[feature] S3 support (#674)

* feat: vendor minio client

* feat: introduce storage package with s3 support

* feat: serve s3 files directly

this saves a lot of bandwith as the files are fetched from the object
store directly

* fix: use explicit local storage in tests

* feat: integrate s3 storage with the main server

* fix: add s3 config to cli tests

* docs: explicitly set values in example config

also adds license header to the storage package

* fix: use better http status code on s3 redirect

HTTP 302 Found is the best fit, as it signifies that the resource
requested was found but not under its presumed URL

307/TemporaryRedirect would mean that this resource is usually located
here, not in this case

303/SeeOther indicates that the redirection does not link to the
requested resource but to another page

* refactor: use context in storage driver interface

6 files changed, 820 insertions, 0 deletions

diff --git a/vendor/golang.org/x/crypto/argon2/argon2.go b/vendor/golang.org/x/crypto/argon2/argon2.go
new file mode 100644
index 000000000..29f0a2de4
--- /dev/null
+++ b/vendor/golang.org/x/crypto/argon2/argon2.go
@@ -0,0 +1,283 @@
+// 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))
+}
diff --git a/vendor/golang.org/x/crypto/argon2/blake2b.go b/vendor/golang.org/x/crypto/argon2/blake2b.go
new file mode 100644
index 000000000..10f46948d
--- /dev/null
+++ b/vendor/golang.org/x/crypto/argon2/blake2b.go
@@ -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])
+}
diff --git a/vendor/golang.org/x/crypto/argon2/blamka_amd64.go b/vendor/golang.org/x/crypto/argon2/blamka_amd64.go
new file mode 100644
index 000000000..a014ac92a
--- /dev/null
+++ b/vendor/golang.org/x/crypto/argon2/blamka_amd64.go
@@ -0,0 +1,61 @@
+// 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.
+
+//go:build amd64 && gc && !purego
+// +build amd64,gc,!purego
+
+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)
+}
diff --git a/vendor/golang.org/x/crypto/argon2/blamka_amd64.s b/vendor/golang.org/x/crypto/argon2/blamka_amd64.s
new file mode 100644
index 000000000..b2cc05150
--- /dev/null
+++ b/vendor/golang.org/x/crypto/argon2/blamka_amd64.s
@@ -0,0 +1,244 @@
+// 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.
+
+//go:build amd64 && gc && !purego
+// +build amd64,gc,!purego
+
+#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
diff --git a/vendor/golang.org/x/crypto/argon2/blamka_generic.go b/vendor/golang.org/x/crypto/argon2/blamka_generic.go
new file mode 100644
index 000000000..a481b2243
--- /dev/null
+++ b/vendor/golang.org/x/crypto/argon2/blamka_generic.go
@@ -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
+}
diff --git a/vendor/golang.org/x/crypto/argon2/blamka_ref.go b/vendor/golang.org/x/crypto/argon2/blamka_ref.go
new file mode 100644
index 000000000..167c59d2d
--- /dev/null
+++ b/vendor/golang.org/x/crypto/argon2/blamka_ref.go
@@ -0,0 +1,16 @@
+// 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.
+
+//go:build !amd64 || purego || !gc
+// +build !amd64 purego !gc
+
+package argon2
+
+func processBlock(out, in1, in2 *block) {
+	processBlockGeneric(out, in1, in2, false)
+}
+
+func processBlockXOR(out, in1, in2 *block) {
+	processBlockGeneric(out, in1, in2, true)
+}