" - [ testGroup "sign" $ map (doModifiedRabinSignTest $ modifiedRabinKey) (zip [katZero..] modifiedRabinSignatureVectors) - , testGroup "verify" $ map (doModifiedRabinVerifyTest $ MRabin.private_pub modifiedRabinKey) (zip [katZero..] modifiedRabinSignatureVectors) + [ testGroup "sign" $ zipWith (doModifiedRabinSignTest modifiedRabinKey) [katZero..] modifiedRabinSignatureVectors + , testGroup "verify" $ zipWith (doModifiedRabinVerifyTest $ MRabin.private_pub modifiedRabinKey) [katZero..] modifiedRabinSignatureVectors ] , testGroup "RW" - [ testGroup "encrypt" $ map (doRwEncryptTest $ RW.private_pub rwKey) (zip [katZero..] rwEncryptionVectors) - , testGroup "decrypt" $ map (doRwDecryptTest $ rwKey) (zip [katZero..] rwEncryptionVectors) - , testGroup "sign" $ map (doRwSignTest $ rwKey) (zip [katZero..] rwSignatureVectors) - , testGroup "verify" $ map (doRwVerifyTest $ RW.private_pub rwKey) (zip [katZero..] rwSignatureVectors) + [ testGroup "encrypt" $ zipWith (doRwEncryptTest $ RW.private_pub rwKey) [katZero..] rwEncryptionVectors + , testGroup "decrypt" $ zipWith (doRwDecryptTest rwKey) [katZero..] rwEncryptionVectors + , testGroup "sign" $ zipWith (doRwSignTest rwKey) [katZero..] rwSignatureVectors + , testGroup "verify" $ zipWith (doRwVerifyTest $ RW.private_pub rwKey) [katZero..] rwSignatureVectors ] ] diff --git a/tests/KAT_RC4.hs b/tests/KAT_RC4.hs --- a/tests/KAT_RC4.hs +++ b/tests/KAT_RC4.hs @@ -27,8 +27,8 @@ ] tests = testGroup "RC4" - $ map toKatTest $ zip is vectors - where toKatTest (i, (key, plainText, cipherText)) = + $ zipWith toKatTest is vectors + where toKatTest i (key, plainText, cipherText) = testCase (show i) (cipherText @=? snd (RC4.combine (RC4.initialize key) plainText)) is :: [Int] is = [1..] diff --git a/tests/KAT_Scrypt.hs b/tests/KAT_Scrypt.hs --- a/tests/KAT_Scrypt.hs +++ b/tests/KAT_Scrypt.hs @@ -28,6 +28,6 @@ ] tests = testGroup "Scrypt" - $ map toCase $ zip [(1::Int)..] vectors - where toCase (i, ((pass,salt,n,r,p,dklen), output)) = + $ zipWith toCase [(1::Int)..] vectors + where toCase i ((pass,salt,n,r,p,dklen), output) = testCase (show i) (output @=? Scrypt.generate (Scrypt.Parameters n r p dklen) pass salt) diff --git a/tests/Number.hs b/tests/Number.hs --- a/tests/Number.hs +++ b/tests/Number.hs @@ -10,6 +10,7 @@ import qualified Crypto.Number.Serialize as BE import qualified Crypto.Number.Serialize.LE as LE import Crypto.Number.Prime +import Crypto.Number.ModArithmetic import Data.Bits serializationVectors :: [(Int, Integer, ByteString)] @@ -55,6 +56,17 @@ , testProperty "as-power-of-2-and-odd" $ \n -> let (e, a1) = asPowerOf2AndOdd n in n == (2^e)*a1 + , testProperty "squareRoot" $ \testDRG (Int0_2901 baseBits') -> do + let baseBits = baseBits' `mod` 500 + bits = 5 + baseBits -- generating lower than 5 bits causes an error .. + p = withTestDRG testDRG $ generatePrime bits + g <- choose (1, p - 1) + let square x = (x * x) `mod` p + r = square <$> squareRoot p g + case jacobi g p of + Just 1 -> return $ Just g `assertEq` r + Just (-1) -> return $ Nothing `assertEq` r + _ -> error "invalid jacobi result" , testProperty "marshalling-be" $ \qaInt -> getQAInteger qaInt == BE.os2ip (BE.i2osp (getQAInteger qaInt) :: Bytes) , testProperty "marshalling-le" $ \qaInt -> @@ -67,9 +79,9 @@ getQAInteger qaInt == BE.os2ip (B.reverse (LE.i2osp (getQAInteger qaInt) :: Bytes)) , testProperty "le-rev-be-40" $ \qaInt -> getQAInteger qaInt == BE.os2ip (B.reverse (LE.i2ospOf_ 40 (getQAInteger qaInt) :: Bytes)) - , testGroup "marshalling-kat-to-bytearray" $ map toSerializationKat $ zip [katZero..] serializationVectors - , testGroup "marshalling-kat-to-integer" $ map toSerializationKatInteger $ zip [katZero..] serializationVectors + , testGroup "marshalling-kat-to-bytearray" $ zipWith toSerializationKat [katZero..] serializationVectors + , testGroup "marshalling-kat-to-integer" $ zipWith toSerializationKatInteger [katZero..] serializationVectors ] where - toSerializationKat (i, (sz, n, ba)) = testCase (show i) (ba @=? BE.i2ospOf_ sz n) - toSerializationKatInteger (i, (_, n, ba)) = testCase (show i) (n @=? BE.os2ip ba) + toSerializationKat i (sz, n, ba) = testCase (show i) (ba @=? BE.i2ospOf_ sz n) + toSerializationKatInteger i (_, n, ba) = testCase (show i) (n @=? BE.os2ip ba) diff --git a/tests/Number/F2m.hs b/tests/Number/F2m.hs --- a/tests/Number/F2m.hs +++ b/tests/Number/F2m.hs @@ -2,6 +2,7 @@ import Imports hiding ((.&.)) import Data.Bits +import Data.Maybe import Crypto.Number.Basic (log2) import Crypto.Number.F2m @@ -52,8 +53,34 @@ squareTests = testGroup "squareF2m" [ testProperty "sqr(a) == a * a" $ \(Positive m) (NonNegative a) -> mulF2m m a a == squareF2m m a + -- disabled because we require @m@ to be a suitable modulus and there is no + -- way to guarantee this + -- , testProperty "sqrt(a) * sqrt(a) = a" + -- $ \(Positive m) (NonNegative aa) -> let a = sqrtF2m m aa in mulF2m m a a == modF2m m aa + , testProperty "sqrt(a) * sqrt(a) = a in GF(2^16)" + $ let m = 65581 :: Integer -- x^16 + x^5 + x^3 + x^2 + 1 + nums = [0 .. 65535 :: Integer] + in nums == [let y = sqrtF2m m x in squareF2m m y | x <- nums] ] +powTests = testGroup "powF2m" + [ testProperty "2 is square" + $ \(Positive m) (NonNegative a) -> powF2m m a 2 == squareF2m m a + , testProperty "1 is identity" + $ \(Positive m) (NonNegative a) -> powF2m m a 1 == modF2m m a + , testProperty "0 is annihilator" + $ \(Positive m) (NonNegative a) -> powF2m m a 0 == modF2m m 1 + , testProperty "(a * b) ^ c == (a ^ c) * (b ^ c)" + $ \(Positive m) (NonNegative a) (NonNegative b) (NonNegative c) + -> powF2m m (mulF2m m a b) c == mulF2m m (powF2m m a c) (powF2m m b c) + , testProperty "a ^ (b + c) == (a ^ b) * (a ^ c)" + $ \(Positive m) (NonNegative a) (NonNegative b) (NonNegative c) + -> powF2m m a (b + c) == mulF2m m (powF2m m a b) (powF2m m a c) + , testProperty "a ^ (b * c) == (a ^ b) ^ c" + $ \(Positive m) (NonNegative a) (NonNegative b) (NonNegative c) + -> powF2m m a (b * c) == powF2m m (powF2m m a b) c + ] + invTests = testGroup "invF2m" [ testProperty "1 / a * a == 1" $ \(Positive m) (NonNegative a) @@ -70,7 +97,7 @@ -> divF2m m a b == (mulF2m m a <$> invF2m m b) , testProperty "a * b / b == a" $ \(Positive m) (NonNegative a) (NonNegative b) - -> invF2m m b == Nothing || divF2m m (mulF2m m a b) b == Just (modF2m m a) + -> isNothing (invF2m m b) || divF2m m (mulF2m m a b) b == Just (modF2m m a) ] tests = testGroup "number.F2m" @@ -78,6 +105,7 @@ , modTests , mulTests , squareTests + , powTests , invTests , divTests ] diff --git a/tests/Padding.hs b/tests/Padding.hs --- a/tests/Padding.hs +++ b/tests/Padding.hs @@ -33,6 +33,6 @@ ] tests = testGroup "Padding" - [ testGroup "Cases" $ map (uncurry testPad) (zip [1..] cases) - , testGroup "ZeroCases" $ map (uncurry testZeroPad) (zip [1..] zeroCases) + [ testGroup "Cases" $ zipWith testPad [1..] cases + , testGroup "ZeroCases" $ zipWith testZeroPad [1..] zeroCases ] diff --git a/tests/Salsa.hs b/tests/Salsa.hs --- a/tests/Salsa.hs +++ b/tests/Salsa.hs @@ -37,7 +37,7 @@ tests = testGroup "Salsa" [ testGroup "KAT" $ - map (\(i,f) -> testCase (show (i :: Int)) f) $ zip [1..] $ map (\(r, k,i,e) -> salsaRunSimple e r k i) vectors + zipWith (\i (r,k,n,e) -> testCase (show (i :: Int)) $ salsaRunSimple e r k n) [1..] vectors , testProperty "generate-combine" salsaGenerateCombine , testProperty "chunking-generate" salsaGenerateChunks , testProperty "chunking-combine" salsaCombineChunks diff --git a/tests/Tests.hs b/tests/Tests.hs --- a/tests/Tests.hs +++ b/tests/Tests.hs @@ -3,12 +3,15 @@ import Imports +import Crypto.System.CPU + import qualified Number import qualified Number.F2m import qualified BCrypt import qualified BCryptPBKDF import qualified ECC import qualified ECC.Edwards25519 +import qualified ECDSA import qualified Hash import qualified Poly1305 import qualified Salsa @@ -31,6 +34,7 @@ import qualified KAT_Scrypt -- symmetric cipher -------------------- import qualified KAT_AES +import qualified KAT_AESGCMSIV import qualified KAT_Blowfish import qualified KAT_CAST5 import qualified KAT_Camellia @@ -43,7 +47,10 @@ import qualified Padding tests = testGroup "cryptonite" - [ Number.tests + [ testGroup "runtime" + [ testCaseInfo "CPU" (return $ show processorOptions) + ] + , Number.tests , Number.F2m.tests , Hash.tests , Padding.tests @@ -72,6 +79,7 @@ ] , testGroup "block-cipher" [ KAT_AES.tests + , KAT_AESGCMSIV.tests , KAT_Blowfish.tests , KAT_CAST5.tests , KAT_Camellia.tests @@ -89,6 +97,7 @@ , KAT_AFIS.tests , ECC.tests , ECC.Edwards25519.tests + , ECDSA.tests ] main = defaultMain tests diff --git a/tests/Utils.hs b/tests/Utils.hs --- a/tests/Utils.hs +++ b/tests/Utils.hs @@ -19,7 +19,7 @@ deriving (Show,Eq) instance Arbitrary TestDRG where - arbitrary = TestDRG `fmap` arbitrary + arbitrary = TestDRG `fmap` arbitrary -- distribution not uniform withTestDRG (TestDRG l) f = fst $ withDRG (drgNewTest l) f diff --git a/tests/XSalsa.hs b/tests/XSalsa.hs --- a/tests/XSalsa.hs +++ b/tests/XSalsa.hs @@ -98,11 +98,31 @@ , "\xB2\xB7\x95\xFE\x6C\x1D\x4C\x83\xC1\x32\x7E\x01\x5A\x67\xD4\x46\x5F\xD8\xE3\x28\x13\x57\x5C\xBA\xB2\x63\xE2\x0E\xF0\x58\x64\xD2\xDC\x17\xE0\xE4\xEB\x81\x43\x6A\xDF\xE9\xF6\x38\xDC\xC1\xC8\xD7\x8F\x6B\x03\x06\xBA\xF9\x38\xE5\xD2\xAB\x0B\x3E\x05\xE7\x35\xCC\x6F\xFF\x2D\x6E\x02\xE3\xD6\x04\x84\xBE\xA7\xC7\xA8\xE1\x3E\x23\x19\x7F\xEA\x7B\x04\xD4\x7D\x48\xF4\xA4\xE5\x94\x41\x74\x53\x94\x92\x80\x0D\x3E\xF5\x1E\x2E\xE5\xE4\xC8\xA0\xBD\xF0\x50\xC2\xDD\x3D\xD7\x4F\xCE\x5E\x7E\x5C\x37\x36\x4F\x75\x47\xA1\x14\x80\xA3\x06\x3B\x9A\x0A\x15\x7B\x15\xB1\x0A\x5A\x95\x4D\xE2\x73\x1C\xED\x05\x5A\xA2\xE2\x76\x7F\x08\x91\xD4\x32\x9C\x42\x6F\x38\x08\xEE\x86\x7B\xED\x0D\xC7\x5B\x59\x22\xB7\xCF\xB8\x95\x70\x0F\xDA\x01\x61\x05\xA4\xC7\xB7\xF0\xBB\x90\xF0\x29\xF6\xBB\xCB\x04\xAC\x36\xAC\x16") ] +-- Test vector from paper "Cryptography in NaCl" +vectorsCB :: [Vector] +vectorsCB = + [ ( 20 + , "\x4A\x5D\x9D\x5B\xA4\xCE\x2D\xE1\x72\x8E\x3B\xF4\x80\x35\x0F\x25\xE0\x7E\x21\xC9\x47\xD1\x9E\x33\x76\xF0\x9B\x3C\x1E\x16\x17\x42" + , "\x69\x69\x6E\xE9\x55\xB6\x2B\x73\xCD\x62\xBD\xA8\x75\xFC\x73\xD6\x82\x19\xE0\x03\x6B\x7A\x0B\x37" + , "\xBE\x07\x5F\xC5\x3C\x81\xF2\xD5\xCF\x14\x13\x16\xEB\xEB\x0C\x7B\x52\x28\xC5\x2A\x4C\x62\xCB\xD4\x4B\x66\x84\x9B\x64\x24\x4F\xFC\xE5\xEC\xBA\xAF\x33\xBD\x75\x1A\x1A\xC7\x28\xD4\x5E\x6C\x61\x29\x6C\xDC\x3C\x01\x23\x35\x61\xF4\x1D\xB6\x6C\xCE\x31\x4A\xDB\x31\x0E\x3B\xE8\x25\x0C\x46\xF0\x6D\xCE\xEA\x3A\x7F\xA1\x34\x80\x57\xE2\xF6\x55\x6A\xD6\xB1\x31\x8A\x02\x4A\x83\x8F\x21\xAF\x1F\xDE\x04\x89\x77\xEB\x48\xF5\x9F\xFD\x49\x24\xCA\x1C\x60\x90\x2E\x52\xF0\xA0\x89\xBC\x76\x89\x70\x40\xE0\x82\xF9\x37\x76\x38\x48\x64\x5E\x07\x05" + , "\x8E\x99\x3B\x9F\x48\x68\x12\x73\xC2\x96\x50\xBA\x32\xFC\x76\xCE\x48\x33\x2E\xA7\x16\x4D\x96\xA4\x47\x6F\xB8\xC5\x31\xA1\x18\x6A\xC0\xDF\xC1\x7C\x98\xDC\xE8\x7B\x4D\xA7\xF0\x11\xEC\x48\xC9\x72\x71\xD2\xC2\x0F\x9B\x92\x8F\xE2\x27\x0D\x6F\xB8\x63\xD5\x17\x38\xB4\x8E\xEE\xE3\x14\xA7\xCC\x8A\xB9\x32\x16\x45\x48\xE5\x26\xAE\x90\x22\x43\x68\x51\x7A\xCF\xEA\xBD\x6B\xB3\x73\x2B\xC0\xE9\xDA\x99\x83\x2B\x61\xCA\x01\xB6\xDE\x56\x24\x4A\x9E\x88\xD5\xF9\xB3\x79\x73\xF6\x22\xA4\x3D\x14\xA6\x59\x9B\x1F\x65\x4C\xB4\x5A\x74\xE3\x55\xA5") + ] + tests = testGroup "XSalsa" [ testGroup "KAT" $ - map (\(i,f) -> testCase (show (i :: Int)) f) $ zip [1..] $ map (\(r, k, i, p, e) -> salsaRunSimple r k i p e) vectors + zipWith (\i (r, k, n, p, e) -> testCase (show (i :: Int)) $ salsaRunSimple r k n p e) [1..] vectors + , testGroup "crypto_box encryption" $ + zipWith (\i (r, k, n, p, e) -> testCase (show (i :: Int)) $ cryptoBoxEnc r k n p e) [1..] vectorsCB ] where salsaRunSimple rounds key nonce plain expected = let salsa = XSalsa.initialize rounds key nonce in fst (XSalsa.combine salsa plain) @?= expected + + cryptoBoxEnc rounds shared nonce plain expected = + let zero = B.replicate 16 0 + (iv0, iv1) = B.splitAt 8 nonce + salsa0 = XSalsa.initialize rounds shared (zero `B.append` iv0) + salsa1 = XSalsa.derive salsa0 iv1 + (_, salsa2) = XSalsa.generate salsa1 32 :: (B.ByteString, XSalsa.State) + in fst (XSalsa.combine salsa2 plain) @?= expected