summaryrefslogtreecommitdiff
path: root/libgo/go/golang.org/x/crypto/internal
diff options
context:
space:
mode:
Diffstat (limited to 'libgo/go/golang.org/x/crypto/internal')
-rw-r--r--libgo/go/golang.org/x/crypto/internal/poly1305/bits_compat.go40
-rw-r--r--libgo/go/golang.org/x/crypto/internal/poly1305/bits_go1.13.go22
-rw-r--r--libgo/go/golang.org/x/crypto/internal/poly1305/mac_noasm.go10
-rw-r--r--libgo/go/golang.org/x/crypto/internal/poly1305/poly1305.go99
-rw-r--r--libgo/go/golang.org/x/crypto/internal/poly1305/sum_amd64.go48
-rw-r--r--libgo/go/golang.org/x/crypto/internal/poly1305/sum_generic.go310
-rw-r--r--libgo/go/golang.org/x/crypto/internal/poly1305/sum_ppc64le.go48
-rw-r--r--libgo/go/golang.org/x/crypto/internal/poly1305/sum_s390x.go76
8 files changed, 653 insertions, 0 deletions
diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/bits_compat.go b/libgo/go/golang.org/x/crypto/internal/poly1305/bits_compat.go
new file mode 100644
index 00000000000..45b5c966b2b
--- /dev/null
+++ b/libgo/go/golang.org/x/crypto/internal/poly1305/bits_compat.go
@@ -0,0 +1,40 @@
+// Copyright 2019 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.
+
+//go:build !go1.13
+// +build !go1.13
+
+package poly1305
+
+// Generic fallbacks for the math/bits intrinsics, copied from
+// src/math/bits/bits.go. They were added in Go 1.12, but Add64 and Sum64 had
+// variable time fallbacks until Go 1.13.
+
+func bitsAdd64(x, y, carry uint64) (sum, carryOut uint64) {
+ sum = x + y + carry
+ carryOut = ((x & y) | ((x | y) &^ sum)) >> 63
+ return
+}
+
+func bitsSub64(x, y, borrow uint64) (diff, borrowOut uint64) {
+ diff = x - y - borrow
+ borrowOut = ((^x & y) | (^(x ^ y) & diff)) >> 63
+ return
+}
+
+func bitsMul64(x, y uint64) (hi, lo uint64) {
+ const mask32 = 1<<32 - 1
+ x0 := x & mask32
+ x1 := x >> 32
+ y0 := y & mask32
+ y1 := y >> 32
+ w0 := x0 * y0
+ t := x1*y0 + w0>>32
+ w1 := t & mask32
+ w2 := t >> 32
+ w1 += x0 * y1
+ hi = x1*y1 + w2 + w1>>32
+ lo = x * y
+ return
+}
diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/bits_go1.13.go b/libgo/go/golang.org/x/crypto/internal/poly1305/bits_go1.13.go
new file mode 100644
index 00000000000..ed52b3418ab
--- /dev/null
+++ b/libgo/go/golang.org/x/crypto/internal/poly1305/bits_go1.13.go
@@ -0,0 +1,22 @@
+// Copyright 2019 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.
+
+//go:build go1.13
+// +build go1.13
+
+package poly1305
+
+import "math/bits"
+
+func bitsAdd64(x, y, carry uint64) (sum, carryOut uint64) {
+ return bits.Add64(x, y, carry)
+}
+
+func bitsSub64(x, y, borrow uint64) (diff, borrowOut uint64) {
+ return bits.Sub64(x, y, borrow)
+}
+
+func bitsMul64(x, y uint64) (hi, lo uint64) {
+ return bits.Mul64(x, y)
+}
diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/mac_noasm.go b/libgo/go/golang.org/x/crypto/internal/poly1305/mac_noasm.go
new file mode 100644
index 00000000000..f184b67d98d
--- /dev/null
+++ b/libgo/go/golang.org/x/crypto/internal/poly1305/mac_noasm.go
@@ -0,0 +1,10 @@
+// Copyright 2018 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.
+
+//go:build (!amd64 && !ppc64le && !s390x) || !gc || purego
+// +build !amd64,!ppc64le,!s390x !gc purego
+
+package poly1305
+
+type mac struct{ macGeneric }
diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/poly1305.go b/libgo/go/golang.org/x/crypto/internal/poly1305/poly1305.go
new file mode 100644
index 00000000000..4aaea810a26
--- /dev/null
+++ b/libgo/go/golang.org/x/crypto/internal/poly1305/poly1305.go
@@ -0,0 +1,99 @@
+// Copyright 2012 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 poly1305 implements Poly1305 one-time message authentication code as
+// specified in https://cr.yp.to/mac/poly1305-20050329.pdf.
+//
+// Poly1305 is a fast, one-time authentication function. It is infeasible for an
+// attacker to generate an authenticator for a message without the key. However, a
+// key must only be used for a single message. Authenticating two different
+// messages with the same key allows an attacker to forge authenticators for other
+// messages with the same key.
+//
+// Poly1305 was originally coupled with AES in order to make Poly1305-AES. AES was
+// used with a fixed key in order to generate one-time keys from an nonce.
+// However, in this package AES isn't used and the one-time key is specified
+// directly.
+package poly1305
+
+import "crypto/subtle"
+
+// TagSize is the size, in bytes, of a poly1305 authenticator.
+const TagSize = 16
+
+// Sum generates an authenticator for msg using a one-time key and puts the
+// 16-byte result into out. Authenticating two different messages with the same
+// key allows an attacker to forge messages at will.
+func Sum(out *[16]byte, m []byte, key *[32]byte) {
+ h := New(key)
+ h.Write(m)
+ h.Sum(out[:0])
+}
+
+// Verify returns true if mac is a valid authenticator for m with the given key.
+func Verify(mac *[16]byte, m []byte, key *[32]byte) bool {
+ var tmp [16]byte
+ Sum(&tmp, m, key)
+ return subtle.ConstantTimeCompare(tmp[:], mac[:]) == 1
+}
+
+// New returns a new MAC computing an authentication
+// tag of all data written to it with the given key.
+// This allows writing the message progressively instead
+// of passing it as a single slice. Common users should use
+// the Sum function instead.
+//
+// The key must be unique for each message, as authenticating
+// two different messages with the same key allows an attacker
+// to forge messages at will.
+func New(key *[32]byte) *MAC {
+ m := &MAC{}
+ initialize(key, &m.macState)
+ return m
+}
+
+// MAC is an io.Writer computing an authentication tag
+// of the data written to it.
+//
+// MAC cannot be used like common hash.Hash implementations,
+// because using a poly1305 key twice breaks its security.
+// Therefore writing data to a running MAC after calling
+// Sum or Verify causes it to panic.
+type MAC struct {
+ mac // platform-dependent implementation
+
+ finalized bool
+}
+
+// Size returns the number of bytes Sum will return.
+func (h *MAC) Size() int { return TagSize }
+
+// Write adds more data to the running message authentication code.
+// It never returns an error.
+//
+// It must not be called after the first call of Sum or Verify.
+func (h *MAC) Write(p []byte) (n int, err error) {
+ if h.finalized {
+ panic("poly1305: write to MAC after Sum or Verify")
+ }
+ return h.mac.Write(p)
+}
+
+// Sum computes the authenticator of all data written to the
+// message authentication code.
+func (h *MAC) Sum(b []byte) []byte {
+ var mac [TagSize]byte
+ h.mac.Sum(&mac)
+ h.finalized = true
+ return append(b, mac[:]...)
+}
+
+// Verify returns whether the authenticator of all data written to
+// the message authentication code matches the expected value.
+func (h *MAC) Verify(expected []byte) bool {
+ var mac [TagSize]byte
+ h.mac.Sum(&mac)
+ h.finalized = true
+ return subtle.ConstantTimeCompare(expected, mac[:]) == 1
+}
diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/sum_amd64.go b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_amd64.go
new file mode 100644
index 00000000000..6d522333f29
--- /dev/null
+++ b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_amd64.go
@@ -0,0 +1,48 @@
+// Copyright 2012 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.
+
+//go:build gc && !purego
+// +build gc,!purego
+
+package poly1305
+
+//go:noescape
+func update(state *macState, msg []byte)
+
+// mac is a wrapper for macGeneric that redirects calls that would have gone to
+// updateGeneric to update.
+//
+// Its Write and Sum methods are otherwise identical to the macGeneric ones, but
+// using function pointers would carry a major performance cost.
+type mac struct{ macGeneric }
+
+func (h *mac) Write(p []byte) (int, error) {
+ nn := len(p)
+ if h.offset > 0 {
+ n := copy(h.buffer[h.offset:], p)
+ if h.offset+n < TagSize {
+ h.offset += n
+ return nn, nil
+ }
+ p = p[n:]
+ h.offset = 0
+ update(&h.macState, h.buffer[:])
+ }
+ if n := len(p) - (len(p) % TagSize); n > 0 {
+ update(&h.macState, p[:n])
+ p = p[n:]
+ }
+ if len(p) > 0 {
+ h.offset += copy(h.buffer[h.offset:], p)
+ }
+ return nn, nil
+}
+
+func (h *mac) Sum(out *[16]byte) {
+ state := h.macState
+ if h.offset > 0 {
+ update(&state, h.buffer[:h.offset])
+ }
+ finalize(out, &state.h, &state.s)
+}
diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/sum_generic.go b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_generic.go
new file mode 100644
index 00000000000..c942a65904f
--- /dev/null
+++ b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_generic.go
@@ -0,0 +1,310 @@
+// Copyright 2018 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.
+
+// This file provides the generic implementation of Sum and MAC. Other files
+// might provide optimized assembly implementations of some of this code.
+
+package poly1305
+
+import "encoding/binary"
+
+// Poly1305 [RFC 7539] is a relatively simple algorithm: the authentication tag
+// for a 64 bytes message is approximately
+//
+// s + m[0:16] * r⁴ + m[16:32] * r³ + m[32:48] * r² + m[48:64] * r mod 2¹³⁰ - 5
+//
+// for some secret r and s. It can be computed sequentially like
+//
+// for len(msg) > 0:
+// h += read(msg, 16)
+// h *= r
+// h %= 2¹³⁰ - 5
+// return h + s
+//
+// All the complexity is about doing performant constant-time math on numbers
+// larger than any available numeric type.
+
+func sumGeneric(out *[TagSize]byte, msg []byte, key *[32]byte) {
+ h := newMACGeneric(key)
+ h.Write(msg)
+ h.Sum(out)
+}
+
+func newMACGeneric(key *[32]byte) macGeneric {
+ m := macGeneric{}
+ initialize(key, &m.macState)
+ return m
+}
+
+// macState holds numbers in saturated 64-bit little-endian limbs. That is,
+// the value of [x0, x1, x2] is x[0] + x[1] * 2⁶⁴ + x[2] * 2¹²⁸.
+type macState struct {
+ // h is the main accumulator. It is to be interpreted modulo 2¹³⁰ - 5, but
+ // can grow larger during and after rounds. It must, however, remain below
+ // 2 * (2¹³⁰ - 5).
+ h [3]uint64
+ // r and s are the private key components.
+ r [2]uint64
+ s [2]uint64
+}
+
+type macGeneric struct {
+ macState
+
+ buffer [TagSize]byte
+ offset int
+}
+
+// Write splits the incoming message into TagSize chunks, and passes them to
+// update. It buffers incomplete chunks.
+func (h *macGeneric) Write(p []byte) (int, error) {
+ nn := len(p)
+ if h.offset > 0 {
+ n := copy(h.buffer[h.offset:], p)
+ if h.offset+n < TagSize {
+ h.offset += n
+ return nn, nil
+ }
+ p = p[n:]
+ h.offset = 0
+ updateGeneric(&h.macState, h.buffer[:])
+ }
+ if n := len(p) - (len(p) % TagSize); n > 0 {
+ updateGeneric(&h.macState, p[:n])
+ p = p[n:]
+ }
+ if len(p) > 0 {
+ h.offset += copy(h.buffer[h.offset:], p)
+ }
+ return nn, nil
+}
+
+// Sum flushes the last incomplete chunk from the buffer, if any, and generates
+// the MAC output. It does not modify its state, in order to allow for multiple
+// calls to Sum, even if no Write is allowed after Sum.
+func (h *macGeneric) Sum(out *[TagSize]byte) {
+ state := h.macState
+ if h.offset > 0 {
+ updateGeneric(&state, h.buffer[:h.offset])
+ }
+ finalize(out, &state.h, &state.s)
+}
+
+// [rMask0, rMask1] is the specified Poly1305 clamping mask in little-endian. It
+// clears some bits of the secret coefficient to make it possible to implement
+// multiplication more efficiently.
+const (
+ rMask0 = 0x0FFFFFFC0FFFFFFF
+ rMask1 = 0x0FFFFFFC0FFFFFFC
+)
+
+// initialize loads the 256-bit key into the two 128-bit secret values r and s.
+func initialize(key *[32]byte, m *macState) {
+ m.r[0] = binary.LittleEndian.Uint64(key[0:8]) & rMask0
+ m.r[1] = binary.LittleEndian.Uint64(key[8:16]) & rMask1
+ m.s[0] = binary.LittleEndian.Uint64(key[16:24])
+ m.s[1] = binary.LittleEndian.Uint64(key[24:32])
+}
+
+// uint128 holds a 128-bit number as two 64-bit limbs, for use with the
+// bits.Mul64 and bits.Add64 intrinsics.
+type uint128 struct {
+ lo, hi uint64
+}
+
+func mul64(a, b uint64) uint128 {
+ hi, lo := bitsMul64(a, b)
+ return uint128{lo, hi}
+}
+
+func add128(a, b uint128) uint128 {
+ lo, c := bitsAdd64(a.lo, b.lo, 0)
+ hi, c := bitsAdd64(a.hi, b.hi, c)
+ if c != 0 {
+ panic("poly1305: unexpected overflow")
+ }
+ return uint128{lo, hi}
+}
+
+func shiftRightBy2(a uint128) uint128 {
+ a.lo = a.lo>>2 | (a.hi&3)<<62
+ a.hi = a.hi >> 2
+ return a
+}
+
+// updateGeneric absorbs msg into the state.h accumulator. For each chunk m of
+// 128 bits of message, it computes
+//
+// h₊ = (h + m) * r mod 2¹³⁰ - 5
+//
+// If the msg length is not a multiple of TagSize, it assumes the last
+// incomplete chunk is the final one.
+func updateGeneric(state *macState, msg []byte) {
+ h0, h1, h2 := state.h[0], state.h[1], state.h[2]
+ r0, r1 := state.r[0], state.r[1]
+
+ for len(msg) > 0 {
+ var c uint64
+
+ // For the first step, h + m, we use a chain of bits.Add64 intrinsics.
+ // The resulting value of h might exceed 2¹³⁰ - 5, but will be partially
+ // reduced at the end of the multiplication below.
+ //
+ // The spec requires us to set a bit just above the message size, not to
+ // hide leading zeroes. For full chunks, that's 1 << 128, so we can just
+ // add 1 to the most significant (2¹²⁸) limb, h2.
+ if len(msg) >= TagSize {
+ h0, c = bitsAdd64(h0, binary.LittleEndian.Uint64(msg[0:8]), 0)
+ h1, c = bitsAdd64(h1, binary.LittleEndian.Uint64(msg[8:16]), c)
+ h2 += c + 1
+
+ msg = msg[TagSize:]
+ } else {
+ var buf [TagSize]byte
+ copy(buf[:], msg)
+ buf[len(msg)] = 1
+
+ h0, c = bitsAdd64(h0, binary.LittleEndian.Uint64(buf[0:8]), 0)
+ h1, c = bitsAdd64(h1, binary.LittleEndian.Uint64(buf[8:16]), c)
+ h2 += c
+
+ msg = nil
+ }
+
+ // Multiplication of big number limbs is similar to elementary school
+ // columnar multiplication. Instead of digits, there are 64-bit limbs.
+ //
+ // We are multiplying a 3 limbs number, h, by a 2 limbs number, r.
+ //
+ // h2 h1 h0 x
+ // r1 r0 =
+ // ----------------
+ // h2r0 h1r0 h0r0 <-- individual 128-bit products
+ // + h2r1 h1r1 h0r1
+ // ------------------------
+ // m3 m2 m1 m0 <-- result in 128-bit overlapping limbs
+ // ------------------------
+ // m3.hi m2.hi m1.hi m0.hi <-- carry propagation
+ // + m3.lo m2.lo m1.lo m0.lo
+ // -------------------------------
+ // t4 t3 t2 t1 t0 <-- final result in 64-bit limbs
+ //
+ // The main difference from pen-and-paper multiplication is that we do
+ // carry propagation in a separate step, as if we wrote two digit sums
+ // at first (the 128-bit limbs), and then carried the tens all at once.
+
+ h0r0 := mul64(h0, r0)
+ h1r0 := mul64(h1, r0)
+ h2r0 := mul64(h2, r0)
+ h0r1 := mul64(h0, r1)
+ h1r1 := mul64(h1, r1)
+ h2r1 := mul64(h2, r1)
+
+ // Since h2 is known to be at most 7 (5 + 1 + 1), and r0 and r1 have their
+ // top 4 bits cleared by rMask{0,1}, we know that their product is not going
+ // to overflow 64 bits, so we can ignore the high part of the products.
+ //
+ // This also means that the product doesn't have a fifth limb (t4).
+ if h2r0.hi != 0 {
+ panic("poly1305: unexpected overflow")
+ }
+ if h2r1.hi != 0 {
+ panic("poly1305: unexpected overflow")
+ }
+
+ m0 := h0r0
+ m1 := add128(h1r0, h0r1) // These two additions don't overflow thanks again
+ m2 := add128(h2r0, h1r1) // to the 4 masked bits at the top of r0 and r1.
+ m3 := h2r1
+
+ t0 := m0.lo
+ t1, c := bitsAdd64(m1.lo, m0.hi, 0)
+ t2, c := bitsAdd64(m2.lo, m1.hi, c)
+ t3, _ := bitsAdd64(m3.lo, m2.hi, c)
+
+ // Now we have the result as 4 64-bit limbs, and we need to reduce it
+ // modulo 2¹³⁰ - 5. The special shape of this Crandall prime lets us do
+ // a cheap partial reduction according to the reduction identity
+ //
+ // c * 2¹³⁰ + n = c * 5 + n mod 2¹³⁰ - 5
+ //
+ // because 2¹³⁰ = 5 mod 2¹³⁰ - 5. Partial reduction since the result is
+ // likely to be larger than 2¹³⁰ - 5, but still small enough to fit the
+ // assumptions we make about h in the rest of the code.
+ //
+ // See also https://speakerdeck.com/gtank/engineering-prime-numbers?slide=23
+
+ // We split the final result at the 2¹³⁰ mark into h and cc, the carry.
+ // Note that the carry bits are effectively shifted left by 2, in other
+ // words, cc = c * 4 for the c in the reduction identity.
+ h0, h1, h2 = t0, t1, t2&maskLow2Bits
+ cc := uint128{t2 & maskNotLow2Bits, t3}
+
+ // To add c * 5 to h, we first add cc = c * 4, and then add (cc >> 2) = c.
+
+ h0, c = bitsAdd64(h0, cc.lo, 0)
+ h1, c = bitsAdd64(h1, cc.hi, c)
+ h2 += c
+
+ cc = shiftRightBy2(cc)
+
+ h0, c = bitsAdd64(h0, cc.lo, 0)
+ h1, c = bitsAdd64(h1, cc.hi, c)
+ h2 += c
+
+ // h2 is at most 3 + 1 + 1 = 5, making the whole of h at most
+ //
+ // 5 * 2¹²⁸ + (2¹²⁸ - 1) = 6 * 2¹²⁸ - 1
+ }
+
+ state.h[0], state.h[1], state.h[2] = h0, h1, h2
+}
+
+const (
+ maskLow2Bits uint64 = 0x0000000000000003
+ maskNotLow2Bits uint64 = ^maskLow2Bits
+)
+
+// select64 returns x if v == 1 and y if v == 0, in constant time.
+func select64(v, x, y uint64) uint64 { return ^(v-1)&x | (v-1)&y }
+
+// [p0, p1, p2] is 2¹³⁰ - 5 in little endian order.
+const (
+ p0 = 0xFFFFFFFFFFFFFFFB
+ p1 = 0xFFFFFFFFFFFFFFFF
+ p2 = 0x0000000000000003
+)
+
+// finalize completes the modular reduction of h and computes
+//
+// out = h + s mod 2¹²⁸
+//
+func finalize(out *[TagSize]byte, h *[3]uint64, s *[2]uint64) {
+ h0, h1, h2 := h[0], h[1], h[2]
+
+ // After the partial reduction in updateGeneric, h might be more than
+ // 2¹³⁰ - 5, but will be less than 2 * (2¹³⁰ - 5). To complete the reduction
+ // in constant time, we compute t = h - (2¹³⁰ - 5), and select h as the
+ // result if the subtraction underflows, and t otherwise.
+
+ hMinusP0, b := bitsSub64(h0, p0, 0)
+ hMinusP1, b := bitsSub64(h1, p1, b)
+ _, b = bitsSub64(h2, p2, b)
+
+ // h = h if h < p else h - p
+ h0 = select64(b, h0, hMinusP0)
+ h1 = select64(b, h1, hMinusP1)
+
+ // Finally, we compute the last Poly1305 step
+ //
+ // tag = h + s mod 2¹²⁸
+ //
+ // by just doing a wide addition with the 128 low bits of h and discarding
+ // the overflow.
+ h0, c := bitsAdd64(h0, s[0], 0)
+ h1, _ = bitsAdd64(h1, s[1], c)
+
+ binary.LittleEndian.PutUint64(out[0:8], h0)
+ binary.LittleEndian.PutUint64(out[8:16], h1)
+}
diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/sum_ppc64le.go b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_ppc64le.go
new file mode 100644
index 00000000000..4a069941a6e
--- /dev/null
+++ b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_ppc64le.go
@@ -0,0 +1,48 @@
+// Copyright 2019 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.
+
+//go:build gc && !purego
+// +build gc,!purego
+
+package poly1305
+
+//go:noescape
+func update(state *macState, msg []byte)
+
+// mac is a wrapper for macGeneric that redirects calls that would have gone to
+// updateGeneric to update.
+//
+// Its Write and Sum methods are otherwise identical to the macGeneric ones, but
+// using function pointers would carry a major performance cost.
+type mac struct{ macGeneric }
+
+func (h *mac) Write(p []byte) (int, error) {
+ nn := len(p)
+ if h.offset > 0 {
+ n := copy(h.buffer[h.offset:], p)
+ if h.offset+n < TagSize {
+ h.offset += n
+ return nn, nil
+ }
+ p = p[n:]
+ h.offset = 0
+ update(&h.macState, h.buffer[:])
+ }
+ if n := len(p) - (len(p) % TagSize); n > 0 {
+ update(&h.macState, p[:n])
+ p = p[n:]
+ }
+ if len(p) > 0 {
+ h.offset += copy(h.buffer[h.offset:], p)
+ }
+ return nn, nil
+}
+
+func (h *mac) Sum(out *[16]byte) {
+ state := h.macState
+ if h.offset > 0 {
+ update(&state, h.buffer[:h.offset])
+ }
+ finalize(out, &state.h, &state.s)
+}
diff --git a/libgo/go/golang.org/x/crypto/internal/poly1305/sum_s390x.go b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_s390x.go
new file mode 100644
index 00000000000..62cc9f84709
--- /dev/null
+++ b/libgo/go/golang.org/x/crypto/internal/poly1305/sum_s390x.go
@@ -0,0 +1,76 @@
+// Copyright 2018 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.
+
+//go:build gc && !purego
+// +build gc,!purego
+
+package poly1305
+
+import (
+ "golang.org/x/sys/cpu"
+)
+
+// updateVX is an assembly implementation of Poly1305 that uses vector
+// instructions. It must only be called if the vector facility (vx) is
+// available.
+//go:noescape
+func updateVX(state *macState, msg []byte)
+
+// mac is a replacement for macGeneric that uses a larger buffer and redirects
+// calls that would have gone to updateGeneric to updateVX if the vector
+// facility is installed.
+//
+// A larger buffer is required for good performance because the vector
+// implementation has a higher fixed cost per call than the generic
+// implementation.
+type mac struct {
+ macState
+
+ buffer [16 * TagSize]byte // size must be a multiple of block size (16)
+ offset int
+}
+
+func (h *mac) Write(p []byte) (int, error) {
+ nn := len(p)
+ if h.offset > 0 {
+ n := copy(h.buffer[h.offset:], p)
+ if h.offset+n < len(h.buffer) {
+ h.offset += n
+ return nn, nil
+ }
+ p = p[n:]
+ h.offset = 0
+ if cpu.S390X.HasVX {
+ updateVX(&h.macState, h.buffer[:])
+ } else {
+ updateGeneric(&h.macState, h.buffer[:])
+ }
+ }
+
+ tail := len(p) % len(h.buffer) // number of bytes to copy into buffer
+ body := len(p) - tail // number of bytes to process now
+ if body > 0 {
+ if cpu.S390X.HasVX {
+ updateVX(&h.macState, p[:body])
+ } else {
+ updateGeneric(&h.macState, p[:body])
+ }
+ }
+ h.offset = copy(h.buffer[:], p[body:]) // copy tail bytes - can be 0
+ return nn, nil
+}
+
+func (h *mac) Sum(out *[TagSize]byte) {
+ state := h.macState
+ remainder := h.buffer[:h.offset]
+
+ // Use the generic implementation if we have 2 or fewer blocks left
+ // to sum. The vector implementation has a higher startup time.
+ if cpu.S390X.HasVX && len(remainder) > 2*TagSize {
+ updateVX(&state, remainder)
+ } else if len(remainder) > 0 {
+ updateGeneric(&state, remainder)
+ }
+ finalize(out, &state.h, &state.s)
+}