Add additional password hash algorithms (closes #5859) (#6023)

pull/6705/head
EpicCoder 4 years ago committed by techknowlogick
parent 1b85b248e4
commit 8d9d6aa903
  1. 2
      custom/conf/app.ini.sample
  2. 1
      docs/content/doc/advanced/config-cheat-sheet.en-us.md
  3. 10
      models/login_source.go
  4. 48
      models/user.go
  5. 38
      models/user_test.go
  6. 2
      modules/setting/setting.go
  7. 285
      vendor/golang.org/x/crypto/argon2/argon2.go
  8. 53
      vendor/golang.org/x/crypto/argon2/blake2b.go
  9. 60
      vendor/golang.org/x/crypto/argon2/blamka_amd64.go
  10. 243
      vendor/golang.org/x/crypto/argon2/blamka_amd64.s
  11. 163
      vendor/golang.org/x/crypto/argon2/blamka_generic.go
  12. 15
      vendor/golang.org/x/crypto/argon2/blamka_ref.go
  13. 289
      vendor/golang.org/x/crypto/blake2b/blake2b.go
  14. 37
      vendor/golang.org/x/crypto/blake2b/blake2bAVX2_amd64.go
  15. 750
      vendor/golang.org/x/crypto/blake2b/blake2bAVX2_amd64.s
  16. 24
      vendor/golang.org/x/crypto/blake2b/blake2b_amd64.go
  17. 281
      vendor/golang.org/x/crypto/blake2b/blake2b_amd64.s
  18. 182
      vendor/golang.org/x/crypto/blake2b/blake2b_generic.go
  19. 11
      vendor/golang.org/x/crypto/blake2b/blake2b_ref.go
  20. 177
      vendor/golang.org/x/crypto/blake2b/blake2x.go
  21. 32
      vendor/golang.org/x/crypto/blake2b/register.go
  22. 213
      vendor/golang.org/x/crypto/scrypt/scrypt.go
  23. 5
      vendor/modules.txt

@ -319,6 +319,8 @@ MIN_PASSWORD_LENGTH = 6
IMPORT_LOCAL_PATHS = false
; Set to true to prevent all users (including admin) from creating custom git hooks
DISABLE_GIT_HOOKS = false
; Password Hash algorithm, either "pbkdf2", "argon2", "scrypt" or "bcrypt"
PASSWORD_HASH_ALGO = pbkdf2
[openid]
;

@ -197,6 +197,7 @@ Values containing `#` or `;` must be quoted using `` ` `` or `"""`.
- `IMPORT_LOCAL_PATHS`: **false**: Set to `false` to prevent all users (including admin) from importing local path on server.
- `INTERNAL_TOKEN`: **\<random at every install if no uri set\>**: Secret used to validate communication within Gitea binary.
- `INTERNAL_TOKEN_URI`: **<empty>**: Instead of defining internal token in the configuration, this configuration option can be used to give Gitea a path to a file that contains the internal token (example value: `file:/etc/gitea/internal_token`)
- `PASSWORD_HASH_ALGO`: **pbkdf2**: The hash algorithm to use \[pbkdf2, argon2, scrypt, bcrypt\].
## OpenID (`openid`)

@ -22,6 +22,7 @@ import (
"code.gitea.io/gitea/modules/auth/oauth2"
"code.gitea.io/gitea/modules/auth/pam"
"code.gitea.io/gitea/modules/log"
"code.gitea.io/gitea/modules/setting"
"code.gitea.io/gitea/modules/util"
)
@ -665,6 +666,15 @@ func UserSignIn(username, password string) (*User, error) {
switch user.LoginType {
case LoginNoType, LoginPlain, LoginOAuth2:
if user.IsPasswordSet() && user.ValidatePassword(password) {
// Update password hash if server password hash algorithm have changed
if user.PasswdHashAlgo != setting.PasswordHashAlgo {
user.HashPassword(password)
if err := UpdateUserCols(user, "passwd", "passwd_hash_algo"); err != nil {
return nil, err
}
}
// WARN: DON'T check user.IsActive, that will be checked on reqSign so that
// user could be hint to resend confirm email.
if user.ProhibitLogin {

@ -33,7 +33,10 @@ import (
"github.com/Unknwon/com"
"github.com/go-xorm/xorm"
"golang.org/x/crypto/argon2"
"golang.org/x/crypto/bcrypt"
"golang.org/x/crypto/pbkdf2"
"golang.org/x/crypto/scrypt"
"golang.org/x/crypto/ssh"
"xorm.io/builder"
"xorm.io/core"
@ -50,6 +53,13 @@ const (
UserTypeOrganization
)
const (
algoBcrypt = "bcrypt"
algoScrypt = "scrypt"
algoArgon2 = "argon2"
algoPbkdf2 = "pbkdf2"
)
const syncExternalUsers = "sync_external_users"
var (
@ -82,6 +92,7 @@ type User struct {
Email string `xorm:"NOT NULL"`
KeepEmailPrivate bool
Passwd string `xorm:"NOT NULL"`
PasswdHashAlgo string `xorm:"NOT NULL DEFAULT 'pbkdf2'"`
// MustChangePassword is an attribute that determines if a user
// is to change his/her password after registration.
@ -430,25 +441,48 @@ func (u *User) NewGitSig() *git.Signature {
}
}
func hashPassword(passwd, salt string) string {
tempPasswd := pbkdf2.Key([]byte(passwd), []byte(salt), 10000, 50, sha256.New)
func hashPassword(passwd, salt, algo string) string {
var tempPasswd []byte
switch algo {
case algoBcrypt:
tempPasswd, _ = bcrypt.GenerateFromPassword([]byte(passwd), bcrypt.DefaultCost)
return string(tempPasswd)
case algoScrypt:
tempPasswd, _ = scrypt.Key([]byte(passwd), []byte(salt), 65536, 16, 2, 50)
case algoArgon2:
tempPasswd = argon2.IDKey([]byte(passwd), []byte(salt), 2, 65536, 8, 50)
case algoPbkdf2:
fallthrough
default:
tempPasswd = pbkdf2.Key([]byte(passwd), []byte(salt), 10000, 50, sha256.New)
}
return fmt.Sprintf("%x", tempPasswd)
}
// HashPassword hashes a password using PBKDF.
// HashPassword hashes a password using the algorithm defined in the config value of PASSWORD_HASH_ALGO.
func (u *User) HashPassword(passwd string) {
u.Passwd = hashPassword(passwd, u.Salt)
u.PasswdHashAlgo = setting.PasswordHashAlgo
u.Passwd = hashPassword(passwd, u.Salt, setting.PasswordHashAlgo)
}
// ValidatePassword checks if given password matches the one belongs to the user.
func (u *User) ValidatePassword(passwd string) bool {
tempHash := hashPassword(passwd, u.Salt)
return subtle.ConstantTimeCompare([]byte(u.Passwd), []byte(tempHash)) == 1
tempHash := hashPassword(passwd, u.Salt, u.PasswdHashAlgo)
if u.PasswdHashAlgo != algoBcrypt && subtle.ConstantTimeCompare([]byte(u.Passwd), []byte(tempHash)) == 1 {
return true
}
if u.PasswdHashAlgo == algoBcrypt && bcrypt.CompareHashAndPassword([]byte(u.Passwd), []byte(passwd)) == nil {
return true
}
return false
}
// IsPasswordSet checks if the password is set or left empty
func (u *User) IsPasswordSet() bool {
return !u.ValidatePassword("")
return len(u.Passwd) > 0
}
// UploadAvatar saves custom avatar for user.

@ -147,21 +147,29 @@ func TestHashPasswordDeterministic(t *testing.T) {
b := make([]byte, 16)
rand.Read(b)
u := &User{Salt: string(b)}
for i := 0; i < 50; i++ {
// generate a random password
rand.Read(b)
pass := string(b)
// save the current password in the user - hash it and store the result
u.HashPassword(pass)
r1 := u.Passwd
// run again
u.HashPassword(pass)
r2 := u.Passwd
// assert equal (given the same salt+pass, the same result is produced)
assert.Equal(t, r1, r2)
algos := []string{"pbkdf2", "argon2", "scrypt", "bcrypt"}
for j := 0; j < len(algos); j++ {
u.PasswdHashAlgo = algos[j]
for i := 0; i < 50; i++ {
// generate a random password
rand.Read(b)
pass := string(b)
// save the current password in the user - hash it and store the result
u.HashPassword(pass)
r1 := u.Passwd
// run again
u.HashPassword(pass)
r2 := u.Passwd
// assert equal (given the same salt+pass, the same result is produced) except bcrypt
if u.PasswdHashAlgo == "bcrypt" {
assert.NotEqual(t, r1, r2)
} else {
assert.Equal(t, r1, r2)
}
}
}
}

@ -154,6 +154,7 @@ var (
MinPasswordLength int
ImportLocalPaths bool
DisableGitHooks bool
PasswordHashAlgo string
// Database settings
UseSQLite3 bool
@ -779,6 +780,7 @@ func NewContext() {
MinPasswordLength = sec.Key("MIN_PASSWORD_LENGTH").MustInt(6)
ImportLocalPaths = sec.Key("IMPORT_LOCAL_PATHS").MustBool(false)
DisableGitHooks = sec.Key("DISABLE_GIT_HOOKS").MustBool(false)
PasswordHashAlgo = sec.Key("PASSWORD_HASH_ALGO").MustString("pbkdf2")
InternalToken = loadInternalToken(sec)
IterateBufferSize = Cfg.Section("database").Key("ITERATE_BUFFER_SIZE").MustInt(50)
LogSQL = Cfg.Section("database").Key("LOG_SQL").MustBool(true)

@ -0,0 +1,285 @@
// Copyright 2017 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// Package argon2 implements the key derivation function Argon2.
// Argon2 was selected as the winner of the Password Hashing Competition and can
// be used to derive cryptographic keys from passwords.
//
// For a detailed specification of Argon2 see [1].
//
// If you aren't sure which function you need, use Argon2id (IDKey) and
// the parameter recommendations for your scenario.
//
//
// Argon2i
//
// Argon2i (implemented by Key) is the side-channel resistant version of Argon2.
// It uses data-independent memory access, which is preferred for password
// hashing and password-based key derivation. Argon2i requires more passes over
// memory than Argon2id to protect from trade-off attacks. The recommended
// parameters (taken from [2]) for non-interactive operations are time=3 and to
// use the maximum available memory.
//
//
// Argon2id
//
// Argon2id (implemented by IDKey) is a hybrid version of Argon2 combining
// Argon2i and Argon2d. It uses data-independent memory access for the first
// half of the first iteration over the memory and data-dependent memory access
// for the rest. Argon2id is side-channel resistant and provides better brute-
// force cost savings due to time-memory tradeoffs than Argon2i. The recommended
// parameters for non-interactive operations (taken from [2]) are time=1 and to
// use the maximum available memory.
//
// [1] https://github.com/P-H-C/phc-winner-argon2/blob/master/argon2-specs.pdf
// [2] https://tools.ietf.org/html/draft-irtf-cfrg-argon2-03#section-9.3
package argon2
import (
"encoding/binary"
"sync"
"golang.org/x/crypto/blake2b"
)
// The Argon2 version implemented by this package.
const Version = 0x13
const (
argon2d = iota
argon2i
argon2id
)
// Key derives a key from the password, salt, and cost parameters using Argon2i
// returning a byte slice of length keyLen that can be used as cryptographic
// key. The CPU cost and parallelism degree must be greater than zero.
//
// For example, you can get a derived key for e.g. AES-256 (which needs a
// 32-byte key) by doing:
//
// key := argon2.Key([]byte("some password"), salt, 3, 32*1024, 4, 32)
//
// The draft RFC recommends[2] time=3, and memory=32*1024 is a sensible number.
// If using that amount of memory (32 MB) is not possible in some contexts then
// the time parameter can be increased to compensate.
//
// The time parameter specifies the number of passes over the memory and the
// memory parameter specifies the size of the memory in KiB. For example
// memory=32*1024 sets the memory cost to ~32 MB. The number of threads can be
// adjusted to the number of available CPUs. The cost parameters should be
// increased as memory latency and CPU parallelism increases. Remember to get a
// good random salt.
func Key(password, salt []byte, time, memory uint32, threads uint8, keyLen uint32) []byte {
return deriveKey(argon2i, password, salt, nil, nil, time, memory, threads, keyLen)
}
// IDKey derives a key from the password, salt, and cost parameters using
// Argon2id returning a byte slice of length keyLen that can be used as
// cryptographic key. The CPU cost and parallelism degree must be greater than
// zero.
//
// For example, you can get a derived key for e.g. AES-256 (which needs a
// 32-byte key) by doing:
//
// key := argon2.IDKey([]byte("some password"), salt, 1, 64*1024, 4, 32)
//
// The draft RFC recommends[2] time=1, and memory=64*1024 is a sensible number.
// If using that amount of memory (64 MB) is not possible in some contexts then
// the time parameter can be increased to compensate.
//
// The time parameter specifies the number of passes over the memory and the
// memory parameter specifies the size of the memory in KiB. For example
// memory=64*1024 sets the memory cost to ~64 MB. The number of threads can be
// adjusted to the numbers of available CPUs. The cost parameters should be
// increased as memory latency and CPU parallelism increases. Remember to get a
// good random salt.
func IDKey(password, salt []byte, time, memory uint32, threads uint8, keyLen uint32) []byte {
return deriveKey(argon2id, password, salt, nil, nil, time, memory, threads, keyLen)
}
func deriveKey(mode int, password, salt, secret, data []byte, time, memory uint32, threads uint8, keyLen uint32) []byte {
if time < 1 {
panic("argon2: number of rounds too small")
}
if threads < 1 {
panic("argon2: parallelism degree too low")
}
h0 := initHash(password, salt, secret, data, time, memory, uint32(threads), keyLen, mode)
memory = memory / (syncPoints * uint32(threads)) * (syncPoints * uint32(threads))
if memory < 2*syncPoints*uint32(threads) {
memory = 2 * syncPoints * uint32(threads)
}
B := initBlocks(&h0, memory, uint32(threads))
processBlocks(B, time, memory, uint32(threads), mode)
return extractKey(B, memory, uint32(threads), keyLen)
}
const (
blockLength = 128
syncPoints = 4
)
type block [blockLength]uint64
func initHash(password, salt, key, data []byte, time, memory, threads, keyLen uint32, mode int) [blake2b.Size + 8]byte {
var (
h0 [blake2b.Size + 8]byte
params [24]byte
tmp [4]byte
)
b2, _ := blake2b.New512(nil)
binary.LittleEndian.PutUint32(params[0:4], threads)
binary.LittleEndian.PutUint32(params[4:8], keyLen)
binary.LittleEndian.PutUint32(params[8:12], memory)
binary.LittleEndian.PutUint32(params[12:16], time)
binary.LittleEndian.PutUint32(params[16:20], uint32(Version))
binary.LittleEndian.PutUint32(params[20:24], uint32(mode))
b2.Write(params[:])
binary.LittleEndian.PutUint32(tmp[:], uint32(len(password)))
b2.Write(tmp[:])
b2.Write(password)
binary.LittleEndian.PutUint32(tmp[:], uint32(len(salt)))
b2.Write(tmp[:])
b2.Write(salt)
binary.LittleEndian.PutUint32(tmp[:], uint32(len(key)))
b2.Write(tmp[:])
b2.Write(key)
binary.LittleEndian.PutUint32(tmp[:], uint32(len(data)))
b2.Write(tmp[:])
b2.Write(data)
b2.Sum(h0[:0])
return h0
}
func initBlocks(h0 *[blake2b.Size + 8]byte, memory, threads uint32) []block {
var block0 [1024]byte
B := make([]block, memory)
for lane := uint32(0); lane < threads; lane++ {
j := lane * (memory / threads)
binary.LittleEndian.PutUint32(h0[blake2b.Size+4:], lane)
binary.LittleEndian.PutUint32(h0[blake2b.Size:], 0)
blake2bHash(block0[:], h0[:])
for i := range B[j+0] {
B[j+0][i] = binary.LittleEndian.Uint64(block0[i*8:])
}
binary.LittleEndian.PutUint32(h0[blake2b.Size:], 1)
blake2bHash(block0[:], h0[:])
for i := range B[j+1] {
B[j+1][i] = binary.LittleEndian.Uint64(block0[i*8:])
}
}
return B
}
func processBlocks(B []block, time, memory, threads uint32, mode int) {
lanes := memory / threads
segments := lanes / syncPoints
processSegment := func(n, slice, lane uint32, wg *sync.WaitGroup) {
var addresses, in, zero block
if mode == argon2i || (mode == argon2id && n == 0 && slice < syncPoints/2) {
in[0] = uint64(n)
in[1] = uint64(lane)
in[2] = uint64(slice)
in[3] = uint64(memory)
in[4] = uint64(time)
in[5] = uint64(mode)
}
index := uint32(0)
if n == 0 && slice == 0 {
index = 2 // we have already generated the first two blocks
if mode == argon2i || mode == argon2id {
in[6]++
processBlock(&addresses, &in, &zero)
processBlock(&addresses, &addresses, &zero)
}
}
offset := lane*lanes + slice*segments + index
var random uint64
for index < segments {
prev := offset - 1
if index == 0 && slice == 0 {
prev += lanes // last block in lane
}
if mode == argon2i || (mode == argon2id && n == 0 && slice < syncPoints/2) {
if index%blockLength == 0 {
in[6]++
processBlock(&addresses, &in, &zero)
processBlock(&addresses, &addresses, &zero)
}
random = addresses[index%blockLength]
} else {
random = B[prev][0]
}
newOffset := indexAlpha(random, lanes, segments, threads, n, slice, lane, index)
processBlockXOR(&B[offset], &B[prev], &B[newOffset])
index, offset = index+1, offset+1
}
wg.Done()
}
for n := uint32(0); n < time; n++ {
for slice := uint32(0); slice < syncPoints; slice++ {
var wg sync.WaitGroup
for lane := uint32(0); lane < threads; lane++ {
wg.Add(1)
go processSegment(n, slice, lane, &wg)
}
wg.Wait()
}
}
}
func extractKey(B []block, memory, threads, keyLen uint32) []byte {
lanes := memory / threads
for lane := uint32(0); lane < threads-1; lane++ {
for i, v := range B[(lane*lanes)+lanes-1] {
B[memory-1][i] ^= v
}
}
var block [1024]byte
for i, v := range B[memory-1] {
binary.LittleEndian.PutUint64(block[i*8:], v)
}
key := make([]byte, keyLen)
blake2bHash(key, block[:])
return key
}
func indexAlpha(rand uint64, lanes, segments, threads, n, slice, lane, index uint32) uint32 {
refLane := uint32(rand>>32) % threads
if n == 0 && slice == 0 {
refLane = lane
}
m, s := 3*segments, ((slice+1)%syncPoints)*segments
if lane == refLane {
m += index
}
if n == 0 {
m, s = slice*segments, 0
if slice == 0 || lane == refLane {
m += index
}
}
if index == 0 || lane == refLane {
m--
}
return phi(rand, uint64(m), uint64(s), refLane, lanes)
}
func phi(rand, m, s uint64, lane, lanes uint32) uint32 {
p := rand & 0xFFFFFFFF
p = (p * p) >> 32
p = (p * m) >> 32
return lane*lanes + uint32((s+m-(p+1))%uint64(lanes))
}

@ -0,0 +1,53 @@
// Copyright 2017 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package argon2
import (
"encoding/binary"
"hash"
"golang.org/x/crypto/blake2b"
)
// blake2bHash computes an arbitrary long hash value of in
// and writes the hash to out.
func blake2bHash(out []byte, in []byte) {
var b2 hash.Hash
if n := len(out); n < blake2b.Size {
b2, _ = blake2b.New(n, nil)
} else {
b2, _ = blake2b.New512(nil)
}
var buffer [blake2b.Size]byte
binary.LittleEndian.PutUint32(buffer[:4], uint32(len(out)))
b2.Write(buffer[:4])
b2.Write(in)
if len(out) <= blake2b.Size {
b2.Sum(out[:0])
return
}
outLen := len(out)
b2.Sum(buffer[:0])
b2.Reset()
copy(out, buffer[:32])
out = out[32:]
for len(out) > blake2b.Size {
b2.Write(buffer[:])
b2.Sum(buffer[:0])
copy(out, buffer[:32])
out = out[32:]
b2.Reset()
}
if outLen%blake2b.Size > 0 { // outLen > 64
r := ((outLen + 31) / 32) - 2 // ⌈τ /32⌉-2
b2, _ = blake2b.New(outLen-32*r, nil)
}
b2.Write(buffer[:])
b2.Sum(out[:0])
}

@ -0,0 +1,60 @@
// Copyright 2017 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// +build amd64,!gccgo,!appengine
package argon2
import "golang.org/x/sys/cpu"
func init() {
useSSE4 = cpu.X86.HasSSE41
}
//go:noescape
func mixBlocksSSE2(out, a, b, c *block)
//go:noescape
func xorBlocksSSE2(out, a, b, c *block)
//go:noescape
func blamkaSSE4(b *block)
func processBlockSSE(out, in1, in2 *block, xor bool) {
var t block
mixBlocksSSE2(&t, in1, in2, &t)
if useSSE4 {
blamkaSSE4(&t)
} else {
for i := 0; i < blockLength; i += 16 {
blamkaGeneric(
&t[i+0], &t[i+1], &t[i+2], &t[i+3],
&t[i+4], &t[i+5], &t[i+6], &t[i+7],
&t[i+8], &t[i+9], &t[i+10], &t[i+11],
&t[i+12], &t[i+13], &t[i+14], &t[i+15],
)
}
for i := 0; i < blockLength/8; i += 2 {
blamkaGeneric(
&t[i], &t[i+1], &t[16+i], &t[16+i+1],
&t[32+i], &t[32+i+1], &t[48+i], &t[48+i+1],
&t[64+i], &t[64+i+1], &t[80+i], &t[80+i+1],
&t[96+i], &t[96+i+1], &t[112+i], &t[112+i+1],
)
}
}
if xor {
xorBlocksSSE2(out, in1, in2, &t)
} else {
mixBlocksSSE2(out, in1, in2, &t)
}
}
func processBlock(out, in1, in2 *block) {
processBlockSSE(out, in1, in2, false)
}
func processBlockXOR(out, in1, in2 *block) {
processBlockSSE(out, in1, in2, true)
}

@ -0,0 +1,243 @@
// Copyright 2017 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// +build amd64,!gccgo,!appengine
#include "textflag.h"
DATA ·c40<>+0x00(SB)/8, $0x0201000706050403
DATA ·c40<>+0x08(SB)/8, $0x0a09080f0e0d0c0b
GLOBL ·c40<>(SB), (NOPTR+RODATA), $16
DATA ·c48<>+0x00(SB)/8, $0x0100070605040302
DATA ·c48<>+0x08(SB)/8, $0x09080f0e0d0c0b0a
GLOBL ·c48<>(SB), (NOPTR+RODATA), $16
#define SHUFFLE(v2, v3, v4, v5, v6, v7, t1, t2) \
MOVO v4, t1; \
MOVO v5, v4; \
MOVO t1, v5; \
MOVO v6, t1; \
PUNPCKLQDQ v6, t2; \
PUNPCKHQDQ v7, v6; \
PUNPCKHQDQ t2, v6; \
PUNPCKLQDQ v7, t2; \
MOVO t1, v7; \
MOVO v2, t1; \
PUNPCKHQDQ t2, v7; \
PUNPCKLQDQ v3, t2; \
PUNPCKHQDQ t2, v2; \
PUNPCKLQDQ t1, t2; \
PUNPCKHQDQ t2, v3
#define SHUFFLE_INV(v2, v3, v4, v5, v6, v7, t1, t2) \
MOVO v4, t1; \
MOVO v5, v4; \
MOVO t1, v5; \
MOVO v2, t1; \
PUNPCKLQDQ v2, t2; \
PUNPCKHQDQ v3, v2; \
PUNPCKHQDQ t2, v2; \
PUNPCKLQDQ v3, t2; \
MOVO t1, v3; \
MOVO v6, t1; \
PUNPCKHQDQ t2, v3; \
PUNPCKLQDQ v7, t2; \
PUNPCKHQDQ t2, v6; \
PUNPCKLQDQ t1, t2; \
PUNPCKHQDQ t2, v7
#define HALF_ROUND(v0, v1, v2, v3, v4, v5, v6, v7, t0, c40, c48) \
MOVO v0, t0; \
PMULULQ v2, t0; \
PADDQ v2, v0; \
PADDQ t0, v0; \
PADDQ t0, v0; \
PXOR v0, v6; \
PSHUFD $0xB1, v6, v6; \
MOVO v4, t0; \
PMULULQ v6, t0; \
PADDQ v6, v4; \
PADDQ t0, v4; \
PADDQ t0, v4; \
PXOR v4, v2; \
PSHUFB c40, v2; \
MOVO v0, t0; \
PMULULQ v2, t0; \
PADDQ v2, v0; \
PADDQ t0, v0; \
PADDQ t0, v0; \
PXOR v0, v6; \
PSHUFB c48, v6; \
MOVO v4, t0; \
PMULULQ v6, t0; \
PADDQ v6, v4; \
PADDQ t0, v4; \
PADDQ t0, v4; \
PXOR v4, v2; \
MOVO v2, t0; \
PADDQ v2, t0; \
PSRLQ $63, v2; \
PXOR t0, v2; \
MOVO v1, t0; \
PMULULQ v3, t0; \
PADDQ v3, v1; \
PADDQ t0, v1; \
PADDQ t0, v1; \
PXOR v1, v7; \
PSHUFD $0xB1, v7, v7; \
MOVO v5, t0; \
PMULULQ v7, t0; \
PADDQ v7, v5; \
PADDQ t0, v5; \
PADDQ t0, v5; \
PXOR v5, v3; \
PSHUFB c40, v3; \
MOVO v1, t0; \
PMULULQ v3, t0; \
PADDQ v3, v1; \
PADDQ t0, v1; \
PADDQ t0, v1; \
PXOR v1, v7; \
PSHUFB c48, v7; \
MOVO v5, t0; \
PMULULQ v7, t0; \
PADDQ v7, v5; \
PADDQ t0, v5; \
PADDQ t0, v5; \
PXOR v5, v3; \
MOVO v3, t0; \
PADDQ v3, t0; \
PSRLQ $63, v3; \
PXOR t0, v3
#define LOAD_MSG_0(block, off) \
MOVOU 8*(off+0)(block), X0; \
MOVOU 8*(off+2)(block), X1; \
MOVOU 8*(off+4)(block), X2; \
MOVOU 8*(off+6)(block), X3; \
MOVOU 8*(off+8)(block), X4; \
MOVOU 8*(off+10)(block), X5; \
MOVOU 8*(off+12)(block), X6; \
MOVOU 8*(off+14)(block), X7
#define STORE_MSG_0(block, off) \
MOVOU X0, 8*(off+0)(block); \
MOVOU X1, 8*(off+2)(block); \
MOVOU X2, 8*(off+4)(block); \
MOVOU X3, 8*(off+6)(block); \
MOVOU X4, 8*(off+8)(block); \
MOVOU X5, 8*(off+10)(block); \
MOVOU X6, 8*(off+12)(block); \
MOVOU X7, 8*(off+14)(block)
#define LOAD_MSG_1(block, off) \
MOVOU 8*off+0*8(block), X0; \
MOVOU 8*off+16*8(block), X1; \
MOVOU 8*off+32*8(block), X2; \
MOVOU 8*off+48*8(block), X3; \
MOVOU 8*off+64*8(block), X4; \
MOVOU 8*off+80*8(block), X5; \
MOVOU 8*off+96*8(block), X6; \
MOVOU 8*off+112*8(block), X7
#define STORE_MSG_1(block, off) \
MOVOU X0, 8*off+0*8(block); \
MOVOU X1, 8*off+16*8(block); \
MOVOU X2, 8*off+32*8(block); \
MOVOU X3, 8*off+48*8(block); \
MOVOU X4, 8*off+64*8(block); \
MOVOU X5, 8*off+80*8(block); \
MOVOU X6, 8*off+96*8(block); \
MOVOU X7, 8*off+112*8(block)
#define BLAMKA_ROUND_0(block, off, t0, t1, c40, c48) \
LOAD_MSG_0(block, off); \
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, t0, c40, c48); \
SHUFFLE(X2, X3, X4, X5, X6, X7, t0, t1); \
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, t0, c40, c48); \
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, t0, t1); \
STORE_MSG_0(block, off)
#define BLAMKA_ROUND_1(block, off, t0, t1, c40, c48) \
LOAD_MSG_1(block, off); \
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, t0, c40, c48); \
SHUFFLE(X2, X3, X4, X5, X6, X7, t0, t1); \
HALF_ROUND(X0, X1, X2, X3, X4, X5, X6, X7, t0, c40, c48); \
SHUFFLE_INV(X2, X3, X4, X5, X6, X7, t0, t1); \
STORE_MSG_1(block, off)
// func blamkaSSE4(b *block)
TEXT ·blamkaSSE4(SB), 4, $0-8
MOVQ b+0(FP), AX
MOVOU ·c40<>(SB), X10
MOVOU ·c48<>(SB), X11
BLAMKA_ROUND_0(AX, 0, X8, X9, X10, X11)
BLAMKA_ROUND_0(AX, 16, X8, X9, X10, X11)
BLAMKA_ROUND_0(AX, 32, X8, X9, X10, X11)
BLAMKA_ROUND_0(AX, 48, X8, X9, X10, X11)
BLAMKA_ROUND_0(AX, 64, X8, X9, X10, X11)
BLAMKA_ROUND_0(AX, 80, X8, X9, X10, X11)
BLAMKA_ROUND_0(AX, 96, X8, X9, X10, X11)
BLAMKA_ROUND_0(AX, 112, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 0, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 2, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 4, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 6, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 8, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 10, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 12, X8, X9, X10, X11)
BLAMKA_ROUND_1(AX, 14, X8, X9, X10, X11)
RET
// func mixBlocksSSE2(out, a, b, c *block)
TEXT ·mixBlocksSSE2(SB), 4, $0-32
MOVQ out+0(FP), DX
MOVQ a+8(FP), AX
MOVQ b+16(FP), BX
MOVQ a+24(FP), CX
MOVQ $128, BP
loop:
MOVOU 0(AX), X0
MOVOU 0(BX), X1
MOVOU 0(CX), X2
PXOR X1, X0
PXOR X2, X0
MOVOU X0, 0(DX)
ADDQ $16, AX
ADDQ $16, BX
ADDQ $16, CX
ADDQ $16, DX
SUBQ $2, BP
JA loop
RET
// func xorBlocksSSE2(out, a, b, c *block)
TEXT ·xorBlocksSSE2(SB), 4, $0-32
MOVQ out+0(FP), DX
MOVQ a+8(FP), AX
MOVQ b+16(FP), BX
MOVQ a+24(FP), CX
MOVQ $128, BP
loop:
MOVOU 0(AX), X0
MOVOU 0(BX), X1
MOVOU 0(CX), X2
MOVOU 0(DX), X3
PXOR X1, X0
PXOR X2, X0
PXOR X3, X0
MOVOU X0, 0(DX)
ADDQ $16, AX
ADDQ $16, BX
ADDQ $16, CX
ADDQ $16, DX
SUBQ $2, BP
JA loop
RET

@ -0,0 +1,163 @@
// Copyright 2017 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package argon2
var useSSE4 bool
func processBlockGeneric(out, in1, in2 *block, xor bool) {
var t block
for i := range t {
t[i] = in1[i] ^ in2[i]
}
for i := 0; i < blockLength; i += 16 {
blamkaGeneric(
&t[i+0], &t[i+1], &t[i+2], &t[i+3],
&t[i+4], &t[i+5], &t[i+6], &t[i+7],
&t[i+8], &t[i+9], &t[i+10], &t[i+11],
&t[i+12], &t[i+13], &t[i+14], &t[i+15],
)
}
for i := 0; i < blockLength/8; i += 2 {
blamkaGeneric(
&t[i], &t[i+1], &t[16+i], &t[16+i+1],
&t[32+i], &t[32+i+1], &t[48+i], &t[48+i+1],
&t[64+i], &t[64+i+1], &t[80+i], &t[80+i+1],
&t[96+i], &t[96+i+1], &t[112+i], &t[112+i+1],
)
}
if xor {
for i := range t {
out[i] ^= in1[i] ^ in2[i] ^ t[i]
}
} else {
for i := range t {
out[i] = in1[i] ^ in2[i] ^ t[i]
}
}
}
func blamkaGeneric(t00, t01, t02, t03, t04, t05, t06, t07, t08, t09, t10, t11, t12, t13, t14, t15 *uint64) {
v00, v01, v02, v03 := *t00, *t01, *t02, *t03
v04, v05, v06, v07 := *t04, *t05, *t06, *t07
v08, v09, v10, v11 := *t08, *t09, *t10, *t11
v12, v13, v14, v15 := *t12, *t13, *t14, *t15
v00 += v04 + 2*uint64(uint32(v00))*uint64(uint32(v04))
v12 ^= v00
v12 = v12>>32 | v12<<32
v08 += v12 + 2*uint64(uint32(v08))*uint64(uint32(v12))
v04 ^= v08
v04 = v04>>24 | v04<<40
v00 += v04 + 2*uint64(uint32(v00))*uint64(uint32(v04))
v12 ^= v00
v12 = v12>>16 | v12<<48
v08 += v12 + 2*uint64(uint32(v08))*uint64(uint32(v12))
v04 ^= v08
v04 = v04>>63 | v04<<1
v01 += v05 + 2*uint64(uint32(v01))*uint64(uint32(v05))
v13 ^= v01
v13 = v13>>32 | v13<<32
v09 += v13 + 2*uint64(uint32(v09))*uint64(uint32(v13))
v05 ^= v09
v05 = v05>>24 | v05<<40
v01 += v05 + 2*uint64(uint32(v01))*uint64(uint32(v05))
v13 ^= v01
v13 = v13>>16 | v13<<48
v09 += v13 + 2*uint64(uint32(v09))*uint64(uint32(v13))
v05 ^= v09
v05 = v05>>63 | v05<<1
v02 += v06 + 2*uint64(uint32(v02))*uint64(uint32(v06))
v14 ^= v02
v14 = v14>>32 | v14<<32
v10 += v14 + 2*uint64(uint32(v10))*uint64(uint32(v14))
v06 ^= v10
v06 = v06>>24 | v06<<40
v02 += v06 + 2*uint64(uint32(v02))*uint64(uint32(v06))
v14 ^= v02
v14 = v14>>16 | v14<<48
v10 += v14 + 2*uint64(uint32(v10))*uint64(uint32(v14))
v06 ^= v10
v06 = v06>>63 | v06<<1
v03 += v07 + 2*uint64(uint32(v03))*uint64(uint32(v07))
v15 ^= v03
v15 = v15>>32 | v15<<32
v11 += v15 + 2*uint64(uint32(v11))*uint64(uint32(v15))
v07 ^= v11
v07 = v07>>24 | v07<<40
v03 += v07 + 2*uint64(uint32(v03))*uint64(uint32(v07))
v15 ^= v03
v15 = v15>>16 | v15<<48
v11 += v15 + 2*uint64(uint32(v11))*uint64(uint32(v15))
v07 ^= v11
v07 = v07>>63 | v07<<1
v00 += v05 + 2*uint64(uint32(v00))*uint64(uint32(v05))
v15 ^= v00
v15 = v15>>32 | v15<<32
v10 += v15 + 2*uint64(uint32(v10))*uint64(uint32(v15))
v05 ^= v10
v05 = v05>>24 | v05<<40
v00 += v05 + 2*uint64(uint32(v00))*uint64(uint32(v05))
v15 ^= v00
v15 = v15>>16 | v15<<48
v10 += v15 + 2*uint64(uint32(v10))*uint64(uint32(v15))
v05 ^= v10
v05 = v05>>63 | v05<<1
v01 += v06 + 2*uint64(uint32(v01))*uint64(uint32(v06))
v12 ^= v01
v12 = v12>>32 | v12<<32
v11 += v12 + 2*uint64(uint32(v11))*uint64(uint32(v12))
v06 ^= v11
v06 = v06>>24 | v06<<40
v01 += v06 + 2*uint64(uint32(v01))*uint64(uint32(v06))
v12 ^= v01
v12 = v12>>16 | v12<<48
v11 += v12 + 2*uint64(uint32(v11))*uint64(uint32(v12))
v06 ^= v11
v06 = v06>>63 | v06<<1
v02 += v07 + 2*uint64(uint32(v02))*uint64(uint32(v07))
v13 ^= v02
v13 = v13>>32 | v13<<32
v08 += v13 + 2*uint64(uint32(v08))*uint64(uint32(v13))
v07 ^= v08
v07 = v07>>24 | v07<<40
v02 += v07 + 2*uint64(uint32(v02))*uint64(uint32(v07))
v13 ^= v02
v13 = v13>>16 | v13<<48
v08 += v13 + 2*uint64(uint32(v08))*uint64(uint32(v13))
v07 ^= v08
v07 = v07>>63 | v07<<1
v03 += v04 + 2*uint64(uint32(v03))*uint64(uint32(v04))
v14 ^= v03
v14 = v14>>32 | v14<<32
v09 += v14 + 2*uint64(uint32(v09))*uint64(uint32(v14))
v04 ^= v09
v04 = v04>>24 | v04<<40
v03 += v04 + 2*uint64(uint32(v03))*uint64(uint32(v04))
v14 ^= v03
v14 = v14>>16 | v14<<48
v09 += v14 + 2*uint64(uint32(v09))*uint64(uint32(v14))
v04 ^= v09
v04 = v04>>63 | v04<<1
*t00, *t01, *t02, *t03 = v00, v01, v02, v03
*t04, *t05, *t06, *t07 = v04, v05, v06, v07
*t08, *t09, *t10, *t11 = v08, v09, v10, v11
*t12, *t13, *t14, *t15 = v12, v13, v14, v15
}

@ -0,0 +1,15 @@
// Copyright 2017 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// +build !amd64 appengine gccgo
package argon2
func processBlock(out, in1, in2 *block) {
processBlockGeneric(out, in1, in2, false)
}
func processBlockXOR(out, in1, in2 *block) {
processBlockGeneric(out, in1, in2, true)
}

@ -0,0 +1,289 @@
// Copyright 2016 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// Package blake2b implements the BLAKE2b hash algorithm defined by RFC 7693
// and the extendable output function (XOF) BLAKE2Xb.
//
// For a detailed specification of BLAKE2b see https://blake2.net/blake2.pdf
// and for BLAKE2Xb see https://blake2.net/blake2x.pdf
//
// If you aren't sure which function you need, use BLAKE2b (Sum512 or New512).
// If you need a secret-key MAC (message authentication code), use the New512
// function with a non-nil key.
//
// BLAKE2X is a construction to compute hash values larger than 64 bytes. It
// can produce hash values between 0 and 4 GiB.
package blake2b
import (
"encoding/binary"
"errors"
"hash"
)
const (
// The blocksize of BLAKE2b in bytes.
BlockSize = 128
// The hash size of BLAKE2b-512 in bytes.
Size = 64
// The hash size of BLAKE2b-384 in bytes.
Size384 = 48
// The hash size of BLAKE2b-256 in bytes.
Size256 = 32
)
var (
useAVX2 bool
useAVX bool
useSSE4 bool
)
var (
errKeySize = errors.New("blake2b: invalid key size")
errHashSize = errors.New("blake2b: invalid hash size")
)
var iv = [8]uint64{
0x6a09e667f3bcc908, 0xbb67ae8584caa73b, 0x3c6ef372fe94f82b, 0xa54ff53a5f1d36f1,
0x510e527fade682d1, 0x9b05688c2b3e6c1f, 0x1f83d9abfb41bd6b, 0x5be0cd19137e2179,
}
// Sum512 returns the BLAKE2b-512 checksum of the data.
func Sum512(data []byte) [Size]byte {
var sum [Size]byte
checkSum(&sum, Size, data)
return sum
}
// Sum384 returns the BLAKE2b-384 checksum of the data.
func Sum384(data []byte) [Size384]byte {
var sum [Size]byte
var sum384 [Size384]byte
checkSum(&sum, Size384, data)
copy(sum384[:], sum[:Size384])
return sum384
}
// Sum256 returns the BLAKE2b-256 checksum of the data.
func Sum256(data []byte) [Size256]byte {
var sum [Size]byte
var sum256 [Size256]byte
checkSum(&sum, Size256, data)
copy(sum256[:], sum[:Size256])
return sum256
}
// New512 returns a new hash.Hash computing the BLAKE2b-512 checksum. A non-nil
// key turns the hash into a MAC. The key must be between zero and 64 bytes long.
func New512(key []byte) (hash.Hash, error) { return newDigest(Size, key) }
// New384 returns a new hash.Hash computing the BLAKE2b-384 checksum. A non-nil
// key turns the hash into a MAC. The key must be between zero and 64 bytes long.
func New384(key []byte) (hash.Hash, error) { return newDigest(Size384, key) }
// New256 returns a new hash.Hash computing the BLAKE2b-256 checksum. A non-nil
// key turns the hash into a MAC. The key must be between zero and 64 bytes long.
func New256(key []byte) (hash.Hash, error) { return newDigest(Size256, key) }
// New returns a new hash.Hash computing the BLAKE2b checksum with a custom length.
// A non-nil key turns the hash into a MAC. The key must be between zero and 64 bytes long.
// The hash size can be a value between 1 and 64 but it is highly recommended to use
// values equal or greater than:
// - 32 if BLAKE2b is used as a hash function (The key is zero bytes long).
// - 16 if BLAKE2b is used as a MAC function (The key is at least 16 bytes long).
// When the key is nil, the returned hash.Hash implements BinaryMarshaler
// and BinaryUnmarshaler for state (de)serialization as documented by hash.Hash.
func New(size int, key []byte) (hash.Hash, error) { return newDigest(size, key) }
func newDigest(hashSize int, key []byte) (*digest, error) {
if hashSize < 1 || hashSize > Size {
return nil, errHashSize
}
if len(key) > Size {
return nil, errKeySize
}
d := &digest{
size: hashSize,
keyLen: len(key),
}
copy(d.key[:], key)
d.Reset()
return d, nil
}
func checkSum(sum *[Size]byte, hashSize int, data []byte) {
h := iv
h[0] ^= uint64(hashSize) | (1 << 16) | (1 << 24)
var c [2]uint64
if length := len(data); length > BlockSize {
n := length &^ (BlockSize - 1)
if length == n {
n -= BlockSize
}
hashBlocks(&h, &c, 0, data[:n])
data = data[n:]
}
var block [BlockSize]byte
offset := copy(block[:], data)
remaining := uint64(BlockSize - offset)
if c[0] < remaining {
c[1]--
}
c[0] -= remaining
hashBlocks(&h, &c, 0xFFFFFFFFFFFFFFFF, block[:])
for i, v := range h[:(hashSize+7)/8] {
binary.LittleEndian.PutUint64(sum[8*i:], v)
}
}
type digest struct {
h [8]uint64
c [2]uint64
size int
block [BlockSize]byte
offset int
key [BlockSize]byte
keyLen int
}
const (
magic = "b2b"
marshaledSize = len(magic) + 8*8 + 2*8 + 1 + BlockSize + 1
)
func (d *digest) MarshalBinary() ([]byte, error) {
if d.keyLen != 0 {
return nil, errors.New("crypto/blake2b: cannot marshal MACs")
}
b := make([]byte, 0, marshaledSize)
b = append(b, magic...)
for i := 0; i < 8; i++ {
b = appendUint64(b, d.h[i])
}
b = appendUint64(b, d.c[0])
b = appendUint64(b, d.c[1])
// Maximum value for size is 64
b = append(b, byte(d.size))
b = append(b, d.block[:]...)
b = append(b, byte(d.offset))
return b, nil
}
func (d *digest) UnmarshalBinary(b []byte) error {
if len(b) < len(magic) || string(b[:len(magic)]) != magic {
return errors.New("crypto/blake2b: invalid hash state identifier")
}
if len(b) != marshaledSize {
return errors.New("crypto/blake2b: invalid hash state size")
}
b = b[len(magic):]
for i := 0; i < 8; i++ {
b, d.h[i] = consumeUint64(b)
}
b, d.c[0] = consumeUint64(b)
b, d.c[1] = consumeUint64(b)
d.size = int(b[0])
b = b[1:]
copy(d.block[:], b[:BlockSize])
b = b[BlockSize:]
d.offset = int(b[0])
return nil
}
func (d *digest) BlockSize() int { return BlockSize }
func (d *digest) Size() int { return d.size }
func (d *digest) Reset() {
d.h = iv
d.h[0] ^= uint64(d.size) | (uint64(d.keyLen) << 8) | (1 << 16) | (1 << 24)
d.offset, d.c[0], d.c[1] = 0, 0, 0
if d.keyLen > 0 {
d.block = d.key
d.offset = BlockSize
}
}
func (d *digest) Write(p []byte) (n int, err error) {
n = len(p)
if d.offset > 0 {
remaining := BlockSize - d.offset
if n <= remaining {
d.offset += copy(d.block[d.offset:], p)
return
}
copy(d.block[d.offset:], p[:remaining])
hashBlocks(&d.h, &d.c, 0, d.block[:])
d.offset = 0
p = p[remaining:]
}
if length := len(p); length > BlockSize {
nn := length &^ (BlockSize - 1)
if length == nn {
nn -= BlockSize
}
hashBlocks(&d.h, &d.c, 0, p[:nn])
p = p[nn:]
}
if len(p) > 0 {
d.offset += copy(d.block[:], p)
}
return
}
func (d *digest) Sum(sum []byte) []byte {
var hash [Size]byte
d.finalize(&hash)
return append(sum, hash[:d.size]...)
}
func (d *digest) finalize(hash *[Size]byte) {
var block [BlockSize]byte