Migrate this project to minio micro services code

This commit is contained in:
Harshavardhana
2015-10-16 11:26:01 -07:00
parent 8c4119cbeb
commit 762b798767
349 changed files with 3704 additions and 76049 deletions
-77
View File
@@ -1,77 +0,0 @@
/*
* Minio Client (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
// NOTE - Rename() not guaranteed to be atomic on all filesystems which are not fully POSIX compatible
// Package atomic provides atomic file write semantics by leveraging Rename's() atomicity.
package atomic
import (
"io/ioutil"
"os"
"path/filepath"
)
// File container provided for atomic file writes
type File struct {
*os.File
file string
}
// Close the file replacing, returns an error if any
func (f *File) Close() error {
// close the embedded fd
if err := f.File.Close(); err != nil {
return err
}
// atomic rename to final destination
if err := os.Rename(f.Name(), f.file); err != nil {
return err
}
return nil
}
// CloseAndPurge removes the temp file, closes the transaction and returns an error if any
func (f *File) CloseAndPurge() error {
// close the embedded fd
if err := f.File.Close(); err != nil {
return err
}
if err := os.Remove(f.Name()); err != nil {
return err
}
return nil
}
// FileCreate creates a new file at filePath for atomic writes, it also creates parent directories if they don't exist
func FileCreate(filePath string) (*File, error) {
// if parent directories do not exist, ioutil.TempFile doesn't create them
// handle such a case with os.MkdirAll()
if err := os.MkdirAll(filepath.Dir(filePath), 0700); err != nil {
return nil, err
}
f, err := ioutil.TempFile(filepath.Dir(filePath), filepath.Base(filePath))
if err != nil {
return nil, err
}
if err := os.Chmod(f.Name(), 0600); err != nil {
if err := os.Remove(f.Name()); err != nil {
return nil, err
}
return nil, err
}
return &File{File: f, file: filePath}, nil
}
-66
View File
@@ -1,66 +0,0 @@
/*
* Minio Client (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package atomic
import (
"io/ioutil"
"os"
"path/filepath"
"testing"
. "gopkg.in/check.v1"
)
func Test(t *testing.T) { TestingT(t) }
type MySuite struct {
root string
}
var _ = Suite(&MySuite{})
func (s *MySuite) SetUpSuite(c *C) {
root, err := ioutil.TempDir("/tmp", "atomic-")
c.Assert(err, IsNil)
s.root = root
}
func (s *MySuite) TearDownSuite(c *C) {
os.RemoveAll(s.root)
}
func (s *MySuite) TestAtomic(c *C) {
f, err := FileCreate(filepath.Join(s.root, "testfile"))
c.Assert(err, IsNil)
_, err = os.Stat(filepath.Join(s.root, "testfile"))
c.Assert(err, Not(IsNil))
err = f.Close()
c.Assert(err, IsNil)
_, err = os.Stat(filepath.Join(s.root, "testfile"))
c.Assert(err, IsNil)
}
func (s *MySuite) TestAtomicPurge(c *C) {
f, err := FileCreate(filepath.Join(s.root, "purgefile"))
c.Assert(err, IsNil)
_, err = os.Stat(filepath.Join(s.root, "purgefile"))
c.Assert(err, Not(IsNil))
err = f.CloseAndPurge()
c.Assert(err, IsNil)
err = f.Close()
c.Assert(err, Not(IsNil))
}
-39
View File
@@ -1,39 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package cpu
// cpuid, cpuidex
func cpuid(op uint32) (eax, ebx, ecx, edx uint32)
func cpuidex(op, op2 uint32) (eax, ebx, ecx, edx uint32)
// HasSSE41 - CPUID instruction verification wrapper for SSE41 extensions
func HasSSE41() bool {
_, _, c, _ := cpuid(1)
return ((c & (1 << 19)) != 0)
}
// HasAVX - CPUID instruction verification wrapper for AVX extensions
func HasAVX() bool {
_, _, c, _ := cpuid(1)
return ((c & (1 << 28)) != 0)
}
// HasAVX2 - CPUID instruction verification wrapper for AVX2 extensions
func HasAVX2() bool {
_, b, _, _ := cpuidex(7, 0)
return ((b & (1 << 5)) != 0)
}
-27
View File
@@ -1,27 +0,0 @@
// Copyright (c) 2015 Klaus Post, released under MIT License. See LICENSE file.
//
// See https://github.com/klauspost/cpuid/blob/master/LICENSE
//
// Using this inside Minio with modifications
//
// func cpuid(op uint32) (eax, ebx, ecx, edx uint32)
TEXT ·cpuid(SB),7,$0
MOVL op+0(FP),AX
CPUID
MOVL AX,eax+8(FP)
MOVL BX,ebx+12(FP)
MOVL CX,ecx+16(FP)
MOVL DX,edx+20(FP)
RET
// func cpuidex(op, op2 uint32) (eax, ebx, ecx, edx uint32)
TEXT ·cpuidex(SB),7,$0
MOVL op+0(FP),AX
MOVL op2+4(FP),CX
CPUID
MOVL AX,eax+8(FP)
MOVL BX,ebx+12(FP)
MOVL CX,ecx+16(FP)
MOVL DX,edx+20(FP)
RET
-76
View File
@@ -1,76 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package cpu_test
import (
"errors"
"os/exec"
"runtime"
"strings"
"testing"
"github.com/minio/minio/pkg/cpu"
. "gopkg.in/check.v1"
)
func Test(t *testing.T) { TestingT(t) }
type MySuite struct{}
var _ = Suite(&MySuite{})
func hasCPUFeatureFromOS(feature string) (bool, error) {
if runtime.GOOS == "linux" {
command := exec.Command("/bin/cat", "/proc/cpuinfo")
output, err := command.Output()
if err != nil {
return false, err
}
if strings.Contains(string(output), feature) {
return true, nil
}
return false, nil
}
return false, errors.New("Not Implemented on this platform")
}
func (s *MySuite) TestHasSSE41(c *C) {
if runtime.GOOS == "linux" {
var flag = cpu.HasSSE41()
osCheck, err := hasCPUFeatureFromOS("sse4_1")
c.Assert(err, IsNil)
c.Check(flag, Equals, osCheck)
}
}
func (s *MySuite) TestHasAVX(c *C) {
if runtime.GOOS == "linux" {
var flag = cpu.HasAVX()
osFlag, err := hasCPUFeatureFromOS("avx")
c.Assert(err, IsNil)
c.Check(osFlag, Equals, flag)
}
}
func (s *MySuite) TestHasAVX2(c *C) {
if runtime.GOOS == "linux" {
var flag = cpu.HasAVX2()
osFlag, err := hasCPUFeatureFromOS("avx2")
c.Assert(err, IsNil)
c.Check(osFlag, Equals, flag)
}
}
-11
View File
@@ -1,11 +0,0 @@
// Package cpu provides wrapper around assembly functions for checking processor
// instruction capabilities for SSE4.1, AVX, AVX2 support
//
// Example
//
// ``cpu.HasSSE41()`` returns true for SSE4.1 instruction support, false otherwise
//
// ``cpu.HasAVX()`` returns true for AVX instruction support, false otherwise
//
// ``cpu.HasAVX2()`` returns true for AVX2 instruction support, false otherwise
package cpu
-44
View File
@@ -1,44 +0,0 @@
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package md5
import (
"crypto/md5"
"io"
)
// Sum - low memory footprint io.Reader based md5sum helper
func Sum(reader io.Reader) ([]byte, error) {
hash := md5.New()
var err error
var length int
for err == nil {
byteBuffer := make([]byte, 1024*1024)
length, err = reader.Read(byteBuffer)
// While hash.Write() wouldn't mind a Nil byteBuffer
// It is necessary for us to verify this and break
if length == 0 {
break
}
byteBuffer = byteBuffer[0:length]
hash.Write(byteBuffer)
}
if err != io.EOF {
return nil, err
}
return hash.Sum(nil), nil
}
-24
View File
@@ -1,24 +0,0 @@
package md5_test
import (
"bytes"
"encoding/hex"
"testing"
"github.com/minio/minio/pkg/crypto/md5"
. "gopkg.in/check.v1"
)
func Test(t *testing.T) { TestingT(t) }
type MySuite struct{}
var _ = Suite(&MySuite{})
func (s *MySuite) TestMd5sum(c *C) {
testString := []byte("Test string")
expectedHash, _ := hex.DecodeString("0fd3dbec9730101bff92acc820befc34")
hash, err := md5.Sum(bytes.NewBuffer(testString))
c.Assert(err, IsNil)
c.Assert(bytes.Equal(expectedHash, hash), Equals, true)
}
-1
View File
@@ -1 +0,0 @@
*.syso
-202
View File
@@ -1,202 +0,0 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-167
View File
@@ -1,167 +0,0 @@
// Copyright 2009 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 of
// Golang project:
// https://github.com/golang/go/blob/master/LICENSE
// Using this part of Minio codebase under the license
// Apache License Version 2.0 with modifications
// Package sha1 implements the SHA1 hash algorithm as defined in RFC 3174.
package sha1
import (
"hash"
"io"
"github.com/minio/minio/pkg/cpu"
)
// The size of a SHA1 checksum in bytes.
const Size = 20
// The blocksize of SHA1 in bytes.
const BlockSize = 64
const (
chunk = 64
init0 = 0x67452301
init1 = 0xEFCDAB89
init2 = 0x98BADCFE
init3 = 0x10325476
init4 = 0xC3D2E1F0
)
// digest represents the partial evaluation of a checksum.
type digest struct {
h [5]uint32
x [chunk]byte
nx int
len uint64
}
// Reset digest
func (d *digest) Reset() {
d.h[0] = init0
d.h[1] = init1
d.h[2] = init2
d.h[3] = init3
d.h[4] = init4
d.nx = 0
d.len = 0
}
// New returns a new hash.Hash computing the SHA1 checksum.
func New() hash.Hash {
d := new(digest)
d.Reset()
return d
}
func block(dig *digest, p []byte) {
switch true {
case cpu.HasSSE41() == true:
blockSSE3(dig, p)
default:
blockGeneric(dig, p)
}
}
// Return output size
func (d *digest) Size() int { return Size }
// Return checksum blocksize
func (d *digest) BlockSize() int { return BlockSize }
// Write to digest
func (d *digest) Write(p []byte) (nn int, err error) {
nn = len(p)
d.len += uint64(nn)
if d.nx > 0 {
n := copy(d.x[d.nx:], p)
d.nx += n
if d.nx == chunk {
block(d, d.x[:])
d.nx = 0
}
p = p[n:]
}
if len(p) >= chunk {
n := len(p) &^ (chunk - 1)
block(d, p[:n])
p = p[n:]
}
if len(p) > 0 {
d.nx = copy(d.x[:], p)
}
return
}
// Return checksum bytes
func (d *digest) Sum(in []byte) []byte {
// Make a copy of d0 so that caller can keep writing and summing.
d0 := *d
hash := d0.checkSum()
return append(in, hash[:]...)
}
// Intermediate checksum function
func (d *digest) checkSum() [Size]byte {
len := d.len
// Padding. Add a 1 bit and 0 bits until 56 bytes mod 64.
var tmp [64]byte
tmp[0] = 0x80
if len%64 < 56 {
d.Write(tmp[0 : 56-len%64])
} else {
d.Write(tmp[0 : 64+56-len%64])
}
// Length in bits.
len <<= 3
for i := uint(0); i < 8; i++ {
tmp[i] = byte(len >> (56 - 8*i))
}
d.Write(tmp[0:8])
if d.nx != 0 {
panic("d.nx != 0")
}
var digest [Size]byte
for i, s := range d.h {
digest[i*4] = byte(s >> 24)
digest[i*4+1] = byte(s >> 16)
digest[i*4+2] = byte(s >> 8)
digest[i*4+3] = byte(s)
}
return digest
}
/// Convenience functions
// Sum1 - single caller sha1 helper
func Sum1(data []byte) [Size]byte {
var d digest
d.Reset()
d.Write(data)
return d.checkSum()
}
// Sum - io.Reader based streaming sha1 helper
func Sum(reader io.Reader) ([]byte, error) {
h := New()
var err error
for err == nil {
length := 0
byteBuffer := make([]byte, 1024*1024)
length, err = reader.Read(byteBuffer)
byteBuffer = byteBuffer[0:length]
h.Write(byteBuffer)
}
if err != io.EOF {
return nil, err
}
return h.Sum(nil), nil
}
-967
View File
@@ -1,967 +0,0 @@
/*
* Implement fast SHA-1 with AVX2 instructions. (x86_64)
*
* This file is provided under a dual BSD/GPLv2 license. When using or
* redistributing this file, you may do so under either license.
*
* GPL LICENSE SUMMARY
*
* Copyright(c) 2014 Intel Corporation.
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of version 2 of the GNU General Public License as
* published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but
* WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*
* Contact Information:
* Ilya Albrekht <ilya.albrekht@intel.com>
* Maxim Locktyukhin <maxim.locktyukhin@intel.com>
* Ronen Zohar <ronen.zohar@intel.com>
* Chandramouli Narayanan <mouli@linux.intel.com>
*
* BSD LICENSE
*
* Copyright(c) 2014 Intel Corporation.
*
* 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 Intel Corporation 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "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 THE COPYRIGHT
* OWNER OR CONTRIBUTORS 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.
*
*/
/*
* SHA-1 implementation with Intel(R) AVX2 instruction set extensions.
*
*This implementation is based on the previous SSSE3 release:
* https://software.intel.com/en-us/articles/improving-the-performance-of-the-secure-hash-algorithm-1
*
*Updates 20-byte SHA-1 record in 'hash' for even number of
*'num_blocks' consecutive 64-byte blocks
*
*/
/*
* Using this part of Minio codebase under the license
* Apache License Version 2.0 with modifications
*
*/
#ifdef HAS_AVX2
#ifndef ENTRY
#define ENTRY(name) \
.globl name ; \
.align 4,0x90 ; \
name:
#endif
#ifndef END
#define END(name) \
.size name, .-name
#endif
#ifndef ENDPROC
#define ENDPROC(name) \
.type name, @function ; \
END(name)
#endif
#define NUM_INVALID 100
#define TYPE_R32 0
#define TYPE_R64 1
#define TYPE_XMM 2
#define TYPE_INVALID 100
.macro R32_NUM opd r32
\opd = NUM_INVALID
.ifc \r32,%eax
\opd = 0
.endif
.ifc \r32,%ecx
\opd = 1
.endif
.ifc \r32,%edx
\opd = 2
.endif
.ifc \r32,%ebx
\opd = 3
.endif
.ifc \r32,%esp
\opd = 4
.endif
.ifc \r32,%ebp
\opd = 5
.endif
.ifc \r32,%esi
\opd = 6
.endif
.ifc \r32,%edi
\opd = 7
.endif
#ifdef X86_64
.ifc \r32,%r8d
\opd = 8
.endif
.ifc \r32,%r9d
\opd = 9
.endif
.ifc \r32,%r10d
\opd = 10
.endif
.ifc \r32,%r11d
\opd = 11
.endif
.ifc \r32,%r12d
\opd = 12
.endif
.ifc \r32,%r13d
\opd = 13
.endif
.ifc \r32,%r14d
\opd = 14
.endif
.ifc \r32,%r15d
\opd = 15
.endif
#endif
.endm
.macro R64_NUM opd r64
\opd = NUM_INVALID
#ifdef X86_64
.ifc \r64,%rax
\opd = 0
.endif
.ifc \r64,%rcx
\opd = 1
.endif
.ifc \r64,%rdx
\opd = 2
.endif
.ifc \r64,%rbx
\opd = 3
.endif
.ifc \r64,%rsp
\opd = 4
.endif
.ifc \r64,%rbp
\opd = 5
.endif
.ifc \r64,%rsi
\opd = 6
.endif
.ifc \r64,%rdi
\opd = 7
.endif
.ifc \r64,%r8
\opd = 8
.endif
.ifc \r64,%r9
\opd = 9
.endif
.ifc \r64,%r10
\opd = 10
.endif
.ifc \r64,%r11
\opd = 11
.endif
.ifc \r64,%r12
\opd = 12
.endif
.ifc \r64,%r13
\opd = 13
.endif
.ifc \r64,%r14
\opd = 14
.endif
.ifc \r64,%r15
\opd = 15
.endif
#endif
.endm
.macro XMM_NUM opd xmm
\opd = NUM_INVALID
.ifc \xmm,%xmm0
\opd = 0
.endif
.ifc \xmm,%xmm1
\opd = 1
.endif
.ifc \xmm,%xmm2
\opd = 2
.endif
.ifc \xmm,%xmm3
\opd = 3
.endif
.ifc \xmm,%xmm4
\opd = 4
.endif
.ifc \xmm,%xmm5
\opd = 5
.endif
.ifc \xmm,%xmm6
\opd = 6
.endif
.ifc \xmm,%xmm7
\opd = 7
.endif
.ifc \xmm,%xmm8
\opd = 8
.endif
.ifc \xmm,%xmm9
\opd = 9
.endif
.ifc \xmm,%xmm10
\opd = 10
.endif
.ifc \xmm,%xmm11
\opd = 11
.endif
.ifc \xmm,%xmm12
\opd = 12
.endif
.ifc \xmm,%xmm13
\opd = 13
.endif
.ifc \xmm,%xmm14
\opd = 14
.endif
.ifc \xmm,%xmm15
\opd = 15
.endif
.endm
.macro TYPE type reg
R32_NUM reg_type_r32 \reg
R64_NUM reg_type_r64 \reg
XMM_NUM reg_type_xmm \reg
.if reg_type_r64 <> NUM_INVALID
\type = TYPE_R64
.elseif reg_type_r32 <> NUM_INVALID
\type = TYPE_R32
.elseif reg_type_xmm <> NUM_INVALID
\type = TYPE_XMM
.else
\type = TYPE_INVALID
.endif
.endm
.macro PFX_OPD_SIZE
.byte 0x66
.endm
.macro PFX_REX opd1 opd2 W=0
.if ((\opd1 | \opd2) & 8) || \W
.byte 0x40 | ((\opd1 & 8) >> 3) | ((\opd2 & 8) >> 1) | (\W << 3)
.endif
.endm
.macro MODRM mod opd1 opd2
.byte \mod | (\opd1 & 7) | ((\opd2 & 7) << 3)
.endm
.macro PSHUFB_XMM xmm1 xmm2
XMM_NUM pshufb_opd1 \xmm1
XMM_NUM pshufb_opd2 \xmm2
PFX_OPD_SIZE
PFX_REX pshufb_opd1 pshufb_opd2
.byte 0x0f, 0x38, 0x00
MODRM 0xc0 pshufb_opd1 pshufb_opd2
.endm
.macro PCLMULQDQ imm8 xmm1 xmm2
XMM_NUM clmul_opd1 \xmm1
XMM_NUM clmul_opd2 \xmm2
PFX_OPD_SIZE
PFX_REX clmul_opd1 clmul_opd2
.byte 0x0f, 0x3a, 0x44
MODRM 0xc0 clmul_opd1 clmul_opd2
.byte \imm8
.endm
.macro PEXTRD imm8 xmm gpr
R32_NUM extrd_opd1 \gpr
XMM_NUM extrd_opd2 \xmm
PFX_OPD_SIZE
PFX_REX extrd_opd1 extrd_opd2
.byte 0x0f, 0x3a, 0x16
MODRM 0xc0 extrd_opd1 extrd_opd2
.byte \imm8
.endm
.macro MOVQ_R64_XMM opd1 opd2
TYPE movq_r64_xmm_opd1_type \opd1
.if movq_r64_xmm_opd1_type == TYPE_XMM
XMM_NUM movq_r64_xmm_opd1 \opd1
R64_NUM movq_r64_xmm_opd2 \opd2
.else
R64_NUM movq_r64_xmm_opd1 \opd1
XMM_NUM movq_r64_xmm_opd2 \opd2
.endif
PFX_OPD_SIZE
PFX_REX movq_r64_xmm_opd1 movq_r64_xmm_opd2 1
.if movq_r64_xmm_opd1_type == TYPE_XMM
.byte 0x0f, 0x7e
.else
.byte 0x0f, 0x6e
.endif
MODRM 0xc0 movq_r64_xmm_opd1 movq_r64_xmm_opd2
.endm
#define CTX %rdi /* arg1 */
#define BUF %rsi /* arg2 */
#define CNT %rdx /* arg3 */
#define REG_A %ecx
#define REG_B %esi
#define REG_C %edi
#define REG_D %eax
#define REG_E %edx
#define REG_TB %ebx
#define REG_TA %r12d
#define REG_RA %rcx
#define REG_RB %rsi
#define REG_RC %rdi
#define REG_RD %rax
#define REG_RE %rdx
#define REG_RTA %r12
#define REG_RTB %rbx
#define REG_T1 %ebp
#define xmm_mov vmovups
#define avx2_zeroupper vzeroupper
#define RND_F1 1
#define RND_F2 2
#define RND_F3 3
.macro REGALLOC
.set A, REG_A
.set B, REG_B
.set C, REG_C
.set D, REG_D
.set E, REG_E
.set TB, REG_TB
.set TA, REG_TA
.set RA, REG_RA
.set RB, REG_RB
.set RC, REG_RC
.set RD, REG_RD
.set RE, REG_RE
.set RTA, REG_RTA
.set RTB, REG_RTB
.set T1, REG_T1
.endm
#define K_BASE %r8
#define HASH_PTR %r9
#define BUFFER_PTR %r10
#define BUFFER_PTR2 %r13
#define BUFFER_END %r11
#define PRECALC_BUF %r14
#define WK_BUF %r15
#define W_TMP %xmm0
#define WY_TMP %ymm0
#define WY_TMP2 %ymm9
# AVX2 variables
#define WY0 %ymm3
#define WY4 %ymm5
#define WY08 %ymm7
#define WY12 %ymm8
#define WY16 %ymm12
#define WY20 %ymm13
#define WY24 %ymm14
#define WY28 %ymm15
#define YMM_SHUFB_BSWAP %ymm10
/*
* Keep 2 iterations precalculated at a time:
* - 80 DWORDs per iteration * 2
*/
#define W_SIZE (80*2*2 +16)
#define WK(t) ((((t) % 80) / 4)*32 + ( (t) % 4)*4 + ((t)/80)*16 )(WK_BUF)
#define PRECALC_WK(t) ((t)*2*2)(PRECALC_BUF)
.macro UPDATE_HASH hash, val
add \hash, \val
mov \val, \hash
.endm
.macro PRECALC_RESET_WY
.set WY_00, WY0
.set WY_04, WY4
.set WY_08, WY08
.set WY_12, WY12
.set WY_16, WY16
.set WY_20, WY20
.set WY_24, WY24
.set WY_28, WY28
.set WY_32, WY_00
.endm
.macro PRECALC_ROTATE_WY
/* Rotate macros */
.set WY_32, WY_28
.set WY_28, WY_24
.set WY_24, WY_20
.set WY_20, WY_16
.set WY_16, WY_12
.set WY_12, WY_08
.set WY_08, WY_04
.set WY_04, WY_00
.set WY_00, WY_32
/* Define register aliases */
.set WY, WY_00
.set WY_minus_04, WY_04
.set WY_minus_08, WY_08
.set WY_minus_12, WY_12
.set WY_minus_16, WY_16
.set WY_minus_20, WY_20
.set WY_minus_24, WY_24
.set WY_minus_28, WY_28
.set WY_minus_32, WY
.endm
.macro PRECALC_00_15
.if (i == 0) # Initialize and rotate registers
PRECALC_RESET_WY
PRECALC_ROTATE_WY
.endif
/* message scheduling pre-compute for rounds 0-15 */
.if ((i & 7) == 0)
/*
* blended AVX2 and ALU instruction scheduling
* 1 vector iteration per 8 rounds
*/
vmovdqu ((i * 2) + PRECALC_OFFSET)(BUFFER_PTR), W_TMP
.elseif ((i & 7) == 1)
vinsertf128 $1, (((i-1) * 2)+PRECALC_OFFSET)(BUFFER_PTR2),\
WY_TMP, WY_TMP
.elseif ((i & 7) == 2)
vpshufb YMM_SHUFB_BSWAP, WY_TMP, WY
.elseif ((i & 7) == 4)
vpaddd K_XMM(K_BASE), WY, WY_TMP
.elseif ((i & 7) == 7)
vmovdqu WY_TMP, PRECALC_WK(i&~7)
PRECALC_ROTATE_WY
.endif
.endm
.macro PRECALC_16_31
/*
* message scheduling pre-compute for rounds 16-31
* calculating last 32 w[i] values in 8 XMM registers
* pre-calculate K+w[i] values and store to mem
* for later load by ALU add instruction
*
* "brute force" vectorization for rounds 16-31 only
* due to w[i]->w[i-3] dependency
*/
.if ((i & 7) == 0)
/*
* blended AVX2 and ALU instruction scheduling
* 1 vector iteration per 8 rounds
*/
/* w[i-14] */
vpalignr $8, WY_minus_16, WY_minus_12, WY
vpsrldq $4, WY_minus_04, WY_TMP /* w[i-3] */
.elseif ((i & 7) == 1)
vpxor WY_minus_08, WY, WY
vpxor WY_minus_16, WY_TMP, WY_TMP
.elseif ((i & 7) == 2)
vpxor WY_TMP, WY, WY
vpslldq $12, WY, WY_TMP2
.elseif ((i & 7) == 3)
vpslld $1, WY, WY_TMP
vpsrld $31, WY, WY
.elseif ((i & 7) == 4)
vpor WY, WY_TMP, WY_TMP
vpslld $2, WY_TMP2, WY
.elseif ((i & 7) == 5)
vpsrld $30, WY_TMP2, WY_TMP2
vpxor WY, WY_TMP, WY_TMP
.elseif ((i & 7) == 7)
vpxor WY_TMP2, WY_TMP, WY
vpaddd K_XMM(K_BASE), WY, WY_TMP
vmovdqu WY_TMP, PRECALC_WK(i&~7)
PRECALC_ROTATE_WY
.endif
.endm
.macro PRECALC_32_79
/*
* in SHA-1 specification:
* w[i] = (w[i-3] ^ w[i-8] ^ w[i-14] ^ w[i-16]) rol 1
* instead we do equal:
* w[i] = (w[i-6] ^ w[i-16] ^ w[i-28] ^ w[i-32]) rol 2
* allows more efficient vectorization
* since w[i]=>w[i-3] dependency is broken
*/
.if ((i & 7) == 0)
/*
* blended AVX2 and ALU instruction scheduling
* 1 vector iteration per 8 rounds
*/
vpalignr $8, WY_minus_08, WY_minus_04, WY_TMP
.elseif ((i & 7) == 1)
/* W is W_minus_32 before xor */
vpxor WY_minus_28, WY, WY
.elseif ((i & 7) == 2)
vpxor WY_minus_16, WY_TMP, WY_TMP
.elseif ((i & 7) == 3)
vpxor WY_TMP, WY, WY
.elseif ((i & 7) == 4)
vpslld $2, WY, WY_TMP
.elseif ((i & 7) == 5)
vpsrld $30, WY, WY
vpor WY, WY_TMP, WY
.elseif ((i & 7) == 7)
vpaddd K_XMM(K_BASE), WY, WY_TMP
vmovdqu WY_TMP, PRECALC_WK(i&~7)
PRECALC_ROTATE_WY
.endif
.endm
.macro PRECALC r, s
.set i, \r
.if (i < 40)
.set K_XMM, 32*0
.elseif (i < 80)
.set K_XMM, 32*1
.elseif (i < 120)
.set K_XMM, 32*2
.else
.set K_XMM, 32*3
.endif
.if (i<32)
PRECALC_00_15 \s
.elseif (i<64)
PRECALC_16_31 \s
.elseif (i < 160)
PRECALC_32_79 \s
.endif
.endm
.macro ROTATE_STATE
.set T_REG, E
.set E, D
.set D, C
.set C, B
.set B, TB
.set TB, A
.set A, T_REG
.set T_REG, RE
.set RE, RD
.set RD, RC
.set RC, RB
.set RB, RTB
.set RTB, RA
.set RA, T_REG
.endm
/* Macro relies on saved ROUND_Fx */
.macro RND_FUN f, r
.if (\f == RND_F1)
ROUND_F1 \r
.elseif (\f == RND_F2)
ROUND_F2 \r
.elseif (\f == RND_F3)
ROUND_F3 \r
.endif
.endm
.macro RR r
.set round_id, (\r % 80)
.if (round_id == 0) /* Precalculate F for first round */
.set ROUND_FUNC, RND_F1
mov B, TB
rorx $(32-30), B, B /* b>>>2 */
andn D, TB, T1
and C, TB
xor T1, TB
.endif
RND_FUN ROUND_FUNC, \r
ROTATE_STATE
.if (round_id == 18)
.set ROUND_FUNC, RND_F2
.elseif (round_id == 38)
.set ROUND_FUNC, RND_F3
.elseif (round_id == 58)
.set ROUND_FUNC, RND_F2
.endif
.set round_id, ( (\r+1) % 80)
RND_FUN ROUND_FUNC, (\r+1)
ROTATE_STATE
.endm
.macro ROUND_F1 r
add WK(\r), E
andn C, A, T1 /* ~b&d */
lea (RE,RTB), E /* Add F from the previous round */
rorx $(32-5), A, TA /* T2 = A >>> 5 */
rorx $(32-30),A, TB /* b>>>2 for next round */
PRECALC (\r) /* msg scheduling for next 2 blocks */
/*
* Calculate F for the next round
* (b & c) ^ andn[b, d]
*/
and B, A /* b&c */
xor T1, A /* F1 = (b&c) ^ (~b&d) */
lea (RE,RTA), E /* E += A >>> 5 */
.endm
.macro ROUND_F2 r
add WK(\r), E
lea (RE,RTB), E /* Add F from the previous round */
/* Calculate F for the next round */
rorx $(32-5), A, TA /* T2 = A >>> 5 */
.if ((round_id) < 79)
rorx $(32-30), A, TB /* b>>>2 for next round */
.endif
PRECALC (\r) /* msg scheduling for next 2 blocks */
.if ((round_id) < 79)
xor B, A
.endif
add TA, E /* E += A >>> 5 */
.if ((round_id) < 79)
xor C, A
.endif
.endm
.macro ROUND_F3 r
add WK(\r), E
PRECALC (\r) /* msg scheduling for next 2 blocks */
lea (RE,RTB), E /* Add F from the previous round */
mov B, T1
or A, T1
rorx $(32-5), A, TA /* T2 = A >>> 5 */
rorx $(32-30), A, TB /* b>>>2 for next round */
/* Calculate F for the next round
* (b and c) or (d and (b or c))
*/
and C, T1
and B, A
or T1, A
add TA, E /* E += A >>> 5 */
.endm
/*
* macro implements 80 rounds of SHA-1, for multiple blocks with s/w pipelining
*/
.macro SHA1_PIPELINED_MAIN_BODY
REGALLOC
mov (HASH_PTR), A
mov 4(HASH_PTR), B
mov 8(HASH_PTR), C
mov 12(HASH_PTR), D
mov 16(HASH_PTR), E
mov %rsp, PRECALC_BUF
lea (2*4*80+32)(%rsp), WK_BUF
# Precalc WK for first 2 blocks
PRECALC_OFFSET = 0
.set i, 0
.rept 160
PRECALC i
.set i, i + 1
.endr
PRECALC_OFFSET = 128
xchg WK_BUF, PRECALC_BUF
.align 32
_loop:
/*
* code loops through more than one block
* we use K_BASE value as a signal of a last block,
* it is set below by: cmovae BUFFER_PTR, K_BASE
*/
cmp K_BASE, BUFFER_PTR
jne _begin
.align 32
jmp _end
.align 32
_begin:
/*
* Do first block
* rounds: 0,2,4,6,8
*/
.set j, 0
.rept 5
RR j
.set j, j+2
.endr
jmp _loop0
_loop0:
/*
* rounds:
* 10,12,14,16,18
* 20,22,24,26,28
* 30,32,34,36,38
* 40,42,44,46,48
* 50,52,54,56,58
*/
.rept 25
RR j
.set j, j+2
.endr
add $(2*64), BUFFER_PTR /* move to next odd-64-byte block */
cmp BUFFER_END, BUFFER_PTR /* is current block the last one? */
cmovae K_BASE, BUFFER_PTR /* signal the last iteration smartly */
/*
* rounds
* 60,62,64,66,68
* 70,72,74,76,78
*/
.rept 10
RR j
.set j, j+2
.endr
UPDATE_HASH (HASH_PTR), A
UPDATE_HASH 4(HASH_PTR), TB
UPDATE_HASH 8(HASH_PTR), C
UPDATE_HASH 12(HASH_PTR), D
UPDATE_HASH 16(HASH_PTR), E
cmp K_BASE, BUFFER_PTR /* is current block the last one? */
je _loop
mov TB, B
/* Process second block */
/*
* rounds
* 0+80, 2+80, 4+80, 6+80, 8+80
* 10+80,12+80,14+80,16+80,18+80
*/
.set j, 0
.rept 10
RR j+80
.set j, j+2
.endr
jmp _loop1
_loop1:
/*
* rounds
* 20+80,22+80,24+80,26+80,28+80
* 30+80,32+80,34+80,36+80,38+80
*/
.rept 10
RR j+80
.set j, j+2
.endr
jmp _loop2
_loop2:
/*
* rounds
* 40+80,42+80,44+80,46+80,48+80
* 50+80,52+80,54+80,56+80,58+80
*/
.rept 10
RR j+80
.set j, j+2
.endr
add $(2*64), BUFFER_PTR2 /* move to next even-64-byte block */
cmp BUFFER_END, BUFFER_PTR2 /* is current block the last one */
cmovae K_BASE, BUFFER_PTR /* signal the last iteration smartly */
jmp _loop3
_loop3:
/*
* rounds
* 60+80,62+80,64+80,66+80,68+80
* 70+80,72+80,74+80,76+80,78+80
*/
.rept 10
RR j+80
.set j, j+2
.endr
UPDATE_HASH (HASH_PTR), A
UPDATE_HASH 4(HASH_PTR), TB
UPDATE_HASH 8(HASH_PTR), C
UPDATE_HASH 12(HASH_PTR), D
UPDATE_HASH 16(HASH_PTR), E
/* Reset state for AVX2 reg permutation */
mov A, TA
mov TB, A
mov C, TB
mov E, C
mov D, B
mov TA, D
REGALLOC
xchg WK_BUF, PRECALC_BUF
jmp _loop
.align 32
_end:
.endm
.section .rodata
#define K1 0x5a827999
#define K2 0x6ed9eba1
#define K3 0x8f1bbcdc
#define K4 0xca62c1d6
.align 128
K_XMM_AR:
.long K1, K1, K1, K1
.long K1, K1, K1, K1
.long K2, K2, K2, K2
.long K2, K2, K2, K2
.long K3, K3, K3, K3
.long K3, K3, K3, K3
.long K4, K4, K4, K4
.long K4, K4, K4, K4
BSWAP_SHUFB_CTL:
.long 0x00010203
.long 0x04050607
.long 0x08090a0b
.long 0x0c0d0e0f
.long 0x00010203
.long 0x04050607
.long 0x08090a0b
.long 0x0c0d0e0f
# void sha1_transform(int32_t *hash, const char* input, size_t num_blocks) ;
.text
ENTRY(sha1_transform)
push %rbx
push %rbp
push %r12
push %r13
push %r14
push %r15
RESERVE_STACK = (W_SIZE*4 + 8+24)
/* Align stack */
mov %rsp, %rbx
and $~(0x20-1), %rsp
push %rbx
sub $RESERVE_STACK, %rsp
avx2_zeroupper
lea K_XMM_AR(%rip), K_BASE
mov CTX, HASH_PTR
mov BUF, BUFFER_PTR
lea 64(BUF), BUFFER_PTR2
shl $6, CNT /* mul by 64 */
add BUF, CNT
add $64, CNT
mov CNT, BUFFER_END
cmp BUFFER_END, BUFFER_PTR2
cmovae K_BASE, BUFFER_PTR2
xmm_mov BSWAP_SHUFB_CTL(%rip), YMM_SHUFB_BSWAP
SHA1_PIPELINED_MAIN_BODY
avx2_zeroupper
add $RESERVE_STACK, %rsp
pop %rsp
pop %r15
pop %r14
pop %r13
pop %r12
pop %rbp
pop %rbx
ret
ENDPROC(sha1_transform)
#endif
-169
View File
@@ -1,169 +0,0 @@
// Copyright 2009 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 of
// Golang project:
// https://github.com/golang/go/blob/master/LICENSE
// Using this part of Minio codebase under the license
// Apache License Version 2.0 with modifications
// Package sha1 implements the SHA1 hash algorithm as defined in RFC 3174.
package sha1
import (
"hash"
"io"
"github.com/minio/minio/pkg/cpu"
)
// The size of a SHA1 checksum in bytes.
const Size = 20
// The blocksize of SHA1 in bytes.
const BlockSize = 64
const (
chunk = 64
init0 = 0x67452301
init1 = 0xEFCDAB89
init2 = 0x98BADCFE
init3 = 0x10325476
init4 = 0xC3D2E1F0
)
// digest represents the partial evaluation of a checksum.
type digest struct {
h [5]uint32
x [chunk]byte
nx int
len uint64
}
// Reset digest
func (d *digest) Reset() {
d.h[0] = init0
d.h[1] = init1
d.h[2] = init2
d.h[3] = init3
d.h[4] = init4
d.nx = 0
d.len = 0
}
// New returns a new hash.Hash computing the SHA1 checksum.
func New() hash.Hash {
d := new(digest)
d.Reset()
return d
}
func block(dig *digest, p []byte) {
switch true {
case cpu.HasAVX2() == true:
blockAVX2(dig, p)
case cpu.HasSSE41() == true:
blockSSE3(dig, p)
default:
blockGeneric(dig, p)
}
}
// Return output size
func (d *digest) Size() int { return Size }
// Return checksum blocksize
func (d *digest) BlockSize() int { return BlockSize }
// Write to digest
func (d *digest) Write(p []byte) (nn int, err error) {
nn = len(p)
d.len += uint64(nn)
if d.nx > 0 {
n := copy(d.x[d.nx:], p)
d.nx += n
if d.nx == chunk {
block(d, d.x[:])
d.nx = 0
}
p = p[n:]
}
if len(p) >= chunk {
n := len(p) &^ (chunk - 1)
block(d, p[:n])
p = p[n:]
}
if len(p) > 0 {
d.nx = copy(d.x[:], p)
}
return
}
// Return checksum bytes
func (d *digest) Sum(in []byte) []byte {
// Make a copy of d0 so that caller can keep writing and summing.
d0 := *d
hash := d0.checkSum()
return append(in, hash[:]...)
}
// Intermediate checksum function
func (d *digest) checkSum() [Size]byte {
len := d.len
// Padding. Add a 1 bit and 0 bits until 56 bytes mod 64.
var tmp [64]byte
tmp[0] = 0x80
if len%64 < 56 {
d.Write(tmp[0 : 56-len%64])
} else {
d.Write(tmp[0 : 64+56-len%64])
}
// Length in bits.
len <<= 3
for i := uint(0); i < 8; i++ {
tmp[i] = byte(len >> (56 - 8*i))
}
d.Write(tmp[0:8])
if d.nx != 0 {
panic("d.nx != 0")
}
var digest [Size]byte
for i, s := range d.h {
digest[i*4] = byte(s >> 24)
digest[i*4+1] = byte(s >> 16)
digest[i*4+2] = byte(s >> 8)
digest[i*4+3] = byte(s)
}
return digest
}
/// Convenience functions
// Sum1 - single caller sha1 helper
func Sum1(data []byte) [Size]byte {
var d digest
d.Reset()
d.Write(data)
return d.checkSum()
}
// Sum - io.Reader based streaming sha1 helper
func Sum(reader io.Reader) ([]byte, error) {
h := New()
var err error
for err == nil {
length := 0
byteBuffer := make([]byte, 1024*1024)
length, err = reader.Read(byteBuffer)
byteBuffer = byteBuffer[0:length]
h.Write(byteBuffer)
}
if err != io.EOF {
return nil, err
}
return h.Sum(nil), nil
}
-579
View File
@@ -1,579 +0,0 @@
;---------------------
; https://software.intel.com/en-us/articles/improving-the-performance-of-the-secure-hash-algorithm-1
;
; License information:
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; This implementation notably advances the performance of SHA-1 algorithm compared to existing
; implementations. We are encouraging all projects utilizing SHA-1 to integrate this new fast
; implementation and are ready to help if issues or concerns arise (you are welcome to leave
; a comment or write an email to the authors). It is provided 'as is' and free for either
; commercial or non-commercial use.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;
; This code implements two interfaces of SHA-1 update function: 1) working on a single
; 64-byte block and 2) working on a buffer of multiple 64-bit blocks. Multiple blocks
; version of code is software pipelined and faster overall, it is a default. Assemble
; with -DINTEL_SHA1_SINGLEBLOCK to select single 64-byte block function interface.
;
; C++ prototypes of implemented functions are below:
;
; #ifndef INTEL_SHA1_SINGLEBLOCK
; // Updates 20-byte SHA-1 record in 'hash' for 'num_blocks' consequtive 64-byte blocks
; extern "C" void sha1_update_intel(int *hash, const char* input, size_t num_blocks );
; #else
; // Updates 20-byte SHA-1 record in 'hash' for one 64-byte block pointed by 'input'
; extern "C" void sha1_update_intel(int *hash, const char* input);
; #endif
;
; Function name 'sha1_update_intel' can be changed in the source or via macro:
; -DINTEL_SHA1_UPDATE_FUNCNAME=my_sha1_update_func_name
;
; It implements both UNIX(default) and Windows ABIs, use -DWIN_ABI on Windows
;
; Code checks CPU for SSSE3 support via CPUID feature flag (CPUID.1.ECX.SSSE3[bit 9]==1),
; and performs dispatch. Since in most cases the functionality on non-SSSE3 supporting CPUs
; is also required, the default (e.g. one being replaced) function can be provided for
; dispatch on such CPUs, the name of old function can be changed in the source or via macro:
; -DINTEL_SHA1_UPDATE_DEFAULT_DISPATCH=default_sha1_update_function_name
;
; Authors: Maxim Locktyukhin and Ronen Zohar at Intel.com
;
%ifndef INTEL_SHA1_UPDATE_DEFAULT_DISPATCH
;; can be replaced with a default SHA-1 update function name
%define INTEL_SHA1_UPDATE_DEFAULT_DISPATCH sha1_intel_non_ssse3_cpu_stub_
%else
extern INTEL_SHA1_UPDATE_DEFAULT_DISPATCH
%endif
;; provide alternative SHA-1 update function's name here
%ifndef INTEL_SHA1_UPDATE_FUNCNAME
%define INTEL_SHA1_UPDATE_FUNCNAME sha1_update_intel
%endif
global INTEL_SHA1_UPDATE_FUNCNAME
%ifndef INTEL_SHA1_SINGLEBLOCK
%assign multiblock 1
%else
%assign multiblock 0
%endif
bits 64
default rel
%ifdef WIN_ABI
%xdefine arg1 rcx
%xdefine arg2 rdx
%xdefine arg3 r8
%else
%xdefine arg1 rdi
%xdefine arg2 rsi
%xdefine arg3 rdx
%endif
%xdefine ctx arg1
%xdefine buf arg2
%xdefine cnt arg3
%macro REGALLOC 0
%xdefine A ecx
%xdefine B esi
%xdefine C edi
%xdefine D ebp
%xdefine E edx
%xdefine T1 eax
%xdefine T2 ebx
%endmacro
%xdefine K_BASE r8
%xdefine HASH_PTR r9
%xdefine BUFFER_PTR r10
%xdefine BUFFER_END r11
%xdefine W_TMP xmm0
%xdefine W_TMP2 xmm9
%xdefine W0 xmm1
%xdefine W4 xmm2
%xdefine W8 xmm3
%xdefine W12 xmm4
%xdefine W16 xmm5
%xdefine W20 xmm6
%xdefine W24 xmm7
%xdefine W28 xmm8
%xdefine XMM_SHUFB_BSWAP xmm10
;; we keep window of 64 w[i]+K pre-calculated values in a circular buffer
%xdefine WK(t) (rsp + (t & 15)*4)
;------------------------------------------------------------------------------
;
; macro implements SHA-1 function's body for single or several 64-byte blocks
; first param: function's name
; second param: =0 - function implements single 64-byte block hash
; =1 - function implements multiple64-byte blocks hash
; 3rd function's argument is a number, greater 0, of 64-byte blocks to calc hash for
;
%macro SHA1_VECTOR_ASM 2
align 4096
%1:
push rbx
push rbp
%ifdef WIN_ABI
push rdi
push rsi
%xdefine stack_size (16*4 + 16*5 + 8)
%else
%xdefine stack_size (16*4 + 8)
%endif
sub rsp, stack_size
%ifdef WIN_ABI
%xdefine xmm_save_base (rsp + 16*4)
xmm_mov [xmm_save_base + 0*16], xmm6
xmm_mov [xmm_save_base + 1*16], xmm7
xmm_mov [xmm_save_base + 2*16], xmm8
xmm_mov [xmm_save_base + 3*16], xmm9
xmm_mov [xmm_save_base + 4*16], xmm10
%endif
mov HASH_PTR, ctx
mov BUFFER_PTR, buf
%if (%2 == 1)
shl cnt, 6 ;; mul by 64
add cnt, buf
mov BUFFER_END, cnt
%endif
lea K_BASE, [K_XMM_AR]
xmm_mov XMM_SHUFB_BSWAP, [bswap_shufb_ctl]
SHA1_PIPELINED_MAIN_BODY %2
%ifdef WIN_ABI
xmm_mov xmm6, [xmm_save_base + 0*16]
xmm_mov xmm7, [xmm_save_base + 1*16]
xmm_mov xmm8, [xmm_save_base + 2*16]
xmm_mov xmm9, [xmm_save_base + 3*16]
xmm_mov xmm10,[xmm_save_base + 4*16]
%endif
add rsp, stack_size
%ifdef WIN_ABI
pop rsi
pop rdi
%endif
pop rbp
pop rbx
ret
%endmacro
;--------------------------------------------
; macro implements 80 rounds of SHA-1, for one 64-byte block or multiple blocks with s/w pipelining
; macro param: =0 - process single 64-byte block
; =1 - multiple blocks
;
%macro SHA1_PIPELINED_MAIN_BODY 1
REGALLOC
mov A, [HASH_PTR ]
mov B, [HASH_PTR+ 4]
mov C, [HASH_PTR+ 8]
mov D, [HASH_PTR+12]
mov E, [HASH_PTR+16]
%assign i 0
%rep W_PRECALC_AHEAD
W_PRECALC i
%assign i i+1
%endrep
%xdefine F F1
%if (%1 == 1) ;; code loops through more than one block
%%_loop:
cmp BUFFER_PTR, K_BASE ;; we use K_BASE value as a signal of a last block,
jne %%_begin ;; it is set below by: cmovae BUFFER_PTR, K_BASE
jmp %%_end
align 32
%%_begin:
%endif
RR A,B,C,D,E,0
RR D,E,A,B,C,2
RR B,C,D,E,A,4
RR E,A,B,C,D,6
RR C,D,E,A,B,8
RR A,B,C,D,E,10
RR D,E,A,B,C,12
RR B,C,D,E,A,14
RR E,A,B,C,D,16
RR C,D,E,A,B,18
%xdefine F F2
RR A,B,C,D,E,20
RR D,E,A,B,C,22
RR B,C,D,E,A,24
RR E,A,B,C,D,26
RR C,D,E,A,B,28
RR A,B,C,D,E,30
RR D,E,A,B,C,32
RR B,C,D,E,A,34
RR E,A,B,C,D,36
RR C,D,E,A,B,38
%xdefine F F3
RR A,B,C,D,E,40
RR D,E,A,B,C,42
RR B,C,D,E,A,44
RR E,A,B,C,D,46
RR C,D,E,A,B,48
RR A,B,C,D,E,50
RR D,E,A,B,C,52
RR B,C,D,E,A,54
RR E,A,B,C,D,56
RR C,D,E,A,B,58
%xdefine F F4
%if (%1 == 1) ;; if code loops through more than one block
add BUFFER_PTR, 64 ;; move to next 64-byte block
cmp BUFFER_PTR, BUFFER_END ;; check if current block is the last one
cmovae BUFFER_PTR, K_BASE ;; smart way to signal the last iteration
%else
%xdefine W_NO_TAIL_PRECALC 1 ;; no software pipelining for single block interface
%endif
RR A,B,C,D,E,60
RR D,E,A,B,C,62
RR B,C,D,E,A,64
RR E,A,B,C,D,66
RR C,D,E,A,B,68
RR A,B,C,D,E,70
RR D,E,A,B,C,72
RR B,C,D,E,A,74
RR E,A,B,C,D,76
RR C,D,E,A,B,78
UPDATE_HASH [HASH_PTR ],A
UPDATE_HASH [HASH_PTR+ 4],B
UPDATE_HASH [HASH_PTR+ 8],C
UPDATE_HASH [HASH_PTR+12],D
UPDATE_HASH [HASH_PTR+16],E
%if (%1 == 1)
jmp %%_loop
align 32
%%_end:
%endif
%xdefine W_NO_TAIL_PRECALC 0
%xdefine F %error
%endmacro
%macro F1 3
mov T1,%2
xor T1,%3
and T1,%1
xor T1,%3
%endmacro
%macro F2 3
mov T1,%3
xor T1,%2
xor T1,%1
%endmacro
%macro F3 3
mov T1,%2
mov T2,%1
or T1,%1
and T2,%2
and T1,%3
or T1,T2
%endmacro
%define F4 F2
%macro UPDATE_HASH 2
add %2, %1
mov %1, %2
%endmacro
%macro W_PRECALC 1
%xdefine i (%1)
%if (i < 20)
%xdefine K_XMM 0
%elif (i < 40)
%xdefine K_XMM 16
%elif (i < 60)
%xdefine K_XMM 32
%else
%xdefine K_XMM 48
%endif
%if (i<16 || (i>=80 && i<(80 + W_PRECALC_AHEAD)))
%if (W_NO_TAIL_PRECALC == 0)
%xdefine i ((%1) % 80) ;; pre-compute for the next iteration
%if (i == 0)
W_PRECALC_RESET
%endif
W_PRECALC_00_15
%endif
%elif (i < 32)
W_PRECALC_16_31
%elif (i < 80) ;; rounds 32-79
W_PRECALC_32_79
%endif
%endmacro
%macro W_PRECALC_RESET 0
%xdefine W W0
%xdefine W_minus_04 W4
%xdefine W_minus_08 W8
%xdefine W_minus_12 W12
%xdefine W_minus_16 W16
%xdefine W_minus_20 W20
%xdefine W_minus_24 W24
%xdefine W_minus_28 W28
%xdefine W_minus_32 W
%endmacro
%macro W_PRECALC_ROTATE 0
%xdefine W_minus_32 W_minus_28
%xdefine W_minus_28 W_minus_24
%xdefine W_minus_24 W_minus_20
%xdefine W_minus_20 W_minus_16
%xdefine W_minus_16 W_minus_12
%xdefine W_minus_12 W_minus_08
%xdefine W_minus_08 W_minus_04
%xdefine W_minus_04 W
%xdefine W W_minus_32
%endmacro
%xdefine W_PRECALC_AHEAD 16
%xdefine W_NO_TAIL_PRECALC 0
%xdefine xmm_mov movdqa
%macro W_PRECALC_00_15 0
;; message scheduling pre-compute for rounds 0-15
%if ((i & 3) == 0) ;; blended SSE and ALU instruction scheduling, 1 vector iteration per 4 rounds
movdqu W_TMP, [BUFFER_PTR + (i * 4)]
%elif ((i & 3) == 1)
pshufb W_TMP, XMM_SHUFB_BSWAP
movdqa W, W_TMP
%elif ((i & 3) == 2)
paddd W_TMP, [K_BASE]
%elif ((i & 3) == 3)
movdqa [WK(i&~3)], W_TMP
W_PRECALC_ROTATE
%endif
%endmacro
%macro W_PRECALC_16_31 0
;; message scheduling pre-compute for rounds 16-31
;; calculating last 32 w[i] values in 8 XMM registers
;; pre-calculate K+w[i] values and store to mem, for later load by ALU add instruction
;;
;; "brute force" vectorization for rounds 16-31 only due to w[i]->w[i-3] dependency
;;
%if ((i & 3) == 0) ;; blended SSE and ALU instruction scheduling, 1 vector iteration per 4 rounds
movdqa W, W_minus_12
palignr W, W_minus_16, 8 ;; w[i-14]
movdqa W_TMP, W_minus_04
psrldq W_TMP, 4 ;; w[i-3]
pxor W, W_minus_08
%elif ((i & 3) == 1)
pxor W_TMP, W_minus_16
pxor W, W_TMP
movdqa W_TMP2, W
movdqa W_TMP, W
pslldq W_TMP2, 12
%elif ((i & 3) == 2)
psrld W, 31
pslld W_TMP, 1
por W_TMP, W
movdqa W, W_TMP2
psrld W_TMP2, 30
pslld W, 2
%elif ((i & 3) == 3)
pxor W_TMP, W
pxor W_TMP, W_TMP2
movdqa W, W_TMP
paddd W_TMP, [K_BASE + K_XMM]
movdqa [WK(i&~3)],W_TMP
W_PRECALC_ROTATE
%endif
%endmacro
%macro W_PRECALC_32_79 0
;; in SHA-1 specification: w[i] = (w[i-3] ^ w[i-8] ^ w[i-14] ^ w[i-16]) rol 1
;; instead we do equal: w[i] = (w[i-6] ^ w[i-16] ^ w[i-28] ^ w[i-32]) rol 2
;; allows more efficient vectorization since w[i]=>w[i-3] dependency is broken
;;
%if ((i & 3) == 0) ;; blended SSE and ALU instruction scheduling, 1 vector iteration per 4 rounds
movdqa W_TMP, W_minus_04
pxor W, W_minus_28 ;; W is W_minus_32 before xor
palignr W_TMP, W_minus_08, 8
%elif ((i & 3) == 1)
pxor W, W_minus_16
pxor W, W_TMP
movdqa W_TMP, W
%elif ((i & 3) == 2)
psrld W, 30
pslld W_TMP, 2
por W_TMP, W
%elif ((i & 3) == 3)
movdqa W, W_TMP
paddd W_TMP, [K_BASE + K_XMM]
movdqa [WK(i&~3)],W_TMP
W_PRECALC_ROTATE
%endif
%endmacro
%macro RR 6 ;; RR does two rounds of SHA-1 back to back with W pre-calculation
;; TEMP = A
;; A = F( i, B, C, D ) + E + ROTATE_LEFT( A, 5 ) + W[i] + K(i)
;; C = ROTATE_LEFT( B, 30 )
;; D = C
;; E = D
;; B = TEMP
W_PRECALC (%6 + W_PRECALC_AHEAD)
F %2, %3, %4 ;; F returns result in T1
add %5, [WK(%6)]
rol %2, 30
mov T2, %1
add %4, [WK(%6 + 1)]
rol T2, 5
add %5, T1
W_PRECALC (%6 + W_PRECALC_AHEAD + 1)
add T2, %5
mov %5, T2
rol T2, 5
add %4, T2
F %1, %2, %3 ;; F returns result in T1
add %4, T1
rol %1, 30
;; write: %1, %2
;; rotate: %1<=%4, %2<=%5, %3<=%1, %4<=%2, %5<=%3
%endmacro
;;----------------------
section .data align=128
%xdefine K1 0x5a827999
%xdefine K2 0x6ed9eba1
%xdefine K3 0x8f1bbcdc
%xdefine K4 0xca62c1d6
align 128
K_XMM_AR:
DD K1, K1, K1, K1
DD K2, K2, K2, K2
DD K3, K3, K3, K3
DD K4, K4, K4, K4
align 16
bswap_shufb_ctl:
DD 00010203h
DD 04050607h
DD 08090a0bh
DD 0c0d0e0fh
;; dispatch pointer, points to the init routine for the first invocation
sha1_update_intel_dispatched:
DQ sha1_update_intel_init_
;;----------------------
section .text align=4096
SHA1_VECTOR_ASM sha1_update_intel_ssse3_, multiblock
align 32
sha1_update_intel_init_: ;; we get here with the first time invocation
call sha1_update_intel_dispacth_init_
INTEL_SHA1_UPDATE_FUNCNAME: ;; we get here after init
jmp qword [sha1_update_intel_dispatched]
;; CPUID feature flag based dispatch
sha1_update_intel_dispacth_init_:
push rax
push rbx
push rcx
push rdx
push rsi
lea rsi, [INTEL_SHA1_UPDATE_DEFAULT_DISPATCH]
mov eax, 1
cpuid
test ecx, 0200h ;; SSSE3 support, CPUID.1.ECX[bit 9]
jz _done
lea rsi, [sha1_update_intel_ssse3_]
_done:
mov [sha1_update_intel_dispatched], rsi
pop rsi
pop rdx
pop rcx
pop rbx
pop rax
ret
;;----------------------
;; in the case a default SHA-1 update function implementation was not provided
;; and code was invoked on a non-SSSE3 supporting CPU, dispatch handles this
;; failure in a safest way - jumps to the stub function with UD2 instruction below
sha1_intel_non_ssse3_cpu_stub_:
ud2 ;; in the case no default SHA-1 was provided non-SSSE3 CPUs safely fail here
ret
; END
;----------------------
-138
View File
@@ -1,138 +0,0 @@
// Copyright 2009 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 of
// Golang project:
// https://github.com/golang/go/blob/master/LICENSE
// Using this part of Minio codebase under the license
// Apache License Version 2.0 with modifications
// SHA1 hash algorithm. See RFC 3174.
package sha1
import (
"crypto/rand"
"fmt"
"io"
"testing"
)
type sha1Test struct {
out string
in string
}
var golden = []sha1Test{
{"da39a3ee5e6b4b0d3255bfef95601890afd80709", ""},
{"86f7e437faa5a7fce15d1ddcb9eaeaea377667b8", "a"},
{"da23614e02469a0d7c7bd1bdab5c9c474b1904dc", "ab"},
{"a9993e364706816aba3e25717850c26c9cd0d89d", "abc"},
{"81fe8bfe87576c3ecb22426f8e57847382917acf", "abcd"},
{"03de6c570bfe24bfc328ccd7ca46b76eadaf4334", "abcde"},
{"1f8ac10f23c5b5bc1167bda84b833e5c057a77d2", "abcdef"},
{"2fb5e13419fc89246865e7a324f476ec624e8740", "abcdefg"},
{"425af12a0743502b322e93a015bcf868e324d56a", "abcdefgh"},
{"c63b19f1e4c8b5f76b25c49b8b87f57d8e4872a1", "abcdefghi"},
{"d68c19a0a345b7eab78d5e11e991c026ec60db63", "abcdefghij"},
{"ebf81ddcbe5bf13aaabdc4d65354fdf2044f38a7", "Discard medicine more than two years old."},
{"e5dea09392dd886ca63531aaa00571dc07554bb6", "He who has a shady past knows that nice guys finish last."},
{"45988f7234467b94e3e9494434c96ee3609d8f8f", "I wouldn't marry him with a ten foot pole."},
{"55dee037eb7460d5a692d1ce11330b260e40c988", "Free! Free!/A trip/to Mars/for 900/empty jars/Burma Shave"},
{"b7bc5fb91080c7de6b582ea281f8a396d7c0aee8", "The days of the digital watch are numbered. -Tom Stoppard"},
{"c3aed9358f7c77f523afe86135f06b95b3999797", "Nepal premier won't resign."},
{"6e29d302bf6e3a5e4305ff318d983197d6906bb9", "For every action there is an equal and opposite government program."},
{"597f6a540010f94c15d71806a99a2c8710e747bd", "His money is twice tainted: 'taint yours and 'taint mine."},
{"6859733b2590a8a091cecf50086febc5ceef1e80", "There is no reason for any individual to have a computer in their home. -Ken Olsen, 1977"},
{"514b2630ec089b8aee18795fc0cf1f4860cdacad", "It's a tiny change to the code and not completely disgusting. - Bob Manchek"},
{"c5ca0d4a7b6676fc7aa72caa41cc3d5df567ed69", "size: a.out: bad magic"},
{"74c51fa9a04eadc8c1bbeaa7fc442f834b90a00a", "The major problem is with sendmail. -Mark Horton"},
{"0b4c4ce5f52c3ad2821852a8dc00217fa18b8b66", "Give me a rock, paper and scissors and I will move the world. CCFestoon"},
{"3ae7937dd790315beb0f48330e8642237c61550a", "If the enemy is within range, then so are you."},
{"410a2b296df92b9a47412b13281df8f830a9f44b", "It's well we cannot hear the screams/That we create in others' dreams."},
{"841e7c85ca1adcddbdd0187f1289acb5c642f7f5", "You remind me of a TV show, but that's all right: I watch it anyway."},
{"163173b825d03b952601376b25212df66763e1db", "C is as portable as Stonehedge!!"},
{"32b0377f2687eb88e22106f133c586ab314d5279", "Even if I could be Shakespeare, I think I should still choose to be Faraday. - A. Huxley"},
{"0885aaf99b569542fd165fa44e322718f4a984e0", "The fugacity of a constituent in a mixture of gases at a given temperature is proportional to its mole fraction. Lewis-Randall Rule"},
{"6627d6904d71420b0bf3886ab629623538689f45", "How can you write a big system without C++? -Paul Glick"},
}
func TestGolden(t *testing.T) {
for i := 0; i < len(golden); i++ {
g := golden[i]
s := fmt.Sprintf("%x", Sum1([]byte(g.in)))
if s != g.out {
t.Fatalf("Sum function: sha1(%s) = %s want %s", g.in, s, g.out)
}
c := New()
for j := 0; j < 3; j++ {
if j < 2 {
io.WriteString(c, g.in)
} else {
io.WriteString(c, g.in[0:len(g.in)/2])
c.Sum(nil)
io.WriteString(c, g.in[len(g.in)/2:])
}
s := fmt.Sprintf("%x", c.Sum(nil))
if s != g.out {
t.Fatalf("sha1[%d](%s) = %s want %s", j, g.in, s, g.out)
}
c.Reset()
}
}
}
func TestSize(t *testing.T) {
c := New()
if got := c.Size(); got != Size {
t.Errorf("Size = %d; want %d", got, Size)
}
}
func TestBlockSize(t *testing.T) {
c := New()
if got := c.BlockSize(); got != BlockSize {
t.Errorf("BlockSize = %d; want %d", got, BlockSize)
}
}
// Tests that blockGeneric (pure Go) and block (in assembly for amd64, 386, arm) match.
func TestBlockGeneric(t *testing.T) {
gen, asm := New().(*digest), New().(*digest)
buf := make([]byte, BlockSize*20) // arbitrary factor
rand.Read(buf)
blockGeneric(gen, buf)
block(asm, buf)
if *gen != *asm {
t.Error("block and blockGeneric resulted in different states")
}
}
var bench = New()
var buf = make([]byte, 1024*1024)
func benchmarkSize(b *testing.B, size int) {
b.SetBytes(int64(size))
sum := make([]byte, bench.Size())
for i := 0; i < b.N; i++ {
bench.Reset()
bench.Write(buf[:size])
bench.Sum(sum[:0])
}
}
func BenchmarkHash8Bytes(b *testing.B) {
benchmarkSize(b, 8)
}
func BenchmarkHash1K(b *testing.B) {
benchmarkSize(b, 1024)
}
func BenchmarkHash8K(b *testing.B) {
benchmarkSize(b, 8192)
}
func BenchmarkHash1M(b *testing.B) {
benchmarkSize(b, 1024*1024)
}
-167
View File
@@ -1,167 +0,0 @@
// Copyright 2009 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 of
// Golang project:
// https://github.com/golang/go/blob/master/LICENSE
// Using this part of Minio codebase under the license
// Apache License Version 2.0 with modifications
// Package sha1 implements the SHA1 hash algorithm as defined in RFC 3174.
package sha1
import (
"hash"
"io"
"github.com/minio/minio/pkg/cpu"
)
// The size of a SHA1 checksum in bytes.
const Size = 20
// The blocksize of SHA1 in bytes.
const BlockSize = 64
const (
chunk = 64
init0 = 0x67452301
init1 = 0xEFCDAB89
init2 = 0x98BADCFE
init3 = 0x10325476
init4 = 0xC3D2E1F0
)
// digest represents the partial evaluation of a checksum.
type digest struct {
h [5]uint32
x [chunk]byte
nx int
len uint64
}
// Reset digest
func (d *digest) Reset() {
d.h[0] = init0
d.h[1] = init1
d.h[2] = init2
d.h[3] = init3
d.h[4] = init4
d.nx = 0
d.len = 0
}
// New returns a new hash.Hash computing the SHA1 checksum.
func New() hash.Hash {
d := new(digest)
d.Reset()
return d
}
func block(dig *digest, p []byte) {
switch true {
case cpu.HasSSE41() == true:
blockSSE3(dig, p)
default:
blockGeneric(dig, p)
}
}
// Return output size
func (d *digest) Size() int { return Size }
// Return checksum blocksize
func (d *digest) BlockSize() int { return BlockSize }
// Write to digest
func (d *digest) Write(p []byte) (nn int, err error) {
nn = len(p)
d.len += uint64(nn)
if d.nx > 0 {
n := copy(d.x[d.nx:], p)
d.nx += n
if d.nx == chunk {
block(d, d.x[:])
d.nx = 0
}
p = p[n:]
}
if len(p) >= chunk {
n := len(p) &^ (chunk - 1)
block(d, p[:n])
p = p[n:]
}
if len(p) > 0 {
d.nx = copy(d.x[:], p)
}
return
}
// Return checksum bytes
func (d *digest) Sum(in []byte) []byte {
// Make a copy of d0 so that caller can keep writing and summing.
d0 := *d
hash := d0.checkSum()
return append(in, hash[:]...)
}
// Intermediate checksum function
func (d *digest) checkSum() [Size]byte {
len := d.len
// Padding. Add a 1 bit and 0 bits until 56 bytes mod 64.
var tmp [64]byte
tmp[0] = 0x80
if len%64 < 56 {
d.Write(tmp[0 : 56-len%64])
} else {
d.Write(tmp[0 : 64+56-len%64])
}
// Length in bits.
len <<= 3
for i := uint(0); i < 8; i++ {
tmp[i] = byte(len >> (56 - 8*i))
}
d.Write(tmp[0:8])
if d.nx != 0 {
panic("d.nx != 0")
}
var digest [Size]byte
for i, s := range d.h {
digest[i*4] = byte(s >> 24)
digest[i*4+1] = byte(s >> 16)
digest[i*4+2] = byte(s >> 8)
digest[i*4+3] = byte(s)
}
return digest
}
/// Convenience functions
// Sum1 - single caller sha1 helper
func Sum1(data []byte) [Size]byte {
var d digest
d.Reset()
d.Write(data)
return d.checkSum()
}
// Sum - io.Reader based streaming sha1 helper
func Sum(reader io.Reader) ([]byte, error) {
h := New()
var err error
for err == nil {
length := 0
byteBuffer := make([]byte, 1024*1024)
length, err = reader.Read(byteBuffer)
byteBuffer = byteBuffer[0:length]
h.Write(byteBuffer)
}
if err != io.EOF {
return nil, err
}
return h.Sum(nil), nil
}
-21
View File
@@ -1,21 +0,0 @@
// !build amd64
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package sha1
//go:generate yasm -f macho64 -DINTEL_SHA1_UPDATE_FUNCNAME=_sha1_update_intel sha1_sse3_amd64.asm -o sha1_sse3_amd64.syso
-21
View File
@@ -1,21 +0,0 @@
// !build amd64
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package sha1
//go:generate yasm -f elf64 sha1_sse3_amd64.asm -o sha1_sse3_amd64.syso
-21
View File
@@ -1,21 +0,0 @@
// !build amd64
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package sha1
//go:generate yasm -f win64 -DWIN_ABI=1 sha1_sse3_amd64.asm -o sha1_sse3_amd64.syso
-29
View File
@@ -1,29 +0,0 @@
// +build amd64
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package sha1
// #include <stdint.h>
// #include <stdlib.h>
// void sha1_update_intel(int32_t *hash, const char* input, size_t num_blocks);
import "C"
import "unsafe"
func blockSSE3(dig *digest, p []byte) {
C.sha1_update_intel((*C.int32_t)(unsafe.Pointer(&dig.h[0])), (*C.char)(unsafe.Pointer(&p[0])), (C.size_t)(len(p)/chunk))
}
-96
View File
@@ -1,96 +0,0 @@
// +build amd64
// Copyright 2009 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 of
// Golang project:
// https://github.com/golang/go/blob/master/LICENSE
// Using this part of Minio codebase under the license
// Apache License Version 2.0 with modifications
package sha1
const (
_K0 = 0x5A827999
_K1 = 0x6ED9EBA1
_K2 = 0x8F1BBCDC
_K3 = 0xCA62C1D6
)
// blockGeneric is a portable, pure Go version of the SHA1 block step.
// It's used by sha1block_generic.go and tests.
func blockGeneric(dig *digest, p []byte) {
var w [16]uint32
h0, h1, h2, h3, h4 := dig.h[0], dig.h[1], dig.h[2], dig.h[3], dig.h[4]
for len(p) >= chunk {
// Can interlace the computation of w with the
// rounds below if needed for speed.
for i := 0; i < 16; i++ {
j := i * 4
w[i] = uint32(p[j])<<24 | uint32(p[j+1])<<16 | uint32(p[j+2])<<8 | uint32(p[j+3])
}
a, b, c, d, e := h0, h1, h2, h3, h4
// Each of the four 20-iteration rounds
// differs only in the computation of f and
// the choice of K (_K0, _K1, etc).
i := 0
for ; i < 16; i++ {
f := b&c | (^b)&d
a5 := a<<5 | a>>(32-5)
b30 := b<<30 | b>>(32-30)
t := a5 + f + e + w[i&0xf] + _K0
a, b, c, d, e = t, a, b30, c, d
}
for ; i < 20; i++ {
tmp := w[(i-3)&0xf] ^ w[(i-8)&0xf] ^ w[(i-14)&0xf] ^ w[(i)&0xf]
w[i&0xf] = tmp<<1 | tmp>>(32-1)
f := b&c | (^b)&d
a5 := a<<5 | a>>(32-5)
b30 := b<<30 | b>>(32-30)
t := a5 + f + e + w[i&0xf] + _K0
a, b, c, d, e = t, a, b30, c, d
}
for ; i < 40; i++ {
tmp := w[(i-3)&0xf] ^ w[(i-8)&0xf] ^ w[(i-14)&0xf] ^ w[(i)&0xf]
w[i&0xf] = tmp<<1 | tmp>>(32-1)
f := b ^ c ^ d
a5 := a<<5 | a>>(32-5)
b30 := b<<30 | b>>(32-30)
t := a5 + f + e + w[i&0xf] + _K1
a, b, c, d, e = t, a, b30, c, d
}
for ; i < 60; i++ {
tmp := w[(i-3)&0xf] ^ w[(i-8)&0xf] ^ w[(i-14)&0xf] ^ w[(i)&0xf]
w[i&0xf] = tmp<<1 | tmp>>(32-1)
f := ((b | c) & d) | (b & c)
a5 := a<<5 | a>>(32-5)
b30 := b<<30 | b>>(32-30)
t := a5 + f + e + w[i&0xf] + _K2
a, b, c, d, e = t, a, b30, c, d
}
for ; i < 80; i++ {
tmp := w[(i-3)&0xf] ^ w[(i-8)&0xf] ^ w[(i-14)&0xf] ^ w[(i)&0xf]
w[i&0xf] = tmp<<1 | tmp>>(32-1)
f := b ^ c ^ d
a5 := a<<5 | a>>(32-5)
b30 := b<<30 | b>>(32-30)
t := a5 + f + e + w[i&0xf] + _K3
a, b, c, d, e = t, a, b30, c, d
}
h0 += a
h1 += b
h2 += c
h3 += d
h4 += e
p = p[chunk:]
}
dig.h[0], dig.h[1], dig.h[2], dig.h[3], dig.h[4] = h0, h1, h2, h3, h4
}
-35
View File
@@ -1,35 +0,0 @@
// +build amd64
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package sha1
// #cgo CFLAGS: -DHAS_AVX2
// #include <stdint.h>
// #include <stdlib.h>
// void sha1_transform(int32_t *hash, const char* input, size_t num_blocks);
// void sha1_update_intel(int32_t *hash, const char* input, size_t num_blocks);
import "C"
import "unsafe"
func blockAVX2(dig *digest, p []byte) {
C.sha1_transform((*C.int32_t)(unsafe.Pointer(&dig.h[0])), (*C.char)(unsafe.Pointer(&p[0])), (C.size_t)(len(p)/chunk))
}
func blockSSE3(dig *digest, p []byte) {
C.sha1_update_intel((*C.int32_t)(unsafe.Pointer(&dig.h[0])), (*C.char)(unsafe.Pointer(&p[0])), (C.size_t)(len(p)/chunk))
}
-29
View File
@@ -1,29 +0,0 @@
// +build amd64
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package sha1
// #include <stdint.h>
// #include <stdlib.h>
// void sha1_update_intel(int32_t *hash, const char* input, size_t num_blocks);
import "C"
import "unsafe"
func blockSSE3(dig *digest, p []byte) {
C.sha1_update_intel((*C.int32_t)(unsafe.Pointer(&dig.h[0])), (*C.char)(unsafe.Pointer(&p[0])), (C.size_t)(len(p)/chunk))
}
-202
View File
@@ -1,202 +0,0 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-759
View File
@@ -1,759 +0,0 @@
########################################################################
# Implement fast SHA-256 with AVX1 instructions. (x86_64)
#
# Copyright (C) 2013 Intel Corporation.
#
# Authors:
# James Guilford <james.guilford@intel.com>
# Kirk Yap <kirk.s.yap@intel.com>
# Tim Chen <tim.c.chen@linux.intel.com>
#
# This software is available to you under a choice of one of two
# licenses. You may choose to be licensed under the terms of the GNU
# General Public License (GPL) Version 2, available from the file
# COPYING in the main directory of this source tree, or the
# OpenIB.org BSD license below:
#
# 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.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
# EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
# NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
# BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
# ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
########################################################################
#
# This code is described in an Intel White-Paper:
# "Fast SHA-256 Implementations on Intel Architecture Processors"
#
# To find it, surf to http://www.intel.com/p/en_US/embedded
# and search for that title.
#
########################################################################
# This code schedules 1 block at a time, with 4 lanes per block
########################################################################
# Using this part of Minio codebase under the license
# Apache License Version 2.0 with modifications
##
#ifdef HAS_AVX
#ifndef ENTRY
#define ENTRY(name) \
.globl name ; \
.align 4,0x90 ; \
name:
#endif
#ifndef END
#define END(name) \
.size name, .-name
#endif
#ifndef ENDPROC
#define ENDPROC(name) \
.type name, @function ; \
END(name)
#endif
#define NUM_INVALID 100
#define TYPE_R32 0
#define TYPE_R64 1
#define TYPE_XMM 2
#define TYPE_INVALID 100
.macro R32_NUM opd r32
\opd = NUM_INVALID
.ifc \r32,%eax
\opd = 0
.endif
.ifc \r32,%ecx
\opd = 1
.endif
.ifc \r32,%edx
\opd = 2
.endif
.ifc \r32,%ebx
\opd = 3
.endif
.ifc \r32,%esp
\opd = 4
.endif
.ifc \r32,%ebp
\opd = 5
.endif
.ifc \r32,%esi
\opd = 6
.endif
.ifc \r32,%edi
\opd = 7
.endif
#ifdef X86_64
.ifc \r32,%r8d
\opd = 8
.endif
.ifc \r32,%r9d
\opd = 9
.endif
.ifc \r32,%r10d
\opd = 10
.endif
.ifc \r32,%r11d
\opd = 11
.endif
.ifc \r32,%r12d
\opd = 12
.endif
.ifc \r32,%r13d
\opd = 13
.endif
.ifc \r32,%r14d
\opd = 14
.endif
.ifc \r32,%r15d
\opd = 15
.endif
#endif
.endm
.macro R64_NUM opd r64
\opd = NUM_INVALID
#ifdef X86_64
.ifc \r64,%rax
\opd = 0
.endif
.ifc \r64,%rcx
\opd = 1
.endif
.ifc \r64,%rdx
\opd = 2
.endif
.ifc \r64,%rbx
\opd = 3
.endif
.ifc \r64,%rsp
\opd = 4
.endif
.ifc \r64,%rbp
\opd = 5
.endif
.ifc \r64,%rsi
\opd = 6
.endif
.ifc \r64,%rdi
\opd = 7
.endif
.ifc \r64,%r8
\opd = 8
.endif
.ifc \r64,%r9
\opd = 9
.endif
.ifc \r64,%r10
\opd = 10
.endif
.ifc \r64,%r11
\opd = 11
.endif
.ifc \r64,%r12
\opd = 12
.endif
.ifc \r64,%r13
\opd = 13
.endif
.ifc \r64,%r14
\opd = 14
.endif
.ifc \r64,%r15
\opd = 15
.endif
#endif
.endm
.macro XMM_NUM opd xmm
\opd = NUM_INVALID
.ifc \xmm,%xmm0
\opd = 0
.endif
.ifc \xmm,%xmm1
\opd = 1
.endif
.ifc \xmm,%xmm2
\opd = 2
.endif
.ifc \xmm,%xmm3
\opd = 3
.endif
.ifc \xmm,%xmm4
\opd = 4
.endif
.ifc \xmm,%xmm5
\opd = 5
.endif
.ifc \xmm,%xmm6
\opd = 6
.endif
.ifc \xmm,%xmm7
\opd = 7
.endif
.ifc \xmm,%xmm8
\opd = 8
.endif
.ifc \xmm,%xmm9
\opd = 9
.endif
.ifc \xmm,%xmm10
\opd = 10
.endif
.ifc \xmm,%xmm11
\opd = 11
.endif
.ifc \xmm,%xmm12
\opd = 12
.endif
.ifc \xmm,%xmm13
\opd = 13
.endif
.ifc \xmm,%xmm14
\opd = 14
.endif
.ifc \xmm,%xmm15
\opd = 15
.endif
.endm
.macro TYPE type reg
R32_NUM reg_type_r32 \reg
R64_NUM reg_type_r64 \reg
XMM_NUM reg_type_xmm \reg
.if reg_type_r64 <> NUM_INVALID
\type = TYPE_R64
.elseif reg_type_r32 <> NUM_INVALID
\type = TYPE_R32
.elseif reg_type_xmm <> NUM_INVALID
\type = TYPE_XMM
.else
\type = TYPE_INVALID
.endif
.endm
.macro PFX_OPD_SIZE
.byte 0x66
.endm
.macro PFX_REX opd1 opd2 W=0
.if ((\opd1 | \opd2) & 8) || \W
.byte 0x40 | ((\opd1 & 8) >> 3) | ((\opd2 & 8) >> 1) | (\W << 3)
.endif
.endm
.macro MODRM mod opd1 opd2
.byte \mod | (\opd1 & 7) | ((\opd2 & 7) << 3)
.endm
.macro PSHUFB_XMM xmm1 xmm2
XMM_NUM pshufb_opd1 \xmm1
XMM_NUM pshufb_opd2 \xmm2
PFX_OPD_SIZE
PFX_REX pshufb_opd1 pshufb_opd2
.byte 0x0f, 0x38, 0x00
MODRM 0xc0 pshufb_opd1 pshufb_opd2
.endm
.macro PCLMULQDQ imm8 xmm1 xmm2
XMM_NUM clmul_opd1 \xmm1
XMM_NUM clmul_opd2 \xmm2
PFX_OPD_SIZE
PFX_REX clmul_opd1 clmul_opd2
.byte 0x0f, 0x3a, 0x44
MODRM 0xc0 clmul_opd1 clmul_opd2
.byte \imm8
.endm
.macro PEXTRD imm8 xmm gpr
R32_NUM extrd_opd1 \gpr
XMM_NUM extrd_opd2 \xmm
PFX_OPD_SIZE
PFX_REX extrd_opd1 extrd_opd2
.byte 0x0f, 0x3a, 0x16
MODRM 0xc0 extrd_opd1 extrd_opd2
.byte \imm8
.endm
.macro MOVQ_R64_XMM opd1 opd2
TYPE movq_r64_xmm_opd1_type \opd1
.if movq_r64_xmm_opd1_type == TYPE_XMM
XMM_NUM movq_r64_xmm_opd1 \opd1
R64_NUM movq_r64_xmm_opd2 \opd2
.else
R64_NUM movq_r64_xmm_opd1 \opd1
XMM_NUM movq_r64_xmm_opd2 \opd2
.endif
PFX_OPD_SIZE
PFX_REX movq_r64_xmm_opd1 movq_r64_xmm_opd2 1
.if movq_r64_xmm_opd1_type == TYPE_XMM
.byte 0x0f, 0x7e
.else
.byte 0x0f, 0x6e
.endif
MODRM 0xc0 movq_r64_xmm_opd1 movq_r64_xmm_opd2
.endm
## assume buffers not aligned
#define VMOVDQ vmovdqu
################################ Define Macros
# addm [mem], reg
# Add reg to mem using reg-mem add and store
.macro addm p1 p2
add \p1, \p2
mov \p2, \p1
.endm
.macro MY_ROR p1 p2
shld $(32-(\p1)), \p2, \p2
.endm
################################
# COPY_XMM_AND_BSWAP xmm, [mem], byte_flip_mask
# Load xmm with mem and byte swap each dword
.macro COPY_XMM_AND_BSWAP p1 p2 p3
VMOVDQ \p2, \p1
vpshufb \p3, \p1, \p1
.endm
################################
X0 = %xmm4
X1 = %xmm5
X2 = %xmm6
X3 = %xmm7
XTMP0 = %xmm0
XTMP1 = %xmm1
XTMP2 = %xmm2
XTMP3 = %xmm3
XTMP4 = %xmm8
XFER = %xmm9
XTMP5 = %xmm11
SHUF_00BA = %xmm10 # shuffle xBxA -> 00BA
SHUF_DC00 = %xmm12 # shuffle xDxC -> DC00
BYTE_FLIP_MASK = %xmm13
NUM_BLKS = %rdx # 3rd arg
CTX = %rsi # 2nd arg
INP = %rdi # 1st arg
SRND = %rdi # clobbers INP
c = %ecx
d = %r8d
e = %edx
TBL = %rbp
a = %eax
b = %ebx
f = %r9d
g = %r10d
h = %r11d
y0 = %r13d
y1 = %r14d
y2 = %r15d
_INP_END_SIZE = 8
_INP_SIZE = 8
_XFER_SIZE = 16
_XMM_SAVE_SIZE = 0
_INP_END = 0
_INP = _INP_END + _INP_END_SIZE
_XFER = _INP + _INP_SIZE
_XMM_SAVE = _XFER + _XFER_SIZE
STACK_SIZE = _XMM_SAVE + _XMM_SAVE_SIZE
# rotate_Xs
# Rotate values of symbols X0...X3
.macro rotate_Xs
X_ = X0
X0 = X1
X1 = X2
X2 = X3
X3 = X_
.endm
# ROTATE_ARGS
# Rotate values of symbols a...h
.macro ROTATE_ARGS
TMP_ = h
h = g
g = f
f = e
e = d
d = c
c = b
b = a
a = TMP_
.endm
.macro FOUR_ROUNDS_AND_SCHED
## compute s0 four at a time and s1 two at a time
## compute W[-16] + W[-7] 4 at a time
mov e, y0 # y0 = e
MY_ROR (25-11), y0 # y0 = e >> (25-11)
mov a, y1 # y1 = a
vpalignr $4, X2, X3, XTMP0 # XTMP0 = W[-7]
MY_ROR (22-13), y1 # y1 = a >> (22-13)
xor e, y0 # y0 = e ^ (e >> (25-11))
mov f, y2 # y2 = f
MY_ROR (11-6), y0 # y0 = (e >> (11-6)) ^ (e >> (25-6))
xor a, y1 # y1 = a ^ (a >> (22-13)
xor g, y2 # y2 = f^g
vpaddd X0, XTMP0, XTMP0 # XTMP0 = W[-7] + W[-16]
xor e, y0 # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
and e, y2 # y2 = (f^g)&e
MY_ROR (13-2), y1 # y1 = (a >> (13-2)) ^ (a >> (22-2))
## compute s0
vpalignr $4, X0, X1, XTMP1 # XTMP1 = W[-15]
xor a, y1 # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
MY_ROR 6, y0 # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
xor g, y2 # y2 = CH = ((f^g)&e)^g
MY_ROR 2, y1 # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
add y0, y2 # y2 = S1 + CH
add _XFER(%rsp), y2 # y2 = k + w + S1 + CH
mov a, y0 # y0 = a
add y2, h # h = h + S1 + CH + k + w
mov a, y2 # y2 = a
vpsrld $7, XTMP1, XTMP2
or c, y0 # y0 = a|c
add h, d # d = d + h + S1 + CH + k + w
and c, y2 # y2 = a&c
vpslld $(32-7), XTMP1, XTMP3
and b, y0 # y0 = (a|c)&b
add y1, h # h = h + S1 + CH + k + w + S0
vpor XTMP2, XTMP3, XTMP3 # XTMP1 = W[-15] MY_ROR 7
or y2, y0 # y0 = MAJ = (a|c)&b)|(a&c)
add y0, h # h = h + S1 + CH + k + w + S0 + MAJ
ROTATE_ARGS
mov e, y0 # y0 = e
mov a, y1 # y1 = a
MY_ROR (25-11), y0 # y0 = e >> (25-11)
xor e, y0 # y0 = e ^ (e >> (25-11))
mov f, y2 # y2 = f
MY_ROR (22-13), y1 # y1 = a >> (22-13)
vpsrld $18, XTMP1, XTMP2 #
xor a, y1 # y1 = a ^ (a >> (22-13)
MY_ROR (11-6), y0 # y0 = (e >> (11-6)) ^ (e >> (25-6))
xor g, y2 # y2 = f^g
vpsrld $3, XTMP1, XTMP4 # XTMP4 = W[-15] >> 3
MY_ROR (13-2), y1 # y1 = (a >> (13-2)) ^ (a >> (22-2))
xor e, y0 # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
and e, y2 # y2 = (f^g)&e
MY_ROR 6, y0 # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
vpslld $(32-18), XTMP1, XTMP1
xor a, y1 # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
xor g, y2 # y2 = CH = ((f^g)&e)^g
vpxor XTMP1, XTMP3, XTMP3 #
add y0, y2 # y2 = S1 + CH
add (1*4 + _XFER)(%rsp), y2 # y2 = k + w + S1 + CH
MY_ROR 2, y1 # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
vpxor XTMP2, XTMP3, XTMP3 # XTMP1 = W[-15] MY_ROR 7 ^ W[-15] MY_ROR
mov a, y0 # y0 = a
add y2, h # h = h + S1 + CH + k + w
mov a, y2 # y2 = a
vpxor XTMP4, XTMP3, XTMP1 # XTMP1 = s0
or c, y0 # y0 = a|c
add h, d # d = d + h + S1 + CH + k + w
and c, y2 # y2 = a&c
## compute low s1
vpshufd $0b11111010, X3, XTMP2 # XTMP2 = W[-2] {BBAA}
and b, y0 # y0 = (a|c)&b
add y1, h # h = h + S1 + CH + k + w + S0
vpaddd XTMP1, XTMP0, XTMP0 # XTMP0 = W[-16] + W[-7] + s0
or y2, y0 # y0 = MAJ = (a|c)&b)|(a&c)
add y0, h # h = h + S1 + CH + k + w + S0 + MAJ
ROTATE_ARGS
mov e, y0 # y0 = e
mov a, y1 # y1 = a
MY_ROR (25-11), y0 # y0 = e >> (25-11)
xor e, y0 # y0 = e ^ (e >> (25-11))
MY_ROR (22-13), y1 # y1 = a >> (22-13)
mov f, y2 # y2 = f
xor a, y1 # y1 = a ^ (a >> (22-13)
MY_ROR (11-6), y0 # y0 = (e >> (11-6)) ^ (e >> (25-6))
vpsrld $10, XTMP2, XTMP4 # XTMP4 = W[-2] >> 10 {BBAA}
xor g, y2 # y2 = f^g
vpsrlq $19, XTMP2, XTMP3 # XTMP3 = W[-2] MY_ROR 19 {xBxA}
xor e, y0 # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
and e, y2 # y2 = (f^g)&e
vpsrlq $17, XTMP2, XTMP2 # XTMP2 = W[-2] MY_ROR 17 {xBxA}
MY_ROR (13-2), y1 # y1 = (a >> (13-2)) ^ (a >> (22-2))
xor a, y1 # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
xor g, y2 # y2 = CH = ((f^g)&e)^g
MY_ROR 6, y0 # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
vpxor XTMP3, XTMP2, XTMP2 #
add y0, y2 # y2 = S1 + CH
MY_ROR 2, y1 # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
add (2*4 + _XFER)(%rsp), y2 # y2 = k + w + S1 + CH
vpxor XTMP2, XTMP4, XTMP4 # XTMP4 = s1 {xBxA}
mov a, y0 # y0 = a
add y2, h # h = h + S1 + CH + k + w
mov a, y2 # y2 = a
vpshufb SHUF_00BA, XTMP4, XTMP4 # XTMP4 = s1 {00BA}
or c, y0 # y0 = a|c
add h, d # d = d + h + S1 + CH + k + w
and c, y2 # y2 = a&c
vpaddd XTMP4, XTMP0, XTMP0 # XTMP0 = {..., ..., W[1], W[0]}
and b, y0 # y0 = (a|c)&b
add y1, h # h = h + S1 + CH + k + w + S0
## compute high s1
vpshufd $0b01010000, XTMP0, XTMP2 # XTMP2 = W[-2] {DDCC}
or y2, y0 # y0 = MAJ = (a|c)&b)|(a&c)
add y0, h # h = h + S1 + CH + k + w + S0 + MAJ
ROTATE_ARGS
mov e, y0 # y0 = e
MY_ROR (25-11), y0 # y0 = e >> (25-11)
mov a, y1 # y1 = a
MY_ROR (22-13), y1 # y1 = a >> (22-13)
xor e, y0 # y0 = e ^ (e >> (25-11))
mov f, y2 # y2 = f
MY_ROR (11-6), y0 # y0 = (e >> (11-6)) ^ (e >> (25-6))
vpsrld $10, XTMP2, XTMP5 # XTMP5 = W[-2] >> 10 {DDCC}
xor a, y1 # y1 = a ^ (a >> (22-13)
xor g, y2 # y2 = f^g
vpsrlq $19, XTMP2, XTMP3 # XTMP3 = W[-2] MY_ROR 19 {xDxC}
xor e, y0 # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
and e, y2 # y2 = (f^g)&e
MY_ROR (13-2), y1 # y1 = (a >> (13-2)) ^ (a >> (22-2))
vpsrlq $17, XTMP2, XTMP2 # XTMP2 = W[-2] MY_ROR 17 {xDxC}
xor a, y1 # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
MY_ROR 6, y0 # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
xor g, y2 # y2 = CH = ((f^g)&e)^g
vpxor XTMP3, XTMP2, XTMP2
MY_ROR 2, y1 # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
add y0, y2 # y2 = S1 + CH
add (3*4 + _XFER)(%rsp), y2 # y2 = k + w + S1 + CH
vpxor XTMP2, XTMP5, XTMP5 # XTMP5 = s1 {xDxC}
mov a, y0 # y0 = a
add y2, h # h = h + S1 + CH + k + w
mov a, y2 # y2 = a
vpshufb SHUF_DC00, XTMP5, XTMP5 # XTMP5 = s1 {DC00}
or c, y0 # y0 = a|c
add h, d # d = d + h + S1 + CH + k + w
and c, y2 # y2 = a&c
vpaddd XTMP0, XTMP5, X0 # X0 = {W[3], W[2], W[1], W[0]}
and b, y0 # y0 = (a|c)&b
add y1, h # h = h + S1 + CH + k + w + S0
or y2, y0 # y0 = MAJ = (a|c)&b)|(a&c)
add y0, h # h = h + S1 + CH + k + w + S0 + MAJ
ROTATE_ARGS
rotate_Xs
.endm
## input is [rsp + _XFER + %1 * 4]
.macro DO_ROUND round
mov e, y0 # y0 = e
MY_ROR (25-11), y0 # y0 = e >> (25-11)
mov a, y1 # y1 = a
xor e, y0 # y0 = e ^ (e >> (25-11))
MY_ROR (22-13), y1 # y1 = a >> (22-13)
mov f, y2 # y2 = f
xor a, y1 # y1 = a ^ (a >> (22-13)
MY_ROR (11-6), y0 # y0 = (e >> (11-6)) ^ (e >> (25-6))
xor g, y2 # y2 = f^g
xor e, y0 # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
MY_ROR (13-2), y1 # y1 = (a >> (13-2)) ^ (a >> (22-2))
and e, y2 # y2 = (f^g)&e
xor a, y1 # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
MY_ROR 6, y0 # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
xor g, y2 # y2 = CH = ((f^g)&e)^g
add y0, y2 # y2 = S1 + CH
MY_ROR 2, y1 # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
offset = \round * 4 + _XFER #
add offset(%rsp), y2 # y2 = k + w + S1 + CH
mov a, y0 # y0 = a
add y2, h # h = h + S1 + CH + k + w
mov a, y2 # y2 = a
or c, y0 # y0 = a|c
add h, d # d = d + h + S1 + CH + k + w
and c, y2 # y2 = a&c
and b, y0 # y0 = (a|c)&b
add y1, h # h = h + S1 + CH + k + w + S0
or y2, y0 # y0 = MAJ = (a|c)&b)|(a&c)
add y0, h # h = h + S1 + CH + k + w + S0 + MAJ
ROTATE_ARGS
.endm
########################################################################
## void sha256_transform_avx(void *input_data, UINT32 digest[8], UINT64 num_blks)
## arg 1 : pointer to input data
## arg 2 : pointer to digest
## arg 3 : Num blocks
########################################################################
.text
ENTRY(sha256_transform_avx)
.align 32
pushq %rbx
pushq %rbp
pushq %r13
pushq %r14
pushq %r15
pushq %r12
mov %rsp, %r12
subq $STACK_SIZE, %rsp # allocate stack space
and $~15, %rsp # align stack pointer
shl $6, NUM_BLKS # convert to bytes
jz done_hash
add INP, NUM_BLKS # pointer to end of data
mov NUM_BLKS, _INP_END(%rsp)
## load initial digest
mov 4*0(CTX), a
mov 4*1(CTX), b
mov 4*2(CTX), c
mov 4*3(CTX), d
mov 4*4(CTX), e
mov 4*5(CTX), f
mov 4*6(CTX), g
mov 4*7(CTX), h
vmovdqa PSHUFFLE_BYTE_FLIP_MASK(%rip), BYTE_FLIP_MASK
vmovdqa _SHUF_00BA(%rip), SHUF_00BA
vmovdqa _SHUF_DC00(%rip), SHUF_DC00
loop0:
lea K256(%rip), TBL
## byte swap first 16 dwords
COPY_XMM_AND_BSWAP X0, 0*16(INP), BYTE_FLIP_MASK
COPY_XMM_AND_BSWAP X1, 1*16(INP), BYTE_FLIP_MASK
COPY_XMM_AND_BSWAP X2, 2*16(INP), BYTE_FLIP_MASK
COPY_XMM_AND_BSWAP X3, 3*16(INP), BYTE_FLIP_MASK
mov INP, _INP(%rsp)
## schedule 48 input dwords, by doing 3 rounds of 16 each
mov $3, SRND
.align 16
loop1:
vpaddd (TBL), X0, XFER
vmovdqa XFER, _XFER(%rsp)
FOUR_ROUNDS_AND_SCHED
vpaddd 1*16(TBL), X0, XFER
vmovdqa XFER, _XFER(%rsp)
FOUR_ROUNDS_AND_SCHED
vpaddd 2*16(TBL), X0, XFER
vmovdqa XFER, _XFER(%rsp)
FOUR_ROUNDS_AND_SCHED
vpaddd 3*16(TBL), X0, XFER
vmovdqa XFER, _XFER(%rsp)
add $4*16, TBL
FOUR_ROUNDS_AND_SCHED
sub $1, SRND
jne loop1
mov $2, SRND
loop2:
vpaddd (TBL), X0, XFER
vmovdqa XFER, _XFER(%rsp)
DO_ROUND 0
DO_ROUND 1
DO_ROUND 2
DO_ROUND 3
vpaddd 1*16(TBL), X1, XFER
vmovdqa XFER, _XFER(%rsp)
add $2*16, TBL
DO_ROUND 0
DO_ROUND 1
DO_ROUND 2
DO_ROUND 3
vmovdqa X2, X0
vmovdqa X3, X1
sub $1, SRND
jne loop2
addm (4*0)(CTX),a
addm (4*1)(CTX),b
addm (4*2)(CTX),c
addm (4*3)(CTX),d
addm (4*4)(CTX),e
addm (4*5)(CTX),f
addm (4*6)(CTX),g
addm (4*7)(CTX),h
mov _INP(%rsp), INP
add $64, INP
cmp _INP_END(%rsp), INP
jne loop0
done_hash:
mov %r12, %rsp
popq %r12
popq %r15
popq %r14
popq %r13
popq %rbp
popq %rbx
ret
ENDPROC(sha256_transform_avx)
.data
.align 64
K256:
.long 0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5
.long 0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5
.long 0xd807aa98,0x12835b01,0x243185be,0x550c7dc3
.long 0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174
.long 0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc
.long 0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da
.long 0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7
.long 0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967
.long 0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13
.long 0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85
.long 0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3
.long 0xd192e819,0xd6990624,0xf40e3585,0x106aa070
.long 0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5
.long 0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3
.long 0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208
.long 0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2
PSHUFFLE_BYTE_FLIP_MASK:
.octa 0x0c0d0e0f08090a0b0405060700010203
# shuffle xBxA -> 00BA
_SHUF_00BA:
.octa 0xFFFFFFFFFFFFFFFF0b0a090803020100
# shuffle xDxC -> DC00
_SHUF_DC00:
.octa 0x0b0a090803020100FFFFFFFFFFFFFFFF
#endif
File diff suppressed because it is too large Load Diff
-772
View File
@@ -1,772 +0,0 @@
########################################################################
# Implement fast SHA-256 with SSSE3 instructions. (x86_64)
#
# Copyright (C) 2013 Intel Corporation.
#
# Authors:
# James Guilford <james.guilford@intel.com>
# Kirk Yap <kirk.s.yap@intel.com>
# Tim Chen <tim.c.chen@linux.intel.com>
#
# This software is available to you under a choice of one of two
# licenses. You may choose to be licensed under the terms of the GNU
# General Public License (GPL) Version 2, available from the file
# COPYING in the main directory of this source tree, or the
# OpenIB.org BSD license below:
#
# 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.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
# EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
# NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
# BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
# ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
#
########################################################################
#
# This code is described in an Intel White-Paper:
# "Fast SHA-256 Implementations on Intel Architecture Processors"
#
# To find it, surf to http://www.intel.com/p/en_US/embedded
# and search for that title.
#
########################################################################
#
# Using this part of Minio codebase under the license
# Apache License Version 2.0 with modifications
##
#ifdef HAS_SSE41
#ifndef ENTRY
#define ENTRY(name) \
.globl name ; \
.align 4,0x90 ; \
name:
#endif
#ifndef END
#define END(name) \
.size name, .-name
#endif
#ifndef ENDPROC
#define ENDPROC(name) \
.type name, @function ; \
END(name)
#endif
#define NUM_INVALID 100
#define TYPE_R32 0
#define TYPE_R64 1
#define TYPE_XMM 2
#define TYPE_INVALID 100
.macro R32_NUM opd r32
\opd = NUM_INVALID
.ifc \r32,%eax
\opd = 0
.endif
.ifc \r32,%ecx
\opd = 1
.endif
.ifc \r32,%edx
\opd = 2
.endif
.ifc \r32,%ebx
\opd = 3
.endif
.ifc \r32,%esp
\opd = 4
.endif
.ifc \r32,%ebp
\opd = 5
.endif
.ifc \r32,%esi
\opd = 6
.endif
.ifc \r32,%edi
\opd = 7
.endif
#ifdef X86_64
.ifc \r32,%r8d
\opd = 8
.endif
.ifc \r32,%r9d
\opd = 9
.endif
.ifc \r32,%r10d
\opd = 10
.endif
.ifc \r32,%r11d
\opd = 11
.endif
.ifc \r32,%r12d
\opd = 12
.endif
.ifc \r32,%r13d
\opd = 13
.endif
.ifc \r32,%r14d
\opd = 14
.endif
.ifc \r32,%r15d
\opd = 15
.endif
#endif
.endm
.macro R64_NUM opd r64
\opd = NUM_INVALID
#ifdef X86_64
.ifc \r64,%rax
\opd = 0
.endif
.ifc \r64,%rcx
\opd = 1
.endif
.ifc \r64,%rdx
\opd = 2
.endif
.ifc \r64,%rbx
\opd = 3
.endif
.ifc \r64,%rsp
\opd = 4
.endif
.ifc \r64,%rbp
\opd = 5
.endif
.ifc \r64,%rsi
\opd = 6
.endif
.ifc \r64,%rdi
\opd = 7
.endif
.ifc \r64,%r8
\opd = 8
.endif
.ifc \r64,%r9
\opd = 9
.endif
.ifc \r64,%r10
\opd = 10
.endif
.ifc \r64,%r11
\opd = 11
.endif
.ifc \r64,%r12
\opd = 12
.endif
.ifc \r64,%r13
\opd = 13
.endif
.ifc \r64,%r14
\opd = 14
.endif
.ifc \r64,%r15
\opd = 15
.endif
#endif
.endm
.macro XMM_NUM opd xmm
\opd = NUM_INVALID
.ifc \xmm,%xmm0
\opd = 0
.endif
.ifc \xmm,%xmm1
\opd = 1
.endif
.ifc \xmm,%xmm2
\opd = 2
.endif
.ifc \xmm,%xmm3
\opd = 3
.endif
.ifc \xmm,%xmm4
\opd = 4
.endif
.ifc \xmm,%xmm5
\opd = 5
.endif
.ifc \xmm,%xmm6
\opd = 6
.endif
.ifc \xmm,%xmm7
\opd = 7
.endif
.ifc \xmm,%xmm8
\opd = 8
.endif
.ifc \xmm,%xmm9
\opd = 9
.endif
.ifc \xmm,%xmm10
\opd = 10
.endif
.ifc \xmm,%xmm11
\opd = 11
.endif
.ifc \xmm,%xmm12
\opd = 12
.endif
.ifc \xmm,%xmm13
\opd = 13
.endif
.ifc \xmm,%xmm14
\opd = 14
.endif
.ifc \xmm,%xmm15
\opd = 15
.endif
.endm
.macro TYPE type reg
R32_NUM reg_type_r32 \reg
R64_NUM reg_type_r64 \reg
XMM_NUM reg_type_xmm \reg
.if reg_type_r64 <> NUM_INVALID
\type = TYPE_R64
.elseif reg_type_r32 <> NUM_INVALID
\type = TYPE_R32
.elseif reg_type_xmm <> NUM_INVALID
\type = TYPE_XMM
.else
\type = TYPE_INVALID
.endif
.endm
.macro PFX_OPD_SIZE
.byte 0x66
.endm
.macro PFX_REX opd1 opd2 W=0
.if ((\opd1 | \opd2) & 8) || \W
.byte 0x40 | ((\opd1 & 8) >> 3) | ((\opd2 & 8) >> 1) | (\W << 3)
.endif
.endm
.macro MODRM mod opd1 opd2
.byte \mod | (\opd1 & 7) | ((\opd2 & 7) << 3)
.endm
.macro PSHUFB_XMM xmm1 xmm2
XMM_NUM pshufb_opd1 \xmm1
XMM_NUM pshufb_opd2 \xmm2
PFX_OPD_SIZE
PFX_REX pshufb_opd1 pshufb_opd2
.byte 0x0f, 0x38, 0x00
MODRM 0xc0 pshufb_opd1 pshufb_opd2
.endm
.macro PCLMULQDQ imm8 xmm1 xmm2
XMM_NUM clmul_opd1 \xmm1
XMM_NUM clmul_opd2 \xmm2
PFX_OPD_SIZE
PFX_REX clmul_opd1 clmul_opd2
.byte 0x0f, 0x3a, 0x44
MODRM 0xc0 clmul_opd1 clmul_opd2
.byte \imm8
.endm
.macro PEXTRD imm8 xmm gpr
R32_NUM extrd_opd1 \gpr
XMM_NUM extrd_opd2 \xmm
PFX_OPD_SIZE
PFX_REX extrd_opd1 extrd_opd2
.byte 0x0f, 0x3a, 0x16
MODRM 0xc0 extrd_opd1 extrd_opd2
.byte \imm8
.endm
.macro MOVQ_R64_XMM opd1 opd2
TYPE movq_r64_xmm_opd1_type \opd1
.if movq_r64_xmm_opd1_type == TYPE_XMM
XMM_NUM movq_r64_xmm_opd1 \opd1
R64_NUM movq_r64_xmm_opd2 \opd2
.else
R64_NUM movq_r64_xmm_opd1 \opd1
XMM_NUM movq_r64_xmm_opd2 \opd2
.endif
PFX_OPD_SIZE
PFX_REX movq_r64_xmm_opd1 movq_r64_xmm_opd2 1
.if movq_r64_xmm_opd1_type == TYPE_XMM
.byte 0x0f, 0x7e
.else
.byte 0x0f, 0x6e
.endif
MODRM 0xc0 movq_r64_xmm_opd1 movq_r64_xmm_opd2
.endm
## assume buffers not aligned
#define MOVDQ movdqu
################################ Define Macros
# addm [mem], reg
# Add reg to mem using reg-mem add and store
.macro addm p1 p2
add \p1, \p2
mov \p2, \p1
.endm
################################
# COPY_XMM_AND_BSWAP xmm, [mem], byte_flip_mask
# Load xmm with mem and byte swap each dword
.macro COPY_XMM_AND_BSWAP p1 p2 p3
MOVDQ \p2, \p1
pshufb \p3, \p1
.endm
################################
X0 = %xmm4
X1 = %xmm5
X2 = %xmm6
X3 = %xmm7
XTMP0 = %xmm0
XTMP1 = %xmm1
XTMP2 = %xmm2
XTMP3 = %xmm3
XTMP4 = %xmm8
XFER = %xmm9
SHUF_00BA = %xmm10 # shuffle xBxA -> 00BA
SHUF_DC00 = %xmm11 # shuffle xDxC -> DC00
BYTE_FLIP_MASK = %xmm12
NUM_BLKS = %rdx # 3rd arg
CTX = %rsi # 2nd arg
INP = %rdi # 1st arg
SRND = %rdi # clobbers INP
c = %ecx
d = %r8d
e = %edx
TBL = %rbp
a = %eax
b = %ebx
f = %r9d
g = %r10d
h = %r11d
y0 = %r13d
y1 = %r14d
y2 = %r15d
_INP_END_SIZE = 8
_INP_SIZE = 8
_XFER_SIZE = 16
_XMM_SAVE_SIZE = 0
_INP_END = 0
_INP = _INP_END + _INP_END_SIZE
_XFER = _INP + _INP_SIZE
_XMM_SAVE = _XFER + _XFER_SIZE
STACK_SIZE = _XMM_SAVE + _XMM_SAVE_SIZE
# rotate_Xs
# Rotate values of symbols X0...X3
.macro rotate_Xs
X_ = X0
X0 = X1
X1 = X2
X2 = X3
X3 = X_
.endm
# ROTATE_ARGS
# Rotate values of symbols a...h
.macro ROTATE_ARGS
TMP_ = h
h = g
g = f
f = e
e = d
d = c
c = b
b = a
a = TMP_
.endm
.macro FOUR_ROUNDS_AND_SCHED
## compute s0 four at a time and s1 two at a time
## compute W[-16] + W[-7] 4 at a time
movdqa X3, XTMP0
mov e, y0 # y0 = e
ror $(25-11), y0 # y0 = e >> (25-11)
mov a, y1 # y1 = a
palignr $4, X2, XTMP0 # XTMP0 = W[-7]
ror $(22-13), y1 # y1 = a >> (22-13)
xor e, y0 # y0 = e ^ (e >> (25-11))
mov f, y2 # y2 = f
ror $(11-6), y0 # y0 = (e >> (11-6)) ^ (e >> (25-6))
movdqa X1, XTMP1
xor a, y1 # y1 = a ^ (a >> (22-13)
xor g, y2 # y2 = f^g
paddd X0, XTMP0 # XTMP0 = W[-7] + W[-16]
xor e, y0 # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
and e, y2 # y2 = (f^g)&e
ror $(13-2), y1 # y1 = (a >> (13-2)) ^ (a >> (22-2))
## compute s0
palignr $4, X0, XTMP1 # XTMP1 = W[-15]
xor a, y1 # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
ror $6, y0 # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
xor g, y2 # y2 = CH = ((f^g)&e)^g
movdqa XTMP1, XTMP2 # XTMP2 = W[-15]
ror $2, y1 # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
add y0, y2 # y2 = S1 + CH
add _XFER(%rsp) , y2 # y2 = k + w + S1 + CH
movdqa XTMP1, XTMP3 # XTMP3 = W[-15]
mov a, y0 # y0 = a
add y2, h # h = h + S1 + CH + k + w
mov a, y2 # y2 = a
pslld $(32-7), XTMP1 #
or c, y0 # y0 = a|c
add h, d # d = d + h + S1 + CH + k + w
and c, y2 # y2 = a&c
psrld $7, XTMP2 #
and b, y0 # y0 = (a|c)&b
add y1, h # h = h + S1 + CH + k + w + S0
por XTMP2, XTMP1 # XTMP1 = W[-15] ror 7
or y2, y0 # y0 = MAJ = (a|c)&b)|(a&c)
add y0, h # h = h + S1 + CH + k + w + S0 + MAJ
#
ROTATE_ARGS #
movdqa XTMP3, XTMP2 # XTMP2 = W[-15]
mov e, y0 # y0 = e
mov a, y1 # y1 = a
movdqa XTMP3, XTMP4 # XTMP4 = W[-15]
ror $(25-11), y0 # y0 = e >> (25-11)
xor e, y0 # y0 = e ^ (e >> (25-11))
mov f, y2 # y2 = f
ror $(22-13), y1 # y1 = a >> (22-13)
pslld $(32-18), XTMP3 #
xor a, y1 # y1 = a ^ (a >> (22-13)
ror $(11-6), y0 # y0 = (e >> (11-6)) ^ (e >> (25-6))
xor g, y2 # y2 = f^g
psrld $18, XTMP2 #
ror $(13-2), y1 # y1 = (a >> (13-2)) ^ (a >> (22-2))
xor e, y0 # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
and e, y2 # y2 = (f^g)&e
ror $6, y0 # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
pxor XTMP3, XTMP1
xor a, y1 # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
xor g, y2 # y2 = CH = ((f^g)&e)^g
psrld $3, XTMP4 # XTMP4 = W[-15] >> 3
add y0, y2 # y2 = S1 + CH
add (1*4 + _XFER)(%rsp), y2 # y2 = k + w + S1 + CH
ror $2, y1 # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
pxor XTMP2, XTMP1 # XTMP1 = W[-15] ror 7 ^ W[-15] ror 18
mov a, y0 # y0 = a
add y2, h # h = h + S1 + CH + k + w
mov a, y2 # y2 = a
pxor XTMP4, XTMP1 # XTMP1 = s0
or c, y0 # y0 = a|c
add h, d # d = d + h + S1 + CH + k + w
and c, y2 # y2 = a&c
## compute low s1
pshufd $0b11111010, X3, XTMP2 # XTMP2 = W[-2] {BBAA}
and b, y0 # y0 = (a|c)&b
add y1, h # h = h + S1 + CH + k + w + S0
paddd XTMP1, XTMP0 # XTMP0 = W[-16] + W[-7] + s0
or y2, y0 # y0 = MAJ = (a|c)&b)|(a&c)
add y0, h # h = h + S1 + CH + k + w + S0 + MAJ
ROTATE_ARGS
movdqa XTMP2, XTMP3 # XTMP3 = W[-2] {BBAA}
mov e, y0 # y0 = e
mov a, y1 # y1 = a
ror $(25-11), y0 # y0 = e >> (25-11)
movdqa XTMP2, XTMP4 # XTMP4 = W[-2] {BBAA}
xor e, y0 # y0 = e ^ (e >> (25-11))
ror $(22-13), y1 # y1 = a >> (22-13)
mov f, y2 # y2 = f
xor a, y1 # y1 = a ^ (a >> (22-13)
ror $(11-6), y0 # y0 = (e >> (11-6)) ^ (e >> (25-6))
psrlq $17, XTMP2 # XTMP2 = W[-2] ror 17 {xBxA}
xor g, y2 # y2 = f^g
psrlq $19, XTMP3 # XTMP3 = W[-2] ror 19 {xBxA}
xor e, y0 # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
and e, y2 # y2 = (f^g)&e
psrld $10, XTMP4 # XTMP4 = W[-2] >> 10 {BBAA}
ror $(13-2), y1 # y1 = (a >> (13-2)) ^ (a >> (22-2))
xor a, y1 # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
xor g, y2 # y2 = CH = ((f^g)&e)^g
ror $6, y0 # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
pxor XTMP3, XTMP2
add y0, y2 # y2 = S1 + CH
ror $2, y1 # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
add (2*4 + _XFER)(%rsp), y2 # y2 = k + w + S1 + CH
pxor XTMP2, XTMP4 # XTMP4 = s1 {xBxA}
mov a, y0 # y0 = a
add y2, h # h = h + S1 + CH + k + w
mov a, y2 # y2 = a
pshufb SHUF_00BA, XTMP4 # XTMP4 = s1 {00BA}
or c, y0 # y0 = a|c
add h, d # d = d + h + S1 + CH + k + w
and c, y2 # y2 = a&c
paddd XTMP4, XTMP0 # XTMP0 = {..., ..., W[1], W[0]}
and b, y0 # y0 = (a|c)&b
add y1, h # h = h + S1 + CH + k + w + S0
## compute high s1
pshufd $0b01010000, XTMP0, XTMP2 # XTMP2 = W[-2] {BBAA}
or y2, y0 # y0 = MAJ = (a|c)&b)|(a&c)
add y0, h # h = h + S1 + CH + k + w + S0 + MAJ
#
ROTATE_ARGS #
movdqa XTMP2, XTMP3 # XTMP3 = W[-2] {DDCC}
mov e, y0 # y0 = e
ror $(25-11), y0 # y0 = e >> (25-11)
mov a, y1 # y1 = a
movdqa XTMP2, X0 # X0 = W[-2] {DDCC}
ror $(22-13), y1 # y1 = a >> (22-13)
xor e, y0 # y0 = e ^ (e >> (25-11))
mov f, y2 # y2 = f
ror $(11-6), y0 # y0 = (e >> (11-6)) ^ (e >> (25-6))
psrlq $17, XTMP2 # XTMP2 = W[-2] ror 17 {xDxC}
xor a, y1 # y1 = a ^ (a >> (22-13)
xor g, y2 # y2 = f^g
psrlq $19, XTMP3 # XTMP3 = W[-2] ror 19 {xDxC}
xor e, y0 # y0 = e ^ (e >> (11-6)) ^ (e >> (25
and e, y2 # y2 = (f^g)&e
ror $(13-2), y1 # y1 = (a >> (13-2)) ^ (a >> (22-2))
psrld $10, X0 # X0 = W[-2] >> 10 {DDCC}
xor a, y1 # y1 = a ^ (a >> (13-2)) ^ (a >> (22
ror $6, y0 # y0 = S1 = (e>>6) & (e>>11) ^ (e>>2
xor g, y2 # y2 = CH = ((f^g)&e)^g
pxor XTMP3, XTMP2 #
ror $2, y1 # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>2
add y0, y2 # y2 = S1 + CH
add (3*4 + _XFER)(%rsp), y2 # y2 = k + w + S1 + CH
pxor XTMP2, X0 # X0 = s1 {xDxC}
mov a, y0 # y0 = a
add y2, h # h = h + S1 + CH + k + w
mov a, y2 # y2 = a
pshufb SHUF_DC00, X0 # X0 = s1 {DC00}
or c, y0 # y0 = a|c
add h, d # d = d + h + S1 + CH + k + w
and c, y2 # y2 = a&c
paddd XTMP0, X0 # X0 = {W[3], W[2], W[1], W[0]}
and b, y0 # y0 = (a|c)&b
add y1, h # h = h + S1 + CH + k + w + S0
or y2, y0 # y0 = MAJ = (a|c)&b)|(a&c)
add y0, h # h = h + S1 + CH + k + w + S0 + MAJ
ROTATE_ARGS
rotate_Xs
.endm
## input is [rsp + _XFER + %1 * 4]
.macro DO_ROUND round
mov e, y0 # y0 = e
ror $(25-11), y0 # y0 = e >> (25-11)
mov a, y1 # y1 = a
xor e, y0 # y0 = e ^ (e >> (25-11))
ror $(22-13), y1 # y1 = a >> (22-13)
mov f, y2 # y2 = f
xor a, y1 # y1 = a ^ (a >> (22-13)
ror $(11-6), y0 # y0 = (e >> (11-6)) ^ (e >> (25-6))
xor g, y2 # y2 = f^g
xor e, y0 # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
ror $(13-2), y1 # y1 = (a >> (13-2)) ^ (a >> (22-2))
and e, y2 # y2 = (f^g)&e
xor a, y1 # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
ror $6, y0 # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
xor g, y2 # y2 = CH = ((f^g)&e)^g
add y0, y2 # y2 = S1 + CH
ror $2, y1 # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
offset = \round * 4 + _XFER
add offset(%rsp), y2 # y2 = k + w + S1 + CH
mov a, y0 # y0 = a
add y2, h # h = h + S1 + CH + k + w
mov a, y2 # y2 = a
or c, y0 # y0 = a|c
add h, d # d = d + h + S1 + CH + k + w
and c, y2 # y2 = a&c
and b, y0 # y0 = (a|c)&b
add y1, h # h = h + S1 + CH + k + w + S0
or y2, y0 # y0 = MAJ = (a|c)&b)|(a&c)
add y0, h # h = h + S1 + CH + k + w + S0 + MAJ
ROTATE_ARGS
.endm
########################################################################
## void sha256_transform_ssse3(void *input_data, UINT32 digest[8], UINT64 num_blks)
## arg 1 : pointer to input data
## arg 2 : pointer to digest
## arg 3 : Num blocks
########################################################################
.text
ENTRY(sha256_transform_ssse3)
.align 32
pushq %rbx
pushq %rbp
pushq %r13
pushq %r14
pushq %r15
pushq %r12
mov %rsp, %r12
subq $STACK_SIZE, %rsp
and $~15, %rsp
shl $6, NUM_BLKS # convert to bytes
jz done_hash
add INP, NUM_BLKS
mov NUM_BLKS, _INP_END(%rsp) # pointer to end of data
## load initial digest
mov 4*0(CTX), a
mov 4*1(CTX), b
mov 4*2(CTX), c
mov 4*3(CTX), d
mov 4*4(CTX), e
mov 4*5(CTX), f
mov 4*6(CTX), g
mov 4*7(CTX), h
movdqa PSHUFFLE_BYTE_FLIP_MASK(%rip), BYTE_FLIP_MASK
movdqa _SHUF_00BA(%rip), SHUF_00BA
movdqa _SHUF_DC00(%rip), SHUF_DC00
loop0:
lea K256(%rip), TBL
## byte swap first 16 dwords
COPY_XMM_AND_BSWAP X0, 0*16(INP), BYTE_FLIP_MASK
COPY_XMM_AND_BSWAP X1, 1*16(INP), BYTE_FLIP_MASK
COPY_XMM_AND_BSWAP X2, 2*16(INP), BYTE_FLIP_MASK
COPY_XMM_AND_BSWAP X3, 3*16(INP), BYTE_FLIP_MASK
mov INP, _INP(%rsp)
## schedule 48 input dwords, by doing 3 rounds of 16 each
mov $3, SRND
.align 16
loop1:
movdqa (TBL), XFER
paddd X0, XFER
movdqa XFER, _XFER(%rsp)
FOUR_ROUNDS_AND_SCHED
movdqa 1*16(TBL), XFER
paddd X0, XFER
movdqa XFER, _XFER(%rsp)
FOUR_ROUNDS_AND_SCHED
movdqa 2*16(TBL), XFER
paddd X0, XFER
movdqa XFER, _XFER(%rsp)
FOUR_ROUNDS_AND_SCHED
movdqa 3*16(TBL), XFER
paddd X0, XFER
movdqa XFER, _XFER(%rsp)
add $4*16, TBL
FOUR_ROUNDS_AND_SCHED
sub $1, SRND
jne loop1
mov $2, SRND
loop2:
paddd (TBL), X0
movdqa X0, _XFER(%rsp)
DO_ROUND 0
DO_ROUND 1
DO_ROUND 2
DO_ROUND 3
paddd 1*16(TBL), X1
movdqa X1, _XFER(%rsp)
add $2*16, TBL
DO_ROUND 0
DO_ROUND 1
DO_ROUND 2
DO_ROUND 3
movdqa X2, X0
movdqa X3, X1
sub $1, SRND
jne loop2
addm (4*0)(CTX),a
addm (4*1)(CTX),b
addm (4*2)(CTX),c
addm (4*3)(CTX),d
addm (4*4)(CTX),e
addm (4*5)(CTX),f
addm (4*6)(CTX),g
addm (4*7)(CTX),h
mov _INP(%rsp), INP
add $64, INP
cmp _INP_END(%rsp), INP
jne loop0
done_hash:
mov %r12, %rsp
popq %r12
popq %r15
popq %r14
popq %r13
popq %rbp
popq %rbx
ret
ENDPROC(sha256_transform_ssse3)
.data
.align 64
K256:
.long 0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5
.long 0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5
.long 0xd807aa98,0x12835b01,0x243185be,0x550c7dc3
.long 0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174
.long 0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc
.long 0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da
.long 0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7
.long 0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967
.long 0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13
.long 0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85
.long 0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3
.long 0xd192e819,0xd6990624,0xf40e3585,0x106aa070
.long 0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5
.long 0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3
.long 0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208
.long 0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2
PSHUFFLE_BYTE_FLIP_MASK:
.octa 0x0c0d0e0f08090a0b0405060700010203
# shuffle xBxA -> 00BA
_SHUF_00BA:
.octa 0xFFFFFFFFFFFFFFFF0b0a090803020100
# shuffle xDxC -> DC00
_SHUF_DC00:
.octa 0x0b0a090803020100FFFFFFFFFFFFFFFF
#endif
-53
View File
@@ -1,53 +0,0 @@
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package sha256
import (
"hash"
"io"
"crypto/sha256"
)
// Sum256 - single caller sha256 helper
func Sum256(data []byte) []byte {
d := sha256.New()
d.Write(data)
return d.Sum(nil)
}
// Sum - io.Reader based streaming sha256 helper
func Sum(reader io.Reader) ([]byte, error) {
d := sha256.New()
var err error
for err == nil {
length := 0
byteBuffer := make([]byte, 1024*1024)
length, err = reader.Read(byteBuffer)
byteBuffer = byteBuffer[0:length]
d.Write(byteBuffer)
}
if err != io.EOF {
return nil, err
}
return d.Sum(nil), nil
}
// New returns a new hash.Hash computing SHA256.
func New() hash.Hash {
return sha256.New()
}
-179
View File
@@ -1,179 +0,0 @@
// Copyright 2009 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 of
// Golang project:
// https://github.com/golang/go/blob/master/LICENSE
// Using this part of Minio codebase under the license
// Apache License Version 2.0 with modifications
// Package sha256 provides SHA256SSE3, SHA256AVX, SHA256AVX2
package sha256
import (
"hash"
"io"
"github.com/minio/minio/pkg/cpu"
)
// The size of a SHA256 checksum in bytes.
const Size = 32
// The blocksize of SHA256 in bytes.
const BlockSize = 64
const (
chunk = 64
init0 = 0x6A09E667
init1 = 0xBB67AE85
init2 = 0x3C6EF372
init3 = 0xA54FF53A
init4 = 0x510E527F
init5 = 0x9B05688C
init6 = 0x1F83D9AB
init7 = 0x5BE0CD19
)
// digest represents the partial evaluation of a checksum.
type digest struct {
h [8]uint32
x [chunk]byte
nx int
len uint64
}
// Reset digest back to default
func (d *digest) Reset() {
d.h[0] = init0
d.h[1] = init1
d.h[2] = init2
d.h[3] = init3
d.h[4] = init4
d.h[5] = init5
d.h[6] = init6
d.h[7] = init7
d.nx = 0
d.len = 0
}
func block(dig *digest, p []byte) {
switch true {
case cpu.HasAVX2() == true:
blockAVX2(dig, p)
case cpu.HasAVX() == true:
blockAVX(dig, p)
case cpu.HasSSE41() == true:
blockSSE(dig, p)
default:
blockGeneric(dig, p)
}
}
// New returns a new hash.Hash computing the SHA256 checksum.
func New() hash.Hash {
d := new(digest)
d.Reset()
return d
}
// Return size of checksum
func (d *digest) Size() int { return Size }
// Return blocksize of checksum
func (d *digest) BlockSize() int { return BlockSize }
// Write to digest
func (d *digest) Write(p []byte) (nn int, err error) {
nn = len(p)
d.len += uint64(nn)
if d.nx > 0 {
n := copy(d.x[d.nx:], p)
d.nx += n
if d.nx == chunk {
block(d, d.x[:])
d.nx = 0
}
p = p[n:]
}
if len(p) >= chunk {
n := len(p) &^ (chunk - 1)
block(d, p[:n])
p = p[n:]
}
if len(p) > 0 {
d.nx = copy(d.x[:], p)
}
return
}
// Return sha256 sum in bytes
func (d *digest) Sum(in []byte) []byte {
// Make a copy of d0 so that caller can keep writing and summing.
d0 := *d
hash := d0.checkSum()
return append(in, hash[:]...)
}
// Intermediate checksum function
func (d *digest) checkSum() [Size]byte {
len := d.len
// Padding. Add a 1 bit and 0 bits until 56 bytes mod 64.
var tmp [64]byte
tmp[0] = 0x80
if len%64 < 56 {
d.Write(tmp[0 : 56-len%64])
} else {
d.Write(tmp[0 : 64+56-len%64])
}
// Length in bits.
len <<= 3
for i := uint(0); i < 8; i++ {
tmp[i] = byte(len >> (56 - 8*i))
}
d.Write(tmp[0:8])
if d.nx != 0 {
panic("d.nx != 0")
}
h := d.h[:]
var digest [Size]byte
for i, s := range h {
digest[i*4] = byte(s >> 24)
digest[i*4+1] = byte(s >> 16)
digest[i*4+2] = byte(s >> 8)
digest[i*4+3] = byte(s)
}
return digest
}
/// Convenience functions
// Sum256 - single caller sha256 helper
func Sum256(data []byte) []byte {
var d digest
d.Reset()
d.Write(data)
return d.Sum(nil)
}
// Sum - io.Reader based streaming sha256 helper
func Sum(reader io.Reader) ([]byte, error) {
h := New()
var err error
for err == nil {
length := 0
byteBuffer := make([]byte, 1024*1024)
length, err = reader.Read(byteBuffer)
byteBuffer = byteBuffer[0:length]
h.Write(byteBuffer)
}
if err != io.EOF {
return nil, err
}
return h.Sum(nil), nil
}
-125
View File
@@ -1,125 +0,0 @@
// Copyright 2009 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 of
// Golang project:
// https://github.com/golang/go/blob/master/LICENSE
// Using this part of Minio codebase under the license
// Apache License Version 2.0 with modifications
// SHA256 hash algorithm. See FIPS 180-2.
package sha256
import (
"fmt"
"io"
"testing"
)
type sha256Test struct {
out string
in string
}
var golden = []sha256Test{
{"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", ""},
{"ca978112ca1bbdcafac231b39a23dc4da786eff8147c4e72b9807785afee48bb", "a"},
{"fb8e20fc2e4c3f248c60c39bd652f3c1347298bb977b8b4d5903b85055620603", "ab"},
{"ba7816bf8f01cfea414140de5dae2223b00361a396177a9cb410ff61f20015ad", "abc"},
{"88d4266fd4e6338d13b845fcf289579d209c897823b9217da3e161936f031589", "abcd"},
{"36bbe50ed96841d10443bcb670d6554f0a34b761be67ec9c4a8ad2c0c44ca42c", "abcde"},
{"bef57ec7f53a6d40beb640a780a639c83bc29ac8a9816f1fc6c5c6dcd93c4721", "abcdef"},
{"7d1a54127b222502f5b79b5fb0803061152a44f92b37e23c6527baf665d4da9a", "abcdefg"},
{"9c56cc51b374c3ba189210d5b6d4bf57790d351c96c47c02190ecf1e430635ab", "abcdefgh"},
{"19cc02f26df43cc571bc9ed7b0c4d29224a3ec229529221725ef76d021c8326f", "abcdefghi"},
{"72399361da6a7754fec986dca5b7cbaf1c810a28ded4abaf56b2106d06cb78b0", "abcdefghij"},
{"a144061c271f152da4d151034508fed1c138b8c976339de229c3bb6d4bbb4fce", "Discard medicine more than two years old."},
{"6dae5caa713a10ad04b46028bf6dad68837c581616a1589a265a11288d4bb5c4", "He who has a shady past knows that nice guys finish last."},
{"ae7a702a9509039ddbf29f0765e70d0001177914b86459284dab8b348c2dce3f", "I wouldn't marry him with a ten foot pole."},
{"6748450b01c568586715291dfa3ee018da07d36bb7ea6f180c1af6270215c64f", "Free! Free!/A trip/to Mars/for 900/empty jars/Burma Shave"},
{"14b82014ad2b11f661b5ae6a99b75105c2ffac278cd071cd6c05832793635774", "The days of the digital watch are numbered. -Tom Stoppard"},
{"7102cfd76e2e324889eece5d6c41921b1e142a4ac5a2692be78803097f6a48d8", "Nepal premier won't resign."},
{"23b1018cd81db1d67983c5f7417c44da9deb582459e378d7a068552ea649dc9f", "For every action there is an equal and opposite government program."},
{"8001f190dfb527261c4cfcab70c98e8097a7a1922129bc4096950e57c7999a5a", "His money is twice tainted: 'taint yours and 'taint mine."},
{"8c87deb65505c3993eb24b7a150c4155e82eee6960cf0c3a8114ff736d69cad5", "There is no reason for any individual to have a computer in their home. -Ken Olsen, 1977"},
{"bfb0a67a19cdec3646498b2e0f751bddc41bba4b7f30081b0b932aad214d16d7", "It's a tiny change to the code and not completely disgusting. - Bob Manchek"},
{"7f9a0b9bf56332e19f5a0ec1ad9c1425a153da1c624868fda44561d6b74daf36", "size: a.out: bad magic"},
{"b13f81b8aad9e3666879af19886140904f7f429ef083286195982a7588858cfc", "The major problem is with sendmail. -Mark Horton"},
{"b26c38d61519e894480c70c8374ea35aa0ad05b2ae3d6674eec5f52a69305ed4", "Give me a rock, paper and scissors and I will move the world. CCFestoon"},
{"049d5e26d4f10222cd841a119e38bd8d2e0d1129728688449575d4ff42b842c1", "If the enemy is within range, then so are you."},
{"0e116838e3cc1c1a14cd045397e29b4d087aa11b0853fc69ec82e90330d60949", "It's well we cannot hear the screams/That we create in others' dreams."},
{"4f7d8eb5bcf11de2a56b971021a444aa4eafd6ecd0f307b5109e4e776cd0fe46", "You remind me of a TV show, but that's all right: I watch it anyway."},
{"61c0cc4c4bd8406d5120b3fb4ebc31ce87667c162f29468b3c779675a85aebce", "C is as portable as Stonehedge!!"},
{"1fb2eb3688093c4a3f80cd87a5547e2ce940a4f923243a79a2a1e242220693ac", "Even if I could be Shakespeare, I think I should still choose to be Faraday. - A. Huxley"},
{"395585ce30617b62c80b93e8208ce866d4edc811a177fdb4b82d3911d8696423", "The fugacity of a constituent in a mixture of gases at a given temperature is proportional to its mole fraction. Lewis-Randall Rule"},
{"4f9b189a13d030838269dce846b16a1ce9ce81fe63e65de2f636863336a98fe6", "How can you write a big system without C++? -Paul Glick"},
}
func TestGolden(t *testing.T) {
for i := 0; i < len(golden); i++ {
g := golden[i]
s := fmt.Sprintf("%x", Sum256([]byte(g.in)))
if s != g.out {
t.Fatalf("Sum256 function: sha256(%s) = %s want %s", g.in, s, g.out)
}
c := New()
for j := 0; j < 3; j++ {
if j < 2 {
io.WriteString(c, g.in)
} else {
io.WriteString(c, g.in[0:len(g.in)/2])
c.Sum(nil)
io.WriteString(c, g.in[len(g.in)/2:])
}
s := fmt.Sprintf("%x", c.Sum(nil))
if s != g.out {
t.Fatalf("sha256[%d](%s) = %s want %s", j, g.in, s, g.out)
}
c.Reset()
}
}
}
func TestSize(t *testing.T) {
c := New()
if got := c.Size(); got != Size {
t.Errorf("Size = %d; want %d", got, Size)
}
}
func TestBlockSize(t *testing.T) {
c := New()
if got := c.BlockSize(); got != BlockSize {
t.Errorf("BlockSize = %d want %d", got, BlockSize)
}
}
var bench = New()
var buf = make([]byte, 1024*1024)
func benchmarkSize(b *testing.B, size int) {
b.SetBytes(int64(size))
sum := make([]byte, bench.Size())
for i := 0; i < b.N; i++ {
bench.Reset()
bench.Write(buf[:size])
bench.Sum(sum[:0])
}
}
func BenchmarkHash8Bytes(b *testing.B) {
benchmarkSize(b, 8)
}
func BenchmarkHash1K(b *testing.B) {
benchmarkSize(b, 1024)
}
func BenchmarkHash8K(b *testing.B) {
benchmarkSize(b, 8192)
}
func BenchmarkHash1M(b *testing.B) {
benchmarkSize(b, 1024*1024)
}
-53
View File
@@ -1,53 +0,0 @@
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package sha256
import (
"hash"
"io"
"crypto/sha256"
)
// Sum256 - single caller sha256 helper
func Sum256(data []byte) []byte {
d := sha256.New()
d.Write(data)
return d.Sum(nil)
}
// Sum - io.Reader based streaming sha256 helper
func Sum(reader io.Reader) ([]byte, error) {
d := sha256.New()
var err error
for err == nil {
length := 0
byteBuffer := make([]byte, 1024*1024)
length, err = reader.Read(byteBuffer)
byteBuffer = byteBuffer[0:length]
d.Write(byteBuffer)
}
if err != io.EOF {
return nil, err
}
return d.Sum(nil), nil
}
// New returns a new hash.Hash computing SHA256.
func New() hash.Hash {
return sha256.New()
}
-162
View File
@@ -1,162 +0,0 @@
// +build amd64
//
// Minio Cloud Storage, (C) 2015 Minio, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
// Software block transform are provided by The Go Authors:
// Copyright 2009 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 of
// Golang project:
// https://github.com/golang/go/blob/master/LICENSE
package sha256
// #cgo CFLAGS: -DHAS_SSE41 -DHAS_AVX -DHAS_AVX2
// #include <stdint.h>
// void sha256_transform_ssse3 (const char *input_data, uint32_t *digest, unsigned long num_blks);
// void sha256_transform_avx (const char *input_data, uint32_t *digest, unsigned long num_blks);
// void sha256_transform_rorx (const char *input_data, uint32_t *digest, unsigned long num_blks);
import "C"
import "unsafe"
func blockSSE(dig *digest, p []byte) {
C.sha256_transform_ssse3((*C.char)(unsafe.Pointer(&p[0])), (*C.uint32_t)(unsafe.Pointer(&dig.h[0])), (C.ulong)(len(p)/64))
}
func blockAVX(dig *digest, p []byte) {
C.sha256_transform_avx((*C.char)(unsafe.Pointer(&p[0])), (*C.uint32_t)(unsafe.Pointer(&dig.h[0])), (C.ulong)(len(p)/64))
}
func blockAVX2(dig *digest, p []byte) {
C.sha256_transform_rorx((*C.char)(unsafe.Pointer(&p[0])), (*C.uint32_t)(unsafe.Pointer(&dig.h[0])), (C.ulong)(len(p)/64))
}
func blockGeneric(dig *digest, p []byte) {
var w [64]uint32
h0, h1, h2, h3, h4, h5, h6, h7 := dig.h[0], dig.h[1], dig.h[2], dig.h[3], dig.h[4], dig.h[5], dig.h[6], dig.h[7]
for len(p) >= chunk {
// Can interlace the computation of w with the
// rounds below if needed for speed.
for i := 0; i < 16; i++ {
j := i * 4
w[i] = uint32(p[j])<<24 | uint32(p[j+1])<<16 | uint32(p[j+2])<<8 | uint32(p[j+3])
}
for i := 16; i < 64; i++ {
v1 := w[i-2]
t1 := (v1>>17 | v1<<(32-17)) ^ (v1>>19 | v1<<(32-19)) ^ (v1 >> 10)
v2 := w[i-15]
t2 := (v2>>7 | v2<<(32-7)) ^ (v2>>18 | v2<<(32-18)) ^ (v2 >> 3)
w[i] = t1 + w[i-7] + t2 + w[i-16]
}
a, b, c, d, e, f, g, h := h0, h1, h2, h3, h4, h5, h6, h7
for i := 0; i < 64; i++ {
t1 := h + ((e>>6 | e<<(32-6)) ^ (e>>11 | e<<(32-11)) ^ (e>>25 | e<<(32-25))) + ((e & f) ^ (^e & g)) + _K[i] + w[i]
t2 := ((a>>2 | a<<(32-2)) ^ (a>>13 | a<<(32-13)) ^ (a>>22 | a<<(32-22))) + ((a & b) ^ (a & c) ^ (b & c))
h = g
g = f
f = e
e = d + t1
d = c
c = b
b = a
a = t1 + t2
}
h0 += a
h1 += b
h2 += c
h3 += d
h4 += e
h5 += f
h6 += g
h7 += h
p = p[chunk:]
}
dig.h[0], dig.h[1], dig.h[2], dig.h[3], dig.h[4], dig.h[5], dig.h[6], dig.h[7] = h0, h1, h2, h3, h4, h5, h6, h7
}
var _K = []uint32{
0x428a2f98,
0x71374491,
0xb5c0fbcf,
0xe9b5dba5,
0x3956c25b,
0x59f111f1,
0x923f82a4,
0xab1c5ed5,
0xd807aa98,
0x12835b01,
0x243185be,
0x550c7dc3,
0x72be5d74,
0x80deb1fe,
0x9bdc06a7,
0xc19bf174,
0xe49b69c1,
0xefbe4786,
0x0fc19dc6,
0x240ca1cc,
0x2de92c6f,
0x4a7484aa,
0x5cb0a9dc,
0x76f988da,
0x983e5152,
0xa831c66d,
0xb00327c8,
0xbf597fc7,
0xc6e00bf3,
0xd5a79147,
0x06ca6351,
0x14292967,
0x27b70a85,
0x2e1b2138,
0x4d2c6dfc,
0x53380d13,
0x650a7354,
0x766a0abb,
0x81c2c92e,
0x92722c85,
0xa2bfe8a1,
0xa81a664b,
0xc24b8b70,
0xc76c51a3,
0xd192e819,
0xd6990624,
0xf40e3585,
0x106aa070,
0x19a4c116,
0x1e376c08,
0x2748774c,
0x34b0bcb5,
0x391c0cb3,
0x4ed8aa4a,
0x5b9cca4f,
0x682e6ff3,
0x748f82ee,
0x78a5636f,
0x84c87814,
0x8cc70208,
0x90befffa,
0xa4506ceb,
0xbef9a3f7,
0xc67178f2,
}
-202
View File
@@ -1,202 +0,0 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-686
View File
@@ -1,686 +0,0 @@
########################################################################
# Implement fast SHA-512 with AVX instructions. (x86_64)
#
# Copyright (C) 2013 Intel Corporation.
#
# Authors:
# James Guilford <james.guilford@intel.com>
# Kirk Yap <kirk.s.yap@intel.com>
# David Cote <david.m.cote@intel.com>
# Tim Chen <tim.c.chen@linux.intel.com>
#
# This software is available to you under a choice of one of two
# licenses. You may choose to be licensed under the terms of the GNU
# General Public License (GPL) Version 2, available from the file
# COPYING in the main directory of this source tree, or the
# OpenIB.org BSD license below:
#
# 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.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
# EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
# NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
# BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
# ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
#
########################################################################
#
# This code is described in an Intel White-Paper:
# "Fast SHA-512 Implementations on Intel Architecture Processors"
#
# To find it, surf to http://www.intel.com/p/en_US/embedded
# and search for that title.
#
########################################################################
# Using this part of Minio codebase under the license
# Apache License Version 2.0 with modifications
##
#ifdef HAS_AVX
#ifndef ENTRY
#define ENTRY(name) \
.globl name ; \
.align 4,0x90 ; \
name:
#endif
#ifndef END
#define END(name) \
.size name, .-name
#endif
#ifndef ENDPROC
#define ENDPROC(name) \
.type name, @function ; \
END(name)
#endif
#define NUM_INVALID 100
#define TYPE_R32 0
#define TYPE_R64 1
#define TYPE_XMM 2
#define TYPE_INVALID 100
.macro R32_NUM opd r32
\opd = NUM_INVALID
.ifc \r32,%eax
\opd = 0
.endif
.ifc \r32,%ecx
\opd = 1
.endif
.ifc \r32,%edx
\opd = 2
.endif
.ifc \r32,%ebx
\opd = 3
.endif
.ifc \r32,%esp
\opd = 4
.endif
.ifc \r32,%ebp
\opd = 5
.endif
.ifc \r32,%esi
\opd = 6
.endif
.ifc \r32,%edi
\opd = 7
.endif
#ifdef X86_64
.ifc \r32,%r8d
\opd = 8
.endif
.ifc \r32,%r9d
\opd = 9
.endif
.ifc \r32,%r10d
\opd = 10
.endif
.ifc \r32,%r11d
\opd = 11
.endif
.ifc \r32,%r12d
\opd = 12
.endif
.ifc \r32,%r13d
\opd = 13
.endif
.ifc \r32,%r14d
\opd = 14
.endif
.ifc \r32,%r15d
\opd = 15
.endif
#endif
.endm
.macro R64_NUM opd r64
\opd = NUM_INVALID
#ifdef X86_64
.ifc \r64,%rax
\opd = 0
.endif
.ifc \r64,%rcx
\opd = 1
.endif
.ifc \r64,%rdx
\opd = 2
.endif
.ifc \r64,%rbx
\opd = 3
.endif
.ifc \r64,%rsp
\opd = 4
.endif
.ifc \r64,%rbp
\opd = 5
.endif
.ifc \r64,%rsi
\opd = 6
.endif
.ifc \r64,%rdi
\opd = 7
.endif
.ifc \r64,%r8
\opd = 8
.endif
.ifc \r64,%r9
\opd = 9
.endif
.ifc \r64,%r10
\opd = 10
.endif
.ifc \r64,%r11
\opd = 11
.endif
.ifc \r64,%r12
\opd = 12
.endif
.ifc \r64,%r13
\opd = 13
.endif
.ifc \r64,%r14
\opd = 14
.endif
.ifc \r64,%r15
\opd = 15
.endif
#endif
.endm
.macro XMM_NUM opd xmm
\opd = NUM_INVALID
.ifc \xmm,%xmm0
\opd = 0
.endif
.ifc \xmm,%xmm1
\opd = 1
.endif
.ifc \xmm,%xmm2
\opd = 2
.endif
.ifc \xmm,%xmm3
\opd = 3
.endif
.ifc \xmm,%xmm4
\opd = 4
.endif
.ifc \xmm,%xmm5
\opd = 5
.endif
.ifc \xmm,%xmm6
\opd = 6
.endif
.ifc \xmm,%xmm7
\opd = 7
.endif
.ifc \xmm,%xmm8
\opd = 8
.endif
.ifc \xmm,%xmm9
\opd = 9
.endif
.ifc \xmm,%xmm10
\opd = 10
.endif
.ifc \xmm,%xmm11
\opd = 11
.endif
.ifc \xmm,%xmm12
\opd = 12
.endif
.ifc \xmm,%xmm13
\opd = 13
.endif
.ifc \xmm,%xmm14
\opd = 14
.endif
.ifc \xmm,%xmm15
\opd = 15
.endif
.endm
.macro TYPE type reg
R32_NUM reg_type_r32 \reg
R64_NUM reg_type_r64 \reg
XMM_NUM reg_type_xmm \reg
.if reg_type_r64 <> NUM_INVALID
\type = TYPE_R64
.elseif reg_type_r32 <> NUM_INVALID
\type = TYPE_R32
.elseif reg_type_xmm <> NUM_INVALID
\type = TYPE_XMM
.else
\type = TYPE_INVALID
.endif
.endm
.macro PFX_OPD_SIZE
.byte 0x66
.endm
.macro PFX_REX opd1 opd2 W=0
.if ((\opd1 | \opd2) & 8) || \W
.byte 0x40 | ((\opd1 & 8) >> 3) | ((\opd2 & 8) >> 1) | (\W << 3)
.endif
.endm
.macro MODRM mod opd1 opd2
.byte \mod | (\opd1 & 7) | ((\opd2 & 7) << 3)
.endm
.macro PSHUFB_XMM xmm1 xmm2
XMM_NUM pshufb_opd1 \xmm1
XMM_NUM pshufb_opd2 \xmm2
PFX_OPD_SIZE
PFX_REX pshufb_opd1 pshufb_opd2
.byte 0x0f, 0x38, 0x00
MODRM 0xc0 pshufb_opd1 pshufb_opd2
.endm
.macro PCLMULQDQ imm8 xmm1 xmm2
XMM_NUM clmul_opd1 \xmm1
XMM_NUM clmul_opd2 \xmm2
PFX_OPD_SIZE
PFX_REX clmul_opd1 clmul_opd2
.byte 0x0f, 0x3a, 0x44
MODRM 0xc0 clmul_opd1 clmul_opd2
.byte \imm8
.endm
.macro PEXTRD imm8 xmm gpr
R32_NUM extrd_opd1 \gpr
XMM_NUM extrd_opd2 \xmm
PFX_OPD_SIZE
PFX_REX extrd_opd1 extrd_opd2
.byte 0x0f, 0x3a, 0x16
MODRM 0xc0 extrd_opd1 extrd_opd2
.byte \imm8
.endm
.macro MOVQ_R64_XMM opd1 opd2
TYPE movq_r64_xmm_opd1_type \opd1
.if movq_r64_xmm_opd1_type == TYPE_XMM
XMM_NUM movq_r64_xmm_opd1 \opd1
R64_NUM movq_r64_xmm_opd2 \opd2
.else
R64_NUM movq_r64_xmm_opd1 \opd1
XMM_NUM movq_r64_xmm_opd2 \opd2
.endif
PFX_OPD_SIZE
PFX_REX movq_r64_xmm_opd1 movq_r64_xmm_opd2 1
.if movq_r64_xmm_opd1_type == TYPE_XMM
.byte 0x0f, 0x7e
.else
.byte 0x0f, 0x6e
.endif
MODRM 0xc0 movq_r64_xmm_opd1 movq_r64_xmm_opd2
.endm
.text
# Virtual Registers
# ARG1
msg = %rdi
# ARG2
digest = %rsi
# ARG3
msglen = %rdx
T1 = %rcx
T2 = %r8
a_64 = %r9
b_64 = %r10
c_64 = %r11
d_64 = %r12
e_64 = %r13
f_64 = %r14
g_64 = %r15
h_64 = %rbx
tmp0 = %rax
# Local variables (stack frame)
# Message Schedule
W_SIZE = 80*8
# W[t] + K[t] | W[t+1] + K[t+1]
WK_SIZE = 2*8
RSPSAVE_SIZE = 1*8
GPRSAVE_SIZE = 5*8
frame_W = 0
frame_WK = frame_W + W_SIZE
frame_RSPSAVE = frame_WK + WK_SIZE
frame_GPRSAVE = frame_RSPSAVE + RSPSAVE_SIZE
frame_size = frame_GPRSAVE + GPRSAVE_SIZE
# Useful QWORD "arrays" for simpler memory references
# MSG, DIGEST, K_t, W_t are arrays
# WK_2(t) points to 1 of 2 qwords at frame.WK depdending on t being odd/even
# Input message (arg1)
#define MSG(i) 8*i(msg)
# Output Digest (arg2)
#define DIGEST(i) 8*i(digest)
# SHA Constants (static mem)
#define K_t(i) 8*i+K512(%rip)
# Message Schedule (stack frame)
#define W_t(i) 8*i+frame_W(%rsp)
# W[t]+K[t] (stack frame)
#define WK_2(i) 8*((i%2))+frame_WK(%rsp)
.macro RotateState
# Rotate symbols a..h right
TMP = h_64
h_64 = g_64
g_64 = f_64
f_64 = e_64
e_64 = d_64
d_64 = c_64
c_64 = b_64
b_64 = a_64
a_64 = TMP
.endm
.macro RORQ p1 p2
# shld is faster than ror on Sandybridge
shld $(64-\p2), \p1, \p1
.endm
.macro SHA512_Round rnd
# Compute Round %%t
mov f_64, T1 # T1 = f
mov e_64, tmp0 # tmp = e
xor g_64, T1 # T1 = f ^ g
RORQ tmp0, 23 # 41 # tmp = e ror 23
and e_64, T1 # T1 = (f ^ g) & e
xor e_64, tmp0 # tmp = (e ror 23) ^ e
xor g_64, T1 # T1 = ((f ^ g) & e) ^ g = CH(e,f,g)
idx = \rnd
add WK_2(idx), T1 # W[t] + K[t] from message scheduler
RORQ tmp0, 4 # 18 # tmp = ((e ror 23) ^ e) ror 4
xor e_64, tmp0 # tmp = (((e ror 23) ^ e) ror 4) ^ e
mov a_64, T2 # T2 = a
add h_64, T1 # T1 = CH(e,f,g) + W[t] + K[t] + h
RORQ tmp0, 14 # 14 # tmp = ((((e ror23)^e)ror4)^e)ror14 = S1(e)
add tmp0, T1 # T1 = CH(e,f,g) + W[t] + K[t] + S1(e)
mov a_64, tmp0 # tmp = a
xor c_64, T2 # T2 = a ^ c
and c_64, tmp0 # tmp = a & c
and b_64, T2 # T2 = (a ^ c) & b
xor tmp0, T2 # T2 = ((a ^ c) & b) ^ (a & c) = Maj(a,b,c)
mov a_64, tmp0 # tmp = a
RORQ tmp0, 5 # 39 # tmp = a ror 5
xor a_64, tmp0 # tmp = (a ror 5) ^ a
add T1, d_64 # e(next_state) = d + T1
RORQ tmp0, 6 # 34 # tmp = ((a ror 5) ^ a) ror 6
xor a_64, tmp0 # tmp = (((a ror 5) ^ a) ror 6) ^ a
lea (T1, T2), h_64 # a(next_state) = T1 + Maj(a,b,c)
RORQ tmp0, 28 # 28 # tmp = ((((a ror5)^a)ror6)^a)ror28 = S0(a)
add tmp0, h_64 # a(next_state) = T1 + Maj(a,b,c) S0(a)
RotateState
.endm
.macro SHA512_2Sched_2Round_avx rnd
# Compute rounds t-2 and t-1
# Compute message schedule QWORDS t and t+1
# Two rounds are computed based on the values for K[t-2]+W[t-2] and
# K[t-1]+W[t-1] which were previously stored at WK_2 by the message
# scheduler.
# The two new schedule QWORDS are stored at [W_t(t)] and [W_t(t+1)].
# They are then added to their respective SHA512 constants at
# [K_t(t)] and [K_t(t+1)] and stored at dqword [WK_2(t)]
# For brievity, the comments following vectored instructions only refer to
# the first of a pair of QWORDS.
# Eg. XMM4=W[t-2] really means XMM4={W[t-2]|W[t-1]}
# The computation of the message schedule and the rounds are tightly
# stitched to take advantage of instruction-level parallelism.
idx = \rnd - 2
vmovdqa W_t(idx), %xmm4 # XMM4 = W[t-2]
idx = \rnd - 15
vmovdqu W_t(idx), %xmm5 # XMM5 = W[t-15]
mov f_64, T1
vpsrlq $61, %xmm4, %xmm0 # XMM0 = W[t-2]>>61
mov e_64, tmp0
vpsrlq $1, %xmm5, %xmm6 # XMM6 = W[t-15]>>1
xor g_64, T1
RORQ tmp0, 23 # 41
vpsrlq $19, %xmm4, %xmm1 # XMM1 = W[t-2]>>19
and e_64, T1
xor e_64, tmp0
vpxor %xmm1, %xmm0, %xmm0 # XMM0 = W[t-2]>>61 ^ W[t-2]>>19
xor g_64, T1
idx = \rnd
add WK_2(idx), T1#
vpsrlq $8, %xmm5, %xmm7 # XMM7 = W[t-15]>>8
RORQ tmp0, 4 # 18
vpsrlq $6, %xmm4, %xmm2 # XMM2 = W[t-2]>>6
xor e_64, tmp0
mov a_64, T2
add h_64, T1
vpxor %xmm7, %xmm6, %xmm6 # XMM6 = W[t-15]>>1 ^ W[t-15]>>8
RORQ tmp0, 14 # 14
add tmp0, T1
vpsrlq $7, %xmm5, %xmm8 # XMM8 = W[t-15]>>7
mov a_64, tmp0
xor c_64, T2
vpsllq $(64-61), %xmm4, %xmm3 # XMM3 = W[t-2]<<3
and c_64, tmp0
and b_64, T2
vpxor %xmm3, %xmm2, %xmm2 # XMM2 = W[t-2]>>6 ^ W[t-2]<<3
xor tmp0, T2
mov a_64, tmp0
vpsllq $(64-1), %xmm5, %xmm9 # XMM9 = W[t-15]<<63
RORQ tmp0, 5 # 39
vpxor %xmm9, %xmm8, %xmm8 # XMM8 = W[t-15]>>7 ^ W[t-15]<<63
xor a_64, tmp0
add T1, d_64
RORQ tmp0, 6 # 34
xor a_64, tmp0
vpxor %xmm8, %xmm6, %xmm6 # XMM6 = W[t-15]>>1 ^ W[t-15]>>8 ^
# W[t-15]>>7 ^ W[t-15]<<63
lea (T1, T2), h_64
RORQ tmp0, 28 # 28
vpsllq $(64-19), %xmm4, %xmm4 # XMM4 = W[t-2]<<25
add tmp0, h_64
RotateState
vpxor %xmm4, %xmm0, %xmm0 # XMM0 = W[t-2]>>61 ^ W[t-2]>>19 ^
# W[t-2]<<25
mov f_64, T1
vpxor %xmm2, %xmm0, %xmm0 # XMM0 = s1(W[t-2])
mov e_64, tmp0
xor g_64, T1
idx = \rnd - 16
vpaddq W_t(idx), %xmm0, %xmm0 # XMM0 = s1(W[t-2]) + W[t-16]
idx = \rnd - 7
vmovdqu W_t(idx), %xmm1 # XMM1 = W[t-7]
RORQ tmp0, 23 # 41
and e_64, T1
xor e_64, tmp0
xor g_64, T1
vpsllq $(64-8), %xmm5, %xmm5 # XMM5 = W[t-15]<<56
idx = \rnd + 1
add WK_2(idx), T1
vpxor %xmm5, %xmm6, %xmm6 # XMM6 = s0(W[t-15])
RORQ tmp0, 4 # 18
vpaddq %xmm6, %xmm0, %xmm0 # XMM0 = s1(W[t-2]) + W[t-16] + s0(W[t-15])
xor e_64, tmp0
vpaddq %xmm1, %xmm0, %xmm0 # XMM0 = W[t] = s1(W[t-2]) + W[t-7] +
# s0(W[t-15]) + W[t-16]
mov a_64, T2
add h_64, T1
RORQ tmp0, 14 # 14
add tmp0, T1
idx = \rnd
vmovdqa %xmm0, W_t(idx) # Store W[t]
vpaddq K_t(idx), %xmm0, %xmm0 # Compute W[t]+K[t]
vmovdqa %xmm0, WK_2(idx) # Store W[t]+K[t] for next rounds
mov a_64, tmp0
xor c_64, T2
and c_64, tmp0
and b_64, T2
xor tmp0, T2
mov a_64, tmp0
RORQ tmp0, 5 # 39
xor a_64, tmp0
add T1, d_64
RORQ tmp0, 6 # 34
xor a_64, tmp0
lea (T1, T2), h_64
RORQ tmp0, 28 # 28
add tmp0, h_64
RotateState
.endm
########################################################################
# void sha512_transform_avx(const void* M, void* D, u64 L)
# Purpose: Updates the SHA512 digest stored at D with the message stored in M.
# The size of the message pointed to by M must be an integer multiple of SHA512
# message blocks.
# L is the message length in SHA512 blocks
########################################################################
ENTRY(sha512_transform_avx)
cmp $0, msglen
je nowork
# Allocate Stack Space
mov %rsp, %rax
sub $frame_size, %rsp
and $~(0x20 - 1), %rsp
mov %rax, frame_RSPSAVE(%rsp)
# Save GPRs
mov %rbx, frame_GPRSAVE(%rsp)
mov %r12, frame_GPRSAVE +8*1(%rsp)
mov %r13, frame_GPRSAVE +8*2(%rsp)
mov %r14, frame_GPRSAVE +8*3(%rsp)
mov %r15, frame_GPRSAVE +8*4(%rsp)
updateblock:
# Load state variables
mov DIGEST(0), a_64
mov DIGEST(1), b_64
mov DIGEST(2), c_64
mov DIGEST(3), d_64
mov DIGEST(4), e_64
mov DIGEST(5), f_64
mov DIGEST(6), g_64
mov DIGEST(7), h_64
t = 0
.rept 80/2 + 1
# (80 rounds) / (2 rounds/iteration) + (1 iteration)
# +1 iteration because the scheduler leads hashing by 1 iteration
.if t < 2
# BSWAP 2 QWORDS
vmovdqa XMM_QWORD_BSWAP(%rip), %xmm1
vmovdqu MSG(t), %xmm0
vpshufb %xmm1, %xmm0, %xmm0 # BSWAP
vmovdqa %xmm0, W_t(t) # Store Scheduled Pair
vpaddq K_t(t), %xmm0, %xmm0 # Compute W[t]+K[t]
vmovdqa %xmm0, WK_2(t) # Store into WK for rounds
.elseif t < 16
# BSWAP 2 QWORDS# Compute 2 Rounds
vmovdqu MSG(t), %xmm0
vpshufb %xmm1, %xmm0, %xmm0 # BSWAP
SHA512_Round t-2 # Round t-2
vmovdqa %xmm0, W_t(t) # Store Scheduled Pair
vpaddq K_t(t), %xmm0, %xmm0 # Compute W[t]+K[t]
SHA512_Round t-1 # Round t-1
vmovdqa %xmm0, WK_2(t)# Store W[t]+K[t] into WK
.elseif t < 79
# Schedule 2 QWORDS# Compute 2 Rounds
SHA512_2Sched_2Round_avx t
.else
# Compute 2 Rounds
SHA512_Round t-2
SHA512_Round t-1
.endif
t = t+2
.endr
# Update digest
add a_64, DIGEST(0)
add b_64, DIGEST(1)
add c_64, DIGEST(2)
add d_64, DIGEST(3)
add e_64, DIGEST(4)
add f_64, DIGEST(5)
add g_64, DIGEST(6)
add h_64, DIGEST(7)
# Advance to next message block
add $16*8, msg
dec msglen
jnz updateblock
# Restore GPRs
mov frame_GPRSAVE(%rsp), %rbx
mov frame_GPRSAVE +8*1(%rsp), %r12
mov frame_GPRSAVE +8*2(%rsp), %r13
mov frame_GPRSAVE +8*3(%rsp), %r14
mov frame_GPRSAVE +8*4(%rsp), %r15
# Restore Stack Pointer
mov frame_RSPSAVE(%rsp), %rsp
nowork:
ret
ENDPROC(sha512_transform_avx)
########################################################################
### Binary Data
.data
.align 16
# Mask for byte-swapping a couple of qwords in an XMM register using (v)pshufb.
XMM_QWORD_BSWAP:
.octa 0x08090a0b0c0d0e0f0001020304050607
# K[t] used in SHA512 hashing
K512:
.quad 0x428a2f98d728ae22,0x7137449123ef65cd
.quad 0xb5c0fbcfec4d3b2f,0xe9b5dba58189dbbc
.quad 0x3956c25bf348b538,0x59f111f1b605d019
.quad 0x923f82a4af194f9b,0xab1c5ed5da6d8118
.quad 0xd807aa98a3030242,0x12835b0145706fbe
.quad 0x243185be4ee4b28c,0x550c7dc3d5ffb4e2
.quad 0x72be5d74f27b896f,0x80deb1fe3b1696b1
.quad 0x9bdc06a725c71235,0xc19bf174cf692694
.quad 0xe49b69c19ef14ad2,0xefbe4786384f25e3
.quad 0x0fc19dc68b8cd5b5,0x240ca1cc77ac9c65
.quad 0x2de92c6f592b0275,0x4a7484aa6ea6e483
.quad 0x5cb0a9dcbd41fbd4,0x76f988da831153b5
.quad 0x983e5152ee66dfab,0xa831c66d2db43210
.quad 0xb00327c898fb213f,0xbf597fc7beef0ee4
.quad 0xc6e00bf33da88fc2,0xd5a79147930aa725
.quad 0x06ca6351e003826f,0x142929670a0e6e70
.quad 0x27b70a8546d22ffc,0x2e1b21385c26c926
.quad 0x4d2c6dfc5ac42aed,0x53380d139d95b3df
.quad 0x650a73548baf63de,0x766a0abb3c77b2a8
.quad 0x81c2c92e47edaee6,0x92722c851482353b
.quad 0xa2bfe8a14cf10364,0xa81a664bbc423001
.quad 0xc24b8b70d0f89791,0xc76c51a30654be30
.quad 0xd192e819d6ef5218,0xd69906245565a910
.quad 0xf40e35855771202a,0x106aa07032bbd1b8
.quad 0x19a4c116b8d2d0c8,0x1e376c085141ab53
.quad 0x2748774cdf8eeb99,0x34b0bcb5e19b48a8
.quad 0x391c0cb3c5c95a63,0x4ed8aa4ae3418acb
.quad 0x5b9cca4f7763e373,0x682e6ff3d6b2b8a3
.quad 0x748f82ee5defb2fc,0x78a5636f43172f60
.quad 0x84c87814a1f0ab72,0x8cc702081a6439ec
.quad 0x90befffa23631e28,0xa4506cebde82bde9
.quad 0xbef9a3f7b2c67915,0xc67178f2e372532b
.quad 0xca273eceea26619c,0xd186b8c721c0c207
.quad 0xeada7dd6cde0eb1e,0xf57d4f7fee6ed178
.quad 0x06f067aa72176fba,0x0a637dc5a2c898a6
.quad 0x113f9804bef90dae,0x1b710b35131c471b
.quad 0x28db77f523047d84,0x32caab7b40c72493
.quad 0x3c9ebe0a15c9bebc,0x431d67c49c100d4c
.quad 0x4cc5d4becb3e42b6,0x597f299cfc657e2a
.quad 0x5fcb6fab3ad6faec,0x6c44198c4a475817
#endif
File diff suppressed because it is too large Load Diff
-686
View File
@@ -1,686 +0,0 @@
########################################################################
# Implement fast SHA-512 with SSSE3 instructions. (x86_64)
#
# Copyright (C) 2013 Intel Corporation.
#
# Authors:
# James Guilford <james.guilford@intel.com>
# Kirk Yap <kirk.s.yap@intel.com>
# David Cote <david.m.cote@intel.com>
# Tim Chen <tim.c.chen@linux.intel.com>
#
# This software is available to you under a choice of one of two
# licenses. You may choose to be licensed under the terms of the GNU
# General Public License (GPL) Version 2, available from the file
# COPYING in the main directory of this source tree, or the
# OpenIB.org BSD license below:
#
# 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.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
# EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
# NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
# BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
# ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
#
########################################################################
#
# This code is described in an Intel White-Paper:
# "Fast SHA-512 Implementations on Intel Architecture Processors"
#
# To find it, surf to http://www.intel.com/p/en_US/embedded
# and search for that title.
#
########################################################################
# Using this part of Minio codebase under the license
# Apache License Version 2.0 with modifications
##
#ifdef HAS_SSE41
#ifndef ENTRY
#define ENTRY(name) \
.globl name ; \
.align 4,0x90 ; \
name:
#endif
#ifndef END
#define END(name) \
.size name, .-name
#endif
#ifndef ENDPROC
#define ENDPROC(name) \
.type name, @function ; \
END(name)
#endif
#define NUM_INVALID 100
#define TYPE_R32 0
#define TYPE_R64 1
#define TYPE_XMM 2
#define TYPE_INVALID 100
.macro R32_NUM opd r32
\opd = NUM_INVALID
.ifc \r32,%eax
\opd = 0
.endif
.ifc \r32,%ecx
\opd = 1
.endif
.ifc \r32,%edx
\opd = 2
.endif
.ifc \r32,%ebx
\opd = 3
.endif
.ifc \r32,%esp
\opd = 4
.endif
.ifc \r32,%ebp
\opd = 5
.endif
.ifc \r32,%esi
\opd = 6
.endif
.ifc \r32,%edi
\opd = 7
.endif
#ifdef X86_64
.ifc \r32,%r8d
\opd = 8
.endif
.ifc \r32,%r9d
\opd = 9
.endif
.ifc \r32,%r10d
\opd = 10
.endif
.ifc \r32,%r11d
\opd = 11
.endif
.ifc \r32,%r12d
\opd = 12
.endif
.ifc \r32,%r13d
\opd = 13
.endif
.ifc \r32,%r14d
\opd = 14
.endif
.ifc \r32,%r15d
\opd = 15
.endif
#endif
.endm
.macro R64_NUM opd r64
\opd = NUM_INVALID
#ifdef X86_64
.ifc \r64,%rax
\opd = 0
.endif
.ifc \r64,%rcx
\opd = 1
.endif
.ifc \r64,%rdx
\opd = 2
.endif
.ifc \r64,%rbx
\opd = 3
.endif
.ifc \r64,%rsp
\opd = 4
.endif
.ifc \r64,%rbp
\opd = 5
.endif
.ifc \r64,%rsi
\opd = 6
.endif
.ifc \r64,%rdi
\opd = 7
.endif
.ifc \r64,%r8
\opd = 8
.endif
.ifc \r64,%r9
\opd = 9
.endif
.ifc \r64,%r10
\opd = 10
.endif
.ifc \r64,%r11
\opd = 11
.endif
.ifc \r64,%r12
\opd = 12
.endif
.ifc \r64,%r13
\opd = 13
.endif
.ifc \r64,%r14
\opd = 14
.endif
.ifc \r64,%r15
\opd = 15
.endif
#endif
.endm
.macro XMM_NUM opd xmm
\opd = NUM_INVALID
.ifc \xmm,%xmm0
\opd = 0
.endif
.ifc \xmm,%xmm1
\opd = 1
.endif
.ifc \xmm,%xmm2
\opd = 2
.endif
.ifc \xmm,%xmm3
\opd = 3
.endif
.ifc \xmm,%xmm4
\opd = 4
.endif
.ifc \xmm,%xmm5
\opd = 5
.endif
.ifc \xmm,%xmm6
\opd = 6
.endif
.ifc \xmm,%xmm7
\opd = 7
.endif
.ifc \xmm,%xmm8
\opd = 8
.endif
.ifc \xmm,%xmm9
\opd = 9
.endif
.ifc \xmm,%xmm10
\opd = 10
.endif
.ifc \xmm,%xmm11
\opd = 11
.endif
.ifc \xmm,%xmm12
\opd = 12
.endif
.ifc \xmm,%xmm13
\opd = 13
.endif
.ifc \xmm,%xmm14
\opd = 14
.endif
.ifc \xmm,%xmm15
\opd = 15
.endif
.endm
.macro TYPE type reg
R32_NUM reg_type_r32 \reg
R64_NUM reg_type_r64 \reg
XMM_NUM reg_type_xmm \reg
.if reg_type_r64 <> NUM_INVALID
\type = TYPE_R64
.elseif reg_type_r32 <> NUM_INVALID
\type = TYPE_R32
.elseif reg_type_xmm <> NUM_INVALID
\type = TYPE_XMM
.else
\type = TYPE_INVALID
.endif
.endm
.macro PFX_OPD_SIZE
.byte 0x66
.endm
.macro PFX_REX opd1 opd2 W=0
.if ((\opd1 | \opd2) & 8) || \W
.byte 0x40 | ((\opd1 & 8) >> 3) | ((\opd2 & 8) >> 1) | (\W << 3)
.endif
.endm
.macro MODRM mod opd1 opd2
.byte \mod | (\opd1 & 7) | ((\opd2 & 7) << 3)
.endm
.macro PSHUFB_XMM xmm1 xmm2
XMM_NUM pshufb_opd1 \xmm1
XMM_NUM pshufb_opd2 \xmm2
PFX_OPD_SIZE
PFX_REX pshufb_opd1 pshufb_opd2
.byte 0x0f, 0x38, 0x00
MODRM 0xc0 pshufb_opd1 pshufb_opd2
.endm
.macro PCLMULQDQ imm8 xmm1 xmm2
XMM_NUM clmul_opd1 \xmm1
XMM_NUM clmul_opd2 \xmm2
PFX_OPD_SIZE
PFX_REX clmul_opd1 clmul_opd2
.byte 0x0f, 0x3a, 0x44
MODRM 0xc0 clmul_opd1 clmul_opd2
.byte \imm8
.endm
.macro PEXTRD imm8 xmm gpr
R32_NUM extrd_opd1 \gpr
XMM_NUM extrd_opd2 \xmm
PFX_OPD_SIZE
PFX_REX extrd_opd1 extrd_opd2
.byte 0x0f, 0x3a, 0x16
MODRM 0xc0 extrd_opd1 extrd_opd2
.byte \imm8
.endm
.macro MOVQ_R64_XMM opd1 opd2
TYPE movq_r64_xmm_opd1_type \opd1
.if movq_r64_xmm_opd1_type == TYPE_XMM
XMM_NUM movq_r64_xmm_opd1 \opd1
R64_NUM movq_r64_xmm_opd2 \opd2
.else
R64_NUM movq_r64_xmm_opd1 \opd1
XMM_NUM movq_r64_xmm_opd2 \opd2
.endif
PFX_OPD_SIZE
PFX_REX movq_r64_xmm_opd1 movq_r64_xmm_opd2 1
.if movq_r64_xmm_opd1_type == TYPE_XMM
.byte 0x0f, 0x7e
.else
.byte 0x0f, 0x6e
.endif
MODRM 0xc0 movq_r64_xmm_opd1 movq_r64_xmm_opd2
.endm
.text
# Virtual Registers
# ARG1
msg = %rdi
# ARG2
digest = %rsi
# ARG3
msglen = %rdx
T1 = %rcx
T2 = %r8
a_64 = %r9
b_64 = %r10
c_64 = %r11
d_64 = %r12
e_64 = %r13
f_64 = %r14
g_64 = %r15
h_64 = %rbx
tmp0 = %rax
# Local variables (stack frame)
W_SIZE = 80*8
WK_SIZE = 2*8
RSPSAVE_SIZE = 1*8
GPRSAVE_SIZE = 5*8
frame_W = 0
frame_WK = frame_W + W_SIZE
frame_RSPSAVE = frame_WK + WK_SIZE
frame_GPRSAVE = frame_RSPSAVE + RSPSAVE_SIZE
frame_size = frame_GPRSAVE + GPRSAVE_SIZE
# Useful QWORD "arrays" for simpler memory references
# MSG, DIGEST, K_t, W_t are arrays
# WK_2(t) points to 1 of 2 qwords at frame.WK depdending on t being odd/even
# Input message (arg1)
#define MSG(i) 8*i(msg)
# Output Digest (arg2)
#define DIGEST(i) 8*i(digest)
# SHA Constants (static mem)
#define K_t(i) 8*i+K512(%rip)
# Message Schedule (stack frame)
#define W_t(i) 8*i+frame_W(%rsp)
# W[t]+K[t] (stack frame)
#define WK_2(i) 8*((i%2))+frame_WK(%rsp)
.macro RotateState
# Rotate symbols a..h right
TMP = h_64
h_64 = g_64
g_64 = f_64
f_64 = e_64
e_64 = d_64
d_64 = c_64
c_64 = b_64
b_64 = a_64
a_64 = TMP
.endm
.macro SHA512_Round rnd
# Compute Round %%t
mov f_64, T1 # T1 = f
mov e_64, tmp0 # tmp = e
xor g_64, T1 # T1 = f ^ g
ror $23, tmp0 # 41 # tmp = e ror 23
and e_64, T1 # T1 = (f ^ g) & e
xor e_64, tmp0 # tmp = (e ror 23) ^ e
xor g_64, T1 # T1 = ((f ^ g) & e) ^ g = CH(e,f,g)
idx = \rnd
add WK_2(idx), T1 # W[t] + K[t] from message scheduler
ror $4, tmp0 # 18 # tmp = ((e ror 23) ^ e) ror 4
xor e_64, tmp0 # tmp = (((e ror 23) ^ e) ror 4) ^ e
mov a_64, T2 # T2 = a
add h_64, T1 # T1 = CH(e,f,g) + W[t] + K[t] + h
ror $14, tmp0 # 14 # tmp = ((((e ror23)^e)ror4)^e)ror14 = S1(e)
add tmp0, T1 # T1 = CH(e,f,g) + W[t] + K[t] + S1(e)
mov a_64, tmp0 # tmp = a
xor c_64, T2 # T2 = a ^ c
and c_64, tmp0 # tmp = a & c
and b_64, T2 # T2 = (a ^ c) & b
xor tmp0, T2 # T2 = ((a ^ c) & b) ^ (a & c) = Maj(a,b,c)
mov a_64, tmp0 # tmp = a
ror $5, tmp0 # 39 # tmp = a ror 5
xor a_64, tmp0 # tmp = (a ror 5) ^ a
add T1, d_64 # e(next_state) = d + T1
ror $6, tmp0 # 34 # tmp = ((a ror 5) ^ a) ror 6
xor a_64, tmp0 # tmp = (((a ror 5) ^ a) ror 6) ^ a
lea (T1, T2), h_64 # a(next_state) = T1 + Maj(a,b,c)
ror $28, tmp0 # 28 # tmp = ((((a ror5)^a)ror6)^a)ror28 = S0(a)
add tmp0, h_64 # a(next_state) = T1 + Maj(a,b,c) S0(a)
RotateState
.endm
.macro SHA512_2Sched_2Round_sse rnd
# Compute rounds t-2 and t-1
# Compute message schedule QWORDS t and t+1
# Two rounds are computed based on the values for K[t-2]+W[t-2] and
# K[t-1]+W[t-1] which were previously stored at WK_2 by the message
# scheduler.
# The two new schedule QWORDS are stored at [W_t(%%t)] and [W_t(%%t+1)].
# They are then added to their respective SHA512 constants at
# [K_t(%%t)] and [K_t(%%t+1)] and stored at dqword [WK_2(%%t)]
# For brievity, the comments following vectored instructions only refer to
# the first of a pair of QWORDS.
# Eg. XMM2=W[t-2] really means XMM2={W[t-2]|W[t-1]}
# The computation of the message schedule and the rounds are tightly
# stitched to take advantage of instruction-level parallelism.
# For clarity, integer instructions (for the rounds calculation) are indented
# by one tab. Vectored instructions (for the message scheduler) are indented
# by two tabs.
mov f_64, T1
idx = \rnd -2
movdqa W_t(idx), %xmm2 # XMM2 = W[t-2]
xor g_64, T1
and e_64, T1
movdqa %xmm2, %xmm0 # XMM0 = W[t-2]
xor g_64, T1
idx = \rnd
add WK_2(idx), T1
idx = \rnd - 15
movdqu W_t(idx), %xmm5 # XMM5 = W[t-15]
mov e_64, tmp0
ror $23, tmp0 # 41
movdqa %xmm5, %xmm3 # XMM3 = W[t-15]
xor e_64, tmp0
ror $4, tmp0 # 18
psrlq $61-19, %xmm0 # XMM0 = W[t-2] >> 42
xor e_64, tmp0
ror $14, tmp0 # 14
psrlq $(8-7), %xmm3 # XMM3 = W[t-15] >> 1
add tmp0, T1
add h_64, T1
pxor %xmm2, %xmm0 # XMM0 = (W[t-2] >> 42) ^ W[t-2]
mov a_64, T2
xor c_64, T2
pxor %xmm5, %xmm3 # XMM3 = (W[t-15] >> 1) ^ W[t-15]
and b_64, T2
mov a_64, tmp0
psrlq $(19-6), %xmm0 # XMM0 = ((W[t-2]>>42)^W[t-2])>>13
and c_64, tmp0
xor tmp0, T2
psrlq $(7-1), %xmm3 # XMM3 = ((W[t-15]>>1)^W[t-15])>>6
mov a_64, tmp0
ror $5, tmp0 # 39
pxor %xmm2, %xmm0 # XMM0 = (((W[t-2]>>42)^W[t-2])>>13)^W[t-2]
xor a_64, tmp0
ror $6, tmp0 # 34
pxor %xmm5, %xmm3 # XMM3 = (((W[t-15]>>1)^W[t-15])>>6)^W[t-15]
xor a_64, tmp0
ror $28, tmp0 # 28
psrlq $6, %xmm0 # XMM0 = ((((W[t-2]>>42)^W[t-2])>>13)^W[t-2])>>6
add tmp0, T2
add T1, d_64
psrlq $1, %xmm3 # XMM3 = (((W[t-15]>>1)^W[t-15])>>6)^W[t-15]>>1
lea (T1, T2), h_64
RotateState
movdqa %xmm2, %xmm1 # XMM1 = W[t-2]
mov f_64, T1
xor g_64, T1
movdqa %xmm5, %xmm4 # XMM4 = W[t-15]
and e_64, T1
xor g_64, T1
psllq $(64-19)-(64-61) , %xmm1 # XMM1 = W[t-2] << 42
idx = \rnd + 1
add WK_2(idx), T1
mov e_64, tmp0
psllq $(64-1)-(64-8), %xmm4 # XMM4 = W[t-15] << 7
ror $23, tmp0 # 41
xor e_64, tmp0
pxor %xmm2, %xmm1 # XMM1 = (W[t-2] << 42)^W[t-2]
ror $4, tmp0 # 18
xor e_64, tmp0
pxor %xmm5, %xmm4 # XMM4 = (W[t-15]<<7)^W[t-15]
ror $14, tmp0 # 14
add tmp0, T1
psllq $(64-61), %xmm1 # XMM1 = ((W[t-2] << 42)^W[t-2])<<3
add h_64, T1
mov a_64, T2
psllq $(64-8), %xmm4 # XMM4 = ((W[t-15]<<7)^W[t-15])<<56
xor c_64, T2
and b_64, T2
pxor %xmm1, %xmm0 # XMM0 = s1(W[t-2])
mov a_64, tmp0
and c_64, tmp0
idx = \rnd - 7
movdqu W_t(idx), %xmm1 # XMM1 = W[t-7]
xor tmp0, T2
pxor %xmm4, %xmm3 # XMM3 = s0(W[t-15])
mov a_64, tmp0
paddq %xmm3, %xmm0 # XMM0 = s1(W[t-2]) + s0(W[t-15])
ror $5, tmp0 # 39
idx =\rnd-16
paddq W_t(idx), %xmm0 # XMM0 = s1(W[t-2]) + s0(W[t-15]) + W[t-16]
xor a_64, tmp0
paddq %xmm1, %xmm0 # XMM0 = s1(W[t-2]) + W[t-7] + s0(W[t-15]) + W[t-16]
ror $6, tmp0 # 34
movdqa %xmm0, W_t(\rnd) # Store scheduled qwords
xor a_64, tmp0
paddq K_t(\rnd), %xmm0 # Compute W[t]+K[t]
ror $28, tmp0 # 28
idx = \rnd
movdqa %xmm0, WK_2(idx) # Store W[t]+K[t] for next rounds
add tmp0, T2
add T1, d_64
lea (T1, T2), h_64
RotateState
.endm
########################################################################
# void sha512_transform_ssse3(const void* M, void* D, u64 L)#
# Purpose: Updates the SHA512 digest stored at D with the message stored in M.
# The size of the message pointed to by M must be an integer multiple of SHA512
# message blocks.
# L is the message length in SHA512 blocks.
########################################################################
ENTRY(sha512_transform_ssse3)
cmp $0, msglen
je nowork
# Allocate Stack Space
mov %rsp, %rax
sub $frame_size, %rsp
and $~(0x20 - 1), %rsp
mov %rax, frame_RSPSAVE(%rsp)
# Save GPRs
mov %rbx, frame_GPRSAVE(%rsp)
mov %r12, frame_GPRSAVE +8*1(%rsp)
mov %r13, frame_GPRSAVE +8*2(%rsp)
mov %r14, frame_GPRSAVE +8*3(%rsp)
mov %r15, frame_GPRSAVE +8*4(%rsp)
updateblock:
# Load state variables
mov DIGEST(0), a_64
mov DIGEST(1), b_64
mov DIGEST(2), c_64
mov DIGEST(3), d_64
mov DIGEST(4), e_64
mov DIGEST(5), f_64
mov DIGEST(6), g_64
mov DIGEST(7), h_64
t = 0
.rept 80/2 + 1
# (80 rounds) / (2 rounds/iteration) + (1 iteration)
# +1 iteration because the scheduler leads hashing by 1 iteration
.if t < 2
# BSWAP 2 QWORDS
movdqa XMM_QWORD_BSWAP(%rip), %xmm1
movdqu MSG(t), %xmm0
pshufb %xmm1, %xmm0 # BSWAP
movdqa %xmm0, W_t(t) # Store Scheduled Pair
paddq K_t(t), %xmm0 # Compute W[t]+K[t]
movdqa %xmm0, WK_2(t) # Store into WK for rounds
.elseif t < 16
# BSWAP 2 QWORDS# Compute 2 Rounds
movdqu MSG(t), %xmm0
pshufb %xmm1, %xmm0 # BSWAP
SHA512_Round t-2 # Round t-2
movdqa %xmm0, W_t(t) # Store Scheduled Pair
paddq K_t(t), %xmm0 # Compute W[t]+K[t]
SHA512_Round t-1 # Round t-1
movdqa %xmm0, WK_2(t) # Store W[t]+K[t] into WK
.elseif t < 79
# Schedule 2 QWORDS# Compute 2 Rounds
SHA512_2Sched_2Round_sse t
.else
# Compute 2 Rounds
SHA512_Round t-2
SHA512_Round t-1
.endif
t = t+2
.endr
# Update digest
add a_64, DIGEST(0)
add b_64, DIGEST(1)
add c_64, DIGEST(2)
add d_64, DIGEST(3)
add e_64, DIGEST(4)
add f_64, DIGEST(5)
add g_64, DIGEST(6)
add h_64, DIGEST(7)
# Advance to next message block
add $16*8, msg
dec msglen
jnz updateblock
# Restore GPRs
mov frame_GPRSAVE(%rsp), %rbx
mov frame_GPRSAVE +8*1(%rsp), %r12
mov frame_GPRSAVE +8*2(%rsp), %r13
mov frame_GPRSAVE +8*3(%rsp), %r14
mov frame_GPRSAVE +8*4(%rsp), %r15
# Restore Stack Pointer
mov frame_RSPSAVE(%rsp), %rsp
nowork:
ret
ENDPROC(sha512_transform_ssse3)
########################################################################
### Binary Data
.data
.align 16
# Mask for byte-swapping a couple of qwords in an XMM register using (v)pshufb.
XMM_QWORD_BSWAP:
.octa 0x08090a0b0c0d0e0f0001020304050607
# K[t] used in SHA512 hashing
K512:
.quad 0x428a2f98d728ae22,0x7137449123ef65cd
.quad 0xb5c0fbcfec4d3b2f,0xe9b5dba58189dbbc
.quad 0x3956c25bf348b538,0x59f111f1b605d019
.quad 0x923f82a4af194f9b,0xab1c5ed5da6d8118
.quad 0xd807aa98a3030242,0x12835b0145706fbe
.quad 0x243185be4ee4b28c,0x550c7dc3d5ffb4e2
.quad 0x72be5d74f27b896f,0x80deb1fe3b1696b1
.quad 0x9bdc06a725c71235,0xc19bf174cf692694
.quad 0xe49b69c19ef14ad2,0xefbe4786384f25e3
.quad 0x0fc19dc68b8cd5b5,0x240ca1cc77ac9c65
.quad 0x2de92c6f592b0275,0x4a7484aa6ea6e483
.quad 0x5cb0a9dcbd41fbd4,0x76f988da831153b5
.quad 0x983e5152ee66dfab,0xa831c66d2db43210
.quad 0xb00327c898fb213f,0xbf597fc7beef0ee4
.quad 0xc6e00bf33da88fc2,0xd5a79147930aa725
.quad 0x06ca6351e003826f,0x142929670a0e6e70
.quad 0x27b70a8546d22ffc,0x2e1b21385c26c926
.quad 0x4d2c6dfc5ac42aed,0x53380d139d95b3df
.quad 0x650a73548baf63de,0x766a0abb3c77b2a8
.quad 0x81c2c92e47edaee6,0x92722c851482353b
.quad 0xa2bfe8a14cf10364,0xa81a664bbc423001
.quad 0xc24b8b70d0f89791,0xc76c51a30654be30
.quad 0xd192e819d6ef5218,0xd69906245565a910
.quad 0xf40e35855771202a,0x106aa07032bbd1b8
.quad 0x19a4c116b8d2d0c8,0x1e376c085141ab53
.quad 0x2748774cdf8eeb99,0x34b0bcb5e19b48a8
.quad 0x391c0cb3c5c95a63,0x4ed8aa4ae3418acb
.quad 0x5b9cca4f7763e373,0x682e6ff3d6b2b8a3
.quad 0x748f82ee5defb2fc,0x78a5636f43172f60
.quad 0x84c87814a1f0ab72,0x8cc702081a6439ec
.quad 0x90befffa23631e28,0xa4506cebde82bde9
.quad 0xbef9a3f7b2c67915,0xc67178f2e372532b
.quad 0xca273eceea26619c,0xd186b8c721c0c207
.quad 0xeada7dd6cde0eb1e,0xf57d4f7fee6ed178
.quad 0x06f067aa72176fba,0x0a637dc5a2c898a6
.quad 0x113f9804bef90dae,0x1b710b35131c471b
.quad 0x28db77f523047d84,0x32caab7b40c72493
.quad 0x3c9ebe0a15c9bebc,0x431d67c49c100d4c
.quad 0x4cc5d4becb3e42b6,0x597f299cfc657e2a
.quad 0x5fcb6fab3ad6faec,0x6c44198c4a475817
#endif
-58
View File
@@ -1,58 +0,0 @@
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package sha512
import (
"hash"
"io"
"crypto/sha512"
)
// The size of a SHA512 checksum in bytes.
const (
Size = sha512.Size
)
// Sum512 - single caller sha512 helper
func Sum512(data []byte) []byte {
d := sha512.New()
d.Write(data)
return d.Sum(nil)
}
// Sum - io.Reader based streaming sha512 helper
func Sum(reader io.Reader) ([]byte, error) {
d := sha512.New()
var err error
for err == nil {
length := 0
byteBuffer := make([]byte, 1024*1024)
length, err = reader.Read(byteBuffer)
byteBuffer = byteBuffer[0:length]
d.Write(byteBuffer)
}
if err != io.EOF {
return nil, err
}
return d.Sum(nil), nil
}
// New returns a new hash.Hash computing SHA512.
func New() hash.Hash {
return sha512.New()
}
-184
View File
@@ -1,184 +0,0 @@
// Copyright 2009 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 of
// Golang project:
// https://github.com/golang/go/blob/master/LICENSE
// Using this part of Minio codebase under the license
// Apache License Version 2.0 with modifications
// Package sha512 implements the SHA512 hash algorithms as defined
// in FIPS 180-2.
package sha512
import (
"hash"
"io"
"github.com/minio/minio/pkg/cpu"
)
// The size of a SHA512 checksum in bytes.
const Size = 64
// The blocksize of SHA512 in bytes.
const BlockSize = 128
const (
chunk = 128
init0 = 0x6a09e667f3bcc908
init1 = 0xbb67ae8584caa73b
init2 = 0x3c6ef372fe94f82b
init3 = 0xa54ff53a5f1d36f1
init4 = 0x510e527fade682d1
init5 = 0x9b05688c2b3e6c1f
init6 = 0x1f83d9abfb41bd6b
init7 = 0x5be0cd19137e2179
)
// digest represents the partial evaluation of a checksum.
type digest struct {
h [8]uint64
x [chunk]byte
nx int
len uint64
}
func block(dig *digest, p []byte) {
switch true {
case cpu.HasAVX2() == true:
blockAVX2(dig, p)
case cpu.HasAVX() == true:
blockAVX(dig, p)
case cpu.HasSSE41() == true:
blockSSE(dig, p)
default:
blockGeneric(dig, p)
}
}
// Reset digest to its default value
func (d *digest) Reset() {
d.h[0] = init0
d.h[1] = init1
d.h[2] = init2
d.h[3] = init3
d.h[4] = init4
d.h[5] = init5
d.h[6] = init6
d.h[7] = init7
d.nx = 0
d.len = 0
}
// New returns a new hash.Hash computing the SHA512 checksum.
func New() hash.Hash {
d := new(digest)
d.Reset()
return d
}
// Return output array byte size
func (d *digest) Size() int { return Size }
// Return blockSize
func (d *digest) BlockSize() int { return BlockSize }
// Write blocks
func (d *digest) Write(p []byte) (nn int, err error) {
nn = len(p)
d.len += uint64(nn)
if d.nx > 0 {
n := copy(d.x[d.nx:], p)
d.nx += n
if d.nx == chunk {
block(d, d.x[:])
d.nx = 0
}
p = p[n:]
}
if len(p) >= chunk {
n := len(p) &^ (chunk - 1)
block(d, p[:n])
p = p[n:]
}
if len(p) > 0 {
d.nx = copy(d.x[:], p)
}
return
}
// Calculate sha512
func (d *digest) Sum(in []byte) []byte {
// Make a copy of d0 so that caller can keep writing and summing.
d0 := *d
hash := d0.checkSum()
return append(in, hash[:]...)
}
// internal checksum calculation, returns [Size]byte
func (d *digest) checkSum() [Size]byte {
// Padding. Add a 1 bit and 0 bits until 112 bytes mod 128.
len := d.len
var tmp [128]byte
tmp[0] = 0x80
if len%128 < 112 {
d.Write(tmp[0 : 112-len%128])
} else {
d.Write(tmp[0 : 128+112-len%128])
}
// Length in bits.
len <<= 3
for i := uint(0); i < 16; i++ {
tmp[i] = byte(len >> (120 - 8*i))
}
d.Write(tmp[0:16])
if d.nx != 0 {
panic("d.nx != 0")
}
h := d.h[:]
var digest [Size]byte
for i, s := range h {
digest[i*8] = byte(s >> 56)
digest[i*8+1] = byte(s >> 48)
digest[i*8+2] = byte(s >> 40)
digest[i*8+3] = byte(s >> 32)
digest[i*8+4] = byte(s >> 24)
digest[i*8+5] = byte(s >> 16)
digest[i*8+6] = byte(s >> 8)
digest[i*8+7] = byte(s)
}
return digest
}
/// Convenience functions
// Sum512 - single caller sha512 helper
func Sum512(data []byte) []byte {
var d digest
d.Reset()
d.Write(data)
return d.Sum(nil)
}
// Sum - io.Reader based streaming sha512 helper
func Sum(reader io.Reader) ([]byte, error) {
h := New()
var err error
for err == nil {
length := 0
byteBuffer := make([]byte, 1024*1024)
length, err = reader.Read(byteBuffer)
byteBuffer = byteBuffer[0:length]
h.Write(byteBuffer)
}
if err != io.EOF {
return nil, err
}
return h.Sum(nil), nil
}
-125
View File
@@ -1,125 +0,0 @@
// Copyright 2009 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 of
// Golang project:
// https://github.com/golang/go/blob/master/LICENSE
// Using this part of Minio codebase under the license
// Apache License Version 2.0 with modifications
// SHA512 hash algorithm. See FIPS 180-2.
package sha512
import (
"fmt"
"io"
"testing"
)
type sha512Test struct {
out string
in string
}
var golden = []sha512Test{
{"cf83e1357eefb8bdf1542850d66d8007d620e4050b5715dc83f4a921d36ce9ce47d0d13c5d85f2b0ff8318d2877eec2f63b931bd47417a81a538327af927da3e", ""},
{"1f40fc92da241694750979ee6cf582f2d5d7d28e18335de05abc54d0560e0f5302860c652bf08d560252aa5e74210546f369fbbbce8c12cfc7957b2652fe9a75", "a"},
{"2d408a0717ec188158278a796c689044361dc6fdde28d6f04973b80896e1823975cdbf12eb63f9e0591328ee235d80e9b5bf1aa6a44f4617ff3caf6400eb172d", "ab"},
{"ddaf35a193617abacc417349ae20413112e6fa4e89a97ea20a9eeee64b55d39a2192992a274fc1a836ba3c23a3feebbd454d4423643ce80e2a9ac94fa54ca49f", "abc"},
{"d8022f2060ad6efd297ab73dcc5355c9b214054b0d1776a136a669d26a7d3b14f73aa0d0ebff19ee333368f0164b6419a96da49e3e481753e7e96b716bdccb6f", "abcd"},
{"878ae65a92e86cac011a570d4c30a7eaec442b85ce8eca0c2952b5e3cc0628c2e79d889ad4d5c7c626986d452dd86374b6ffaa7cd8b67665bef2289a5c70b0a1", "abcde"},
{"e32ef19623e8ed9d267f657a81944b3d07adbb768518068e88435745564e8d4150a0a703be2a7d88b61e3d390c2bb97e2d4c311fdc69d6b1267f05f59aa920e7", "abcdef"},
{"d716a4188569b68ab1b6dfac178e570114cdf0ea3a1cc0e31486c3e41241bc6a76424e8c37ab26f096fc85ef9886c8cb634187f4fddff645fb099f1ff54c6b8c", "abcdefg"},
{"a3a8c81bc97c2560010d7389bc88aac974a104e0e2381220c6e084c4dccd1d2d17d4f86db31c2a851dc80e6681d74733c55dcd03dd96f6062cdda12a291ae6ce", "abcdefgh"},
{"f22d51d25292ca1d0f68f69aedc7897019308cc9db46efb75a03dd494fc7f126c010e8ade6a00a0c1a5f1b75d81e0ed5a93ce98dc9b833db7839247b1d9c24fe", "abcdefghi"},
{"ef6b97321f34b1fea2169a7db9e1960b471aa13302a988087357c520be957ca119c3ba68e6b4982c019ec89de3865ccf6a3cda1fe11e59f98d99f1502c8b9745", "abcdefghij"},
{"2210d99af9c8bdecda1b4beff822136753d8342505ddce37f1314e2cdbb488c6016bdaa9bd2ffa513dd5de2e4b50f031393d8ab61f773b0e0130d7381e0f8a1d", "Discard medicine more than two years old."},
{"a687a8985b4d8d0a24f115fe272255c6afaf3909225838546159c1ed685c211a203796ae8ecc4c81a5b6315919b3a64f10713da07e341fcdbb08541bf03066ce", "He who has a shady past knows that nice guys finish last."},
{"8ddb0392e818b7d585ab22769a50df660d9f6d559cca3afc5691b8ca91b8451374e42bcdabd64589ed7c91d85f626596228a5c8572677eb98bc6b624befb7af8", "I wouldn't marry him with a ten foot pole."},
{"26ed8f6ca7f8d44b6a8a54ae39640fa8ad5c673f70ee9ce074ba4ef0d483eea00bab2f61d8695d6b34df9c6c48ae36246362200ed820448bdc03a720366a87c6", "Free! Free!/A trip/to Mars/for 900/empty jars/Burma Shave"},
{"e5a14bf044be69615aade89afcf1ab0389d5fc302a884d403579d1386a2400c089b0dbb387ed0f463f9ee342f8244d5a38cfbc0e819da9529fbff78368c9a982", "The days of the digital watch are numbered. -Tom Stoppard"},
{"420a1faa48919e14651bed45725abe0f7a58e0f099424c4e5a49194946e38b46c1f8034b18ef169b2e31050d1648e0b982386595f7df47da4b6fd18e55333015", "Nepal premier won't resign."},
{"d926a863beadb20134db07683535c72007b0e695045876254f341ddcccde132a908c5af57baa6a6a9c63e6649bba0c213dc05fadcf9abccea09f23dcfb637fbe", "For every action there is an equal and opposite government program."},
{"9a98dd9bb67d0da7bf83da5313dff4fd60a4bac0094f1b05633690ffa7f6d61de9a1d4f8617937d560833a9aaa9ccafe3fd24db418d0e728833545cadd3ad92d", "His money is twice tainted: 'taint yours and 'taint mine."},
{"d7fde2d2351efade52f4211d3746a0780a26eec3df9b2ed575368a8a1c09ec452402293a8ea4eceb5a4f60064ea29b13cdd86918cd7a4faf366160b009804107", "There is no reason for any individual to have a computer in their home. -Ken Olsen, 1977"},
{"b0f35ffa2697359c33a56f5c0cf715c7aeed96da9905ca2698acadb08fbc9e669bf566b6bd5d61a3e86dc22999bcc9f2224e33d1d4f32a228cf9d0349e2db518", "It's a tiny change to the code and not completely disgusting. - Bob Manchek"},
{"3d2e5f91778c9e66f7e061293aaa8a8fc742dd3b2e4f483772464b1144189b49273e610e5cccd7a81a19ca1fa70f16b10f1a100a4d8c1372336be8484c64b311", "size: a.out: bad magic"},
{"b2f68ff58ac015efb1c94c908b0d8c2bf06f491e4de8e6302c49016f7f8a33eac3e959856c7fddbc464de618701338a4b46f76dbfaf9a1e5262b5f40639771c7", "The major problem is with sendmail. -Mark Horton"},
{"d8c92db5fdf52cf8215e4df3b4909d29203ff4d00e9ad0b64a6a4e04dec5e74f62e7c35c7fb881bd5de95442123df8f57a489b0ae616bd326f84d10021121c57", "Give me a rock, paper and scissors and I will move the world. CCFestoon"},
{"19a9f8dc0a233e464e8566ad3ca9b91e459a7b8c4780985b015776e1bf239a19bc233d0556343e2b0a9bc220900b4ebf4f8bdf89ff8efeaf79602d6849e6f72e", "If the enemy is within range, then so are you."},
{"00b4c41f307bde87301cdc5b5ab1ae9a592e8ecbb2021dd7bc4b34e2ace60741cc362560bec566ba35178595a91932b8d5357e2c9cec92d393b0fa7831852476", "It's well we cannot hear the screams/That we create in others' dreams."},
{"91eccc3d5375fd026e4d6787874b1dce201cecd8a27dbded5065728cb2d09c58a3d467bb1faf353bf7ba567e005245d5321b55bc344f7c07b91cb6f26c959be7", "You remind me of a TV show, but that's all right: I watch it anyway."},
{"fabbbe22180f1f137cfdc9556d2570e775d1ae02a597ded43a72a40f9b485d500043b7be128fb9fcd982b83159a0d99aa855a9e7cc4240c00dc01a9bdf8218d7", "C is as portable as Stonehedge!!"},
{"2ecdec235c1fa4fc2a154d8fba1dddb8a72a1ad73838b51d792331d143f8b96a9f6fcb0f34d7caa351fe6d88771c4f105040e0392f06e0621689d33b2f3ba92e", "Even if I could be Shakespeare, I think I should still choose to be Faraday. - A. Huxley"},
{"7ad681f6f96f82f7abfa7ecc0334e8fa16d3dc1cdc45b60b7af43fe4075d2357c0c1d60e98350f1afb1f2fe7a4d7cd2ad55b88e458e06b73c40b437331f5dab4", "The fugacity of a constituent in a mixture of gases at a given temperature is proportional to its mole fraction. Lewis-Randall Rule"},
{"833f9248ab4a3b9e5131f745fda1ffd2dd435b30e965957e78291c7ab73605fd1912b0794e5c233ab0a12d205a39778d19b83515d6a47003f19cdee51d98c7e0", "How can you write a big system without C++? -Paul Glick"},
}
func TestGolden(t *testing.T) {
for i := 0; i < len(golden); i++ {
g := golden[i]
s := fmt.Sprintf("%x", Sum512([]byte(g.in)))
if s != g.out {
t.Fatalf("Sum512 function: sha512(%s) = %s want %s", g.in, s, g.out)
}
c := New()
for j := 0; j < 3; j++ {
if j < 2 {
io.WriteString(c, g.in)
} else {
io.WriteString(c, g.in[0:len(g.in)/2])
c.Sum(nil)
io.WriteString(c, g.in[len(g.in)/2:])
}
s := fmt.Sprintf("%x", c.Sum(nil))
if s != g.out {
t.Fatalf("sha512[%d](%s) = %s want %s", j, g.in, s, g.out)
}
c.Reset()
}
}
}
func TestSize(t *testing.T) {
c := New()
if got := c.Size(); got != Size {
t.Errorf("Size = %d; want %d", got, Size)
}
}
func TestBlockSize(t *testing.T) {
c := New()
if got := c.BlockSize(); got != BlockSize {
t.Errorf("BlockSize = %d; want %d", got, BlockSize)
}
}
var bench = New()
var buf = make([]byte, 1024*1024)
func benchmarkSize(b *testing.B, size int) {
b.SetBytes(int64(size))
sum := make([]byte, bench.Size())
for i := 0; i < b.N; i++ {
bench.Reset()
bench.Write(buf[:size])
bench.Sum(sum[:0])
}
}
func BenchmarkHash8Bytes(b *testing.B) {
benchmarkSize(b, 8)
}
func BenchmarkHash1K(b *testing.B) {
benchmarkSize(b, 1024)
}
func BenchmarkHash8K(b *testing.B) {
benchmarkSize(b, 8192)
}
func BenchmarkHash1M(b *testing.B) {
benchmarkSize(b, 1024*1024)
}
-58
View File
@@ -1,58 +0,0 @@
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package sha512
import (
"hash"
"io"
"crypto/sha512"
)
// The size of a SHA512 checksum in bytes.
const (
Size = sha512.Size
)
// Sum512 - single caller sha512 helper
func Sum512(data []byte) []byte {
d := sha512.New()
d.Write(data)
return d.Sum(nil)
}
// Sum - io.Reader based streaming sha512 helper
func Sum(reader io.Reader) ([]byte, error) {
d := sha512.New()
var err error
for err == nil {
length := 0
byteBuffer := make([]byte, 1024*1024)
length, err = reader.Read(byteBuffer)
byteBuffer = byteBuffer[0:length]
d.Write(byteBuffer)
}
if err != io.EOF {
return nil, err
}
return d.Sum(nil), nil
}
// New returns a new hash.Hash computing SHA512.
func New() hash.Hash {
return sha512.New()
}
-180
View File
@@ -1,180 +0,0 @@
// +build amd64
//
// Minio Cloud Storage, (C) 2015 Minio, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
// Software block transform are provided by The Go Authors:
// Copyright 2009 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 of
// Golang project:
// https://github.com/golang/go/blob/master/LICENSE
package sha512
// #cgo CFLAGS: -DHAS_SSE41 -DHAS_AVX -DHAS_AVX2
// #include <stdint.h>
// void sha512_transform_ssse3 (const void* M, void* D, uint64_t L);
// void sha512_transform_avx (const void* M, void* D, uint64_t L);
// void sha512_transform_rorx (const void* M, void* D, uint64_t L);
import "C"
import (
"unsafe"
)
func blockSSE(dig *digest, p []byte) {
C.sha512_transform_ssse3(unsafe.Pointer(&p[0]), unsafe.Pointer(&dig.h[0]), (C.uint64_t)(len(p)/chunk))
}
func blockAVX(dig *digest, p []byte) {
C.sha512_transform_avx(unsafe.Pointer(&p[0]), unsafe.Pointer(&dig.h[0]), (C.uint64_t)(len(p)/chunk))
}
func blockAVX2(dig *digest, p []byte) {
C.sha512_transform_rorx(unsafe.Pointer(&p[0]), unsafe.Pointer(&dig.h[0]), (C.uint64_t)(len(p)/chunk))
}
func blockGeneric(dig *digest, p []byte) {
var w [80]uint64
h0, h1, h2, h3, h4, h5, h6, h7 := dig.h[0], dig.h[1], dig.h[2], dig.h[3], dig.h[4], dig.h[5], dig.h[6], dig.h[7]
for len(p) >= chunk {
for i := 0; i < 16; i++ {
j := i * 8
w[i] = uint64(p[j])<<56 | uint64(p[j+1])<<48 | uint64(p[j+2])<<40 | uint64(p[j+3])<<32 |
uint64(p[j+4])<<24 | uint64(p[j+5])<<16 | uint64(p[j+6])<<8 | uint64(p[j+7])
}
for i := 16; i < 80; i++ {
v1 := w[i-2]
t1 := (v1>>19 | v1<<(64-19)) ^ (v1>>61 | v1<<(64-61)) ^ (v1 >> 6)
v2 := w[i-15]
t2 := (v2>>1 | v2<<(64-1)) ^ (v2>>8 | v2<<(64-8)) ^ (v2 >> 7)
w[i] = t1 + w[i-7] + t2 + w[i-16]
}
a, b, c, d, e, f, g, h := h0, h1, h2, h3, h4, h5, h6, h7
for i := 0; i < 80; i++ {
t1 := h + ((e>>14 | e<<(64-14)) ^ (e>>18 | e<<(64-18)) ^ (e>>41 | e<<(64-41))) + ((e & f) ^ (^e & g)) + _K[i] + w[i]
t2 := ((a>>28 | a<<(64-28)) ^ (a>>34 | a<<(64-34)) ^ (a>>39 | a<<(64-39))) + ((a & b) ^ (a & c) ^ (b & c))
h = g
g = f
f = e
e = d + t1
d = c
c = b
b = a
a = t1 + t2
}
h0 += a
h1 += b
h2 += c
h3 += d
h4 += e
h5 += f
h6 += g
h7 += h
p = p[chunk:]
}
dig.h[0], dig.h[1], dig.h[2], dig.h[3], dig.h[4], dig.h[5], dig.h[6], dig.h[7] = h0, h1, h2, h3, h4, h5, h6, h7
}
var _K = []uint64{
0x428a2f98d728ae22,
0x7137449123ef65cd,
0xb5c0fbcfec4d3b2f,
0xe9b5dba58189dbbc,
0x3956c25bf348b538,
0x59f111f1b605d019,
0x923f82a4af194f9b,
0xab1c5ed5da6d8118,
0xd807aa98a3030242,
0x12835b0145706fbe,
0x243185be4ee4b28c,
0x550c7dc3d5ffb4e2,
0x72be5d74f27b896f,
0x80deb1fe3b1696b1,
0x9bdc06a725c71235,
0xc19bf174cf692694,
0xe49b69c19ef14ad2,
0xefbe4786384f25e3,
0x0fc19dc68b8cd5b5,
0x240ca1cc77ac9c65,
0x2de92c6f592b0275,
0x4a7484aa6ea6e483,
0x5cb0a9dcbd41fbd4,
0x76f988da831153b5,
0x983e5152ee66dfab,
0xa831c66d2db43210,
0xb00327c898fb213f,
0xbf597fc7beef0ee4,
0xc6e00bf33da88fc2,
0xd5a79147930aa725,
0x06ca6351e003826f,
0x142929670a0e6e70,
0x27b70a8546d22ffc,
0x2e1b21385c26c926,
0x4d2c6dfc5ac42aed,
0x53380d139d95b3df,
0x650a73548baf63de,
0x766a0abb3c77b2a8,
0x81c2c92e47edaee6,
0x92722c851482353b,
0xa2bfe8a14cf10364,
0xa81a664bbc423001,
0xc24b8b70d0f89791,
0xc76c51a30654be30,
0xd192e819d6ef5218,
0xd69906245565a910,
0xf40e35855771202a,
0x106aa07032bbd1b8,
0x19a4c116b8d2d0c8,
0x1e376c085141ab53,
0x2748774cdf8eeb99,
0x34b0bcb5e19b48a8,
0x391c0cb3c5c95a63,
0x4ed8aa4ae3418acb,
0x5b9cca4f7763e373,
0x682e6ff3d6b2b8a3,
0x748f82ee5defb2fc,
0x78a5636f43172f60,
0x84c87814a1f0ab72,
0x8cc702081a6439ec,
0x90befffa23631e28,
0xa4506cebde82bde9,
0xbef9a3f7b2c67915,
0xc67178f2e372532b,
0xca273eceea26619c,
0xd186b8c721c0c207,
0xeada7dd6cde0eb1e,
0xf57d4f7fee6ed178,
0x06f067aa72176fba,
0x0a637dc5a2c898a6,
0x113f9804bef90dae,
0x1b710b35131c471b,
0x28db77f523047d84,
0x32caab7b40c72493,
0x3c9ebe0a15c9bebc,
0x431d67c49c100d4c,
0x4cc5d4becb3e42b6,
0x597f299cfc657e2a,
0x5fcb6fab3ad6faec,
0x6c44198c4a475817,
}
-202
View File
@@ -1,202 +0,0 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-3
View File
@@ -1,3 +0,0 @@
# Donut
donut - Donut (do not delete) on disk format implementation released under [Apache license v2](./LICENSE).
-47
View File
@@ -1,47 +0,0 @@
package donut
// BucketACL - bucket level access control
type BucketACL string
// different types of ACL's currently supported for buckets
const (
BucketPrivate = BucketACL("private")
BucketPublicRead = BucketACL("public-read")
BucketPublicReadWrite = BucketACL("public-read-write")
)
func (b BucketACL) String() string {
return string(b)
}
// IsPrivate - is acl Private
func (b BucketACL) IsPrivate() bool {
return b == BucketACL("private")
}
// IsPublicRead - is acl PublicRead
func (b BucketACL) IsPublicRead() bool {
return b == BucketACL("public-read")
}
// IsPublicReadWrite - is acl PublicReadWrite
func (b BucketACL) IsPublicReadWrite() bool {
return b == BucketACL("public-read-write")
}
// IsValidBucketACL - is provided acl string supported
func IsValidBucketACL(acl string) bool {
switch acl {
case "private":
fallthrough
case "public-read":
fallthrough
case "public-read-write":
return true
case "":
// by default its "private"
return true
default:
return false
}
}
-639
View File
@@ -1,639 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package donut
import (
"bytes"
"fmt"
"hash"
"io"
"path/filepath"
"sort"
"strings"
"sync"
"time"
"crypto/md5"
"encoding/hex"
"encoding/json"
"github.com/minio/minio/pkg/crypto/sha256"
"github.com/minio/minio/pkg/crypto/sha512"
"github.com/minio/minio/pkg/donut/disk"
"github.com/minio/minio/pkg/probe"
signv4 "github.com/minio/minio/pkg/signature"
)
const (
blockSize = 10 * 1024 * 1024
)
// internal struct carrying bucket specific information
type bucket struct {
name string
acl string
time time.Time
donutName string
nodes map[string]node
lock *sync.Mutex
}
// newBucket - instantiate a new bucket
func newBucket(bucketName, aclType, donutName string, nodes map[string]node) (bucket, BucketMetadata, *probe.Error) {
if strings.TrimSpace(bucketName) == "" || strings.TrimSpace(donutName) == "" {
return bucket{}, BucketMetadata{}, probe.NewError(InvalidArgument{})
}
b := bucket{}
t := time.Now().UTC()
b.name = bucketName
b.acl = aclType
b.time = t
b.donutName = donutName
b.nodes = nodes
b.lock = new(sync.Mutex)
metadata := BucketMetadata{}
metadata.Version = bucketMetadataVersion
metadata.Name = bucketName
metadata.ACL = BucketACL(aclType)
metadata.Created = t
metadata.Metadata = make(map[string]string)
metadata.BucketObjects = make(map[string]struct{})
return b, metadata, nil
}
// getBucketName -
func (b bucket) getBucketName() string {
return b.name
}
// getBucketMetadataReaders -
func (b bucket) getBucketMetadataReaders() (map[int]io.ReadCloser, *probe.Error) {
readers := make(map[int]io.ReadCloser)
var disks map[int]disk.Disk
var err *probe.Error
for _, node := range b.nodes {
disks, err = node.ListDisks()
if err != nil {
return nil, err.Trace()
}
}
var bucketMetaDataReader io.ReadCloser
for order, disk := range disks {
bucketMetaDataReader, err = disk.Open(filepath.Join(b.donutName, bucketMetadataConfig))
if err != nil {
continue
}
readers[order] = bucketMetaDataReader
}
if err != nil {
return nil, err.Trace()
}
return readers, nil
}
// getBucketMetadata -
func (b bucket) getBucketMetadata() (*AllBuckets, *probe.Error) {
metadata := new(AllBuckets)
var readers map[int]io.ReadCloser
{
var err *probe.Error
readers, err = b.getBucketMetadataReaders()
if err != nil {
return nil, err.Trace()
}
}
for _, reader := range readers {
defer reader.Close()
}
var err error
for _, reader := range readers {
jenc := json.NewDecoder(reader)
if err = jenc.Decode(metadata); err == nil {
return metadata, nil
}
}
return nil, probe.NewError(err)
}
// GetObjectMetadata - get metadata for an object
func (b bucket) GetObjectMetadata(objectName string) (ObjectMetadata, *probe.Error) {
b.lock.Lock()
defer b.lock.Unlock()
return b.readObjectMetadata(normalizeObjectName(objectName))
}
// ListObjects - list all objects
func (b bucket) ListObjects(prefix, marker, delimiter string, maxkeys int) (ListObjectsResults, *probe.Error) {
b.lock.Lock()
defer b.lock.Unlock()
if maxkeys <= 0 {
maxkeys = 1000
}
var isTruncated bool
var objects []string
bucketMetadata, err := b.getBucketMetadata()
if err != nil {
return ListObjectsResults{}, err.Trace()
}
for objectName := range bucketMetadata.Buckets[b.getBucketName()].Multiparts {
if strings.HasPrefix(objectName, strings.TrimSpace(prefix)) {
if objectName > marker {
objects = append(objects, objectName)
}
}
}
for objectName := range bucketMetadata.Buckets[b.getBucketName()].BucketObjects {
if strings.HasPrefix(objectName, strings.TrimSpace(prefix)) {
if objectName > marker {
objects = append(objects, objectName)
}
}
}
if strings.TrimSpace(prefix) != "" {
objects = TrimPrefix(objects, prefix)
}
var prefixes []string
var filteredObjects []string
filteredObjects = objects
if strings.TrimSpace(delimiter) != "" {
filteredObjects = HasNoDelimiter(objects, delimiter)
prefixes = HasDelimiter(objects, delimiter)
prefixes = SplitDelimiter(prefixes, delimiter)
prefixes = SortUnique(prefixes)
}
var results []string
var commonPrefixes []string
for _, commonPrefix := range prefixes {
commonPrefixes = append(commonPrefixes, prefix+commonPrefix)
}
filteredObjects = RemoveDuplicates(filteredObjects)
sort.Strings(filteredObjects)
for _, objectName := range filteredObjects {
if len(results) >= maxkeys {
isTruncated = true
break
}
results = append(results, prefix+objectName)
}
results = RemoveDuplicates(results)
commonPrefixes = RemoveDuplicates(commonPrefixes)
sort.Strings(commonPrefixes)
listObjects := ListObjectsResults{}
listObjects.Objects = make(map[string]ObjectMetadata)
listObjects.CommonPrefixes = commonPrefixes
listObjects.IsTruncated = isTruncated
for _, objectName := range results {
objMetadata, err := b.readObjectMetadata(normalizeObjectName(objectName))
if err != nil {
return ListObjectsResults{}, err.Trace()
}
listObjects.Objects[objectName] = objMetadata
}
return listObjects, nil
}
// ReadObject - open an object to read
func (b bucket) ReadObject(objectName string) (reader io.ReadCloser, size int64, err *probe.Error) {
b.lock.Lock()
defer b.lock.Unlock()
reader, writer := io.Pipe()
// get list of objects
bucketMetadata, err := b.getBucketMetadata()
if err != nil {
return nil, 0, err.Trace()
}
// check if object exists
if _, ok := bucketMetadata.Buckets[b.getBucketName()].BucketObjects[objectName]; !ok {
return nil, 0, probe.NewError(ObjectNotFound{Object: objectName})
}
objMetadata, err := b.readObjectMetadata(normalizeObjectName(objectName))
if err != nil {
return nil, 0, err.Trace()
}
// read and reply back to GetObject() request in a go-routine
go b.readObjectData(normalizeObjectName(objectName), writer, objMetadata)
return reader, objMetadata.Size, nil
}
// WriteObject - write a new object into bucket
func (b bucket) WriteObject(objectName string, objectData io.Reader, size int64, expectedMD5Sum string, metadata map[string]string, signature *signv4.Signature) (ObjectMetadata, *probe.Error) {
b.lock.Lock()
defer b.lock.Unlock()
if objectName == "" || objectData == nil {
return ObjectMetadata{}, probe.NewError(InvalidArgument{})
}
writers, err := b.getObjectWriters(normalizeObjectName(objectName), "data")
if err != nil {
return ObjectMetadata{}, err.Trace()
}
sumMD5 := md5.New()
sum512 := sha512.New()
var sum256 hash.Hash
var mwriter io.Writer
if signature != nil {
sum256 = sha256.New()
mwriter = io.MultiWriter(sumMD5, sum256, sum512)
} else {
mwriter = io.MultiWriter(sumMD5, sum512)
}
objMetadata := ObjectMetadata{}
objMetadata.Version = objectMetadataVersion
objMetadata.Created = time.Now().UTC()
// if total writers are only '1' do not compute erasure
switch len(writers) == 1 {
case true:
mw := io.MultiWriter(writers[0], mwriter)
totalLength, err := io.Copy(mw, objectData)
if err != nil {
CleanupWritersOnError(writers)
return ObjectMetadata{}, probe.NewError(err)
}
objMetadata.Size = totalLength
case false:
// calculate data and parity dictated by total number of writers
k, m, err := b.getDataAndParity(len(writers))
if err != nil {
CleanupWritersOnError(writers)
return ObjectMetadata{}, err.Trace()
}
// write encoded data with k, m and writers
chunkCount, totalLength, err := b.writeObjectData(k, m, writers, objectData, size, mwriter)
if err != nil {
CleanupWritersOnError(writers)
return ObjectMetadata{}, err.Trace()
}
/// donutMetadata section
objMetadata.BlockSize = blockSize
objMetadata.ChunkCount = chunkCount
objMetadata.DataDisks = k
objMetadata.ParityDisks = m
objMetadata.Size = int64(totalLength)
}
objMetadata.Bucket = b.getBucketName()
objMetadata.Object = objectName
dataMD5sum := sumMD5.Sum(nil)
dataSHA512sum := sum512.Sum(nil)
if signature != nil {
ok, err := signature.DoesSignatureMatch(hex.EncodeToString(sum256.Sum(nil)))
if err != nil {
// error occurred while doing signature calculation, we return and also cleanup any temporary writers.
CleanupWritersOnError(writers)
return ObjectMetadata{}, err.Trace()
}
if !ok {
// purge all writers, when control flow reaches here
//
// Signature mismatch occurred all temp files to be removed and all data purged.
CleanupWritersOnError(writers)
return ObjectMetadata{}, probe.NewError(signv4.DoesNotMatch{})
}
}
objMetadata.MD5Sum = hex.EncodeToString(dataMD5sum)
objMetadata.SHA512Sum = hex.EncodeToString(dataSHA512sum)
// Verify if the written object is equal to what is expected, only if it is requested as such
if strings.TrimSpace(expectedMD5Sum) != "" {
if err := b.isMD5SumEqual(strings.TrimSpace(expectedMD5Sum), objMetadata.MD5Sum); err != nil {
return ObjectMetadata{}, err.Trace()
}
}
objMetadata.Metadata = metadata
// write object specific metadata
if err := b.writeObjectMetadata(normalizeObjectName(objectName), objMetadata); err != nil {
// purge all writers, when control flow reaches here
CleanupWritersOnError(writers)
return ObjectMetadata{}, err.Trace()
}
// close all writers, when control flow reaches here
for _, writer := range writers {
writer.Close()
}
return objMetadata, nil
}
// isMD5SumEqual - returns error if md5sum mismatches, other its `nil`
func (b bucket) isMD5SumEqual(expectedMD5Sum, actualMD5Sum string) *probe.Error {
if strings.TrimSpace(expectedMD5Sum) != "" && strings.TrimSpace(actualMD5Sum) != "" {
expectedMD5SumBytes, err := hex.DecodeString(expectedMD5Sum)
if err != nil {
return probe.NewError(err)
}
actualMD5SumBytes, err := hex.DecodeString(actualMD5Sum)
if err != nil {
return probe.NewError(err)
}
if !bytes.Equal(expectedMD5SumBytes, actualMD5SumBytes) {
return probe.NewError(BadDigest{})
}
return nil
}
return probe.NewError(InvalidArgument{})
}
// writeObjectMetadata - write additional object metadata
func (b bucket) writeObjectMetadata(objectName string, objMetadata ObjectMetadata) *probe.Error {
if objMetadata.Object == "" {
return probe.NewError(InvalidArgument{})
}
objMetadataWriters, err := b.getObjectWriters(objectName, objectMetadataConfig)
if err != nil {
return err.Trace()
}
for _, objMetadataWriter := range objMetadataWriters {
jenc := json.NewEncoder(objMetadataWriter)
if err := jenc.Encode(&objMetadata); err != nil {
// Close writers and purge all temporary entries
CleanupWritersOnError(objMetadataWriters)
return probe.NewError(err)
}
}
for _, objMetadataWriter := range objMetadataWriters {
objMetadataWriter.Close()
}
return nil
}
// readObjectMetadata - read object metadata
func (b bucket) readObjectMetadata(objectName string) (ObjectMetadata, *probe.Error) {
if objectName == "" {
return ObjectMetadata{}, probe.NewError(InvalidArgument{})
}
objMetadata := ObjectMetadata{}
objMetadataReaders, err := b.getObjectReaders(objectName, objectMetadataConfig)
if err != nil {
return ObjectMetadata{}, err.Trace()
}
for _, objMetadataReader := range objMetadataReaders {
defer objMetadataReader.Close()
}
{
var err error
for _, objMetadataReader := range objMetadataReaders {
jdec := json.NewDecoder(objMetadataReader)
if err = jdec.Decode(&objMetadata); err == nil {
return objMetadata, nil
}
}
return ObjectMetadata{}, probe.NewError(err)
}
}
// TODO - This a temporary normalization of objectNames, need to find a better way
//
// normalizedObjectName - all objectNames with "/" get normalized to a simple objectName
//
// example:
// user provided value - "this/is/my/deep/directory/structure"
// donut normalized value - "this-is-my-deep-directory-structure"
//
func normalizeObjectName(objectName string) string {
// replace every '/' with '-'
return strings.Replace(objectName, "/", "-", -1)
}
// getDataAndParity - calculate k, m (data and parity) values from number of disks
func (b bucket) getDataAndParity(totalWriters int) (k uint8, m uint8, err *probe.Error) {
if totalWriters <= 1 {
return 0, 0, probe.NewError(InvalidArgument{})
}
quotient := totalWriters / 2 // not using float or abs to let integer round off to lower value
// quotient cannot be bigger than (255 / 2) = 127
if quotient > 127 {
return 0, 0, probe.NewError(ParityOverflow{})
}
remainder := totalWriters % 2 // will be 1 for odd and 0 for even numbers
k = uint8(quotient + remainder)
m = uint8(quotient)
return k, m, nil
}
// writeObjectData -
func (b bucket) writeObjectData(k, m uint8, writers []io.WriteCloser, objectData io.Reader, size int64, hashWriter io.Writer) (int, int, *probe.Error) {
encoder, err := newEncoder(k, m)
if err != nil {
return 0, 0, err.Trace()
}
chunkSize := int64(10 * 1024 * 1024)
chunkCount := 0
totalLength := 0
var e error
for e == nil {
var length int
inputData := make([]byte, chunkSize)
length, e = objectData.Read(inputData)
if length != 0 {
encodedBlocks, err := encoder.Encode(inputData[0:length])
if err != nil {
return 0, 0, err.Trace()
}
if _, err := hashWriter.Write(inputData[0:length]); err != nil {
return 0, 0, probe.NewError(err)
}
for blockIndex, block := range encodedBlocks {
errCh := make(chan error, 1)
go func(writer io.Writer, reader io.Reader, errCh chan<- error) {
defer close(errCh)
_, err := io.Copy(writer, reader)
errCh <- err
}(writers[blockIndex], bytes.NewReader(block), errCh)
if err := <-errCh; err != nil {
// Returning error is fine here CleanupErrors() would cleanup writers
return 0, 0, probe.NewError(err)
}
}
totalLength += length
chunkCount = chunkCount + 1
}
}
if e != io.EOF {
return 0, 0, probe.NewError(e)
}
return chunkCount, totalLength, nil
}
// readObjectData -
func (b bucket) readObjectData(objectName string, writer *io.PipeWriter, objMetadata ObjectMetadata) {
readers, err := b.getObjectReaders(objectName, "data")
if err != nil {
writer.CloseWithError(probe.WrapError(err))
return
}
for _, reader := range readers {
defer reader.Close()
}
var expected512Sum, expectedMd5sum []byte
{
var err error
expectedMd5sum, err = hex.DecodeString(objMetadata.MD5Sum)
if err != nil {
writer.CloseWithError(probe.WrapError(probe.NewError(err)))
return
}
expected512Sum, err = hex.DecodeString(objMetadata.SHA512Sum)
if err != nil {
writer.CloseWithError(probe.WrapError(probe.NewError(err)))
return
}
}
hasher := md5.New()
sum512hasher := sha256.New()
mwriter := io.MultiWriter(writer, hasher, sum512hasher)
switch len(readers) > 1 {
case true:
encoder, err := newEncoder(objMetadata.DataDisks, objMetadata.ParityDisks)
if err != nil {
writer.CloseWithError(probe.WrapError(err))
return
}
totalLeft := objMetadata.Size
for i := 0; i < objMetadata.ChunkCount; i++ {
decodedData, err := b.decodeEncodedData(totalLeft, int64(objMetadata.BlockSize), readers, encoder, writer)
if err != nil {
writer.CloseWithError(probe.WrapError(err))
return
}
if _, err := io.Copy(mwriter, bytes.NewReader(decodedData)); err != nil {
writer.CloseWithError(probe.WrapError(probe.NewError(err)))
return
}
totalLeft = totalLeft - int64(objMetadata.BlockSize)
}
case false:
_, err := io.Copy(writer, readers[0])
if err != nil {
writer.CloseWithError(probe.WrapError(probe.NewError(err)))
return
}
}
// check if decodedData md5sum matches
if !bytes.Equal(expectedMd5sum, hasher.Sum(nil)) {
writer.CloseWithError(probe.WrapError(probe.NewError(ChecksumMismatch{})))
return
}
if !bytes.Equal(expected512Sum, sum512hasher.Sum(nil)) {
writer.CloseWithError(probe.WrapError(probe.NewError(ChecksumMismatch{})))
return
}
writer.Close()
return
}
// decodeEncodedData -
func (b bucket) decodeEncodedData(totalLeft, blockSize int64, readers map[int]io.ReadCloser, encoder encoder, writer *io.PipeWriter) ([]byte, *probe.Error) {
var curBlockSize int64
if blockSize < totalLeft {
curBlockSize = blockSize
} else {
curBlockSize = totalLeft
}
curChunkSize, err := encoder.GetEncodedBlockLen(int(curBlockSize))
if err != nil {
return nil, err.Trace()
}
encodedBytes := make([][]byte, encoder.k+encoder.m)
errCh := make(chan error, len(readers))
var errRet error
var readCnt int
for i, reader := range readers {
go func(reader io.Reader, i int) {
encodedBytes[i] = make([]byte, curChunkSize)
_, err := io.ReadFull(reader, encodedBytes[i])
if err != nil {
encodedBytes[i] = nil
errCh <- err
return
}
errCh <- nil
}(reader, i)
// read through errCh for any errors
err := <-errCh
if err != nil {
errRet = err
} else {
readCnt++
}
}
if readCnt < int(encoder.k) {
return nil, probe.NewError(errRet)
}
decodedData, err := encoder.Decode(encodedBytes, int(curBlockSize))
if err != nil {
return nil, err.Trace()
}
return decodedData, nil
}
// getObjectReaders -
func (b bucket) getObjectReaders(objectName, objectMeta string) (map[int]io.ReadCloser, *probe.Error) {
readers := make(map[int]io.ReadCloser)
var disks map[int]disk.Disk
var err *probe.Error
nodeSlice := 0
for _, node := range b.nodes {
disks, err = node.ListDisks()
if err != nil {
return nil, err.Trace()
}
for order, disk := range disks {
var objectSlice io.ReadCloser
bucketSlice := fmt.Sprintf("%s$%d$%d", b.name, nodeSlice, order)
objectPath := filepath.Join(b.donutName, bucketSlice, objectName, objectMeta)
objectSlice, err = disk.Open(objectPath)
if err == nil {
readers[order] = objectSlice
}
}
nodeSlice = nodeSlice + 1
}
if err != nil {
return nil, err.Trace()
}
return readers, nil
}
// getObjectWriters -
func (b bucket) getObjectWriters(objectName, objectMeta string) ([]io.WriteCloser, *probe.Error) {
var writers []io.WriteCloser
nodeSlice := 0
for _, node := range b.nodes {
disks, err := node.ListDisks()
if err != nil {
return nil, err.Trace()
}
writers = make([]io.WriteCloser, len(disks))
for order, disk := range disks {
bucketSlice := fmt.Sprintf("%s$%d$%d", b.name, nodeSlice, order)
objectPath := filepath.Join(b.donutName, bucketSlice, objectName, objectMeta)
objectSlice, err := disk.CreateFile(objectPath)
if err != nil {
return nil, err.Trace()
}
writers[order] = objectSlice
}
nodeSlice = nodeSlice + 1
}
return writers, nil
}
-204
View File
@@ -1,204 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
// Package data implements in memory caching methods for data
package data
import (
"container/list"
"sync"
"time"
)
var noExpiration = time.Duration(0)
// Cache holds the required variables to compose an in memory cache system
// which also provides expiring key mechanism and also maxSize
type Cache struct {
// Mutex is used for handling the concurrent
// read/write requests for cache
sync.Mutex
// items hold the cached objects
items *list.List
// reverseItems holds the time that related item's updated at
reverseItems map[interface{}]*list.Element
// maxSize is a total size for overall cache
maxSize uint64
// currentSize is a current size in memory
currentSize uint64
// OnEvicted - callback function for eviction
OnEvicted func(a ...interface{})
// totalEvicted counter to keep track of total expirations
totalEvicted int
}
// Stats current cache statistics
type Stats struct {
Bytes uint64
Items int
Evicted int
}
type element struct {
key interface{}
value []byte
}
// NewCache creates an inmemory cache
//
// maxSize is used for expiring objects before we run out of memory
// expiration is used for expiration of a key from cache
func NewCache(maxSize uint64) *Cache {
return &Cache{
items: list.New(),
reverseItems: make(map[interface{}]*list.Element),
maxSize: maxSize,
}
}
// SetMaxSize set a new max size
func (r *Cache) SetMaxSize(maxSize uint64) {
r.Lock()
defer r.Unlock()
r.maxSize = maxSize
return
}
// Stats get current cache statistics
func (r *Cache) Stats() Stats {
return Stats{
Bytes: r.currentSize,
Items: r.items.Len(),
Evicted: r.totalEvicted,
}
}
// Get returns a value of a given key if it exists
func (r *Cache) Get(key interface{}) ([]byte, bool) {
r.Lock()
defer r.Unlock()
ele, hit := r.reverseItems[key]
if !hit {
return nil, false
}
r.items.MoveToFront(ele)
return ele.Value.(*element).value, true
}
// Len returns length of the value of a given key, returns zero if key doesn't exist
func (r *Cache) Len(key interface{}) int {
r.Lock()
defer r.Unlock()
_, ok := r.reverseItems[key]
if !ok {
return 0
}
return len(r.reverseItems[key].Value.(*element).value)
}
// Append will append new data to an existing key,
// if key doesn't exist it behaves like Set()
func (r *Cache) Append(key interface{}, value []byte) bool {
r.Lock()
defer r.Unlock()
valueLen := uint64(len(value))
if r.maxSize > 0 {
// check if the size of the object is not bigger than the
// capacity of the cache
if valueLen > r.maxSize {
return false
}
// remove random key if only we reach the maxSize threshold
for (r.currentSize + valueLen) > r.maxSize {
r.doDeleteOldest()
break
}
}
ele, hit := r.reverseItems[key]
if !hit {
ele := r.items.PushFront(&element{key, value})
r.currentSize += valueLen
r.reverseItems[key] = ele
return true
}
r.items.MoveToFront(ele)
r.currentSize += valueLen
ele.Value.(*element).value = append(ele.Value.(*element).value, value...)
return true
}
// Set will persist a value to the cache
func (r *Cache) Set(key interface{}, value []byte) bool {
r.Lock()
defer r.Unlock()
valueLen := uint64(len(value))
if r.maxSize > 0 {
// check if the size of the object is not bigger than the
// capacity of the cache
if valueLen > r.maxSize {
return false
}
// remove random key if only we reach the maxSize threshold
for (r.currentSize + valueLen) > r.maxSize {
r.doDeleteOldest()
}
}
if _, hit := r.reverseItems[key]; hit {
return false
}
ele := r.items.PushFront(&element{key, value})
r.currentSize += valueLen
r.reverseItems[key] = ele
return true
}
// Delete deletes a given key if exists
func (r *Cache) Delete(key interface{}) {
r.Lock()
defer r.Unlock()
ele, ok := r.reverseItems[key]
if !ok {
return
}
if ele != nil {
r.currentSize -= uint64(len(r.reverseItems[key].Value.(*element).value))
r.items.Remove(ele)
delete(r.reverseItems, key)
r.totalEvicted++
if r.OnEvicted != nil {
r.OnEvicted(key)
}
}
}
func (r *Cache) doDeleteOldest() {
ele := r.items.Back()
if ele != nil {
r.currentSize -= uint64(len(r.reverseItems[ele.Value.(*element).key].Value.(*element).value))
delete(r.reverseItems, ele.Value.(*element).key)
r.items.Remove(ele)
r.totalEvicted++
if r.OnEvicted != nil {
r.OnEvicted(ele.Value.(*element).key)
}
}
}
-45
View File
@@ -1,45 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package data
import (
"testing"
. "gopkg.in/check.v1"
)
func Test(t *testing.T) { TestingT(t) }
type MySuite struct{}
var _ = Suite(&MySuite{})
func (s *MySuite) TestCache(c *C) {
cache := NewCache(1000)
data := []byte("Hello, world!")
ok := cache.Set("filename", data)
c.Assert(ok, Equals, true)
storedata, ok := cache.Get("filename")
c.Assert(ok, Equals, true)
c.Assert(data, DeepEquals, storedata)
cache.Delete("filename")
_, ok = cache.Get("filename")
c.Assert(ok, Equals, false)
}
-110
View File
@@ -1,110 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
// Package metadata implements in memory caching methods for metadata information
package metadata
import (
"sync"
"time"
)
var noExpiration = time.Duration(0)
// Cache holds the required variables to compose an in memory cache system
// which also provides expiring key mechanism and also maxSize
type Cache struct {
// Mutex is used for handling the concurrent
// read/write requests for cache
sync.Mutex
// items hold the cached objects
items map[string]interface{}
// updatedAt holds the time that related item's updated at
updatedAt map[string]time.Time
}
// Stats current cache statistics
type Stats struct {
Items int
}
// NewCache creates an inmemory cache
//
func NewCache() *Cache {
return &Cache{
items: make(map[string]interface{}),
updatedAt: map[string]time.Time{},
}
}
// Stats get current cache statistics
func (r *Cache) Stats() Stats {
return Stats{
Items: len(r.items),
}
}
// GetAll returs all the items
func (r *Cache) GetAll() map[string]interface{} {
r.Lock()
defer r.Unlock()
// copy
items := r.items
return items
}
// Get returns a value of a given key if it exists
func (r *Cache) Get(key string) interface{} {
r.Lock()
defer r.Unlock()
value, ok := r.items[key]
if !ok {
return nil
}
return value
}
// Exists returns true if key exists
func (r *Cache) Exists(key string) bool {
r.Lock()
defer r.Unlock()
_, ok := r.items[key]
return ok
}
// Set will persist a value to the cache
func (r *Cache) Set(key string, value interface{}) bool {
r.Lock()
defer r.Unlock()
r.items[key] = value
return true
}
// Delete deletes a given key if exists
func (r *Cache) Delete(key string) {
r.Lock()
defer r.Unlock()
r.doDelete(key)
}
func (r *Cache) doDelete(key string) {
if _, ok := r.items[key]; ok {
delete(r.items, key)
delete(r.updatedAt, key)
}
}
-46
View File
@@ -1,46 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package metadata
import (
"testing"
. "gopkg.in/check.v1"
)
func Test(t *testing.T) { TestingT(t) }
type MySuite struct{}
var _ = Suite(&MySuite{})
func (s *MySuite) TestCache(c *C) {
cache := NewCache()
data := []byte("Hello, world!")
ok := cache.Set("filename", data)
c.Assert(ok, Equals, true)
storedata := cache.Get("filename")
c.Assert(ok, Equals, true)
c.Assert(data, DeepEquals, storedata)
cache.Delete("filename")
ok = cache.Exists("filename")
c.Assert(ok, Equals, false)
}
-190
View File
@@ -1,190 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package donut
import (
"bufio"
"bytes"
"io"
"regexp"
"sort"
"strings"
"unicode/utf8"
"github.com/minio/minio/pkg/atomic"
)
// IsValidDonut - verify donut name is correct
func IsValidDonut(donutName string) bool {
if len(donutName) < 3 || len(donutName) > 63 {
return false
}
if donutName[0] == '.' || donutName[len(donutName)-1] == '.' {
return false
}
if match, _ := regexp.MatchString("\\.\\.", donutName); match == true {
return false
}
// We don't support donutNames with '.' in them
match, _ := regexp.MatchString("^[a-zA-Z][a-zA-Z0-9\\-]+[a-zA-Z0-9]$", donutName)
return match
}
// IsValidBucket - verify bucket name in accordance with
// - http://docs.aws.amazon.com/AmazonS3/latest/dev/UsingBucket.html
func IsValidBucket(bucket string) bool {
if len(bucket) < 3 || len(bucket) > 63 {
return false
}
if bucket[0] == '.' || bucket[len(bucket)-1] == '.' {
return false
}
if match, _ := regexp.MatchString("\\.\\.", bucket); match == true {
return false
}
// We don't support buckets with '.' in them
match, _ := regexp.MatchString("^[a-zA-Z][a-zA-Z0-9\\-]+[a-zA-Z0-9]$", bucket)
return match
}
// IsValidObjectName - verify object name in accordance with
// - http://docs.aws.amazon.com/AmazonS3/latest/dev/UsingMetadata.html
func IsValidObjectName(object string) bool {
if strings.TrimSpace(object) == "" {
return false
}
if len(object) > 1024 || len(object) == 0 {
return false
}
if !utf8.ValidString(object) {
return false
}
return true
}
// IsValidPrefix - verify prefix name is correct, an empty prefix is valid
func IsValidPrefix(prefix string) bool {
if strings.TrimSpace(prefix) == "" {
return true
}
return IsValidObjectName(prefix)
}
// ProxyWriter implements io.Writer to trap written bytes
type ProxyWriter struct {
writer io.Writer
writtenBytes []byte
}
func (r *ProxyWriter) Write(p []byte) (n int, err error) {
n, err = r.writer.Write(p)
if err != nil {
return
}
r.writtenBytes = append(r.writtenBytes, p[0:n]...)
return
}
// NewProxyWriter - wrap around a given writer with ProxyWriter
func NewProxyWriter(w io.Writer) *ProxyWriter {
return &ProxyWriter{writer: w, writtenBytes: nil}
}
// Delimiter delims the string at delimiter
func Delimiter(object, delimiter string) string {
readBuffer := bytes.NewBufferString(object)
reader := bufio.NewReader(readBuffer)
stringReader := strings.NewReader(delimiter)
delimited, _ := stringReader.ReadByte()
delimitedStr, _ := reader.ReadString(delimited)
return delimitedStr
}
// RemoveDuplicates removes duplicate elements from a slice
func RemoveDuplicates(slice []string) []string {
newSlice := []string{}
seen := make(map[string]struct{})
for _, val := range slice {
if _, ok := seen[val]; !ok {
newSlice = append(newSlice, val)
seen[val] = struct{}{} // avoiding byte allocation
}
}
return newSlice
}
// TrimPrefix trims off a prefix string from all the elements in a given slice
func TrimPrefix(objects []string, prefix string) []string {
var results []string
for _, object := range objects {
results = append(results, strings.TrimPrefix(object, prefix))
}
return results
}
// HasNoDelimiter provides a new slice from an input slice which has elements without delimiter
func HasNoDelimiter(objects []string, delim string) []string {
var results []string
for _, object := range objects {
if !strings.Contains(object, delim) {
results = append(results, object)
}
}
return results
}
// HasDelimiter provides a new slice from an input slice which has elements with a delimiter
func HasDelimiter(objects []string, delim string) []string {
var results []string
for _, object := range objects {
if strings.Contains(object, delim) {
results = append(results, object)
}
}
return results
}
// SplitDelimiter provides a new slice from an input slice by splitting a delimiter
func SplitDelimiter(objects []string, delim string) []string {
var results []string
for _, object := range objects {
parts := strings.Split(object, delim)
results = append(results, parts[0]+delim)
}
return results
}
// SortUnique sort a slice in lexical order, removing duplicate elements
func SortUnique(objects []string) []string {
objectMap := make(map[string]string)
for _, v := range objects {
objectMap[v] = v
}
var results []string
for k := range objectMap {
results = append(results, k)
}
sort.Strings(results)
return results
}
// CleanupWritersOnError purge writers on error
func CleanupWritersOnError(writers []io.WriteCloser) {
for _, writer := range writers {
writer.(*atomic.File).CloseAndPurge()
}
}
-80
View File
@@ -1,80 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package donut
import (
"os/user"
"path/filepath"
"github.com/minio/minio/pkg/probe"
"github.com/minio/minio/pkg/quick"
)
// getDonutConfigPath get donut config file path
func getDonutConfigPath() (string, *probe.Error) {
if customConfigPath != "" {
return customConfigPath, nil
}
u, err := user.Current()
if err != nil {
return "", probe.NewError(err)
}
donutConfigPath := filepath.Join(u.HomeDir, ".minio", "donut.json")
return donutConfigPath, nil
}
// internal variable only accessed via get/set methods
var customConfigPath string
// SetDonutConfigPath - set custom donut config path
func SetDonutConfigPath(configPath string) {
customConfigPath = configPath
}
// SaveConfig save donut config
func SaveConfig(a *Config) *probe.Error {
donutConfigPath, err := getDonutConfigPath()
if err != nil {
return err.Trace()
}
qc, err := quick.New(a)
if err != nil {
return err.Trace()
}
if err := qc.Save(donutConfigPath); err != nil {
return err.Trace()
}
return nil
}
// LoadConfig load donut config
func LoadConfig() (*Config, *probe.Error) {
donutConfigPath, err := getDonutConfigPath()
if err != nil {
return nil, err.Trace()
}
a := &Config{}
a.Version = "0.0.1"
qc, err := quick.New(a)
if err != nil {
return nil, err.Trace()
}
if err := qc.Load(donutConfigPath); err != nil {
return nil, err.Trace()
}
return qc.Data().(*Config), nil
}
-222
View File
@@ -1,222 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or impliedisk.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package disk
import (
"fmt"
"os"
"path/filepath"
"strconv"
"strings"
"sync"
"syscall"
"github.com/minio/minio/pkg/atomic"
"github.com/minio/minio/pkg/probe"
)
// Disk container for disk parameters
type Disk struct {
lock *sync.Mutex
path string
fsInfo map[string]string
}
// New - instantiate new disk
func New(diskPath string) (Disk, *probe.Error) {
if diskPath == "" {
return Disk{}, probe.NewError(InvalidArgument{})
}
st, err := os.Stat(diskPath)
if err != nil {
return Disk{}, probe.NewError(err)
}
if !st.IsDir() {
return Disk{}, probe.NewError(syscall.ENOTDIR)
}
s := syscall.Statfs_t{}
err = syscall.Statfs(diskPath, &s)
if err != nil {
return Disk{}, probe.NewError(err)
}
disk := Disk{
lock: &sync.Mutex{},
path: diskPath,
fsInfo: make(map[string]string),
}
if fsType := getFSType(s.Type); fsType != "UNKNOWN" {
disk.fsInfo["FSType"] = fsType
disk.fsInfo["MountPoint"] = disk.path
return disk, nil
}
return Disk{}, probe.NewError(UnsupportedFilesystem{Type: strconv.FormatInt(int64(s.Type), 10)})
}
// IsUsable - is disk usable, alive
func (disk Disk) IsUsable() bool {
_, err := os.Stat(disk.path)
if err != nil {
return false
}
return true
}
// GetPath - get root disk path
func (disk Disk) GetPath() string {
return disk.path
}
// GetFSInfo - get disk filesystem and its usage information
func (disk Disk) GetFSInfo() map[string]string {
disk.lock.Lock()
defer disk.lock.Unlock()
s := syscall.Statfs_t{}
err := syscall.Statfs(disk.path, &s)
if err != nil {
return nil
}
disk.fsInfo["Total"] = formatBytes(int64(s.Bsize) * int64(s.Blocks))
disk.fsInfo["Free"] = formatBytes(int64(s.Bsize) * int64(s.Bfree))
disk.fsInfo["TotalB"] = strconv.FormatInt(int64(s.Bsize)*int64(s.Blocks), 10)
disk.fsInfo["FreeB"] = strconv.FormatInt(int64(s.Bsize)*int64(s.Bfree), 10)
return disk.fsInfo
}
// MakeDir - make a directory inside disk root path
func (disk Disk) MakeDir(dirname string) *probe.Error {
disk.lock.Lock()
defer disk.lock.Unlock()
if err := os.MkdirAll(filepath.Join(disk.path, dirname), 0700); err != nil {
return probe.NewError(err)
}
return nil
}
// ListDir - list a directory inside disk root path, get only directories
func (disk Disk) ListDir(dirname string) ([]os.FileInfo, *probe.Error) {
disk.lock.Lock()
defer disk.lock.Unlock()
dir, err := os.Open(filepath.Join(disk.path, dirname))
if err != nil {
return nil, probe.NewError(err)
}
defer dir.Close()
contents, err := dir.Readdir(-1)
if err != nil {
return nil, probe.NewError(err)
}
var directories []os.FileInfo
for _, content := range contents {
// Include only directories, ignore everything else
if content.IsDir() {
directories = append(directories, content)
}
}
return directories, nil
}
// ListFiles - list a directory inside disk root path, get only files
func (disk Disk) ListFiles(dirname string) ([]os.FileInfo, *probe.Error) {
disk.lock.Lock()
defer disk.lock.Unlock()
dir, err := os.Open(filepath.Join(disk.path, dirname))
if err != nil {
return nil, probe.NewError(err)
}
defer dir.Close()
contents, err := dir.Readdir(-1)
if err != nil {
return nil, probe.NewError(err)
}
var files []os.FileInfo
for _, content := range contents {
// Include only regular files, ignore everything else
if content.Mode().IsRegular() {
files = append(files, content)
}
}
return files, nil
}
// CreateFile - create a file inside disk root path, replies with custome disk.File which provides atomic writes
func (disk Disk) CreateFile(filename string) (*atomic.File, *probe.Error) {
disk.lock.Lock()
defer disk.lock.Unlock()
if filename == "" {
return nil, probe.NewError(InvalidArgument{})
}
f, err := atomic.FileCreate(filepath.Join(disk.path, filename))
if err != nil {
return nil, probe.NewError(err)
}
return f, nil
}
// Open - read a file inside disk root path
func (disk Disk) Open(filename string) (*os.File, *probe.Error) {
disk.lock.Lock()
defer disk.lock.Unlock()
if filename == "" {
return nil, probe.NewError(InvalidArgument{})
}
dataFile, err := os.Open(filepath.Join(disk.path, filename))
if err != nil {
return nil, probe.NewError(err)
}
return dataFile, nil
}
// OpenFile - Use with caution
func (disk Disk) OpenFile(filename string, flags int, perm os.FileMode) (*os.File, *probe.Error) {
disk.lock.Lock()
defer disk.lock.Unlock()
if filename == "" {
return nil, probe.NewError(InvalidArgument{})
}
dataFile, err := os.OpenFile(filepath.Join(disk.path, filename), flags, perm)
if err != nil {
return nil, probe.NewError(err)
}
return dataFile, nil
}
// formatBytes - Convert bytes to human readable string. Like a 2 MB, 64.2 KB, 52 B
func formatBytes(i int64) (result string) {
switch {
case i > (1024 * 1024 * 1024 * 1024):
result = fmt.Sprintf("%.02f TB", float64(i)/1024/1024/1024/1024)
case i > (1024 * 1024 * 1024):
result = fmt.Sprintf("%.02f GB", float64(i)/1024/1024/1024)
case i > (1024 * 1024):
result = fmt.Sprintf("%.02f MB", float64(i)/1024/1024)
case i > 1024:
result = fmt.Sprintf("%.02f KB", float64(i)/1024)
default:
result = fmt.Sprintf("%d B", i)
}
result = strings.Trim(result, " ")
return
}
-34
View File
@@ -1,34 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package disk
import "strconv"
// fsType2StrinMap - list of filesystems supported by donut
var fsType2StringMap = map[string]string{
"11": "HFS",
}
// getFSType - get filesystem type
func getFSType(fsType uint32) string {
fsTypeHex := strconv.FormatUint(uint64(fsType), 16)
fsTypeString, ok := fsType2StringMap[fsTypeHex]
if ok == false {
return "UNKNOWN"
}
return fsTypeString
}
-45
View File
@@ -1,45 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package disk
import "strconv"
// fsType2StringMap - list of filesystems supported by donut on linux
var fsType2StringMap = map[string]string{
"1021994": "TMPFS",
"137d": "EXT",
"4244": "HFS",
"4d44": "MSDOS",
"52654973": "REISERFS",
"5346544e": "NTFS",
"58465342": "XFS",
"61756673": "AUFS",
"6969": "NFS",
"ef51": "EXT2OLD",
"ef53": "EXT4",
"f15f": "ecryptfs",
}
// getFSType - get filesystem type
func getFSType(fsType int64) string {
fsTypeHex := strconv.FormatInt(fsType, 16)
fsTypeString, ok := fsType2StringMap[fsTypeHex]
if ok == false {
return "UNKNOWN"
}
return fsTypeString
}
-84
View File
@@ -1,84 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or impliedisk.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package disk
import (
"io/ioutil"
"os"
"path/filepath"
"testing"
. "gopkg.in/check.v1"
)
func TestDisk(t *testing.T) { TestingT(t) }
type MyDiskSuite struct {
path string
disk Disk
}
var _ = Suite(&MyDiskSuite{})
func (s *MyDiskSuite) SetUpSuite(c *C) {
path, err := ioutil.TempDir(os.TempDir(), "disk-")
c.Assert(err, IsNil)
s.path = path
d, perr := New(s.path)
c.Assert(perr, IsNil)
s.disk = d
}
func (s *MyDiskSuite) TearDownSuite(c *C) {
os.RemoveAll(s.path)
}
func (s *MyDiskSuite) TestDiskInfo(c *C) {
c.Assert(s.path, Equals, s.disk.GetPath())
fsInfo := s.disk.GetFSInfo()
c.Assert(fsInfo["MountPoint"], Equals, s.disk.GetPath())
c.Assert(fsInfo["FSType"], Not(Equals), "UNKNOWN")
}
func (s *MyDiskSuite) TestDiskCreateDir(c *C) {
c.Assert(s.disk.MakeDir("hello"), IsNil)
}
func (s *MyDiskSuite) TestDiskCreateFile(c *C) {
f, err := s.disk.CreateFile("hello1")
c.Assert(err, IsNil)
c.Assert(f.Name(), Not(Equals), filepath.Join(s.path, "hello1"))
// close renames the file
f.Close()
// Open should be a success
_, err = s.disk.Open("hello1")
c.Assert(err, IsNil)
}
func (s *MyDiskSuite) TestDiskOpen(c *C) {
f1, err := s.disk.CreateFile("hello2")
c.Assert(err, IsNil)
c.Assert(f1.Name(), Not(Equals), filepath.Join(s.path, "hello2"))
// close renames the file
f1.Close()
f2, err := s.disk.Open("hello2")
c.Assert(err, IsNil)
c.Assert(f2.Name(), Equals, filepath.Join(s.path, "hello2"))
defer f2.Close()
}
-33
View File
@@ -1,33 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or impliedisk.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package disk
// InvalidArgument invalid argument
type InvalidArgument struct{}
func (e InvalidArgument) Error() string {
return "Invalid argument"
}
// UnsupportedFilesystem unsupported filesystem type
type UnsupportedFilesystem struct {
Type string
}
func (e UnsupportedFilesystem) Error() string {
return "Unsupported filesystem: " + e.Type
}
-55
View File
@@ -1,55 +0,0 @@
##### Users Collection
```js
"minio": {
"version": 1,
"users": [{
"secretAccessKey": String,
"accessKeyId": String,
"status": String // enum: ok, disabled, deleted
}],
"hosts": [{
"address": String,
"uuid": String,
"status": String, // enum: ok, disabled, deleted, busy, offline.
"disks": [{
"disk": String,
"uuid": String,
"status": String // ok, offline, disabled, busy.
}]
}]
}
```
##### Bucket Collection
```js
"buckets": {
"bucket": String, // index
"deleted": Boolean,
"permissions": String
}
```
##### Object Collection
```js
"objects": {
"key": String, // index
"createdAt": Date,
"hosts[16]": [{
"host": String,
"disk": String,
}],
"deleted": Boolean
}
```
```js
"meta": {
"key": String, // index
"type": String // content-type
// type speific meta
}
```
-680
View File
@@ -1,680 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package donut
import (
"encoding/base64"
"encoding/hex"
"encoding/json"
"encoding/xml"
"fmt"
"io"
"io/ioutil"
"math/rand"
"os"
"path/filepath"
"sort"
"strconv"
"strings"
"time"
"github.com/minio/minio/pkg/crypto/sha256"
"github.com/minio/minio/pkg/crypto/sha512"
"github.com/minio/minio/pkg/donut/disk"
"github.com/minio/minio/pkg/probe"
signv4 "github.com/minio/minio/pkg/signature"
)
// config files used inside Donut
const (
// bucket, object metadata
bucketMetadataConfig = "bucketMetadata.json"
objectMetadataConfig = "objectMetadata.json"
// versions
objectMetadataVersion = "1.0.0"
bucketMetadataVersion = "1.0.0"
)
/// v1 API functions
// makeBucket - make a new bucket
func (donut API) makeBucket(bucket string, acl BucketACL) *probe.Error {
if bucket == "" || strings.TrimSpace(bucket) == "" {
return probe.NewError(InvalidArgument{})
}
return donut.makeDonutBucket(bucket, acl.String())
}
// getBucketMetadata - get bucket metadata
func (donut API) getBucketMetadata(bucketName string) (BucketMetadata, *probe.Error) {
if err := donut.listDonutBuckets(); err != nil {
return BucketMetadata{}, err.Trace()
}
if _, ok := donut.buckets[bucketName]; !ok {
return BucketMetadata{}, probe.NewError(BucketNotFound{Bucket: bucketName})
}
metadata, err := donut.getDonutBucketMetadata()
if err != nil {
return BucketMetadata{}, err.Trace()
}
return metadata.Buckets[bucketName], nil
}
// setBucketMetadata - set bucket metadata
func (donut API) setBucketMetadata(bucketName string, bucketMetadata map[string]string) *probe.Error {
if err := donut.listDonutBuckets(); err != nil {
return err.Trace()
}
metadata, err := donut.getDonutBucketMetadata()
if err != nil {
return err.Trace()
}
oldBucketMetadata := metadata.Buckets[bucketName]
acl, ok := bucketMetadata["acl"]
if !ok {
return probe.NewError(InvalidArgument{})
}
oldBucketMetadata.ACL = BucketACL(acl)
metadata.Buckets[bucketName] = oldBucketMetadata
return donut.setDonutBucketMetadata(metadata)
}
// listBuckets - return list of buckets
func (donut API) listBuckets() (map[string]BucketMetadata, *probe.Error) {
if err := donut.listDonutBuckets(); err != nil {
return nil, err.Trace()
}
metadata, err := donut.getDonutBucketMetadata()
if err != nil {
// intentionally left out the error when Donut is empty
// but we need to revisit this area in future - since we need
// to figure out between acceptable and unacceptable errors
return make(map[string]BucketMetadata), nil
}
if metadata == nil {
return make(map[string]BucketMetadata), nil
}
return metadata.Buckets, nil
}
// listObjects - return list of objects
func (donut API) listObjects(bucket, prefix, marker, delimiter string, maxkeys int) (ListObjectsResults, *probe.Error) {
if err := donut.listDonutBuckets(); err != nil {
return ListObjectsResults{}, err.Trace()
}
if _, ok := donut.buckets[bucket]; !ok {
return ListObjectsResults{}, probe.NewError(BucketNotFound{Bucket: bucket})
}
listObjects, err := donut.buckets[bucket].ListObjects(prefix, marker, delimiter, maxkeys)
if err != nil {
return ListObjectsResults{}, err.Trace()
}
return listObjects, nil
}
// putObject - put object
func (donut API) putObject(bucket, object, expectedMD5Sum string, reader io.Reader, size int64, metadata map[string]string, signature *signv4.Signature) (ObjectMetadata, *probe.Error) {
if bucket == "" || strings.TrimSpace(bucket) == "" {
return ObjectMetadata{}, probe.NewError(InvalidArgument{})
}
if object == "" || strings.TrimSpace(object) == "" {
return ObjectMetadata{}, probe.NewError(InvalidArgument{})
}
if err := donut.listDonutBuckets(); err != nil {
return ObjectMetadata{}, err.Trace()
}
if _, ok := donut.buckets[bucket]; !ok {
return ObjectMetadata{}, probe.NewError(BucketNotFound{Bucket: bucket})
}
bucketMeta, err := donut.getDonutBucketMetadata()
if err != nil {
return ObjectMetadata{}, err.Trace()
}
if _, ok := bucketMeta.Buckets[bucket].BucketObjects[object]; ok {
return ObjectMetadata{}, probe.NewError(ObjectExists{Object: object})
}
objMetadata, err := donut.buckets[bucket].WriteObject(object, reader, size, expectedMD5Sum, metadata, signature)
if err != nil {
return ObjectMetadata{}, err.Trace()
}
bucketMeta.Buckets[bucket].BucketObjects[object] = struct{}{}
if err := donut.setDonutBucketMetadata(bucketMeta); err != nil {
return ObjectMetadata{}, err.Trace()
}
return objMetadata, nil
}
// putObject - put object
func (donut API) putObjectPart(bucket, object, expectedMD5Sum, uploadID string, partID int, reader io.Reader, size int64, metadata map[string]string, signature *signv4.Signature) (PartMetadata, *probe.Error) {
if bucket == "" || strings.TrimSpace(bucket) == "" {
return PartMetadata{}, probe.NewError(InvalidArgument{})
}
if object == "" || strings.TrimSpace(object) == "" {
return PartMetadata{}, probe.NewError(InvalidArgument{})
}
if err := donut.listDonutBuckets(); err != nil {
return PartMetadata{}, err.Trace()
}
if _, ok := donut.buckets[bucket]; !ok {
return PartMetadata{}, probe.NewError(BucketNotFound{Bucket: bucket})
}
bucketMeta, err := donut.getDonutBucketMetadata()
if err != nil {
return PartMetadata{}, err.Trace()
}
if _, ok := bucketMeta.Buckets[bucket].Multiparts[object]; !ok {
return PartMetadata{}, probe.NewError(InvalidUploadID{UploadID: uploadID})
}
if _, ok := bucketMeta.Buckets[bucket].BucketObjects[object]; ok {
return PartMetadata{}, probe.NewError(ObjectExists{Object: object})
}
objectPart := object + "/" + "multipart" + "/" + strconv.Itoa(partID)
objmetadata, err := donut.buckets[bucket].WriteObject(objectPart, reader, size, expectedMD5Sum, metadata, signature)
if err != nil {
return PartMetadata{}, err.Trace()
}
partMetadata := PartMetadata{
PartNumber: partID,
LastModified: objmetadata.Created,
ETag: objmetadata.MD5Sum,
Size: objmetadata.Size,
}
multipartSession := bucketMeta.Buckets[bucket].Multiparts[object]
multipartSession.Parts[strconv.Itoa(partID)] = partMetadata
bucketMeta.Buckets[bucket].Multiparts[object] = multipartSession
if err := donut.setDonutBucketMetadata(bucketMeta); err != nil {
return PartMetadata{}, err.Trace()
}
return partMetadata, nil
}
// getObject - get object
func (donut API) getObject(bucket, object string) (reader io.ReadCloser, size int64, err *probe.Error) {
if bucket == "" || strings.TrimSpace(bucket) == "" {
return nil, 0, probe.NewError(InvalidArgument{})
}
if object == "" || strings.TrimSpace(object) == "" {
return nil, 0, probe.NewError(InvalidArgument{})
}
if err := donut.listDonutBuckets(); err != nil {
return nil, 0, err.Trace()
}
if _, ok := donut.buckets[bucket]; !ok {
return nil, 0, probe.NewError(BucketNotFound{Bucket: bucket})
}
return donut.buckets[bucket].ReadObject(object)
}
// getObjectMetadata - get object metadata
func (donut API) getObjectMetadata(bucket, object string) (ObjectMetadata, *probe.Error) {
if err := donut.listDonutBuckets(); err != nil {
return ObjectMetadata{}, err.Trace()
}
if _, ok := donut.buckets[bucket]; !ok {
return ObjectMetadata{}, probe.NewError(BucketNotFound{Bucket: bucket})
}
bucketMeta, err := donut.getDonutBucketMetadata()
if err != nil {
return ObjectMetadata{}, err.Trace()
}
if _, ok := bucketMeta.Buckets[bucket].BucketObjects[object]; !ok {
return ObjectMetadata{}, probe.NewError(ObjectNotFound{Object: object})
}
objectMetadata, err := donut.buckets[bucket].GetObjectMetadata(object)
if err != nil {
return ObjectMetadata{}, err.Trace()
}
return objectMetadata, nil
}
// newMultipartUpload - new multipart upload request
func (donut API) newMultipartUpload(bucket, object, contentType string) (string, *probe.Error) {
if err := donut.listDonutBuckets(); err != nil {
return "", err.Trace()
}
if _, ok := donut.buckets[bucket]; !ok {
return "", probe.NewError(BucketNotFound{Bucket: bucket})
}
allbuckets, err := donut.getDonutBucketMetadata()
if err != nil {
return "", err.Trace()
}
bucketMetadata := allbuckets.Buckets[bucket]
multiparts := make(map[string]MultiPartSession)
if len(bucketMetadata.Multiparts) > 0 {
multiparts = bucketMetadata.Multiparts
}
id := []byte(strconv.Itoa(rand.Int()) + bucket + object + time.Now().String())
uploadIDSum := sha512.Sum512(id)
uploadID := base64.URLEncoding.EncodeToString(uploadIDSum[:])[:47]
multipartSession := MultiPartSession{
UploadID: uploadID,
Initiated: time.Now().UTC(),
Parts: make(map[string]PartMetadata),
TotalParts: 0,
}
multiparts[object] = multipartSession
bucketMetadata.Multiparts = multiparts
allbuckets.Buckets[bucket] = bucketMetadata
if err := donut.setDonutBucketMetadata(allbuckets); err != nil {
return "", err.Trace()
}
return uploadID, nil
}
// listObjectParts list all object parts
func (donut API) listObjectParts(bucket, object string, resources ObjectResourcesMetadata) (ObjectResourcesMetadata, *probe.Error) {
if bucket == "" || strings.TrimSpace(bucket) == "" {
return ObjectResourcesMetadata{}, probe.NewError(InvalidArgument{})
}
if object == "" || strings.TrimSpace(object) == "" {
return ObjectResourcesMetadata{}, probe.NewError(InvalidArgument{})
}
if err := donut.listDonutBuckets(); err != nil {
return ObjectResourcesMetadata{}, err.Trace()
}
if _, ok := donut.buckets[bucket]; !ok {
return ObjectResourcesMetadata{}, probe.NewError(BucketNotFound{Bucket: bucket})
}
allBuckets, err := donut.getDonutBucketMetadata()
if err != nil {
return ObjectResourcesMetadata{}, err.Trace()
}
bucketMetadata := allBuckets.Buckets[bucket]
if _, ok := bucketMetadata.Multiparts[object]; !ok {
return ObjectResourcesMetadata{}, probe.NewError(InvalidUploadID{UploadID: resources.UploadID})
}
if bucketMetadata.Multiparts[object].UploadID != resources.UploadID {
return ObjectResourcesMetadata{}, probe.NewError(InvalidUploadID{UploadID: resources.UploadID})
}
objectResourcesMetadata := resources
objectResourcesMetadata.Bucket = bucket
objectResourcesMetadata.Key = object
var parts []*PartMetadata
var startPartNumber int
switch {
case objectResourcesMetadata.PartNumberMarker == 0:
startPartNumber = 1
default:
startPartNumber = objectResourcesMetadata.PartNumberMarker
}
for i := startPartNumber; i <= bucketMetadata.Multiparts[object].TotalParts; i++ {
if len(parts) > objectResourcesMetadata.MaxParts {
sort.Sort(partNumber(parts))
objectResourcesMetadata.IsTruncated = true
objectResourcesMetadata.Part = parts
objectResourcesMetadata.NextPartNumberMarker = i
return objectResourcesMetadata, nil
}
part, ok := bucketMetadata.Multiparts[object].Parts[strconv.Itoa(i)]
if !ok {
return ObjectResourcesMetadata{}, probe.NewError(InvalidPart{})
}
parts = append(parts, &part)
}
sort.Sort(partNumber(parts))
objectResourcesMetadata.Part = parts
return objectResourcesMetadata, nil
}
// completeMultipartUpload complete an incomplete multipart upload
func (donut API) completeMultipartUpload(bucket, object, uploadID string, data io.Reader, signature *signv4.Signature) (ObjectMetadata, *probe.Error) {
if bucket == "" || strings.TrimSpace(bucket) == "" {
return ObjectMetadata{}, probe.NewError(InvalidArgument{})
}
if object == "" || strings.TrimSpace(object) == "" {
return ObjectMetadata{}, probe.NewError(InvalidArgument{})
}
if err := donut.listDonutBuckets(); err != nil {
return ObjectMetadata{}, err.Trace()
}
if _, ok := donut.buckets[bucket]; !ok {
return ObjectMetadata{}, probe.NewError(BucketNotFound{Bucket: bucket})
}
allBuckets, err := donut.getDonutBucketMetadata()
if err != nil {
return ObjectMetadata{}, err.Trace()
}
bucketMetadata := allBuckets.Buckets[bucket]
if _, ok := bucketMetadata.Multiparts[object]; !ok {
return ObjectMetadata{}, probe.NewError(InvalidUploadID{UploadID: uploadID})
}
if bucketMetadata.Multiparts[object].UploadID != uploadID {
return ObjectMetadata{}, probe.NewError(InvalidUploadID{UploadID: uploadID})
}
var partBytes []byte
{
var err error
partBytes, err = ioutil.ReadAll(data)
if err != nil {
return ObjectMetadata{}, probe.NewError(err)
}
}
if signature != nil {
ok, err := signature.DoesSignatureMatch(hex.EncodeToString(sha256.Sum256(partBytes)[:]))
if err != nil {
return ObjectMetadata{}, err.Trace()
}
if !ok {
return ObjectMetadata{}, probe.NewError(signv4.DoesNotMatch{})
}
}
parts := &CompleteMultipartUpload{}
if err := xml.Unmarshal(partBytes, parts); err != nil {
return ObjectMetadata{}, probe.NewError(MalformedXML{})
}
if !sort.IsSorted(completedParts(parts.Part)) {
return ObjectMetadata{}, probe.NewError(InvalidPartOrder{})
}
for _, part := range parts.Part {
if strings.Trim(part.ETag, "\"") != bucketMetadata.Multiparts[object].Parts[strconv.Itoa(part.PartNumber)].ETag {
return ObjectMetadata{}, probe.NewError(InvalidPart{})
}
}
var finalETagBytes []byte
var finalSize int64
totalParts := strconv.Itoa(bucketMetadata.Multiparts[object].TotalParts)
for _, part := range bucketMetadata.Multiparts[object].Parts {
partETagBytes, err := hex.DecodeString(part.ETag)
if err != nil {
return ObjectMetadata{}, probe.NewError(err)
}
finalETagBytes = append(finalETagBytes, partETagBytes...)
finalSize += part.Size
}
finalETag := hex.EncodeToString(finalETagBytes)
objMetadata := ObjectMetadata{}
objMetadata.MD5Sum = finalETag + "-" + totalParts
objMetadata.Object = object
objMetadata.Bucket = bucket
objMetadata.Size = finalSize
objMetadata.Created = bucketMetadata.Multiparts[object].Parts[totalParts].LastModified
return objMetadata, nil
}
// listMultipartUploads list all multipart uploads
func (donut API) listMultipartUploads(bucket string, resources BucketMultipartResourcesMetadata) (BucketMultipartResourcesMetadata, *probe.Error) {
if err := donut.listDonutBuckets(); err != nil {
return BucketMultipartResourcesMetadata{}, err.Trace()
}
if _, ok := donut.buckets[bucket]; !ok {
return BucketMultipartResourcesMetadata{}, probe.NewError(BucketNotFound{Bucket: bucket})
}
allbuckets, err := donut.getDonutBucketMetadata()
if err != nil {
return BucketMultipartResourcesMetadata{}, err.Trace()
}
bucketMetadata := allbuckets.Buckets[bucket]
var uploads []*UploadMetadata
for key, session := range bucketMetadata.Multiparts {
if strings.HasPrefix(key, resources.Prefix) {
if len(uploads) > resources.MaxUploads {
sort.Sort(byKey(uploads))
resources.Upload = uploads
resources.NextKeyMarker = key
resources.NextUploadIDMarker = session.UploadID
resources.IsTruncated = true
return resources, nil
}
// uploadIDMarker is ignored if KeyMarker is empty
switch {
case resources.KeyMarker != "" && resources.UploadIDMarker == "":
if key > resources.KeyMarker {
upload := new(UploadMetadata)
upload.Key = key
upload.UploadID = session.UploadID
upload.Initiated = session.Initiated
uploads = append(uploads, upload)
}
case resources.KeyMarker != "" && resources.UploadIDMarker != "":
if session.UploadID > resources.UploadIDMarker {
if key >= resources.KeyMarker {
upload := new(UploadMetadata)
upload.Key = key
upload.UploadID = session.UploadID
upload.Initiated = session.Initiated
uploads = append(uploads, upload)
}
}
default:
upload := new(UploadMetadata)
upload.Key = key
upload.UploadID = session.UploadID
upload.Initiated = session.Initiated
uploads = append(uploads, upload)
}
}
}
sort.Sort(byKey(uploads))
resources.Upload = uploads
return resources, nil
}
// abortMultipartUpload - abort a incomplete multipart upload
func (donut API) abortMultipartUpload(bucket, object, uploadID string) *probe.Error {
if err := donut.listDonutBuckets(); err != nil {
return err.Trace()
}
if _, ok := donut.buckets[bucket]; !ok {
return probe.NewError(BucketNotFound{Bucket: bucket})
}
allbuckets, err := donut.getDonutBucketMetadata()
if err != nil {
return err.Trace()
}
bucketMetadata := allbuckets.Buckets[bucket]
if _, ok := bucketMetadata.Multiparts[object]; !ok {
return probe.NewError(InvalidUploadID{UploadID: uploadID})
}
if bucketMetadata.Multiparts[object].UploadID != uploadID {
return probe.NewError(InvalidUploadID{UploadID: uploadID})
}
delete(bucketMetadata.Multiparts, object)
allbuckets.Buckets[bucket] = bucketMetadata
if err := donut.setDonutBucketMetadata(allbuckets); err != nil {
return err.Trace()
}
return nil
}
//// internal functions
// getBucketMetadataWriters -
func (donut API) getBucketMetadataWriters() ([]io.WriteCloser, *probe.Error) {
var writers []io.WriteCloser
for _, node := range donut.nodes {
disks, err := node.ListDisks()
if err != nil {
return nil, err.Trace()
}
writers = make([]io.WriteCloser, len(disks))
for order, disk := range disks {
bucketMetaDataWriter, err := disk.CreateFile(filepath.Join(donut.config.DonutName, bucketMetadataConfig))
if err != nil {
return nil, err.Trace()
}
writers[order] = bucketMetaDataWriter
}
}
return writers, nil
}
// getBucketMetadataReaders - readers are returned in map rather than slice
func (donut API) getBucketMetadataReaders() (map[int]io.ReadCloser, *probe.Error) {
readers := make(map[int]io.ReadCloser)
disks := make(map[int]disk.Disk)
var err *probe.Error
for _, node := range donut.nodes {
nDisks := make(map[int]disk.Disk)
nDisks, err = node.ListDisks()
if err != nil {
return nil, err.Trace()
}
for k, v := range nDisks {
disks[k] = v
}
}
var bucketMetaDataReader io.ReadCloser
for order, disk := range disks {
bucketMetaDataReader, err = disk.Open(filepath.Join(donut.config.DonutName, bucketMetadataConfig))
if err != nil {
continue
}
readers[order] = bucketMetaDataReader
}
if err != nil {
return nil, err.Trace()
}
return readers, nil
}
// setDonutBucketMetadata -
func (donut API) setDonutBucketMetadata(metadata *AllBuckets) *probe.Error {
writers, err := donut.getBucketMetadataWriters()
if err != nil {
return err.Trace()
}
for _, writer := range writers {
jenc := json.NewEncoder(writer)
if err := jenc.Encode(metadata); err != nil {
CleanupWritersOnError(writers)
return probe.NewError(err)
}
}
for _, writer := range writers {
writer.Close()
}
return nil
}
// getDonutBucketMetadata -
func (donut API) getDonutBucketMetadata() (*AllBuckets, *probe.Error) {
metadata := &AllBuckets{}
readers, err := donut.getBucketMetadataReaders()
if err != nil {
return nil, err.Trace()
}
for _, reader := range readers {
defer reader.Close()
}
{
var err error
for _, reader := range readers {
jenc := json.NewDecoder(reader)
if err = jenc.Decode(metadata); err == nil {
return metadata, nil
}
}
return nil, probe.NewError(err)
}
}
// makeDonutBucket -
func (donut API) makeDonutBucket(bucketName, acl string) *probe.Error {
if err := donut.listDonutBuckets(); err != nil {
return err.Trace()
}
if _, ok := donut.buckets[bucketName]; ok {
return probe.NewError(BucketExists{Bucket: bucketName})
}
bkt, bucketMetadata, err := newBucket(bucketName, acl, donut.config.DonutName, donut.nodes)
if err != nil {
return err.Trace()
}
nodeNumber := 0
donut.buckets[bucketName] = bkt
for _, node := range donut.nodes {
disks := make(map[int]disk.Disk)
disks, err = node.ListDisks()
if err != nil {
return err.Trace()
}
for order, disk := range disks {
bucketSlice := fmt.Sprintf("%s$%d$%d", bucketName, nodeNumber, order)
err := disk.MakeDir(filepath.Join(donut.config.DonutName, bucketSlice))
if err != nil {
return err.Trace()
}
}
nodeNumber = nodeNumber + 1
}
var metadata *AllBuckets
metadata, err = donut.getDonutBucketMetadata()
if err != nil {
if os.IsNotExist(err.ToGoError()) {
metadata = new(AllBuckets)
metadata.Buckets = make(map[string]BucketMetadata)
metadata.Buckets[bucketName] = bucketMetadata
err = donut.setDonutBucketMetadata(metadata)
if err != nil {
return err.Trace()
}
return nil
}
return err.Trace()
}
metadata.Buckets[bucketName] = bucketMetadata
err = donut.setDonutBucketMetadata(metadata)
if err != nil {
return err.Trace()
}
return nil
}
// listDonutBuckets -
func (donut API) listDonutBuckets() *probe.Error {
var disks map[int]disk.Disk
var err *probe.Error
for _, node := range donut.nodes {
disks, err = node.ListDisks()
if err != nil {
return err.Trace()
}
}
var dirs []os.FileInfo
for _, disk := range disks {
dirs, err = disk.ListDir(donut.config.DonutName)
if err == nil {
break
}
}
// if all disks are missing then return error
if err != nil {
return err.Trace()
}
for _, dir := range dirs {
splitDir := strings.Split(dir.Name(), "$")
if len(splitDir) < 3 {
return probe.NewError(CorruptedBackend{Backend: dir.Name()})
}
bucketName := splitDir[0]
// we dont need this once we cache from makeDonutBucket()
bkt, _, err := newBucket(bucketName, "private", donut.config.DonutName, donut.nodes)
if err != nil {
return err.Trace()
}
donut.buckets[bucketName] = bkt
}
return nil
}
-290
View File
@@ -1,290 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or impliedd.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package donut
import (
"bytes"
"crypto/md5"
"encoding/base64"
"encoding/hex"
"io/ioutil"
"os"
"path/filepath"
"strconv"
"testing"
. "gopkg.in/check.v1"
)
func TestDonut(t *testing.T) { TestingT(t) }
type MyDonutSuite struct {
root string
}
var _ = Suite(&MyDonutSuite{})
// create a dummy TestNodeDiskMap
func createTestNodeDiskMap(p string) map[string][]string {
nodes := make(map[string][]string)
nodes["localhost"] = make([]string, 16)
for i := 0; i < len(nodes["localhost"]); i++ {
diskPath := filepath.Join(p, strconv.Itoa(i))
if _, err := os.Stat(diskPath); err != nil {
if os.IsNotExist(err) {
os.MkdirAll(diskPath, 0700)
}
}
nodes["localhost"][i] = diskPath
}
return nodes
}
var dd Interface
func (s *MyDonutSuite) SetUpSuite(c *C) {
root, err := ioutil.TempDir(os.TempDir(), "donut-")
c.Assert(err, IsNil)
s.root = root
conf := new(Config)
conf.Version = "0.0.1"
conf.DonutName = "test"
conf.NodeDiskMap = createTestNodeDiskMap(root)
conf.MaxSize = 100000
SetDonutConfigPath(filepath.Join(root, "donut.json"))
perr := SaveConfig(conf)
c.Assert(perr, IsNil)
dd, perr = New()
c.Assert(perr, IsNil)
// testing empty donut
buckets, perr := dd.ListBuckets()
c.Assert(perr, IsNil)
c.Assert(len(buckets), Equals, 0)
}
func (s *MyDonutSuite) TearDownSuite(c *C) {
os.RemoveAll(s.root)
}
// test make bucket without name
func (s *MyDonutSuite) TestBucketWithoutNameFails(c *C) {
// fail to create new bucket without a name
err := dd.MakeBucket("", "private", nil, nil)
c.Assert(err, Not(IsNil))
err = dd.MakeBucket(" ", "private", nil, nil)
c.Assert(err, Not(IsNil))
}
// test empty bucket
func (s *MyDonutSuite) TestEmptyBucket(c *C) {
c.Assert(dd.MakeBucket("foo1", "private", nil, nil), IsNil)
// check if bucket is empty
var resources BucketResourcesMetadata
resources.Maxkeys = 1
objectsMetadata, resources, err := dd.ListObjects("foo1", resources)
c.Assert(err, IsNil)
c.Assert(len(objectsMetadata), Equals, 0)
c.Assert(resources.CommonPrefixes, DeepEquals, []string{})
c.Assert(resources.IsTruncated, Equals, false)
}
// test bucket list
func (s *MyDonutSuite) TestMakeBucketAndList(c *C) {
// create bucket
err := dd.MakeBucket("foo2", "private", nil, nil)
c.Assert(err, IsNil)
// check bucket exists
buckets, err := dd.ListBuckets()
c.Assert(err, IsNil)
c.Assert(len(buckets), Equals, 5)
c.Assert(buckets[0].ACL, Equals, BucketACL("private"))
}
// test re-create bucket
func (s *MyDonutSuite) TestMakeBucketWithSameNameFails(c *C) {
err := dd.MakeBucket("foo3", "private", nil, nil)
c.Assert(err, IsNil)
err = dd.MakeBucket("foo3", "private", nil, nil)
c.Assert(err, Not(IsNil))
}
// test make multiple buckets
func (s *MyDonutSuite) TestCreateMultipleBucketsAndList(c *C) {
// add a second bucket
err := dd.MakeBucket("foo4", "private", nil, nil)
c.Assert(err, IsNil)
err = dd.MakeBucket("bar1", "private", nil, nil)
c.Assert(err, IsNil)
buckets, err := dd.ListBuckets()
c.Assert(err, IsNil)
c.Assert(len(buckets), Equals, 2)
c.Assert(buckets[0].Name, Equals, "bar1")
c.Assert(buckets[1].Name, Equals, "foo4")
err = dd.MakeBucket("foobar1", "private", nil, nil)
c.Assert(err, IsNil)
buckets, err = dd.ListBuckets()
c.Assert(err, IsNil)
c.Assert(len(buckets), Equals, 3)
c.Assert(buckets[2].Name, Equals, "foobar1")
}
// test object create without bucket
func (s *MyDonutSuite) TestNewObjectFailsWithoutBucket(c *C) {
_, err := dd.CreateObject("unknown", "obj", "", 0, nil, nil, nil)
c.Assert(err, Not(IsNil))
}
// test create object metadata
func (s *MyDonutSuite) TestNewObjectMetadata(c *C) {
data := "Hello World"
hasher := md5.New()
hasher.Write([]byte(data))
expectedMd5Sum := base64.StdEncoding.EncodeToString(hasher.Sum(nil))
reader := ioutil.NopCloser(bytes.NewReader([]byte(data)))
err := dd.MakeBucket("foo6", "private", nil, nil)
c.Assert(err, IsNil)
objectMetadata, err := dd.CreateObject("foo6", "obj", expectedMd5Sum, int64(len(data)), reader, map[string]string{"contentType": "application/json"}, nil)
c.Assert(err, IsNil)
c.Assert(objectMetadata.MD5Sum, Equals, hex.EncodeToString(hasher.Sum(nil)))
c.Assert(objectMetadata.Metadata["contentType"], Equals, "application/json")
}
// test create object fails without name
func (s *MyDonutSuite) TestNewObjectFailsWithEmptyName(c *C) {
_, err := dd.CreateObject("foo", "", "", 0, nil, nil, nil)
c.Assert(err, Not(IsNil))
}
// test create object
func (s *MyDonutSuite) TestNewObjectCanBeWritten(c *C) {
err := dd.MakeBucket("foo", "private", nil, nil)
c.Assert(err, IsNil)
data := "Hello World"
hasher := md5.New()
hasher.Write([]byte(data))
expectedMd5Sum := base64.StdEncoding.EncodeToString(hasher.Sum(nil))
reader := ioutil.NopCloser(bytes.NewReader([]byte(data)))
actualMetadata, err := dd.CreateObject("foo", "obj", expectedMd5Sum, int64(len(data)), reader, map[string]string{"contentType": "application/octet-stream"}, nil)
c.Assert(err, IsNil)
c.Assert(actualMetadata.MD5Sum, Equals, hex.EncodeToString(hasher.Sum(nil)))
var buffer bytes.Buffer
size, err := dd.GetObject(&buffer, "foo", "obj", 0, 0)
c.Assert(err, IsNil)
c.Assert(size, Equals, int64(len(data)))
c.Assert(buffer.Bytes(), DeepEquals, []byte(data))
actualMetadata, err = dd.GetObjectMetadata("foo", "obj")
c.Assert(err, IsNil)
c.Assert(hex.EncodeToString(hasher.Sum(nil)), Equals, actualMetadata.MD5Sum)
c.Assert(int64(len(data)), Equals, actualMetadata.Size)
}
// test list objects
func (s *MyDonutSuite) TestMultipleNewObjects(c *C) {
c.Assert(dd.MakeBucket("foo5", "private", nil, nil), IsNil)
one := ioutil.NopCloser(bytes.NewReader([]byte("one")))
_, err := dd.CreateObject("foo5", "obj1", "", int64(len("one")), one, nil, nil)
c.Assert(err, IsNil)
two := ioutil.NopCloser(bytes.NewReader([]byte("two")))
_, err = dd.CreateObject("foo5", "obj2", "", int64(len("two")), two, nil, nil)
c.Assert(err, IsNil)
var buffer1 bytes.Buffer
size, err := dd.GetObject(&buffer1, "foo5", "obj1", 0, 0)
c.Assert(err, IsNil)
c.Assert(size, Equals, int64(len([]byte("one"))))
c.Assert(buffer1.Bytes(), DeepEquals, []byte("one"))
var buffer2 bytes.Buffer
size, err = dd.GetObject(&buffer2, "foo5", "obj2", 0, 0)
c.Assert(err, IsNil)
c.Assert(size, Equals, int64(len([]byte("two"))))
c.Assert(buffer2.Bytes(), DeepEquals, []byte("two"))
/// test list of objects
// test list objects with prefix and delimiter
var resources BucketResourcesMetadata
resources.Prefix = "o"
resources.Delimiter = "1"
resources.Maxkeys = 10
objectsMetadata, resources, err := dd.ListObjects("foo5", resources)
c.Assert(err, IsNil)
c.Assert(resources.IsTruncated, Equals, false)
c.Assert(resources.CommonPrefixes[0], Equals, "obj1")
// test list objects with only delimiter
resources.Prefix = ""
resources.Delimiter = "1"
resources.Maxkeys = 10
objectsMetadata, resources, err = dd.ListObjects("foo5", resources)
c.Assert(err, IsNil)
c.Assert(objectsMetadata[0].Object, Equals, "obj2")
c.Assert(resources.IsTruncated, Equals, false)
c.Assert(resources.CommonPrefixes[0], Equals, "obj1")
// test list objects with only prefix
resources.Prefix = "o"
resources.Delimiter = ""
resources.Maxkeys = 10
objectsMetadata, resources, err = dd.ListObjects("foo5", resources)
c.Assert(err, IsNil)
c.Assert(resources.IsTruncated, Equals, false)
c.Assert(objectsMetadata[0].Object, Equals, "obj1")
c.Assert(objectsMetadata[1].Object, Equals, "obj2")
three := ioutil.NopCloser(bytes.NewReader([]byte("three")))
_, err = dd.CreateObject("foo5", "obj3", "", int64(len("three")), three, nil, nil)
c.Assert(err, IsNil)
var buffer bytes.Buffer
size, err = dd.GetObject(&buffer, "foo5", "obj3", 0, 0)
c.Assert(err, IsNil)
c.Assert(size, Equals, int64(len([]byte("three"))))
c.Assert(buffer.Bytes(), DeepEquals, []byte("three"))
// test list objects with maxkeys
resources.Prefix = "o"
resources.Delimiter = ""
resources.Maxkeys = 2
objectsMetadata, resources, err = dd.ListObjects("foo5", resources)
c.Assert(err, IsNil)
c.Assert(resources.IsTruncated, Equals, true)
c.Assert(len(objectsMetadata), Equals, 2)
}
-636
View File
@@ -1,636 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package donut
import (
"bytes"
"crypto/md5"
"encoding/base64"
"encoding/hex"
"io"
"io/ioutil"
"log"
"runtime/debug"
"sort"
"strconv"
"strings"
"sync"
"time"
"github.com/minio/minio/pkg/crypto/sha256"
"github.com/minio/minio/pkg/donut/cache/data"
"github.com/minio/minio/pkg/donut/cache/metadata"
"github.com/minio/minio/pkg/probe"
"github.com/minio/minio/pkg/quick"
signv4 "github.com/minio/minio/pkg/signature"
)
// total Number of buckets allowed
const (
totalBuckets = 100
)
// Config donut config
type Config struct {
Version string `json:"version"`
MaxSize uint64 `json:"max-size"`
DonutName string `json:"donut-name"`
NodeDiskMap map[string][]string `json:"node-disk-map"`
}
// API - local variables
type API struct {
config *Config
lock *sync.Mutex
objects *data.Cache
multiPartObjects map[string]*data.Cache
storedBuckets *metadata.Cache
nodes map[string]node
buckets map[string]bucket
}
// storedBucket saved bucket
type storedBucket struct {
bucketMetadata BucketMetadata
objectMetadata map[string]ObjectMetadata
partMetadata map[string]map[int]PartMetadata
multiPartSession map[string]MultiPartSession
}
// New instantiate a new donut
func New() (Interface, *probe.Error) {
var conf *Config
var err *probe.Error
conf, err = LoadConfig()
if err != nil {
conf = &Config{
Version: "0.0.1",
MaxSize: 512000000,
NodeDiskMap: nil,
DonutName: "",
}
if err := quick.CheckData(conf); err != nil {
return nil, err.Trace()
}
}
a := API{config: conf}
a.storedBuckets = metadata.NewCache()
a.nodes = make(map[string]node)
a.buckets = make(map[string]bucket)
a.objects = data.NewCache(a.config.MaxSize)
a.multiPartObjects = make(map[string]*data.Cache)
a.objects.OnEvicted = a.evictedObject
a.lock = new(sync.Mutex)
if len(a.config.NodeDiskMap) > 0 {
for k, v := range a.config.NodeDiskMap {
if len(v) == 0 {
return nil, probe.NewError(InvalidDisksArgument{})
}
err := a.AttachNode(k, v)
if err != nil {
return nil, err.Trace()
}
}
/// Initialization, populate all buckets into memory
buckets, err := a.listBuckets()
if err != nil {
return nil, err.Trace()
}
for k, v := range buckets {
var newBucket = storedBucket{}
newBucket.bucketMetadata = v
newBucket.objectMetadata = make(map[string]ObjectMetadata)
newBucket.multiPartSession = make(map[string]MultiPartSession)
newBucket.partMetadata = make(map[string]map[int]PartMetadata)
a.storedBuckets.Set(k, newBucket)
}
a.Heal()
}
return a, nil
}
/// V2 API functions
// GetObject - GET object from cache buffer
func (donut API) GetObject(w io.Writer, bucket string, object string, start, length int64) (int64, *probe.Error) {
donut.lock.Lock()
defer donut.lock.Unlock()
if !IsValidBucket(bucket) {
return 0, probe.NewError(BucketNameInvalid{Bucket: bucket})
}
if !IsValidObjectName(object) {
return 0, probe.NewError(ObjectNameInvalid{Object: object})
}
if start < 0 {
return 0, probe.NewError(InvalidRange{
Start: start,
Length: length,
})
}
if !donut.storedBuckets.Exists(bucket) {
return 0, probe.NewError(BucketNotFound{Bucket: bucket})
}
objectKey := bucket + "/" + object
data, ok := donut.objects.Get(objectKey)
var written int64
if !ok {
if len(donut.config.NodeDiskMap) > 0 {
reader, size, err := donut.getObject(bucket, object)
if err != nil {
return 0, err.Trace()
}
if start > 0 {
if _, err := io.CopyN(ioutil.Discard, reader, start); err != nil {
return 0, probe.NewError(err)
}
}
// new proxy writer to capture data read from disk
pw := NewProxyWriter(w)
{
var err error
if length > 0 {
written, err = io.CopyN(pw, reader, length)
if err != nil {
return 0, probe.NewError(err)
}
} else {
written, err = io.CopyN(pw, reader, size)
if err != nil {
return 0, probe.NewError(err)
}
}
}
/// cache object read from disk
ok := donut.objects.Append(objectKey, pw.writtenBytes)
pw.writtenBytes = nil
go debug.FreeOSMemory()
if !ok {
return 0, probe.NewError(InternalError{})
}
return written, nil
}
return 0, probe.NewError(ObjectNotFound{Object: object})
}
var err error
if start == 0 && length == 0 {
written, err = io.CopyN(w, bytes.NewBuffer(data), int64(donut.objects.Len(objectKey)))
if err != nil {
return 0, probe.NewError(err)
}
return written, nil
}
written, err = io.CopyN(w, bytes.NewBuffer(data[start:]), length)
if err != nil {
return 0, probe.NewError(err)
}
return written, nil
}
// GetBucketMetadata -
func (donut API) GetBucketMetadata(bucket string) (BucketMetadata, *probe.Error) {
donut.lock.Lock()
defer donut.lock.Unlock()
if !IsValidBucket(bucket) {
return BucketMetadata{}, probe.NewError(BucketNameInvalid{Bucket: bucket})
}
if !donut.storedBuckets.Exists(bucket) {
if len(donut.config.NodeDiskMap) > 0 {
bucketMetadata, err := donut.getBucketMetadata(bucket)
if err != nil {
return BucketMetadata{}, err.Trace()
}
storedBucket := donut.storedBuckets.Get(bucket).(storedBucket)
storedBucket.bucketMetadata = bucketMetadata
donut.storedBuckets.Set(bucket, storedBucket)
}
return BucketMetadata{}, probe.NewError(BucketNotFound{Bucket: bucket})
}
return donut.storedBuckets.Get(bucket).(storedBucket).bucketMetadata, nil
}
// SetBucketMetadata -
func (donut API) SetBucketMetadata(bucket string, metadata map[string]string) *probe.Error {
donut.lock.Lock()
defer donut.lock.Unlock()
if !IsValidBucket(bucket) {
return probe.NewError(BucketNameInvalid{Bucket: bucket})
}
if !donut.storedBuckets.Exists(bucket) {
return probe.NewError(BucketNotFound{Bucket: bucket})
}
if len(donut.config.NodeDiskMap) > 0 {
if err := donut.setBucketMetadata(bucket, metadata); err != nil {
return err.Trace()
}
}
storedBucket := donut.storedBuckets.Get(bucket).(storedBucket)
storedBucket.bucketMetadata.ACL = BucketACL(metadata["acl"])
donut.storedBuckets.Set(bucket, storedBucket)
return nil
}
// isMD5SumEqual - returns error if md5sum mismatches, success its `nil`
func isMD5SumEqual(expectedMD5Sum, actualMD5Sum string) *probe.Error {
if strings.TrimSpace(expectedMD5Sum) != "" && strings.TrimSpace(actualMD5Sum) != "" {
expectedMD5SumBytes, err := hex.DecodeString(expectedMD5Sum)
if err != nil {
return probe.NewError(err)
}
actualMD5SumBytes, err := hex.DecodeString(actualMD5Sum)
if err != nil {
return probe.NewError(err)
}
if !bytes.Equal(expectedMD5SumBytes, actualMD5SumBytes) {
return probe.NewError(BadDigest{})
}
return nil
}
return probe.NewError(InvalidArgument{})
}
// CreateObject - create an object
func (donut API) CreateObject(bucket, key, expectedMD5Sum string, size int64, data io.Reader, metadata map[string]string, signature *signv4.Signature) (ObjectMetadata, *probe.Error) {
donut.lock.Lock()
defer donut.lock.Unlock()
contentType := metadata["contentType"]
objectMetadata, err := donut.createObject(bucket, key, contentType, expectedMD5Sum, size, data, signature)
// free
debug.FreeOSMemory()
return objectMetadata, err.Trace()
}
// createObject - PUT object to cache buffer
func (donut API) createObject(bucket, key, contentType, expectedMD5Sum string, size int64, data io.Reader, signature *signv4.Signature) (ObjectMetadata, *probe.Error) {
if len(donut.config.NodeDiskMap) == 0 {
if size > int64(donut.config.MaxSize) {
generic := GenericObjectError{Bucket: bucket, Object: key}
return ObjectMetadata{}, probe.NewError(EntityTooLarge{
GenericObjectError: generic,
Size: strconv.FormatInt(size, 10),
MaxSize: strconv.FormatUint(donut.config.MaxSize, 10),
})
}
}
if !IsValidBucket(bucket) {
return ObjectMetadata{}, probe.NewError(BucketNameInvalid{Bucket: bucket})
}
if !IsValidObjectName(key) {
return ObjectMetadata{}, probe.NewError(ObjectNameInvalid{Object: key})
}
if !donut.storedBuckets.Exists(bucket) {
return ObjectMetadata{}, probe.NewError(BucketNotFound{Bucket: bucket})
}
storedBucket := donut.storedBuckets.Get(bucket).(storedBucket)
// get object key
objectKey := bucket + "/" + key
if _, ok := storedBucket.objectMetadata[objectKey]; ok == true {
return ObjectMetadata{}, probe.NewError(ObjectExists{Object: key})
}
if contentType == "" {
contentType = "application/octet-stream"
}
contentType = strings.TrimSpace(contentType)
if strings.TrimSpace(expectedMD5Sum) != "" {
expectedMD5SumBytes, err := base64.StdEncoding.DecodeString(strings.TrimSpace(expectedMD5Sum))
if err != nil {
// pro-actively close the connection
return ObjectMetadata{}, probe.NewError(InvalidDigest{Md5: expectedMD5Sum})
}
expectedMD5Sum = hex.EncodeToString(expectedMD5SumBytes)
}
if len(donut.config.NodeDiskMap) > 0 {
objMetadata, err := donut.putObject(
bucket,
key,
expectedMD5Sum,
data,
size,
map[string]string{
"contentType": contentType,
"contentLength": strconv.FormatInt(size, 10),
},
signature,
)
if err != nil {
return ObjectMetadata{}, err.Trace()
}
storedBucket.objectMetadata[objectKey] = objMetadata
donut.storedBuckets.Set(bucket, storedBucket)
return objMetadata, nil
}
// calculate md5
hash := md5.New()
sha256hash := sha256.New()
var err error
var totalLength int64
for err == nil {
var length int
byteBuffer := make([]byte, 1024*1024)
length, err = data.Read(byteBuffer)
if length != 0 {
hash.Write(byteBuffer[0:length])
sha256hash.Write(byteBuffer[0:length])
ok := donut.objects.Append(objectKey, byteBuffer[0:length])
if !ok {
return ObjectMetadata{}, probe.NewError(InternalError{})
}
totalLength += int64(length)
go debug.FreeOSMemory()
}
}
if size != 0 {
if totalLength != size {
// Delete perhaps the object is already saved, due to the nature of append()
donut.objects.Delete(objectKey)
return ObjectMetadata{}, probe.NewError(IncompleteBody{Bucket: bucket, Object: key})
}
}
if err != io.EOF {
return ObjectMetadata{}, probe.NewError(err)
}
md5SumBytes := hash.Sum(nil)
md5Sum := hex.EncodeToString(md5SumBytes)
// Verify if the written object is equal to what is expected, only if it is requested as such
if strings.TrimSpace(expectedMD5Sum) != "" {
if err := isMD5SumEqual(strings.TrimSpace(expectedMD5Sum), md5Sum); err != nil {
// Delete perhaps the object is already saved, due to the nature of append()
donut.objects.Delete(objectKey)
return ObjectMetadata{}, probe.NewError(BadDigest{})
}
}
if signature != nil {
ok, err := signature.DoesSignatureMatch(hex.EncodeToString(sha256hash.Sum(nil)))
if err != nil {
// Delete perhaps the object is already saved, due to the nature of append()
donut.objects.Delete(objectKey)
return ObjectMetadata{}, err.Trace()
}
if !ok {
// Delete perhaps the object is already saved, due to the nature of append()
donut.objects.Delete(objectKey)
return ObjectMetadata{}, probe.NewError(signv4.DoesNotMatch{})
}
}
m := make(map[string]string)
m["contentType"] = contentType
newObject := ObjectMetadata{
Bucket: bucket,
Object: key,
Metadata: m,
Created: time.Now().UTC(),
MD5Sum: md5Sum,
Size: int64(totalLength),
}
storedBucket.objectMetadata[objectKey] = newObject
donut.storedBuckets.Set(bucket, storedBucket)
return newObject, nil
}
// MakeBucket - create bucket in cache
func (donut API) MakeBucket(bucketName, acl string, location io.Reader, signature *signv4.Signature) *probe.Error {
donut.lock.Lock()
defer donut.lock.Unlock()
// do not have to parse location constraint, using this just for signature verification
locationSum := "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
if location != nil {
locationConstraintBytes, err := ioutil.ReadAll(location)
if err != nil {
return probe.NewError(InternalError{})
}
locationSum = hex.EncodeToString(sha256.Sum256(locationConstraintBytes)[:])
}
if signature != nil {
ok, err := signature.DoesSignatureMatch(locationSum)
if err != nil {
return err.Trace()
}
if !ok {
return probe.NewError(signv4.DoesNotMatch{})
}
}
if donut.storedBuckets.Stats().Items == totalBuckets {
return probe.NewError(TooManyBuckets{Bucket: bucketName})
}
if !IsValidBucket(bucketName) {
return probe.NewError(BucketNameInvalid{Bucket: bucketName})
}
if !IsValidBucketACL(acl) {
return probe.NewError(InvalidACL{ACL: acl})
}
if donut.storedBuckets.Exists(bucketName) {
return probe.NewError(BucketExists{Bucket: bucketName})
}
if strings.TrimSpace(acl) == "" {
// default is private
acl = "private"
}
if len(donut.config.NodeDiskMap) > 0 {
if err := donut.makeBucket(bucketName, BucketACL(acl)); err != nil {
return err.Trace()
}
}
var newBucket = storedBucket{}
newBucket.objectMetadata = make(map[string]ObjectMetadata)
newBucket.multiPartSession = make(map[string]MultiPartSession)
newBucket.partMetadata = make(map[string]map[int]PartMetadata)
newBucket.bucketMetadata = BucketMetadata{}
newBucket.bucketMetadata.Name = bucketName
newBucket.bucketMetadata.Created = time.Now().UTC()
newBucket.bucketMetadata.ACL = BucketACL(acl)
donut.storedBuckets.Set(bucketName, newBucket)
return nil
}
// ListObjects - list objects from cache
func (donut API) ListObjects(bucket string, resources BucketResourcesMetadata) ([]ObjectMetadata, BucketResourcesMetadata, *probe.Error) {
donut.lock.Lock()
defer donut.lock.Unlock()
if !IsValidBucket(bucket) {
return nil, BucketResourcesMetadata{IsTruncated: false}, probe.NewError(BucketNameInvalid{Bucket: bucket})
}
if !IsValidPrefix(resources.Prefix) {
return nil, BucketResourcesMetadata{IsTruncated: false}, probe.NewError(ObjectNameInvalid{Object: resources.Prefix})
}
if !donut.storedBuckets.Exists(bucket) {
return nil, BucketResourcesMetadata{IsTruncated: false}, probe.NewError(BucketNotFound{Bucket: bucket})
}
var results []ObjectMetadata
var keys []string
if len(donut.config.NodeDiskMap) > 0 {
listObjects, err := donut.listObjects(
bucket,
resources.Prefix,
resources.Marker,
resources.Delimiter,
resources.Maxkeys,
)
if err != nil {
return nil, BucketResourcesMetadata{IsTruncated: false}, err.Trace()
}
resources.CommonPrefixes = listObjects.CommonPrefixes
resources.IsTruncated = listObjects.IsTruncated
for key := range listObjects.Objects {
keys = append(keys, key)
}
sort.Strings(keys)
for _, key := range keys {
results = append(results, listObjects.Objects[key])
}
if resources.IsTruncated && resources.Delimiter != "" {
resources.NextMarker = results[len(results)-1].Object
}
return results, resources, nil
}
storedBucket := donut.storedBuckets.Get(bucket).(storedBucket)
for key := range storedBucket.objectMetadata {
if strings.HasPrefix(key, bucket+"/") {
key = key[len(bucket)+1:]
if strings.HasPrefix(key, resources.Prefix) {
if key > resources.Marker {
keys = append(keys, key)
}
}
}
}
if strings.TrimSpace(resources.Prefix) != "" {
keys = TrimPrefix(keys, resources.Prefix)
}
var prefixes []string
var filteredKeys []string
filteredKeys = keys
if strings.TrimSpace(resources.Delimiter) != "" {
filteredKeys = HasNoDelimiter(keys, resources.Delimiter)
prefixes = HasDelimiter(keys, resources.Delimiter)
prefixes = SplitDelimiter(prefixes, resources.Delimiter)
prefixes = SortUnique(prefixes)
}
for _, commonPrefix := range prefixes {
resources.CommonPrefixes = append(resources.CommonPrefixes, resources.Prefix+commonPrefix)
}
filteredKeys = RemoveDuplicates(filteredKeys)
sort.Strings(filteredKeys)
for _, key := range filteredKeys {
if len(results) == resources.Maxkeys {
resources.IsTruncated = true
if resources.IsTruncated && resources.Delimiter != "" {
resources.NextMarker = results[len(results)-1].Object
}
return results, resources, nil
}
object := storedBucket.objectMetadata[bucket+"/"+resources.Prefix+key]
results = append(results, object)
}
resources.CommonPrefixes = RemoveDuplicates(resources.CommonPrefixes)
sort.Strings(resources.CommonPrefixes)
return results, resources, nil
}
// byBucketName is a type for sorting bucket metadata by bucket name
type byBucketName []BucketMetadata
func (b byBucketName) Len() int { return len(b) }
func (b byBucketName) Swap(i, j int) { b[i], b[j] = b[j], b[i] }
func (b byBucketName) Less(i, j int) bool { return b[i].Name < b[j].Name }
// ListBuckets - List buckets from cache
func (donut API) ListBuckets() ([]BucketMetadata, *probe.Error) {
donut.lock.Lock()
defer donut.lock.Unlock()
var results []BucketMetadata
if len(donut.config.NodeDiskMap) > 0 {
buckets, err := donut.listBuckets()
if err != nil {
return nil, err.Trace()
}
for _, bucketMetadata := range buckets {
results = append(results, bucketMetadata)
}
sort.Sort(byBucketName(results))
return results, nil
}
for _, bucket := range donut.storedBuckets.GetAll() {
results = append(results, bucket.(storedBucket).bucketMetadata)
}
sort.Sort(byBucketName(results))
return results, nil
}
// GetObjectMetadata - get object metadata from cache
func (donut API) GetObjectMetadata(bucket, key string) (ObjectMetadata, *probe.Error) {
donut.lock.Lock()
defer donut.lock.Unlock()
// check if bucket exists
if !IsValidBucket(bucket) {
return ObjectMetadata{}, probe.NewError(BucketNameInvalid{Bucket: bucket})
}
if !IsValidObjectName(key) {
return ObjectMetadata{}, probe.NewError(ObjectNameInvalid{Object: key})
}
if !donut.storedBuckets.Exists(bucket) {
return ObjectMetadata{}, probe.NewError(BucketNotFound{Bucket: bucket})
}
storedBucket := donut.storedBuckets.Get(bucket).(storedBucket)
objectKey := bucket + "/" + key
if objMetadata, ok := storedBucket.objectMetadata[objectKey]; ok == true {
return objMetadata, nil
}
if len(donut.config.NodeDiskMap) > 0 {
objMetadata, err := donut.getObjectMetadata(bucket, key)
if err != nil {
return ObjectMetadata{}, err.Trace()
}
// update
storedBucket.objectMetadata[objectKey] = objMetadata
donut.storedBuckets.Set(bucket, storedBucket)
return objMetadata, nil
}
return ObjectMetadata{}, probe.NewError(ObjectNotFound{Object: key})
}
// evictedObject callback function called when an item is evicted from memory
func (donut API) evictedObject(a ...interface{}) {
cacheStats := donut.objects.Stats()
log.Printf("CurrentSize: %d, CurrentItems: %d, TotalEvicted: %d",
cacheStats.Bytes, cacheStats.Items, cacheStats.Evicted)
key := a[0].(string)
// loop through all buckets
for _, bucket := range donut.storedBuckets.GetAll() {
delete(bucket.(storedBucket).objectMetadata, key)
}
debug.FreeOSMemory()
}
-265
View File
@@ -1,265 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or impliedc.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package donut
import (
"bytes"
"crypto/md5"
"encoding/base64"
"encoding/hex"
"io/ioutil"
"os"
"path/filepath"
"testing"
. "gopkg.in/check.v1"
)
func TestCache(t *testing.T) { TestingT(t) }
type MyCacheSuite struct {
root string
}
var _ = Suite(&MyCacheSuite{})
var dc Interface
func (s *MyCacheSuite) SetUpSuite(c *C) {
root, err := ioutil.TempDir(os.TempDir(), "donut-")
c.Assert(err, IsNil)
s.root = root
SetDonutConfigPath(filepath.Join(root, "donut.json"))
dc, _ = New()
// testing empty cache
var buckets []BucketMetadata
buckets, perr := dc.ListBuckets()
c.Assert(perr, IsNil)
c.Assert(len(buckets), Equals, 0)
}
func (s *MyCacheSuite) TearDownSuite(c *C) {
os.RemoveAll(s.root)
}
// test make bucket without name
func (s *MyCacheSuite) TestBucketWithoutNameFails(c *C) {
// fail to create new bucket without a name
err := dc.MakeBucket("", "private", nil, nil)
c.Assert(err, Not(IsNil))
err = dc.MakeBucket(" ", "private", nil, nil)
c.Assert(err, Not(IsNil))
}
// test empty bucket
func (s *MyCacheSuite) TestEmptyBucket(c *C) {
c.Assert(dc.MakeBucket("foo1", "private", nil, nil), IsNil)
// check if bucket is empty
var resources BucketResourcesMetadata
resources.Maxkeys = 1
objectsMetadata, resources, err := dc.ListObjects("foo1", resources)
c.Assert(err, IsNil)
c.Assert(len(objectsMetadata), Equals, 0)
c.Assert(resources.CommonPrefixes, DeepEquals, []string{})
c.Assert(resources.IsTruncated, Equals, false)
}
// test bucket list
func (s *MyCacheSuite) TestMakeBucketAndList(c *C) {
// create bucket
err := dc.MakeBucket("foo2", "private", nil, nil)
c.Assert(err, IsNil)
// check bucket exists
buckets, err := dc.ListBuckets()
c.Assert(err, IsNil)
c.Assert(len(buckets), Equals, 5)
c.Assert(buckets[0].ACL, Equals, BucketACL("private"))
}
// test re-create bucket
func (s *MyCacheSuite) TestMakeBucketWithSameNameFails(c *C) {
err := dc.MakeBucket("foo3", "private", nil, nil)
c.Assert(err, IsNil)
err = dc.MakeBucket("foo3", "private", nil, nil)
c.Assert(err, Not(IsNil))
}
// test make multiple buckets
func (s *MyCacheSuite) TestCreateMultipleBucketsAndList(c *C) {
// add a second bucket
err := dc.MakeBucket("foo4", "private", nil, nil)
c.Assert(err, IsNil)
err = dc.MakeBucket("bar1", "private", nil, nil)
c.Assert(err, IsNil)
buckets, err := dc.ListBuckets()
c.Assert(err, IsNil)
c.Assert(len(buckets), Equals, 2)
c.Assert(buckets[0].Name, Equals, "bar1")
c.Assert(buckets[1].Name, Equals, "foo4")
err = dc.MakeBucket("foobar1", "private", nil, nil)
c.Assert(err, IsNil)
buckets, err = dc.ListBuckets()
c.Assert(err, IsNil)
c.Assert(len(buckets), Equals, 3)
c.Assert(buckets[2].Name, Equals, "foobar1")
}
// test object create without bucket
func (s *MyCacheSuite) TestNewObjectFailsWithoutBucket(c *C) {
_, err := dc.CreateObject("unknown", "obj", "", 0, nil, nil, nil)
c.Assert(err, Not(IsNil))
}
// test create object metadata
func (s *MyCacheSuite) TestNewObjectMetadata(c *C) {
data := "Hello World"
hasher := md5.New()
hasher.Write([]byte(data))
expectedMd5Sum := base64.StdEncoding.EncodeToString(hasher.Sum(nil))
reader := ioutil.NopCloser(bytes.NewReader([]byte(data)))
err := dc.MakeBucket("foo6", "private", nil, nil)
c.Assert(err, IsNil)
objectMetadata, err := dc.CreateObject("foo6", "obj", expectedMd5Sum, int64(len(data)), reader, map[string]string{"contentType": "application/json"}, nil)
c.Assert(err, IsNil)
c.Assert(objectMetadata.MD5Sum, Equals, hex.EncodeToString(hasher.Sum(nil)))
c.Assert(objectMetadata.Metadata["contentType"], Equals, "application/json")
}
// test create object fails without name
func (s *MyCacheSuite) TestNewObjectFailsWithEmptyName(c *C) {
_, err := dc.CreateObject("foo", "", "", 0, nil, nil, nil)
c.Assert(err, Not(IsNil))
}
// test create object
func (s *MyCacheSuite) TestNewObjectCanBeWritten(c *C) {
err := dc.MakeBucket("foo", "private", nil, nil)
c.Assert(err, IsNil)
data := "Hello World"
hasher := md5.New()
hasher.Write([]byte(data))
expectedMd5Sum := base64.StdEncoding.EncodeToString(hasher.Sum(nil))
reader := ioutil.NopCloser(bytes.NewReader([]byte(data)))
actualMetadata, err := dc.CreateObject("foo", "obj", expectedMd5Sum, int64(len(data)), reader, map[string]string{"contentType": "application/octet-stream"}, nil)
c.Assert(err, IsNil)
c.Assert(actualMetadata.MD5Sum, Equals, hex.EncodeToString(hasher.Sum(nil)))
var buffer bytes.Buffer
size, err := dc.GetObject(&buffer, "foo", "obj", 0, 0)
c.Assert(err, IsNil)
c.Assert(size, Equals, int64(len(data)))
c.Assert(buffer.Bytes(), DeepEquals, []byte(data))
actualMetadata, err = dc.GetObjectMetadata("foo", "obj")
c.Assert(err, IsNil)
c.Assert(hex.EncodeToString(hasher.Sum(nil)), Equals, actualMetadata.MD5Sum)
c.Assert(int64(len(data)), Equals, actualMetadata.Size)
}
// test list objects
func (s *MyCacheSuite) TestMultipleNewObjects(c *C) {
c.Assert(dc.MakeBucket("foo5", "private", nil, nil), IsNil)
one := ioutil.NopCloser(bytes.NewReader([]byte("one")))
_, err := dc.CreateObject("foo5", "obj1", "", int64(len("one")), one, nil, nil)
c.Assert(err, IsNil)
two := ioutil.NopCloser(bytes.NewReader([]byte("two")))
_, err = dc.CreateObject("foo5", "obj2", "", int64(len("two")), two, nil, nil)
c.Assert(err, IsNil)
var buffer1 bytes.Buffer
size, err := dc.GetObject(&buffer1, "foo5", "obj1", 0, 0)
c.Assert(err, IsNil)
c.Assert(size, Equals, int64(len([]byte("one"))))
c.Assert(buffer1.Bytes(), DeepEquals, []byte("one"))
var buffer2 bytes.Buffer
size, err = dc.GetObject(&buffer2, "foo5", "obj2", 0, 0)
c.Assert(err, IsNil)
c.Assert(size, Equals, int64(len([]byte("two"))))
c.Assert(buffer2.Bytes(), DeepEquals, []byte("two"))
/// test list of objects
// test list objects with prefix and delimiter
var resources BucketResourcesMetadata
resources.Prefix = "o"
resources.Delimiter = "1"
resources.Maxkeys = 10
objectsMetadata, resources, err := dc.ListObjects("foo5", resources)
c.Assert(err, IsNil)
c.Assert(resources.IsTruncated, Equals, false)
c.Assert(resources.CommonPrefixes[0], Equals, "obj1")
// test list objects with only delimiter
resources.Prefix = ""
resources.Delimiter = "1"
resources.Maxkeys = 10
objectsMetadata, resources, err = dc.ListObjects("foo5", resources)
c.Assert(err, IsNil)
c.Assert(objectsMetadata[0].Object, Equals, "obj2")
c.Assert(resources.IsTruncated, Equals, false)
c.Assert(resources.CommonPrefixes[0], Equals, "obj1")
// test list objects with only prefix
resources.Prefix = "o"
resources.Delimiter = ""
resources.Maxkeys = 10
objectsMetadata, resources, err = dc.ListObjects("foo5", resources)
c.Assert(err, IsNil)
c.Assert(resources.IsTruncated, Equals, false)
c.Assert(objectsMetadata[0].Object, Equals, "obj1")
c.Assert(objectsMetadata[1].Object, Equals, "obj2")
three := ioutil.NopCloser(bytes.NewReader([]byte("three")))
_, err = dc.CreateObject("foo5", "obj3", "", int64(len("three")), three, nil, nil)
c.Assert(err, IsNil)
var buffer bytes.Buffer
size, err = dc.GetObject(&buffer, "foo5", "obj3", 0, 0)
c.Assert(err, IsNil)
c.Assert(size, Equals, int64(len([]byte("three"))))
c.Assert(buffer.Bytes(), DeepEquals, []byte("three"))
// test list objects with maxkeys
resources.Prefix = "o"
resources.Delimiter = ""
resources.Maxkeys = 2
objectsMetadata, resources, err = dc.ListObjects("foo5", resources)
c.Assert(err, IsNil)
c.Assert(resources.IsTruncated, Equals, true)
c.Assert(len(objectsMetadata), Equals, 2)
}
-71
View File
@@ -1,71 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package donut
import (
encoding "github.com/minio/minio/pkg/erasure"
"github.com/minio/minio/pkg/probe"
)
// encoder internal struct
type encoder struct {
encoder *encoding.Erasure
k, m uint8
}
// newEncoder - instantiate a new encoder
func newEncoder(k, m uint8) (encoder, *probe.Error) {
e := encoder{}
params, err := encoding.ValidateParams(k, m)
if err != nil {
return encoder{}, probe.NewError(err)
}
e.encoder = encoding.NewErasure(params)
e.k = k
e.m = m
return e, nil
}
// TODO - think again if this is needed
// GetEncodedBlockLen - wrapper around erasure function with the same name
func (e encoder) GetEncodedBlockLen(dataLength int) (int, *probe.Error) {
if dataLength <= 0 {
return 0, probe.NewError(InvalidArgument{})
}
return encoding.GetEncodedBlockLen(dataLength, e.k), nil
}
// Encode - erasure code input bytes
func (e encoder) Encode(data []byte) ([][]byte, *probe.Error) {
if data == nil {
return nil, probe.NewError(InvalidArgument{})
}
encodedData, err := e.encoder.Encode(data)
if err != nil {
return nil, probe.NewError(err)
}
return encodedData, nil
}
// Decode - erasure decode input encoded bytes
func (e encoder) Decode(encodedData [][]byte, dataLength int) ([]byte, *probe.Error) {
decodedData, err := e.encoder.Decode(encodedData, dataLength)
if err != nil {
return nil, probe.NewError(err)
}
return decodedData, nil
}
-69
View File
@@ -1,69 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package donut
import (
"encoding/json"
"fmt"
"path/filepath"
"github.com/minio/minio/pkg/donut/disk"
"github.com/minio/minio/pkg/probe"
)
// healBuckets heal bucket slices
func (donut API) healBuckets() *probe.Error {
if err := donut.listDonutBuckets(); err != nil {
return err.Trace()
}
bucketMetadata, err := donut.getDonutBucketMetadata()
if err != nil {
return err.Trace()
}
disks := make(map[int]disk.Disk)
for _, node := range donut.nodes {
nDisks, err := node.ListDisks()
if err != nil {
return err.Trace()
}
for k, v := range nDisks {
disks[k] = v
}
}
for order, disk := range disks {
if disk.IsUsable() {
disk.MakeDir(donut.config.DonutName)
bucketMetadataWriter, err := disk.CreateFile(filepath.Join(donut.config.DonutName, bucketMetadataConfig))
if err != nil {
return err.Trace()
}
defer bucketMetadataWriter.Close()
jenc := json.NewEncoder(bucketMetadataWriter)
if err := jenc.Encode(bucketMetadata); err != nil {
return probe.NewError(err)
}
for bucket := range bucketMetadata.Buckets {
bucketSlice := fmt.Sprintf("%s$0$%d", bucket, order) // TODO handle node slices
err := disk.MakeDir(filepath.Join(donut.config.DonutName, bucketSlice))
if err != nil {
return err.Trace()
}
}
}
}
return nil
}
-72
View File
@@ -1,72 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package donut
import (
"io"
"github.com/minio/minio/pkg/probe"
signv4 "github.com/minio/minio/pkg/signature"
)
// Collection of Donut specification interfaces
// Interface is a collection of cloud storage and management interface
type Interface interface {
CloudStorage
Management
}
// CloudStorage is a donut cloud storage interface
type CloudStorage interface {
// Storage service operations
GetBucketMetadata(bucket string) (BucketMetadata, *probe.Error)
SetBucketMetadata(bucket string, metadata map[string]string) *probe.Error
ListBuckets() ([]BucketMetadata, *probe.Error)
MakeBucket(bucket string, ACL string, location io.Reader, signature *signv4.Signature) *probe.Error
// Bucket operations
ListObjects(string, BucketResourcesMetadata) ([]ObjectMetadata, BucketResourcesMetadata, *probe.Error)
// Object operations
GetObject(w io.Writer, bucket, object string, start, length int64) (int64, *probe.Error)
GetObjectMetadata(bucket, object string) (ObjectMetadata, *probe.Error)
// bucket, object, expectedMD5Sum, size, reader, metadata, signature
CreateObject(string, string, string, int64, io.Reader, map[string]string, *signv4.Signature) (ObjectMetadata, *probe.Error)
Multipart
}
// Multipart API
type Multipart interface {
NewMultipartUpload(bucket, key, contentType string) (string, *probe.Error)
AbortMultipartUpload(bucket, key, uploadID string) *probe.Error
CreateObjectPart(string, string, string, int, string, string, int64, io.Reader, *signv4.Signature) (string, *probe.Error)
CompleteMultipartUpload(bucket, key, uploadID string, data io.Reader, signature *signv4.Signature) (ObjectMetadata, *probe.Error)
ListMultipartUploads(string, BucketMultipartResourcesMetadata) (BucketMultipartResourcesMetadata, *probe.Error)
ListObjectParts(string, string, ObjectResourcesMetadata) (ObjectResourcesMetadata, *probe.Error)
}
// Management is a donut management system interface
type Management interface {
Heal() *probe.Error
Rebalance() *probe.Error
Info() (map[string][]string, *probe.Error)
AttachNode(hostname string, disks []string) *probe.Error
DetachNode(hostname string) *probe.Error
}
-81
View File
@@ -1,81 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package donut
import (
"github.com/minio/minio/pkg/donut/disk"
"github.com/minio/minio/pkg/probe"
)
// Info - return info about donut configuration
func (donut API) Info() (nodeDiskMap map[string][]string, err *probe.Error) {
nodeDiskMap = make(map[string][]string)
for nodeName, n := range donut.nodes {
disks, err := n.ListDisks()
if err != nil {
return nil, err.Trace()
}
diskList := make([]string, len(disks))
for diskOrder, disk := range disks {
diskList[diskOrder] = disk.GetPath()
}
nodeDiskMap[nodeName] = diskList
}
return nodeDiskMap, nil
}
// AttachNode - attach node
func (donut API) AttachNode(hostname string, disks []string) *probe.Error {
if hostname == "" || len(disks) == 0 {
return probe.NewError(InvalidArgument{})
}
n, err := newNode(hostname)
if err != nil {
return err.Trace()
}
donut.nodes[hostname] = n
for i, d := range disks {
newDisk, err := disk.New(d)
if err != nil {
continue
}
if err := newDisk.MakeDir(donut.config.DonutName); err != nil {
return err.Trace()
}
if err := n.AttachDisk(newDisk, i); err != nil {
return err.Trace()
}
}
return nil
}
// DetachNode - detach node
func (donut API) DetachNode(hostname string) *probe.Error {
delete(donut.nodes, hostname)
return nil
}
// Rebalance - rebalance an existing donut with new disks and nodes
func (donut API) Rebalance() *probe.Error {
return probe.NewError(APINotImplemented{API: "management.Rebalance"})
}
// Heal - heal your donuts
func (donut API) Heal() *probe.Error {
// TODO handle data heal
return donut.healBuckets()
}
-513
View File
@@ -1,513 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package donut
import (
"bytes"
"crypto/md5"
"crypto/sha512"
"encoding/base64"
"encoding/hex"
"encoding/xml"
"io"
"io/ioutil"
"math/rand"
"runtime/debug"
"sort"
"strconv"
"strings"
"time"
"github.com/minio/minio/pkg/crypto/sha256"
"github.com/minio/minio/pkg/donut/cache/data"
"github.com/minio/minio/pkg/probe"
signv4 "github.com/minio/minio/pkg/signature"
)
/// V2 API functions
// NewMultipartUpload - initiate a new multipart session
func (donut API) NewMultipartUpload(bucket, key, contentType string) (string, *probe.Error) {
donut.lock.Lock()
defer donut.lock.Unlock()
if !IsValidBucket(bucket) {
return "", probe.NewError(BucketNameInvalid{Bucket: bucket})
}
if !IsValidObjectName(key) {
return "", probe.NewError(ObjectNameInvalid{Object: key})
}
// if len(donut.config.NodeDiskMap) > 0 {
// return donut.newMultipartUpload(bucket, key, contentType)
// }
if !donut.storedBuckets.Exists(bucket) {
return "", probe.NewError(BucketNotFound{Bucket: bucket})
}
storedBucket := donut.storedBuckets.Get(bucket).(storedBucket)
objectKey := bucket + "/" + key
if _, ok := storedBucket.objectMetadata[objectKey]; ok == true {
return "", probe.NewError(ObjectExists{Object: key})
}
id := []byte(strconv.Itoa(rand.Int()) + bucket + key + time.Now().UTC().String())
uploadIDSum := sha512.Sum512(id)
uploadID := base64.URLEncoding.EncodeToString(uploadIDSum[:])[:47]
storedBucket.multiPartSession[key] = MultiPartSession{
UploadID: uploadID,
Initiated: time.Now().UTC(),
TotalParts: 0,
}
storedBucket.partMetadata[key] = make(map[int]PartMetadata)
multiPartCache := data.NewCache(0)
multiPartCache.OnEvicted = donut.evictedPart
donut.multiPartObjects[uploadID] = multiPartCache
donut.storedBuckets.Set(bucket, storedBucket)
return uploadID, nil
}
// AbortMultipartUpload - abort an incomplete multipart session
func (donut API) AbortMultipartUpload(bucket, key, uploadID string) *probe.Error {
donut.lock.Lock()
defer donut.lock.Unlock()
if !IsValidBucket(bucket) {
return probe.NewError(BucketNameInvalid{Bucket: bucket})
}
if !IsValidObjectName(key) {
return probe.NewError(ObjectNameInvalid{Object: key})
}
// TODO: multipart support for donut is broken, since we haven't finalized the format in which
// it can be stored, disabling this for now until we get the underlying layout stable.
//
// if len(donut.config.NodeDiskMap) > 0 {
// return donut.abortMultipartUpload(bucket, key, uploadID)
// }
if !donut.storedBuckets.Exists(bucket) {
return probe.NewError(BucketNotFound{Bucket: bucket})
}
storedBucket := donut.storedBuckets.Get(bucket).(storedBucket)
if storedBucket.multiPartSession[key].UploadID != uploadID {
return probe.NewError(InvalidUploadID{UploadID: uploadID})
}
donut.cleanupMultipartSession(bucket, key, uploadID)
return nil
}
// CreateObjectPart - create a part in a multipart session
func (donut API) CreateObjectPart(bucket, key, uploadID string, partID int, contentType, expectedMD5Sum string, size int64, data io.Reader, signature *signv4.Signature) (string, *probe.Error) {
donut.lock.Lock()
etag, err := donut.createObjectPart(bucket, key, uploadID, partID, "", expectedMD5Sum, size, data, signature)
donut.lock.Unlock()
// possible free
debug.FreeOSMemory()
return etag, err.Trace()
}
// createObject - internal wrapper function called by CreateObjectPart
func (donut API) createObjectPart(bucket, key, uploadID string, partID int, contentType, expectedMD5Sum string, size int64, data io.Reader, signature *signv4.Signature) (string, *probe.Error) {
if !IsValidBucket(bucket) {
return "", probe.NewError(BucketNameInvalid{Bucket: bucket})
}
if !IsValidObjectName(key) {
return "", probe.NewError(ObjectNameInvalid{Object: key})
}
// TODO: multipart support for donut is broken, since we haven't finalized the format in which
// it can be stored, disabling this for now until we get the underlying layout stable.
//
/*
if len(donut.config.NodeDiskMap) > 0 {
metadata := make(map[string]string)
if contentType == "" {
contentType = "application/octet-stream"
}
contentType = strings.TrimSpace(contentType)
metadata["contentType"] = contentType
if strings.TrimSpace(expectedMD5Sum) != "" {
expectedMD5SumBytes, err := base64.StdEncoding.DecodeString(strings.TrimSpace(expectedMD5Sum))
if err != nil {
// pro-actively close the connection
return "", probe.NewError(InvalidDigest{Md5: expectedMD5Sum})
}
expectedMD5Sum = hex.EncodeToString(expectedMD5SumBytes)
}
partMetadata, err := donut.putObjectPart(bucket, key, expectedMD5Sum, uploadID, partID, data, size, metadata, signature)
if err != nil {
return "", err.Trace()
}
return partMetadata.ETag, nil
}
*/
if !donut.storedBuckets.Exists(bucket) {
return "", probe.NewError(BucketNotFound{Bucket: bucket})
}
strBucket := donut.storedBuckets.Get(bucket).(storedBucket)
// Verify upload id
if strBucket.multiPartSession[key].UploadID != uploadID {
return "", probe.NewError(InvalidUploadID{UploadID: uploadID})
}
// get object key
parts := strBucket.partMetadata[key]
if _, ok := parts[partID]; ok {
return parts[partID].ETag, nil
}
if contentType == "" {
contentType = "application/octet-stream"
}
contentType = strings.TrimSpace(contentType)
if strings.TrimSpace(expectedMD5Sum) != "" {
expectedMD5SumBytes, err := base64.StdEncoding.DecodeString(strings.TrimSpace(expectedMD5Sum))
if err != nil {
// pro-actively close the connection
return "", probe.NewError(InvalidDigest{Md5: expectedMD5Sum})
}
expectedMD5Sum = hex.EncodeToString(expectedMD5SumBytes)
}
// calculate md5
hash := md5.New()
sha256hash := sha256.New()
var totalLength int64
var err error
for err == nil {
var length int
byteBuffer := make([]byte, 1024*1024)
length, err = data.Read(byteBuffer) // do not read error return error here, we will handle this error later
if length != 0 {
hash.Write(byteBuffer[0:length])
sha256hash.Write(byteBuffer[0:length])
ok := donut.multiPartObjects[uploadID].Append(partID, byteBuffer[0:length])
if !ok {
return "", probe.NewError(InternalError{})
}
totalLength += int64(length)
go debug.FreeOSMemory()
}
}
if totalLength != size {
donut.multiPartObjects[uploadID].Delete(partID)
return "", probe.NewError(IncompleteBody{Bucket: bucket, Object: key})
}
if err != io.EOF {
return "", probe.NewError(err)
}
md5SumBytes := hash.Sum(nil)
md5Sum := hex.EncodeToString(md5SumBytes)
// Verify if the written object is equal to what is expected, only if it is requested as such
if strings.TrimSpace(expectedMD5Sum) != "" {
if err := isMD5SumEqual(strings.TrimSpace(expectedMD5Sum), md5Sum); err != nil {
return "", err.Trace()
}
}
if signature != nil {
{
ok, err := signature.DoesSignatureMatch(hex.EncodeToString(sha256hash.Sum(nil)))
if err != nil {
return "", err.Trace()
}
if !ok {
return "", probe.NewError(signv4.DoesNotMatch{})
}
}
}
newPart := PartMetadata{
PartNumber: partID,
LastModified: time.Now().UTC(),
ETag: md5Sum,
Size: totalLength,
}
parts[partID] = newPart
strBucket.partMetadata[key] = parts
multiPartSession := strBucket.multiPartSession[key]
multiPartSession.TotalParts++
strBucket.multiPartSession[key] = multiPartSession
donut.storedBuckets.Set(bucket, strBucket)
return md5Sum, nil
}
// cleanupMultipartSession invoked during an abort or complete multipart session to cleanup session from memory
func (donut API) cleanupMultipartSession(bucket, key, uploadID string) {
storedBucket := donut.storedBuckets.Get(bucket).(storedBucket)
for i := 1; i <= storedBucket.multiPartSession[key].TotalParts; i++ {
donut.multiPartObjects[uploadID].Delete(i)
}
delete(storedBucket.multiPartSession, key)
delete(storedBucket.partMetadata, key)
donut.storedBuckets.Set(bucket, storedBucket)
}
func (donut API) mergeMultipart(parts *CompleteMultipartUpload, uploadID string, fullObjectWriter *io.PipeWriter) {
for _, part := range parts.Part {
recvMD5 := part.ETag
object, ok := donut.multiPartObjects[uploadID].Get(part.PartNumber)
if ok == false {
fullObjectWriter.CloseWithError(probe.WrapError(probe.NewError(InvalidPart{})))
return
}
calcMD5Bytes := md5.Sum(object)
// complete multi part request header md5sum per part is hex encoded
recvMD5Bytes, err := hex.DecodeString(strings.Trim(recvMD5, "\""))
if err != nil {
fullObjectWriter.CloseWithError(probe.WrapError(probe.NewError(InvalidDigest{Md5: recvMD5})))
return
}
if !bytes.Equal(recvMD5Bytes, calcMD5Bytes[:]) {
fullObjectWriter.CloseWithError(probe.WrapError(probe.NewError(BadDigest{})))
return
}
if _, err := io.Copy(fullObjectWriter, bytes.NewReader(object)); err != nil {
fullObjectWriter.CloseWithError(probe.WrapError(probe.NewError(err)))
return
}
object = nil
}
fullObjectWriter.Close()
return
}
// CompleteMultipartUpload - complete a multipart upload and persist the data
func (donut API) CompleteMultipartUpload(bucket, key, uploadID string, data io.Reader, signature *signv4.Signature) (ObjectMetadata, *probe.Error) {
donut.lock.Lock()
defer donut.lock.Unlock()
size := int64(donut.multiPartObjects[uploadID].Stats().Bytes)
fullObjectReader, err := donut.completeMultipartUploadV2(bucket, key, uploadID, data, signature)
if err != nil {
return ObjectMetadata{}, err.Trace()
}
objectMetadata, err := donut.createObject(bucket, key, "", "", size, fullObjectReader, nil)
if err != nil {
// No need to call internal cleanup functions here, caller should call AbortMultipartUpload()
// which would in-turn cleanup properly in accordance with S3 Spec
return ObjectMetadata{}, err.Trace()
}
donut.cleanupMultipartSession(bucket, key, uploadID)
return objectMetadata, nil
}
func (donut API) completeMultipartUploadV2(bucket, key, uploadID string, data io.Reader, signature *signv4.Signature) (io.Reader, *probe.Error) {
if !IsValidBucket(bucket) {
return nil, probe.NewError(BucketNameInvalid{Bucket: bucket})
}
if !IsValidObjectName(key) {
return nil, probe.NewError(ObjectNameInvalid{Object: key})
}
// TODO: multipart support for donut is broken, since we haven't finalized the format in which
// it can be stored, disabling this for now until we get the underlying layout stable.
//
// if len(donut.config.NodeDiskMap) > 0 {
// donut.lock.Unlock()
// return donut.completeMultipartUpload(bucket, key, uploadID, data, signature)
// }
if !donut.storedBuckets.Exists(bucket) {
return nil, probe.NewError(BucketNotFound{Bucket: bucket})
}
storedBucket := donut.storedBuckets.Get(bucket).(storedBucket)
// Verify upload id
if storedBucket.multiPartSession[key].UploadID != uploadID {
return nil, probe.NewError(InvalidUploadID{UploadID: uploadID})
}
partBytes, err := ioutil.ReadAll(data)
if err != nil {
return nil, probe.NewError(err)
}
if signature != nil {
ok, err := signature.DoesSignatureMatch(hex.EncodeToString(sha256.Sum256(partBytes)[:]))
if err != nil {
return nil, err.Trace()
}
if !ok {
return nil, probe.NewError(signv4.DoesNotMatch{})
}
}
parts := &CompleteMultipartUpload{}
if err := xml.Unmarshal(partBytes, parts); err != nil {
return nil, probe.NewError(MalformedXML{})
}
if !sort.IsSorted(completedParts(parts.Part)) {
return nil, probe.NewError(InvalidPartOrder{})
}
fullObjectReader, fullObjectWriter := io.Pipe()
go donut.mergeMultipart(parts, uploadID, fullObjectWriter)
return fullObjectReader, nil
}
// byKey is a sortable interface for UploadMetadata slice
type byKey []*UploadMetadata
func (a byKey) Len() int { return len(a) }
func (a byKey) Swap(i, j int) { a[i], a[j] = a[j], a[i] }
func (a byKey) Less(i, j int) bool { return a[i].Key < a[j].Key }
// ListMultipartUploads - list incomplete multipart sessions for a given bucket
func (donut API) ListMultipartUploads(bucket string, resources BucketMultipartResourcesMetadata) (BucketMultipartResourcesMetadata, *probe.Error) {
// TODO handle delimiter, low priority
donut.lock.Lock()
defer donut.lock.Unlock()
if !IsValidBucket(bucket) {
return BucketMultipartResourcesMetadata{}, probe.NewError(BucketNameInvalid{Bucket: bucket})
}
// TODO: multipart support for donut is broken, since we haven't finalized the format in which
// it can be stored, disabling this for now until we get the underlying layout stable.
//
// if len(donut.config.NodeDiskMap) > 0 {
// return donut.listMultipartUploads(bucket, resources)
// }
if !donut.storedBuckets.Exists(bucket) {
return BucketMultipartResourcesMetadata{}, probe.NewError(BucketNotFound{Bucket: bucket})
}
storedBucket := donut.storedBuckets.Get(bucket).(storedBucket)
var uploads []*UploadMetadata
for key, session := range storedBucket.multiPartSession {
if strings.HasPrefix(key, resources.Prefix) {
if len(uploads) > resources.MaxUploads {
sort.Sort(byKey(uploads))
resources.Upload = uploads
resources.NextKeyMarker = key
resources.NextUploadIDMarker = session.UploadID
resources.IsTruncated = true
return resources, nil
}
// uploadIDMarker is ignored if KeyMarker is empty
switch {
case resources.KeyMarker != "" && resources.UploadIDMarker == "":
if key > resources.KeyMarker {
upload := new(UploadMetadata)
upload.Key = key
upload.UploadID = session.UploadID
upload.Initiated = session.Initiated
uploads = append(uploads, upload)
}
case resources.KeyMarker != "" && resources.UploadIDMarker != "":
if session.UploadID > resources.UploadIDMarker {
if key >= resources.KeyMarker {
upload := new(UploadMetadata)
upload.Key = key
upload.UploadID = session.UploadID
upload.Initiated = session.Initiated
uploads = append(uploads, upload)
}
}
default:
upload := new(UploadMetadata)
upload.Key = key
upload.UploadID = session.UploadID
upload.Initiated = session.Initiated
uploads = append(uploads, upload)
}
}
}
sort.Sort(byKey(uploads))
resources.Upload = uploads
return resources, nil
}
// partNumber is a sortable interface for Part slice
type partNumber []*PartMetadata
func (a partNumber) Len() int { return len(a) }
func (a partNumber) Swap(i, j int) { a[i], a[j] = a[j], a[i] }
func (a partNumber) Less(i, j int) bool { return a[i].PartNumber < a[j].PartNumber }
// ListObjectParts - list parts from incomplete multipart session for a given object
func (donut API) ListObjectParts(bucket, key string, resources ObjectResourcesMetadata) (ObjectResourcesMetadata, *probe.Error) {
// Verify upload id
donut.lock.Lock()
defer donut.lock.Unlock()
if !IsValidBucket(bucket) {
return ObjectResourcesMetadata{}, probe.NewError(BucketNameInvalid{Bucket: bucket})
}
if !IsValidObjectName(key) {
return ObjectResourcesMetadata{}, probe.NewError(ObjectNameInvalid{Object: key})
}
// TODO: multipart support for donut is broken, since we haven't finalized the format in which
// it can be stored, disabling this for now until we get the underlying layout stable.
//
// if len(donut.config.NodeDiskMap) > 0 {
// return donut.listObjectParts(bucket, key, resources)
// }
if !donut.storedBuckets.Exists(bucket) {
return ObjectResourcesMetadata{}, probe.NewError(BucketNotFound{Bucket: bucket})
}
storedBucket := donut.storedBuckets.Get(bucket).(storedBucket)
if _, ok := storedBucket.multiPartSession[key]; ok == false {
return ObjectResourcesMetadata{}, probe.NewError(ObjectNotFound{Object: key})
}
if storedBucket.multiPartSession[key].UploadID != resources.UploadID {
return ObjectResourcesMetadata{}, probe.NewError(InvalidUploadID{UploadID: resources.UploadID})
}
storedParts := storedBucket.partMetadata[key]
objectResourcesMetadata := resources
objectResourcesMetadata.Bucket = bucket
objectResourcesMetadata.Key = key
var parts []*PartMetadata
var startPartNumber int
switch {
case objectResourcesMetadata.PartNumberMarker == 0:
startPartNumber = 1
default:
startPartNumber = objectResourcesMetadata.PartNumberMarker
}
for i := startPartNumber; i <= storedBucket.multiPartSession[key].TotalParts; i++ {
if len(parts) > objectResourcesMetadata.MaxParts {
sort.Sort(partNumber(parts))
objectResourcesMetadata.IsTruncated = true
objectResourcesMetadata.Part = parts
objectResourcesMetadata.NextPartNumberMarker = i
return objectResourcesMetadata, nil
}
part, ok := storedParts[i]
if !ok {
return ObjectResourcesMetadata{}, probe.NewError(InvalidPart{})
}
parts = append(parts, &part)
}
sort.Sort(partNumber(parts))
objectResourcesMetadata.Part = parts
return objectResourcesMetadata, nil
}
// evictedPart - call back function called by caching module during individual cache evictions
func (donut API) evictedPart(a ...interface{}) {
// loop through all buckets
buckets := donut.storedBuckets.GetAll()
for bucketName, bucket := range buckets {
b := bucket.(storedBucket)
donut.storedBuckets.Set(bucketName, b)
}
debug.FreeOSMemory()
}
-76
View File
@@ -1,76 +0,0 @@
/*
* Minio Cloud Storage, (C) 2015 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package donut
import (
"github.com/minio/minio/pkg/donut/disk"
"github.com/minio/minio/pkg/probe"
)
// node struct internal
type node struct {
hostname string
disks map[int]disk.Disk
}
// newNode - instantiates a new node
func newNode(hostname string) (node, *probe.Error) {
if hostname == "" {
return node{}, probe.NewError(InvalidArgument{})
}
disks := make(map[int]disk.Disk)
n := node{
hostname: hostname,
disks: disks,
}
return n, nil
}
// GetHostname - return hostname
func (n node) GetHostname() string {
return n.hostname
}
// ListDisks - return number of disks
func (n node) ListDisks() (map[int]disk.Disk, *probe.Error) {
return n.disks, nil
}
// AttachDisk - attach a disk
func (n node) AttachDisk(disk disk.Disk, diskOrder int) *probe.Error {
if diskOrder < 0 {
return probe.NewError(InvalidArgument{})
}
n.disks[diskOrder] = disk
return nil
}
// DetachDisk - detach a disk
func (n node) DetachDisk(diskOrder int) *probe.Error {
delete(n.disks, diskOrder)
return nil
}
// SaveConfig - save node configuration
func (n node) SaveConfig() *probe.Error {
return probe.NewError(NotImplemented{Function: "SaveConfig"})
}
// LoadConfig - load node configuration from saved configs
func (n node) LoadConfig() *probe.Error {
return probe.NewError(NotImplemented{Function: "LoadConfig"})
}
-1
View File
@@ -1 +0,0 @@
*.syso
-64
View File
@@ -1,64 +0,0 @@
## Ubuntu (Kylin) 14.04
### Build Dependencies
This installation document assumes Ubuntu 14.04+ on x86-64 platform.
##### Install Git, GCC, yasm
```sh
$ sudo apt-get install git build-essential yasm
```
##### Install Go 1.5+
Download Go 1.5+ from [https://golang.org/dl/](https://golang.org/dl/).
```sh
$ wget https://storage.googleapis.com/golang/go1.5.linux-amd64.tar.gz
$ mkdir -p ${HOME}/bin/
$ mkdir -p ${HOME}/go/
$ tar -C ${HOME}/bin/ -xzf go1.5.linux-amd64.tar.gz
```
##### Setup GOROOT and GOPATH
Add the following exports to your ``~/.bashrc``. Environment variable GOROOT specifies the location of your golang binaries
and GOPATH specifies the location of your project workspace.
```sh
$ export GOROOT=${HOME}/bin/go
$ export GOPATH=${HOME}/go
$ export PATH=$PATH:${HOME}/bin/go/bin:${GOPATH}/bin
```
## OS X (Yosemite) 10.10
### Build Dependencies
This installation document assumes OS X Yosemite 10.10+ on x86-64 platform.
##### Install brew
```sh
$ ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"
```
##### Install Git, Python
```sh
$ brew install git python yasm
```
##### Install Go 1.5+
Install golang binaries using `brew`
```sh
$ brew install go
$ mkdir -p $HOME/go
```
##### Setup GOROOT and GOPATH
Add the following exports to your ``~/.bashrc``. Environment variable GOROOT specifies the location of your golang binaries
and GOPATH specifies the location of your project workspace.
```sh
$ export GOPATH=${HOME}/go
$ export GOVERSION=$(brew list go | head -n 1 | cut -d '/' -f 6)
$ export GOROOT=$(brew --prefix)/Cellar/go/${GOVERSION}/libexec
$ export PATH=$PATH:${GOPATH}/bin
```
-26
View File
@@ -1,26 +0,0 @@
Copyright(c) 2011-2014 Intel Corporation All rights reserved.
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 Intel Corporation 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"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 THE COPYRIGHT
OWNER OR CONTRIBUTORS 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.
-202
View File
@@ -1,202 +0,0 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-25
View File
@@ -1,25 +0,0 @@
## Introduction
Erasure is an open source Golang library written on top of ISAL (Intel Intelligent Storage Library) released under [Apache license v2](./LICENSE)
### Developers
* [Get Source](./CONTRIBUTING.md)
* [Build Dependencies](./BUILDDEPS.md)
* [Development Workflow](./CONTRIBUTING.md#developer-guidelines)
* [Developer discussions and bugs](https://github.com/minio/minio/issues)
### Supported platforms
| Name | Supported |
| ------------- | ------------- |
| Linux | Yes |
| Windows | Not yet |
| Mac OSX | Yes |
### Supported architectures
| Arch | Supported |
| ------------- | ------------- |
| x86-64 | Yes |
| arm64 | Not yet|
| i386 | Never |
-49
View File
@@ -1,49 +0,0 @@
================================================================================
v2.10 Intel Intelligent Storage Acceleration Library Release Notes
Open Source Version
================================================================================
================================================================================
RELEASE NOTE CONTENTS
================================================================================
1. KNOWN ISSUES
2. FIXED ISSUES
3. CHANGE LOG & FEATURES ADDED
================================================================================
1. KNOWN ISSUES
================================================================================
* Only erasure code unit included in open source version at this time.
* Perf tests do not run in Windows environment.
* Leaving <unit>/bin directories from builds in unit directories will cause the
top-level make build to fail. Build only in top-level or ensure unit
directories are clean of objects and /bin.
* 32-bit lib is not supported in Windows.
================================================================================
2. FIXED ISSUES
================================================================================
v2.10
* Fix for windows register save overlap in gf_{3-6}vect_dot_prod_sse.asm. Only
affects windows versions of erasure code. GP register saves/restore were
pushed to same stack area as XMM.
================================================================================
3. CHANGE LOG & FEATURES ADDED
================================================================================
v2.10
* Erasure code updates
- New AVX and AVX2 support functions.
- Changes min len requirement on gf_vect_dot_prod() to 32 from 16.
- Tests include both source and parity recovery with ec_encode_data().
- New encoding examples with Vandermonde or Cauchy matrix.
v2.8
* First open release of erasure code unit that is part of ISA-L.
-3
View File
@@ -1,3 +0,0 @@
v1.0 - Erasure Golang Package
============================
- First release, supports only amd64 or x86-64 architecture
-62
View File
@@ -1,62 +0,0 @@
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package erasure
// #include <stdint.h>
import "C"
import (
"fmt"
"unsafe"
)
// intSlice2CIntArray converts Go int slice to C int array
func intSlice2CIntArray(srcErrList []int) *C.int32_t {
if len(srcErrList) == 0 {
return (*C.int32_t)(unsafe.Pointer(nil))
}
var sizeErrInt = int(unsafe.Sizeof(srcErrList[0]))
switch sizeInt {
case sizeErrInt:
return (*C.int32_t)(unsafe.Pointer(&srcErrList[0]))
case sizeInt8:
int8Array := make([]int8, len(srcErrList))
for i, v := range srcErrList {
int8Array[i] = int8(v)
}
return (*C.int32_t)(unsafe.Pointer(&int8Array[0]))
case sizeInt16:
int16Array := make([]int16, len(srcErrList))
for i, v := range srcErrList {
int16Array[i] = int16(v)
}
return (*C.int32_t)(unsafe.Pointer(&int16Array[0]))
case sizeInt32:
int32Array := make([]int32, len(srcErrList))
for i, v := range srcErrList {
int32Array[i] = int32(v)
}
return (*C.int32_t)(unsafe.Pointer(&int32Array[0]))
case sizeInt64:
int64Array := make([]int64, len(srcErrList))
for i, v := range srcErrList {
int64Array[i] = int64(v)
}
return (*C.int32_t)(unsafe.Pointer(&int64Array[0]))
default:
panic(fmt.Sprintf("Unsupported: %d", sizeInt))
}
}
-66
View File
@@ -1,66 +0,0 @@
// Package erasure is a Go wrapper for the Intel Intelligent Storage
// Acceleration Library (Intel ISA-L). Intel ISA-L is a CPU optimized
// implementation of erasure coding algorithms.
//
// For more information on Intel ISA-L, please visit:
// https://01.org/intel%C2%AE-storage-acceleration-library-open-source-version
//
// Usage:
//
// Encode encodes a block of data. The input is the original data. The output
// is a 2 tuple containing (k + m) chunks of erasure encoded data and the
// length of the original object.
//
// Decode decodes 2 tuple data containing (k + m) chunks back into its original form.
// Additionally original block length should also be provided as input.
//
// Decoded data is exactly similar in length and content as the original data.
//
// Encoding data may be performed in 3 steps.
//
// 1. Create a parse set of encoder parameters
// 2. Create a new encoder
// 3. Encode data
//
// Decoding data is also performed in 3 steps.
//
// 1. Create a parse set of encoder parameters for validation
// 2. Create a new encoder
// 3. Decode data
//
// Erasure parameters contain three configurable elements:
// ValidateParams(k, m, technique int) (ErasureParams, error)
// k - Number of rows in matrix
// m - Number of colums in matrix
// technique - Matrix type, can be either Cauchy (recommended) or Vandermonde
// constraints: k + m < Galois Field (2^8)
//
// Choosing right parity and matrix technique is left for application to decide.
//
// But here are the few points to keep in mind
//
// Matrix Type:
// - Vandermonde is most commonly used method for choosing coefficients in erasure
// encoding but does not guarantee invertable for every sub matrix.
// - Whereas Cauchy is our recommended method for choosing coefficients in erasure coding.
// Since any sub-matrix of a Cauchy matrix is invertable.
//
// Total blocks:
// - Data blocks and Parity blocks should not be greater than 'Galois Field' (2^8)
//
// Example
//
// Creating and using an encoder
// var bytes []byte
// params := erasure.ValidateParams(10, 5)
// encoder := erasure.NewErasure(params)
// encodedData, length := encoder.Encode(bytes)
//
// Creating and using a decoder
// var encodedData [][]byte
// var length int
// params := erasure.ValidateParams(10, 5)
// encoder := erasure.NewErasure(params)
// originalData, err := encoder.Decode(encodedData, length)
//
package erasure
File diff suppressed because it is too large Load Diff
Binary file not shown.
-41
View File
@@ -1,41 +0,0 @@
/**********************************************************************
Copyright(c) 2011-2015 Intel Corporation All rights reserved.
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 Intel Corporation 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"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 THE COPYRIGHT
OWNER OR CONTRIBUTORS 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 _ISAL_H_
#define _ISAL_H_
#define ISAL_MAJOR_VERSION 2
#define ISAL_MINOR_VERSION 13
#define ISAL_PATCH_VERSION 0
#define ISAL_MAKE_VERSION(maj, min, patch) ((maj) * 0x10000 + (min) * 0x100 + (patch))
#define ISAL_VERSION ISAL_MAKE_VERSION(ISAL_MAJOR_VERSION, ISAL_MINOR_VERSION, ISAL_PATCH_VERSION)
#include "ec_code.h"
#include "gf_vect_mul.h"
#endif //_ISAL_H_
-348
View File
@@ -1,348 +0,0 @@
/**********************************************************************
Copyright(c) 2011-2015 Intel Corporation All rights reserved.
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 Intel Corporation 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"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 THE COPYRIGHT
OWNER OR CONTRIBUTORS 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.
**********************************************************************/
#include <limits.h>
#include <string.h> // for memset
#include "ec_code.h"
#include "ec_base.h" // for GF tables
#include "ec_types.h"
unsigned char gf_mul(unsigned char a, unsigned char b)
{
#ifndef GF_LARGE_TABLES
int i;
if ((a == 0) || (b == 0))
return 0;
return gff_base[(i = gflog_base[a] + gflog_base[b]) > 254 ? i - 255 : i];
#else
return gf_mul_table_base[b * 256 + a];
#endif
}
unsigned char gf_inv(unsigned char a)
{
#ifndef GF_LARGE_TABLES
if (a == 0)
return 0;
return gff_base[255 - gflog_base[a]];
#else
return gf_inv_table_base[a];
#endif
}
void gf_gen_rs_matrix(unsigned char *a, int m, int k)
{
int i, j;
unsigned char p, gen = 1;
memset(a, 0, k * m);
for (i = 0; i < k; i++)
a[k * i + i] = 1;
for (i = k; i < m; i++) {
p = 1;
for (j = 0; j < k; j++) {
a[k * i + j] = p;
p = gf_mul(p, gen);
}
gen = gf_mul(gen, 2);
}
}
void gf_gen_cauchy1_matrix(unsigned char *a, int m, int k)
{
int i, j;
unsigned char *p;
// Identity matrix in high position
memset(a, 0, k * m);
for (i = 0; i < k; i++)
a[k * i + i] = 1;
// For the rest choose 1/(i + j) | i != j
p = &a[k * k];
for (i = k; i < m; i++)
for (j = 0; j < k; j++)
*p++ = gf_inv(i ^ j);
}
int gf_invert_matrix(unsigned char *in_mat, unsigned char *out_mat, const int n)
{
int i, j, k;
unsigned char temp;
// Set out_mat[] to the identity matrix
for (i = 0; i < n * n; i++) // memset(out_mat, 0, n*n)
out_mat[i] = 0;
for (i = 0; i < n; i++)
out_mat[i * n + i] = 1;
// Inverse
for (i = 0; i < n; i++) {
// Check for 0 in pivot element
if (in_mat[i * n + i] == 0) {
// Find a row with non-zero in current column and swap
for (j = i + 1; j < n; j++)
if (in_mat[j * n + i])
break;
if (j == n) // Couldn't find means it's singular
return -1;
for (k = 0; k < n; k++) { // Swap rows i,j
temp = in_mat[i * n + k];
in_mat[i * n + k] = in_mat[j * n + k];
in_mat[j * n + k] = temp;
temp = out_mat[i * n + k];
out_mat[i * n + k] = out_mat[j * n + k];
out_mat[j * n + k] = temp;
}
}
temp = gf_inv(in_mat[i * n + i]); // 1/pivot
for (j = 0; j < n; j++) { // Scale row i by 1/pivot
in_mat[i * n + j] = gf_mul(in_mat[i * n + j], temp);
out_mat[i * n + j] = gf_mul(out_mat[i * n + j], temp);
}
for (j = 0; j < n; j++) {
if (j == i)
continue;
temp = in_mat[j * n + i];
for (k = 0; k < n; k++) {
out_mat[j * n + k] ^= gf_mul(temp, out_mat[i * n + k]);
in_mat[j * n + k] ^= gf_mul(temp, in_mat[i * n + k]);
}
}
}
return 0;
}
// Calculates const table gftbl in GF(2^8) from single input A
// gftbl(A) = {A{00}, A{01}, A{02}, ... , A{0f} }, {A{00}, A{10}, A{20}, ... , A{f0} }
void gf_vect_mul_init(unsigned char c, unsigned char *tbl)
{
unsigned char c2 = (c << 1) ^ ((c & 0x80) ? 0x1d : 0); //Mult by GF{2}
unsigned char c4 = (c2 << 1) ^ ((c2 & 0x80) ? 0x1d : 0); //Mult by GF{2}
unsigned char c8 = (c4 << 1) ^ ((c4 & 0x80) ? 0x1d : 0); //Mult by GF{2}
#if __WORDSIZE == 64 || _WIN64 || __x86_64__
unsigned long long v1, v2, v4, v8, *t;
unsigned long long v10, v20, v40, v80;
unsigned char c17, c18, c20, c24;
t = (unsigned long long *)tbl;
v1 = c * 0x0100010001000100ull;
v2 = c2 * 0x0101000001010000ull;
v4 = c4 * 0x0101010100000000ull;
v8 = c8 * 0x0101010101010101ull;
v4 = v1 ^ v2 ^ v4;
t[0] = v4;
t[1] = v8 ^ v4;
c17 = (c8 << 1) ^ ((c8 & 0x80) ? 0x1d : 0); //Mult by GF{2}
c18 = (c17 << 1) ^ ((c17 & 0x80) ? 0x1d : 0); //Mult by GF{2}
c20 = (c18 << 1) ^ ((c18 & 0x80) ? 0x1d : 0); //Mult by GF{2}
c24 = (c20 << 1) ^ ((c20 & 0x80) ? 0x1d : 0); //Mult by GF{2}
v10 = c17 * 0x0100010001000100ull;
v20 = c18 * 0x0101000001010000ull;
v40 = c20 * 0x0101010100000000ull;
v80 = c24 * 0x0101010101010101ull;
v40 = v10 ^ v20 ^ v40;
t[2] = v40;
t[3] = v80 ^ v40;
#else // 32-bit or other
unsigned char c3, c5, c6, c7, c9, c10, c11, c12, c13, c14, c15;
unsigned char c17, c18, c19, c20, c21, c22, c23, c24, c25, c26, c27, c28, c29, c30,
c31;
c3 = c2 ^ c;
c5 = c4 ^ c;
c6 = c4 ^ c2;
c7 = c4 ^ c3;
c9 = c8 ^ c;
c10 = c8 ^ c2;
c11 = c8 ^ c3;
c12 = c8 ^ c4;
c13 = c8 ^ c5;
c14 = c8 ^ c6;
c15 = c8 ^ c7;
tbl[0] = 0;
tbl[1] = c;
tbl[2] = c2;
tbl[3] = c3;
tbl[4] = c4;
tbl[5] = c5;
tbl[6] = c6;
tbl[7] = c7;
tbl[8] = c8;
tbl[9] = c9;
tbl[10] = c10;
tbl[11] = c11;
tbl[12] = c12;
tbl[13] = c13;
tbl[14] = c14;
tbl[15] = c15;
c17 = (c8 << 1) ^ ((c8 & 0x80) ? 0x1d : 0); //Mult by GF{2}
c18 = (c17 << 1) ^ ((c17 & 0x80) ? 0x1d : 0); //Mult by GF{2}
c19 = c18 ^ c17;
c20 = (c18 << 1) ^ ((c18 & 0x80) ? 0x1d : 0); //Mult by GF{2}
c21 = c20 ^ c17;
c22 = c20 ^ c18;
c23 = c20 ^ c19;
c24 = (c20 << 1) ^ ((c20 & 0x80) ? 0x1d : 0); //Mult by GF{2}
c25 = c24 ^ c17;
c26 = c24 ^ c18;
c27 = c24 ^ c19;
c28 = c24 ^ c20;
c29 = c24 ^ c21;
c30 = c24 ^ c22;
c31 = c24 ^ c23;
tbl[16] = 0;
tbl[17] = c17;
tbl[18] = c18;
tbl[19] = c19;
tbl[20] = c20;
tbl[21] = c21;
tbl[22] = c22;
tbl[23] = c23;
tbl[24] = c24;
tbl[25] = c25;
tbl[26] = c26;
tbl[27] = c27;
tbl[28] = c28;
tbl[29] = c29;
tbl[30] = c30;
tbl[31] = c31;
#endif //__WORDSIZE == 64 || _WIN64 || __x86_64__
}
void gf_vect_dot_prod_base(int len, int vlen, unsigned char *v,
unsigned char **src, unsigned char *dest)
{
int i, j;
unsigned char s;
for (i = 0; i < len; i++) {
s = 0;
for (j = 0; j < vlen; j++)
s ^= gf_mul(src[j][i], v[j * 32 + 1]);
dest[i] = s;
}
}
void gf_vect_mad_base(int len, int vec, int vec_i,
unsigned char *v, unsigned char *src, unsigned char *dest)
{
int i;
unsigned char s;
for (i = 0; i < len; i++) {
s = dest[i];
s ^= gf_mul(src[i], v[vec_i * 32 + 1]);
dest[i] = s;
}
}
void ec_encode_data_base(int len, int srcs, int dests, unsigned char *v,
unsigned char **src, unsigned char **dest)
{
int i, j, l;
unsigned char s;
for (l = 0; l < dests; l++) {
for (i = 0; i < len; i++) {
s = 0;
for (j = 0; j < srcs; j++)
s ^= gf_mul(src[j][i], v[j * 32 + l * srcs * 32 + 1]);
dest[l][i] = s;
}
}
}
void ec_encode_data_update_base(int len, int k, int rows, int vec_i, unsigned char *v,
unsigned char *data, unsigned char **dest)
{
int i, l;
unsigned char s;
for (l = 0; l < rows; l++) {
for (i = 0; i < len; i++) {
s = dest[l][i];
s ^= gf_mul(data[i], v[vec_i * 32 + l * k * 32 + 1]);
dest[l][i] = s;
}
}
}
void gf_vect_mul_base(int len, unsigned char *a, unsigned char *src, unsigned char *dest)
{
//2nd element of table array is ref value used to fill it in
unsigned char c = a[1];
while (len-- > 0)
*dest++ = gf_mul(c, *src++);
}
struct slver {
UINT16 snum;
UINT8 ver;
UINT8 core;
};
// Version info
struct slver gf_vect_mul_init_slver_00020035;
struct slver gf_vect_mul_init_slver = { 0x0035, 0x02, 0x00 };
struct slver ec_encode_data_base_slver_00010135;
struct slver ec_encode_data_base_slver = { 0x0135, 0x01, 0x00 };
struct slver gf_vect_mul_base_slver_00010136;
struct slver gf_vect_mul_base_slver = { 0x0136, 0x01, 0x00 };
struct slver gf_vect_dot_prod_base_slver_00010137;
struct slver gf_vect_dot_prod_base_slver = { 0x0137, 0x01, 0x00 };
File diff suppressed because it is too large Load Diff
-933
View File
@@ -1,933 +0,0 @@
/**********************************************************************
Copyright(c) 2011-2015 Intel Corporation All rights reserved.
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 Intel Corporation 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"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 THE COPYRIGHT
OWNER OR CONTRIBUTORS 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 _ERASURE_CODE_H_
#define _ERASURE_CODE_H_
/**
* @file erasure_code.h
* @brief Interface to functions supporting erasure code encode and decode.
*
* This file defines the interface to optimized functions used in erasure
* codes. Encode and decode of erasures in GF(2^8) are made by calculating the
* dot product of the symbols (bytes in GF(2^8)) across a set of buffers and a
* set of coefficients. Values for the coefficients are determined by the type
* of erasure code. Using a general dot product means that any sequence of
* coefficients may be used including erasure codes based on random
* coefficients.
* Multiple versions of dot product are supplied to calculate 1-6 output
* vectors in one pass.
* Base GF multiply and divide functions can be sped up by defining
* GF_LARGE_TABLES at the expense of memory size.
*
*/
#include "gf_vect_mul.h"
#ifdef __cplusplus
extern "C" {
#endif
/**
* @brief Initialize tables for fast Erasure Code encode and decode.
*
* Generates the expanded tables needed for fast encode or decode for erasure
* codes on blocks of data. 32bytes is generated for each input coefficient.
*
* @param k The number of vector sources or rows in the generator matrix
* for coding.
* @param rows The number of output vectors to concurrently encode/decode.
* @param a Pointer to sets of arrays of input coefficients used to encode
* or decode data.
* @param gftbls Pointer to start of space for concatenated output tables
* generated from input coefficients. Must be of size 32*k*rows.
* @returns none
*/
void ec_init_tables(int k, int rows, unsigned char* a, unsigned char* gftbls);
/**
* @brief Generate or decode erasure codes on blocks of data, runs appropriate version.
*
* Given a list of source data blocks, generate one or multiple blocks of
* encoded data as specified by a matrix of GF(2^8) coefficients. When given a
* suitable set of coefficients, this function will perform the fast generation
* or decoding of Reed-Solomon type erasure codes.
*
* This function determines what instruction sets are enabled and
* selects the appropriate version at runtime.
*
* @param len Length of each block of data (vector) of source or dest data.
* @param k The number of vector sources or rows in the generator matrix
* for coding.
* @param rows The number of output vectors to concurrently encode/decode.
* @param gftbls Pointer to array of input tables generated from coding
* coefficients in ec_init_tables(). Must be of size 32*k*rows
* @param data Array of pointers to source input buffers.
* @param coding Array of pointers to coded output buffers.
* @returns none
*/
void ec_encode_data(int len, int k, int rows, unsigned char *gftbls, unsigned char **data,
unsigned char **coding);
/**
* @brief Generate or decode erasure codes on blocks of data.
*
* Arch specific version of ec_encode_data() with same parameters.
* @requires SSE4.1
*/
void ec_encode_data_sse(int len, int k, int rows, unsigned char *gftbls, unsigned char **data,
unsigned char **coding);
/**
* @brief Generate or decode erasure codes on blocks of data.
*
* Arch specific version of ec_encode_data() with same parameters.
* @requires AVX
*/
void ec_encode_data_avx(int len, int k, int rows, unsigned char *gftbls, unsigned char **data,
unsigned char **coding);
/**
* @brief Generate or decode erasure codes on blocks of data.
*
* Arch specific version of ec_encode_data() with same parameters.
* @requires AVX2
*/
void ec_encode_data_avx2(int len, int k, int rows, unsigned char *gftbls, unsigned char **data,
unsigned char **coding);
/**
* @brief Generate or decode erasure codes on blocks of data, runs baseline version.
*
* Baseline version of ec_encode_data() with same parameters.
*/
void ec_encode_data_base(int len, int srcs, int dests, unsigned char *v, unsigned char **src,
unsigned char **dest);
/**
* @brief Generate update for encode or decode of erasure codes from single source, runs appropriate version.
*
* Given one source data block, update one or multiple blocks of encoded data as
* specified by a matrix of GF(2^8) coefficients. When given a suitable set of
* coefficients, this function will perform the fast generation or decoding of
* Reed-Solomon type erasure codes from one input source at a time.
*
* This function determines what instruction sets are enabled and selects the
* appropriate version at runtime.
*
* @param len Length of each block of data (vector) of source or dest data.
* @param k The number of vector sources or rows in the generator matrix
* for coding.
* @param rows The number of output vectors to concurrently encode/decode.
* @param vec_i The vector index corresponding to the single input source.
* @param g_tbls Pointer to array of input tables generated from coding
* coefficients in ec_init_tables(). Must be of size 32*k*rows
* @param data Pointer to single input source used to update output parity.
* @param coding Array of pointers to coded output buffers.
* @returns none
*/
void ec_encode_data_update(int len, int k, int rows, int vec_i, unsigned char *g_tbls,
unsigned char *data, unsigned char **coding);
/**
* @brief Generate update for encode or decode of erasure codes from single source.
*
* Arch specific version of ec_encode_data_update() with same parameters.
* @requires SSE4.1
*/
void ec_encode_data_update_sse(int len, int k, int rows, int vec_i, unsigned char *g_tbls,
unsigned char *data, unsigned char **coding);
/**
* @brief Generate update for encode or decode of erasure codes from single source.
*
* Arch specific version of ec_encode_data_update() with same parameters.
* @requires AVX
*/
void ec_encode_data_update_avx(int len, int k, int rows, int vec_i, unsigned char *g_tbls,
unsigned char *data, unsigned char **coding);
/**
* @brief Generate update for encode or decode of erasure codes from single source.
*
* Arch specific version of ec_encode_data_update() with same parameters.
* @requires AVX2
*/
void ec_encode_data_update_avx2(int len, int k, int rows, int vec_i, unsigned char *g_tbls,
unsigned char *data, unsigned char **coding);
/**
* @brief Generate update for encode or decode of erasure codes from single source.
*
* Baseline version of ec_encode_data_update().
*/
void ec_encode_data_update_base(int len, int k, int rows, int vec_i, unsigned char *v,
unsigned char *data, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product.
*
* Does a GF(2^8) dot product across each byte of the input array and a constant
* set of coefficients to produce each byte of the output. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 32*vlen byte constant array based on the input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 32*vlen byte array of pre-calculated constants based
* on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Pointer to destination data array.
* @returns none
*/
void gf_vect_dot_prod_sse(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char *dest);
/**
* @brief GF(2^8) vector dot product.
*
* Does a GF(2^8) dot product across each byte of the input array and a constant
* set of coefficients to produce each byte of the output. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 32*vlen byte constant array based on the input coefficients.
* @requires AVX
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 32*vlen byte array of pre-calculated constants based
* on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Pointer to destination data array.
* @returns none
*/
void gf_vect_dot_prod_avx(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char *dest);
/**
* @brief GF(2^8) vector dot product.
*
* Does a GF(2^8) dot product across each byte of the input array and a constant
* set of coefficients to produce each byte of the output. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 32*vlen byte constant array based on the input coefficients.
* @requires AVX2
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 32*vlen byte array of pre-calculated constants based
* on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Pointer to destination data array.
* @returns none
*/
void gf_vect_dot_prod_avx2(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char *dest);
/**
* @brief GF(2^8) vector dot product with two outputs.
*
* Vector dot product optimized to calculate two ouputs at a time. Does two
* GF(2^8) dot products across each byte of the input array and two constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 2*32*vlen byte constant array based on the two sets of input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 2*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_2vect_dot_prod_sse(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with two outputs.
*
* Vector dot product optimized to calculate two ouputs at a time. Does two
* GF(2^8) dot products across each byte of the input array and two constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 2*32*vlen byte constant array based on the two sets of input coefficients.
* @requires AVX
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 2*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_2vect_dot_prod_avx(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with two outputs.
*
* Vector dot product optimized to calculate two ouputs at a time. Does two
* GF(2^8) dot products across each byte of the input array and two constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 2*32*vlen byte constant array based on the two sets of input coefficients.
* @requires AVX2
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 2*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_2vect_dot_prod_avx2(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with three outputs.
*
* Vector dot product optimized to calculate three ouputs at a time. Does three
* GF(2^8) dot products across each byte of the input array and three constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 3*32*vlen byte constant array based on the three sets of input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 3*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_3vect_dot_prod_sse(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with three outputs.
*
* Vector dot product optimized to calculate three ouputs at a time. Does three
* GF(2^8) dot products across each byte of the input array and three constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 3*32*vlen byte constant array based on the three sets of input coefficients.
* @requires AVX
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 3*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_3vect_dot_prod_avx(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with three outputs.
*
* Vector dot product optimized to calculate three ouputs at a time. Does three
* GF(2^8) dot products across each byte of the input array and three constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 3*32*vlen byte constant array based on the three sets of input coefficients.
* @requires AVX2
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 3*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_3vect_dot_prod_avx2(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with four outputs.
*
* Vector dot product optimized to calculate four ouputs at a time. Does four
* GF(2^8) dot products across each byte of the input array and four constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 4*32*vlen byte constant array based on the four sets of input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 4*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_4vect_dot_prod_sse(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with four outputs.
*
* Vector dot product optimized to calculate four ouputs at a time. Does four
* GF(2^8) dot products across each byte of the input array and four constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 4*32*vlen byte constant array based on the four sets of input coefficients.
* @requires AVX
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 4*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_4vect_dot_prod_avx(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with four outputs.
*
* Vector dot product optimized to calculate four ouputs at a time. Does four
* GF(2^8) dot products across each byte of the input array and four constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 4*32*vlen byte constant array based on the four sets of input coefficients.
* @requires AVX2
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 4*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_4vect_dot_prod_avx2(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with five outputs.
*
* Vector dot product optimized to calculate five ouputs at a time. Does five
* GF(2^8) dot products across each byte of the input array and five constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 5*32*vlen byte constant array based on the five sets of input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 5*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_5vect_dot_prod_sse(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with five outputs.
*
* Vector dot product optimized to calculate five ouputs at a time. Does five
* GF(2^8) dot products across each byte of the input array and five constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 5*32*vlen byte constant array based on the five sets of input coefficients.
* @requires AVX
*
* @param len Length of each vector in bytes. Must >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 5*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_5vect_dot_prod_avx(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with five outputs.
*
* Vector dot product optimized to calculate five ouputs at a time. Does five
* GF(2^8) dot products across each byte of the input array and five constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 5*32*vlen byte constant array based on the five sets of input coefficients.
* @requires AVX2
*
* @param len Length of each vector in bytes. Must >= 32.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 5*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_5vect_dot_prod_avx2(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with six outputs.
*
* Vector dot product optimized to calculate six ouputs at a time. Does six
* GF(2^8) dot products across each byte of the input array and six constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 6*32*vlen byte constant array based on the six sets of input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 6*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_6vect_dot_prod_sse(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with six outputs.
*
* Vector dot product optimized to calculate six ouputs at a time. Does six
* GF(2^8) dot products across each byte of the input array and six constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 6*32*vlen byte constant array based on the six sets of input coefficients.
* @requires AVX
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 6*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_6vect_dot_prod_avx(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product with six outputs.
*
* Vector dot product optimized to calculate six ouputs at a time. Does six
* GF(2^8) dot products across each byte of the input array and six constant
* sets of coefficients to produce each byte of the outputs. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 6*32*vlen byte constant array based on the six sets of input coefficients.
* @requires AVX2
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 6*32*vlen byte array of pre-calculated constants
* based on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Array of pointers to destination data buffers.
* @returns none
*/
void gf_6vect_dot_prod_avx2(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char **dest);
/**
* @brief GF(2^8) vector dot product, runs baseline version.
*
* Does a GF(2^8) dot product across each byte of the input array and a constant
* set of coefficients to produce each byte of the output. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 32*vlen byte constant array based on the input coefficients.
*
* @param len Length of each vector in bytes. Must be >= 16.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 32*vlen byte array of pre-calculated constants based
* on the array of input coefficients. Only elements 32*CONST*j + 1
* of this array are used, where j = (0, 1, 2...) and CONST is the
* number of elements in the array of input coefficients. The
* elements used correspond to the original input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Pointer to destination data array.
* @returns none
*/
void gf_vect_dot_prod_base(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char *dest);
/**
* @brief GF(2^8) vector dot product, runs appropriate version.
*
* Does a GF(2^8) dot product across each byte of the input array and a constant
* set of coefficients to produce each byte of the output. Can be used for
* erasure coding encode and decode. Function requires pre-calculation of a
* 32*vlen byte constant array based on the input coefficients.
*
* This function determines what instruction sets are enabled and
* selects the appropriate version at runtime.
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vlen Number of vector sources.
* @param gftbls Pointer to 32*vlen byte array of pre-calculated constants based
* on the array of input coefficients.
* @param src Array of pointers to source inputs.
* @param dest Pointer to destination data array.
* @returns none
*/
void gf_vect_dot_prod(int len, int vlen, unsigned char *gftbls,
unsigned char **src, unsigned char *dest);
/**
* @brief GF(2^8) vector multiply accumulate, runs appropriate version.
*
* Does a GF(2^8) multiply across each byte of input source with expanded
* constant and add to destination array. Can be used for erasure coding encode
* and decode update when only one source is available at a time. Function
* requires pre-calculation of a 32*vec byte constant array based on the input
* coefficients.
*
* This function determines what instruction sets are enabled and selects the
* appropriate version at runtime.
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vec The number of vector sources or rows in the generator matrix
* for coding.
* @param vec_i The vector index corresponding to the single input source.
* @param gftbls Pointer to array of input tables generated from coding
* coefficients in ec_init_tables(). Must be of size 32*vec.
* @param src Array of pointers to source inputs.
* @param dest Pointer to destination data array.
* @returns none
*/
void gf_vect_mad(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char *dest);
/**
* @brief GF(2^8) vector multiply accumulate, arch specific version.
*
* Arch specific version of gf_vect_mad() with same parameters.
* @requires SSE4.1
*/
void gf_vect_mad_sse(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char *dest);
/**
* @brief GF(2^8) vector multiply accumulate, arch specific version.
*
* Arch specific version of gf_vect_mad() with same parameters.
* @requires AVX
*/
void gf_vect_mad_avx(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char *dest);
/**
* @brief GF(2^8) vector multiply accumulate, arch specific version.
*
* Arch specific version of gf_vect_mad() with same parameters.
* @requires AVX2
*/
void gf_vect_mad_avx2(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char *dest);
/**
* @brief GF(2^8) vector multiply accumulate, baseline version.
*
* Baseline version of gf_vect_mad() with same parameters.
*/
void gf_vect_mad_base(int len, int vec, int vec_i, unsigned char *v, unsigned char *src,
unsigned char *dest);
/**
* @brief GF(2^8) vector multiply with 2 accumulate. SSE version.
*
* Does a GF(2^8) multiply across each byte of input source with expanded
* constants and add to destination arrays. Can be used for erasure coding
* encode and decode update when only one source is available at a
* time. Function requires pre-calculation of a 32*vec byte constant array based
* on the input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vec The number of vector sources or rows in the generator matrix
* for coding.
* @param vec_i The vector index corresponding to the single input source.
* @param gftbls Pointer to array of input tables generated from coding
* coefficients in ec_init_tables(). Must be of size 32*vec.
* @param src Pointer to source input array.
* @param dest Array of pointers to destination input/outputs.
* @returns none
*/
void gf_2vect_mad_sse(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 2 accumulate. AVX version of gf_2vect_mad_sse().
* @requires AVX
*/
void gf_2vect_mad_avx(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 2 accumulate. AVX2 version of gf_2vect_mad_sse().
* @requires AVX2
*/
void gf_2vect_mad_avx2(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 3 accumulate. SSE version.
*
* Does a GF(2^8) multiply across each byte of input source with expanded
* constants and add to destination arrays. Can be used for erasure coding
* encode and decode update when only one source is available at a
* time. Function requires pre-calculation of a 32*vec byte constant array based
* on the input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vec The number of vector sources or rows in the generator matrix
* for coding.
* @param vec_i The vector index corresponding to the single input source.
* @param gftbls Pointer to array of input tables generated from coding
* coefficients in ec_init_tables(). Must be of size 32*vec.
* @param src Pointer to source input array.
* @param dest Array of pointers to destination input/outputs.
* @returns none
*/
void gf_3vect_mad_sse(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 3 accumulate. AVX version of gf_3vect_mad_sse().
* @requires AVX
*/
void gf_3vect_mad_avx(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 3 accumulate. AVX2 version of gf_3vect_mad_sse().
* @requires AVX2
*/
void gf_3vect_mad_avx2(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 4 accumulate. SSE version.
*
* Does a GF(2^8) multiply across each byte of input source with expanded
* constants and add to destination arrays. Can be used for erasure coding
* encode and decode update when only one source is available at a
* time. Function requires pre-calculation of a 32*vec byte constant array based
* on the input coefficients.
* @requires SSE4.1
*
* @param len Length of each vector in bytes. Must be >= 32.
* @param vec The number of vector sources or rows in the generator matrix
* for coding.
* @param vec_i The vector index corresponding to the single input source.
* @param gftbls Pointer to array of input tables generated from coding
* coefficients in ec_init_tables(). Must be of size 32*vec.
* @param src Pointer to source input array.
* @param dest Array of pointers to destination input/outputs.
* @returns none
*/
void gf_4vect_mad_sse(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 4 accumulate. AVX version of gf_4vect_mad_sse().
* @requires AVX
*/
void gf_4vect_mad_avx(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 4 accumulate. AVX2 version of gf_4vect_mad_sse().
* @requires AVX2
*/
void gf_4vect_mad_avx2(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 5 accumulate. SSE version.
* @requires SSE4.1
*/
void gf_5vect_mad_sse(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 5 accumulate. AVX version.
* @requires AVX
*/
void gf_5vect_mad_avx(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 5 accumulate. AVX2 version.
* @requires AVX2
*/
void gf_5vect_mad_avx2(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 6 accumulate. SSE version.
* @requires SSE4.1
*/
void gf_6vect_mad_sse(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 6 accumulate. AVX version.
* @requires AVX
*/
void gf_6vect_mad_avx(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**
* @brief GF(2^8) vector multiply with 6 accumulate. AVX2 version.
* @requires AVX2
*/
void gf_6vect_mad_avx2(int len, int vec, int vec_i, unsigned char *gftbls, unsigned char *src,
unsigned char **dest);
/**********************************************************************
* The remaining are lib support functions used in GF(2^8) operations.
*/
/**
* @brief Single element GF(2^8) multiply.
*
* @param a Multiplicand a
* @param b Multiplicand b
* @returns Product of a and b in GF(2^8)
*/
unsigned char gf_mul(unsigned char a, unsigned char b);
/**
* @brief Single element GF(2^8) inverse.
*
* @param a Input element
* @returns Field element b such that a x b = {1}
*/
unsigned char gf_inv(unsigned char a);
/**
* @brief Generate a matrix of coefficients to be used for encoding.
*
* Vandermonde matrix example of encoding coefficients where high portion of
* matrix is identity matrix I and lower portion is constructed as 2^{i*(j-k+1)}
* i:{0,k-1} j:{k,m-1}. Commonly used method for choosing coefficients in
* erasure encoding but does not guarantee invertable for every sub matrix. For
* large k it is possible to find cases where the decode matrix chosen from
* sources and parity not in erasure are not invertable. Users may want to
* adjust for k > 5.
*
* @param a [mxk] array to hold coefficients
* @param m number of rows in matrix corresponding to srcs + parity.
* @param k number of columns in matrix corresponding to srcs.
* @returns none
*/
void gf_gen_rs_matrix(unsigned char *a, int m, int k);
/**
* @brief Generate a Cauchy matrix of coefficients to be used for encoding.
*
* Cauchy matrix example of encoding coefficients where high portion of matrix
* is identity matrix I and lower portion is constructed as 1/(i + j) | i != j,
* i:{0,k-1} j:{k,m-1}. Any sub-matrix of a Cauchy matrix should be invertable.
*
* @param a [mxk] array to hold coefficients
* @param m number of rows in matrix corresponding to srcs + parity.
* @param k number of columns in matrix corresponding to srcs.
* @returns none
*/
void gf_gen_cauchy1_matrix(unsigned char *a, int m, int k);
/**
* @brief Invert a matrix in GF(2^8)
*
* @param in input matrix
* @param out output matrix such that [in] x [out] = [I] - identity matrix
* @param n size of matrix [nxn]
* @returns 0 successful, other fail on singular input matrix
*/
int gf_invert_matrix(unsigned char *in, unsigned char *out, const int n);
/*************************************************************/
#ifdef __cplusplus
}
#endif
#endif //_ERASURE_CODE_H_
-267
View File
@@ -1,267 +0,0 @@
/**********************************************************************
Copyright(c) 2011-2015 Intel Corporation All rights reserved.
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 Intel Corporation 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"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 THE COPYRIGHT
OWNER OR CONTRIBUTORS 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.
**********************************************************************/
#include <limits.h>
#include "ec_code.h"
#include "ec_types.h"
void ec_init_tables(int k, int rows, unsigned char *a, unsigned char *g_tbls)
{
int i, j;
for (i = 0; i < rows; i++) {
for (j = 0; j < k; j++) {
gf_vect_mul_init(*a++, g_tbls);
g_tbls += 32;
}
}
}
void ec_encode_data_sse(int len, int k, int rows, unsigned char *g_tbls, unsigned char **data,
unsigned char **coding)
{
if (len < 16) {
ec_encode_data_base(len, k, rows, g_tbls, data, coding);
return;
}
while (rows >= 4) {
gf_4vect_dot_prod_sse(len, k, g_tbls, data, coding);
g_tbls += 4 * k * 32;
coding += 4;
rows -= 4;
}
switch (rows) {
case 3:
gf_3vect_dot_prod_sse(len, k, g_tbls, data, coding);
break;
case 2:
gf_2vect_dot_prod_sse(len, k, g_tbls, data, coding);
break;
case 1:
gf_vect_dot_prod_sse(len, k, g_tbls, data, *coding);
break;
case 0:
break;
}
}
void ec_encode_data_avx(int len, int k, int rows, unsigned char *g_tbls, unsigned char **data,
unsigned char **coding)
{
if (len < 16) {
ec_encode_data_base(len, k, rows, g_tbls, data, coding);
return;
}
while (rows >= 4) {
gf_4vect_dot_prod_avx(len, k, g_tbls, data, coding);
g_tbls += 4 * k * 32;
coding += 4;
rows -= 4;
}
switch (rows) {
case 3:
gf_3vect_dot_prod_avx(len, k, g_tbls, data, coding);
break;
case 2:
gf_2vect_dot_prod_avx(len, k, g_tbls, data, coding);
break;
case 1:
gf_vect_dot_prod_avx(len, k, g_tbls, data, *coding);
break;
case 0:
break;
}
}
void ec_encode_data_avx2(int len, int k, int rows, unsigned char *g_tbls, unsigned char **data,
unsigned char **coding)
{
if (len < 32) {
ec_encode_data_base(len, k, rows, g_tbls, data, coding);
return;
}
while (rows >= 4) {
gf_4vect_dot_prod_avx2(len, k, g_tbls, data, coding);
g_tbls += 4 * k * 32;
coding += 4;
rows -= 4;
}
switch (rows) {
case 3:
gf_3vect_dot_prod_avx2(len, k, g_tbls, data, coding);
break;
case 2:
gf_2vect_dot_prod_avx2(len, k, g_tbls, data, coding);
break;
case 1:
gf_vect_dot_prod_avx2(len, k, g_tbls, data, *coding);
break;
case 0:
break;
}
}
#if __WORDSIZE == 64 || _WIN64 || __x86_64__
void ec_encode_data_update_sse(int len, int k, int rows, int vec_i, unsigned char *g_tbls,
unsigned char *data, unsigned char **coding)
{
if (len < 16) {
ec_encode_data_update_base(len, k, rows, vec_i, g_tbls, data, coding);
return;
}
while (rows > 6) {
gf_6vect_mad_sse(len, k, vec_i, g_tbls, data, coding);
g_tbls += 6 * k * 32;
coding += 6;
rows -= 6;
}
switch (rows) {
case 6:
gf_6vect_mad_sse(len, k, vec_i, g_tbls, data, coding);
break;
case 5:
gf_5vect_mad_sse(len, k, vec_i, g_tbls, data, coding);
break;
case 4:
gf_4vect_mad_sse(len, k, vec_i, g_tbls, data, coding);
break;
case 3:
gf_3vect_mad_sse(len, k, vec_i, g_tbls, data, coding);
break;
case 2:
gf_2vect_mad_sse(len, k, vec_i, g_tbls, data, coding);
break;
case 1:
gf_vect_mad_sse(len, k, vec_i, g_tbls, data, *coding);
break;
case 0:
break;
}
}
void ec_encode_data_update_avx(int len, int k, int rows, int vec_i, unsigned char *g_tbls,
unsigned char *data, unsigned char **coding)
{
if (len < 16) {
ec_encode_data_update_base(len, k, rows, vec_i, g_tbls, data, coding);
return;
}
while (rows > 6) {
gf_6vect_mad_avx(len, k, vec_i, g_tbls, data, coding);
g_tbls += 6 * k * 32;
coding += 6;
rows -= 6;
}
switch (rows) {
case 6:
gf_6vect_mad_avx(len, k, vec_i, g_tbls, data, coding);
break;
case 5:
gf_5vect_mad_avx(len, k, vec_i, g_tbls, data, coding);
break;
case 4:
gf_4vect_mad_avx(len, k, vec_i, g_tbls, data, coding);
break;
case 3:
gf_3vect_mad_avx(len, k, vec_i, g_tbls, data, coding);
break;
case 2:
gf_2vect_mad_avx(len, k, vec_i, g_tbls, data, coding);
break;
case 1:
gf_vect_mad_avx(len, k, vec_i, g_tbls, data, *coding);
break;
case 0:
break;
}
}
void ec_encode_data_update_avx2(int len, int k, int rows, int vec_i, unsigned char *g_tbls,
unsigned char *data, unsigned char **coding)
{
if (len < 32) {
ec_encode_data_update_base(len, k, rows, vec_i, g_tbls, data, coding);
return;
}
while (rows > 6) {
gf_6vect_mad_avx2(len, k, vec_i, g_tbls, data, coding);
g_tbls += 6 * k * 32;
coding += 6;
rows -= 6;
}
switch (rows) {
case 6:
gf_6vect_mad_avx2(len, k, vec_i, g_tbls, data, coding);
break;
case 5:
gf_5vect_mad_avx2(len, k, vec_i, g_tbls, data, coding);
break;
case 4:
gf_4vect_mad_avx2(len, k, vec_i, g_tbls, data, coding);
break;
case 3:
gf_3vect_mad_avx2(len, k, vec_i, g_tbls, data, coding);
break;
case 2:
gf_2vect_mad_avx2(len, k, vec_i, g_tbls, data, coding);
break;
case 1:
gf_vect_mad_avx2(len, k, vec_i, g_tbls, data, *coding);
break;
case 0:
break;
}
}
#endif //__WORDSIZE == 64 || _WIN64 || __x86_64__
struct slver {
UINT16 snum;
UINT8 ver;
UINT8 core;
};
// Version info
struct slver ec_init_tables_slver_00010068;
struct slver ec_init_tables_slver = { 0x0068, 0x01, 0x00 };
struct slver ec_encode_data_sse_slver_00020069;
struct slver ec_encode_data_sse_slver = { 0x0069, 0x02, 0x00 };
-39
View File
@@ -1,39 +0,0 @@
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef __COMMON_H__
#define __COMMON_H__
#include <stdint.h>
int32_t minio_init_encoder (int k, int m,
unsigned char **encode_matrix,
unsigned char **encode_tbls);
int32_t minio_init_decoder (int32_t *error_index,
int k, int n, int errs,
unsigned char *encoding_matrix,
unsigned char **decode_matrix,
unsigned char **decode_tbls,
uint32_t **decode_index);
int32_t minio_get_source_target (int errs, int k, int m,
int32_t *error_index,
uint32_t *decode_index,
unsigned char **buffs,
unsigned char ***source,
unsigned char ***target);
#endif /* __COMMON_H__ */
-142
View File
@@ -1,142 +0,0 @@
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "ec.h"
#include "ec_minio_common.h"
static
int32_t _minio_src_index_in_error (int r, int32_t *error_index, int errs)
{
int i;
for (i = 0; i < errs; i++) {
if (error_index[i] == r) {
// true
return 1;
}
}
// false
return 0;
}
// Separate out source data and target buffers
int32_t minio_get_source_target (int errs, int k, int m,
int32_t *error_index,
uint32_t *decode_index,
unsigned char **buffs,
unsigned char ***source,
unsigned char ***target)
{
int i;
unsigned char *tmp_source[k];
unsigned char *tmp_target[m];
if (k < 0 || m < 0) {
return -1;
}
memset (tmp_source, 0, k);
memset (tmp_target, 0, m);
for (i = 0; i < k; i++) {
tmp_source[i] = (unsigned char *) buffs[decode_index[i]];
}
for (i = 0; i < m; i++) {
if (i < errs)
tmp_target[i] = (unsigned char *) buffs[error_index[i]];
}
*source = tmp_source;
*target = tmp_target;
return 0;
}
/*
Generate decode matrix during the decoding phase
*/
int minio_init_decoder (int32_t *error_index,
int k, int n, int errs,
unsigned char *encode_matrix,
unsigned char **decode_matrix,
unsigned char **decode_tbls,
uint32_t **decode_index)
{
int i, j, r, l;
uint32_t *tmp_decode_index = (uint32_t *) malloc(sizeof(uint32_t) * k);
unsigned char *input_matrix;
unsigned char *inverse_matrix;
unsigned char *tmp_decode_matrix;
unsigned char *tmp_decode_tbls;
input_matrix = (unsigned char *) malloc(sizeof(unsigned char) * k * n);
inverse_matrix = (unsigned char *) malloc(sizeof(unsigned char) * k * n);
tmp_decode_matrix = (unsigned char *) malloc(sizeof(unsigned char) * k * n);;
tmp_decode_tbls = (unsigned char *) malloc(sizeof(unsigned char) * k * n * 32);
for (i = 0, r = 0; i < k; i++, r++) {
while (_minio_src_index_in_error(r, error_index, errs))
r++;
for (j = 0; j < k; j++) {
input_matrix[k * i + j] = encode_matrix[k * r + j];
}
tmp_decode_index[i] = r;
}
// Not all vandermonde matrix can be inverted
if (gf_invert_matrix(input_matrix, inverse_matrix, k) < 0) {
free(tmp_decode_matrix);
free(tmp_decode_tbls);
free(tmp_decode_index);
return -1;
}
for (l = 0; l < errs; l++) {
if (error_index[l] < k) {
// decoding matrix elements for data chunks
for (j = 0; j < k; j++) {
tmp_decode_matrix[k * l + j] =
inverse_matrix[k *
error_index[l] + j];
}
} else {
// decoding matrix element for coding chunks
for (i = 0; i < k; i++) {
unsigned char s = 0;
for (j = 0; j < k; j++) {
s ^= gf_mul(inverse_matrix[j * k + i],
encode_matrix[k *
error_index[l] + j]);
}
tmp_decode_matrix[k * l + i] = s;
}
}
}
ec_init_tables (k, errs, tmp_decode_matrix, tmp_decode_tbls);
*decode_matrix = tmp_decode_matrix;
*decode_tbls = tmp_decode_tbls;
*decode_index = tmp_decode_index;
return 0;
}
-55
View File
@@ -1,55 +0,0 @@
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <stdlib.h>
#include <stdio.h>
#include "ec.h"
#include "ec_minio_common.h"
/*
Generate encode matrix during the encoding phase
*/
int32_t minio_init_encoder (int k, int m, unsigned char **encode_matrix, unsigned char **encode_tbls)
{
unsigned char *tmp_matrix;
unsigned char *tmp_tbls;
tmp_matrix = (unsigned char *) malloc (k * (k + m));
tmp_tbls = (unsigned char *) malloc (k * (k + m) * 32);
if (k < 5) {
/*
Commonly used method for choosing coefficients in erasure
encoding but does not guarantee invertable for every sub
matrix. For large k it is possible to find cases where the
decode matrix chosen from sources and parity not in erasure
are not invertable. Users may want to adjust for k > 5.
-- Intel
*/
gf_gen_rs_matrix (tmp_matrix, k + m, k);
} else {
gf_gen_cauchy1_matrix (tmp_matrix, k + m, k);
}
ec_init_tables(k, m, &tmp_matrix[k * k], tmp_tbls);
*encode_matrix = tmp_matrix;
*encode_tbls = tmp_tbls;
return 0;
}
-404
View File
@@ -1,404 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; 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 Intel Corporation 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "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 THE COPYRIGHT
; OWNER OR CONTRIBUTORS 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.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
%ifidn __OUTPUT_FORMAT__, elf64
%define WRT_OPT wrt ..plt
%else
%define WRT_OPT
%endif
%include "ec_reg_sizes.asm"
%ifidn __OUTPUT_FORMAT__, elf32
[bits 32]
%define def_wrd dd
%define wrd_sz dword
%define arg1 esi
%define arg2 eax
%define arg3 ebx
%define arg4 ecx
%define arg5 edx
%else
default rel
[bits 64]
%define def_wrd dq
%define wrd_sz qword
%define arg1 rsi
%define arg2 rax
%define arg3 rbx
%define arg4 rcx
%define arg5 rdx
extern ec_encode_data_update_sse
extern ec_encode_data_update_avx
extern ec_encode_data_update_avx2
extern gf_vect_mul_sse
extern gf_vect_mul_avx
extern gf_vect_mad_sse
extern gf_vect_mad_avx
extern gf_vect_mad_avx2
%endif
extern gf_vect_mul_base
extern ec_encode_data_base
extern ec_encode_data_update_base
extern gf_vect_dot_prod_base
extern gf_vect_mad_base
extern gf_vect_dot_prod_sse
extern gf_vect_dot_prod_avx
extern gf_vect_dot_prod_avx2
extern ec_encode_data_sse
extern ec_encode_data_avx
extern ec_encode_data_avx2
section .data
;;; *_mbinit are initial values for *_dispatched; is updated on first call.
;;; Therefore, *_dispatch_init is only executed on first call.
ec_encode_data_dispatched:
def_wrd ec_encode_data_mbinit
gf_vect_mul_dispatched:
def_wrd gf_vect_mul_mbinit
gf_vect_dot_prod_dispatched:
def_wrd gf_vect_dot_prod_mbinit
ec_encode_data_update_dispatched:
def_wrd ec_encode_data_update_mbinit
gf_vect_mad_dispatched:
def_wrd gf_vect_mad_mbinit
section .text
;;;;
; ec_encode_data multibinary function
;;;;
global ec_encode_data:function
ec_encode_data_mbinit:
call ec_encode_data_dispatch_init
ec_encode_data:
jmp wrd_sz [ec_encode_data_dispatched]
ec_encode_data_dispatch_init:
push arg1
push arg2
push arg3
push arg4
push arg5
lea arg1, [ec_encode_data_base WRT_OPT] ; Default
mov eax, 1
cpuid
lea arg3, [ec_encode_data_sse WRT_OPT]
test ecx, FLAG_CPUID1_ECX_SSE4_1
cmovne arg1, arg3
and ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
cmp ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
lea arg3, [ec_encode_data_avx WRT_OPT]
jne _done_ec_encode_data_init
mov arg1, arg3
;; Try for AVX2
xor ecx, ecx
mov eax, 7
cpuid
test ebx, FLAG_CPUID1_EBX_AVX2
lea arg3, [ec_encode_data_avx2 WRT_OPT]
cmovne arg1, arg3
;; Does it have xmm and ymm support
xor ecx, ecx
xgetbv
and eax, FLAG_XGETBV_EAX_XMM_YMM
cmp eax, FLAG_XGETBV_EAX_XMM_YMM
je _done_ec_encode_data_init
lea arg1, [ec_encode_data_sse WRT_OPT]
_done_ec_encode_data_init:
pop arg5
pop arg4
pop arg3
pop arg2
mov [ec_encode_data_dispatched], arg1
pop arg1
ret
;;;;
; gf_vect_mul multibinary function
;;;;
global gf_vect_mul:function
gf_vect_mul_mbinit:
call gf_vect_mul_dispatch_init
gf_vect_mul:
jmp wrd_sz [gf_vect_mul_dispatched]
gf_vect_mul_dispatch_init:
push arg1
%ifidn __OUTPUT_FORMAT__, elf32 ;; 32-bit check
lea arg1, [gf_vect_mul_base]
%else
push rax
push rbx
push rcx
push rdx
lea arg1, [gf_vect_mul_base WRT_OPT] ; Default
mov eax, 1
cpuid
test ecx, FLAG_CPUID1_ECX_SSE4_2
lea rbx, [gf_vect_mul_sse WRT_OPT]
je _done_gf_vect_mul_dispatch_init
mov arg1, rbx
;; Try for AVX
and ecx, (FLAG_CPUID1_ECX_OSXSAVE | FLAG_CPUID1_ECX_AVX)
cmp ecx, (FLAG_CPUID1_ECX_OSXSAVE | FLAG_CPUID1_ECX_AVX)
jne _done_gf_vect_mul_dispatch_init
;; Does it have xmm and ymm support
xor ecx, ecx
xgetbv
and eax, FLAG_XGETBV_EAX_XMM_YMM
cmp eax, FLAG_XGETBV_EAX_XMM_YMM
jne _done_gf_vect_mul_dispatch_init
lea arg1, [gf_vect_mul_avx WRT_OPT]
_done_gf_vect_mul_dispatch_init:
pop rdx
pop rcx
pop rbx
pop rax
%endif ;; END 32-bit check
mov [gf_vect_mul_dispatched], arg1
pop arg1
ret
;;;;
; ec_encode_data_update multibinary function
;;;;
global ec_encode_data_update:function
ec_encode_data_update_mbinit:
call ec_encode_data_update_dispatch_init
ec_encode_data_update:
jmp wrd_sz [ec_encode_data_update_dispatched]
ec_encode_data_update_dispatch_init:
push arg1
%ifidn __OUTPUT_FORMAT__, elf32 ;; 32-bit check
lea arg1, [ec_encode_data_update_base]
%else
push rax
push rbx
push rcx
push rdx
lea arg1, [ec_encode_data_update_base WRT_OPT] ; Default
mov eax, 1
cpuid
lea rbx, [ec_encode_data_update_sse WRT_OPT]
test ecx, FLAG_CPUID1_ECX_SSE4_1
cmovne arg1, rbx
and ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
cmp ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
lea rbx, [ec_encode_data_update_avx WRT_OPT]
jne _done_ec_encode_data_update_init
mov rsi, rbx
;; Try for AVX2
xor ecx, ecx
mov eax, 7
cpuid
test ebx, FLAG_CPUID1_EBX_AVX2
lea rbx, [ec_encode_data_update_avx2 WRT_OPT]
cmovne rsi, rbx
;; Does it have xmm and ymm support
xor ecx, ecx
xgetbv
and eax, FLAG_XGETBV_EAX_XMM_YMM
cmp eax, FLAG_XGETBV_EAX_XMM_YMM
je _done_ec_encode_data_update_init
lea rsi, [ec_encode_data_update_sse WRT_OPT]
_done_ec_encode_data_update_init:
pop rdx
pop rcx
pop rbx
pop rax
%endif ;; END 32-bit check
mov [ec_encode_data_update_dispatched], arg1
pop arg1
ret
;;;;
; gf_vect_dot_prod multibinary function
;;;;
global gf_vect_dot_prod:function
gf_vect_dot_prod_mbinit:
call gf_vect_dot_prod_dispatch_init
gf_vect_dot_prod:
jmp wrd_sz [gf_vect_dot_prod_dispatched]
gf_vect_dot_prod_dispatch_init:
push arg1
push arg2
push arg3
push arg4
push arg5
lea arg1, [gf_vect_dot_prod_base WRT_OPT] ; Default
mov eax, 1
cpuid
lea arg3, [gf_vect_dot_prod_sse WRT_OPT]
test ecx, FLAG_CPUID1_ECX_SSE4_1
cmovne arg1, arg3
and ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
cmp ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
lea arg3, [gf_vect_dot_prod_avx WRT_OPT]
jne _done_gf_vect_dot_prod_init
mov arg1, arg3
;; Try for AVX2
xor ecx, ecx
mov eax, 7
cpuid
test ebx, FLAG_CPUID1_EBX_AVX2
lea arg3, [gf_vect_dot_prod_avx2 WRT_OPT]
cmovne arg1, arg3
;; Does it have xmm and ymm support
xor ecx, ecx
xgetbv
and eax, FLAG_XGETBV_EAX_XMM_YMM
cmp eax, FLAG_XGETBV_EAX_XMM_YMM
je _done_gf_vect_dot_prod_init
lea arg1, [gf_vect_dot_prod_sse WRT_OPT]
_done_gf_vect_dot_prod_init:
pop arg5
pop arg4
pop arg3
pop arg2
mov [gf_vect_dot_prod_dispatched], arg1
pop arg1
ret
;;;;
; gf_vect_mad multibinary function
;;;;
global gf_vect_mad:function
gf_vect_mad_mbinit:
call gf_vect_mad_dispatch_init
gf_vect_mad:
jmp wrd_sz [gf_vect_mad_dispatched]
gf_vect_mad_dispatch_init:
push arg1
%ifidn __OUTPUT_FORMAT__, elf32 ;; 32-bit check
lea arg1, [gf_vect_mad_base]
%else
push rax
push rbx
push rcx
push rdx
lea arg1, [gf_vect_mad_base WRT_OPT] ; Default
mov eax, 1
cpuid
lea rbx, [gf_vect_mad_sse WRT_OPT]
test ecx, FLAG_CPUID1_ECX_SSE4_1
cmovne arg1, rbx
and ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
cmp ecx, (FLAG_CPUID1_ECX_AVX | FLAG_CPUID1_ECX_OSXSAVE)
lea rbx, [gf_vect_mad_avx WRT_OPT]
jne _done_gf_vect_mad_init
mov rsi, rbx
;; Try for AVX2
xor ecx, ecx
mov eax, 7
cpuid
test ebx, FLAG_CPUID1_EBX_AVX2
lea rbx, [gf_vect_mad_avx2 WRT_OPT]
cmovne rsi, rbx
;; Does it have xmm and ymm support
xor ecx, ecx
xgetbv
and eax, FLAG_XGETBV_EAX_XMM_YMM
cmp eax, FLAG_XGETBV_EAX_XMM_YMM
je _done_gf_vect_mad_init
lea rsi, [gf_vect_mad_sse WRT_OPT]
_done_gf_vect_mad_init:
pop rdx
pop rcx
pop rbx
pop rax
%endif ;; END 32-bit check
mov [gf_vect_mad_dispatched], arg1
pop arg1
ret
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion ec_encode_data, 00, 03, 0133
slversion gf_vect_mul, 00, 02, 0134
slversion ec_encode_data_update, 00, 02, 0212
slversion gf_vect_dot_prod, 00, 02, 0138
slversion gf_vect_mad, 00, 01, 0213
-107
View File
@@ -1,107 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; 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 Intel Corporation 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "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 THE COPYRIGHT
; OWNER OR CONTRIBUTORS 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.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
%define EFLAGS_HAS_CPUID (1<<21)
%define FLAG_CPUID1_ECX_CLMUL (1<<1)
%define FLAG_CPUID1_EDX_SSE2 (1<<26)
%define FLAG_CPUID1_ECX_SSE3 (1)
%define FLAG_CPUID1_ECX_SSE4_1 (1<<19)
%define FLAG_CPUID1_ECX_SSE4_2 (1<<20)
%define FLAG_CPUID1_ECX_POPCNT (1<<23)
%define FLAG_CPUID1_ECX_AESNI (1<<25)
%define FLAG_CPUID1_ECX_OSXSAVE (1<<27)
%define FLAG_CPUID1_ECX_AVX (1<<28)
%define FLAG_CPUID1_EBX_AVX2 (1<<5)
%define FLAG_XGETBV_EAX_XMM_YMM 0x6
%define FLAG_CPUID1_EAX_AVOTON 0x000406d0
; define d and w variants for registers
%define raxd eax
%define raxw ax
%define raxb al
%define rbxd ebx
%define rbxw bx
%define rbxb bl
%define rcxd ecx
%define rcxw cx
%define rcxb cl
%define rdxd edx
%define rdxw dx
%define rdxb dl
%define rsid esi
%define rsiw si
%define rsib sil
%define rdid edi
%define rdiw di
%define rdib dil
%define rbpd ebp
%define rbpw bp
%define rbpb bpl
%define ymm0x xmm0
%define ymm1x xmm1
%define ymm2x xmm2
%define ymm3x xmm3
%define ymm4x xmm4
%define ymm5x xmm5
%define ymm6x xmm6
%define ymm7x xmm7
%define ymm8x xmm8
%define ymm9x xmm9
%define ymm10x xmm10
%define ymm11x xmm11
%define ymm12x xmm12
%define ymm13x xmm13
%define ymm14x xmm14
%define ymm15x xmm15
%define DWORD(reg) reg %+ d
%define WORD(reg) reg %+ w
%define BYTE(reg) reg %+ b
%define XWORD(reg) reg %+ x
%ifidn __OUTPUT_FORMAT__,elf32
section .note.GNU-stack noalloc noexec nowrite progbits
section .text
%endif
%ifidn __OUTPUT_FORMAT__,elf64
section .note.GNU-stack noalloc noexec nowrite progbits
section .text
%endif
-80
View File
@@ -1,80 +0,0 @@
/**********************************************************************
Copyright(c) 2011-2015 Intel Corporation All rights reserved.
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 Intel Corporation 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"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 THE COPYRIGHT
OWNER OR CONTRIBUTORS 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.
**********************************************************************/
/**
* @file types.h
* @brief Defines standard width types.
*
*/
#ifndef __TYPES_H
#define __TYPES_H
#ifdef __cplusplus
extern "C" {
#endif
#if !defined(__unix__) && !defined(__APPLE__)
#ifdef __MINGW32__
# include <_mingw.h>
#endif
typedef unsigned __int64 UINT64;
typedef __int64 INT64;
typedef unsigned __int32 UINT32;
typedef unsigned __int16 UINT16;
typedef unsigned char UINT8;
#else
typedef unsigned long int UINT64;
typedef long int INT64;
typedef unsigned int UINT32;
typedef unsigned short int UINT16;
typedef unsigned char UINT8;
#endif
#if defined(__unix__) || defined(__APPLE__)
# define DECLARE_ALIGNED(decl, alignval) decl __attribute__((aligned(alignval)))
# define __forceinline static inline
#else
# define DECLARE_ALIGNED(decl, alignval) __declspec(align(alignval)) decl
# define posix_memalign(p, algn, len) (NULL == (*((char**)(p)) = (void*) _aligned_malloc(len, algn)))
#endif
#ifdef DEBUG
# define DEBUG_PRINT(x) printf x
#else
# define DEBUG_PRINT(x) do {} while (0)
#endif
#ifdef __cplusplus
}
#endif
#endif //__TYPES_H
-123
View File
@@ -1,123 +0,0 @@
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package erasure
// #cgo CFLAGS: -O0
// #include <stdlib.h>
// #include "ec.h"
// #include "ec_minio_common.h"
import "C"
import (
"errors"
"fmt"
"unsafe"
)
// Decode decodes erasure coded blocks of data into its original
// form. Erasure coded data contains K data blocks and M parity
// blocks. Decode can withstand data loss up to any M number of blocks.
//
// "encodedDataBlocks" is an array of K data blocks and M parity
// blocks. Data blocks are position and order dependent. Missing blocks
// are set to "nil". There must be at least "K" number of data|parity
// blocks.
//
// "dataLen" is the length of original source data
func (e *Erasure) Decode(encodedDataBlocks [][]byte, dataLen int) (decodedData []byte, err error) {
e.mutex.Lock()
defer e.mutex.Unlock()
var source, target **C.uchar
k := int(e.params.K)
m := int(e.params.M)
n := k + m
// We need the data and parity blocks preserved in the same order. Missing blocks are set to nil.
if len(encodedDataBlocks) != n {
msg := fmt.Sprintf("Encoded data blocks slice must of length [%d]", n)
return nil, errors.New(msg)
}
// Length of a single encoded block
encodedBlockLen := GetEncodedBlockLen(dataLen, uint8(k))
// Keep track of errors per block.
missingEncodedBlocks := make([]int, n+1)
var missingEncodedBlocksCount int
// Check for the missing encoded blocks
for i := range encodedDataBlocks {
if encodedDataBlocks[i] == nil || len(encodedDataBlocks[i]) == 0 {
missingEncodedBlocks[missingEncodedBlocksCount] = i
missingEncodedBlocksCount++
}
}
// Cannot reconstruct original data. Need at least M number of data or parity blocks.
if missingEncodedBlocksCount > m {
return nil, fmt.Errorf("Cannot reconstruct original data. Need at least [%d] data or parity blocks", m)
}
// Convert from Go int slice to C int array
missingEncodedBlocksC := intSlice2CIntArray(missingEncodedBlocks[:missingEncodedBlocksCount])
// Allocate buffer for the missing blocks
for i := range encodedDataBlocks {
if encodedDataBlocks[i] == nil || len(encodedDataBlocks[i]) == 0 {
encodedDataBlocks[i] = make([]byte, encodedBlockLen)
}
}
// If not already initialized, recompute and cache
if e.decodeMatrix == nil || e.decodeTbls == nil || e.decodeIndex == nil {
var decodeMatrix, decodeTbls *C.uchar
var decodeIndex *C.uint32_t
C.minio_init_decoder(missingEncodedBlocksC, C.int(k), C.int(n), C.int(missingEncodedBlocksCount),
e.encodeMatrix, &decodeMatrix, &decodeTbls, &decodeIndex)
// cache this for future needs
e.decodeMatrix = decodeMatrix
e.decodeTbls = decodeTbls
e.decodeIndex = decodeIndex
}
// Make a slice of pointers to encoded blocks. Necessary to bridge to the C world.
pointers := make([]*byte, n)
for i := range encodedDataBlocks {
pointers[i] = &encodedDataBlocks[i][0]
}
// Get pointers to source "data" and target "parity" blocks from the output byte array.
ret := C.minio_get_source_target(C.int(missingEncodedBlocksCount), C.int(k), C.int(m), missingEncodedBlocksC,
e.decodeIndex, (**C.uchar)(unsafe.Pointer(&pointers[0])), &source, &target)
if int(ret) == -1 {
return nil, errors.New("Unable to decode data")
}
// Decode data
C.ec_encode_data(C.int(encodedBlockLen), C.int(k), C.int(missingEncodedBlocksCount), e.decodeTbls,
source, target)
// Allocate buffer to output buffer
decodedData = make([]byte, 0, encodedBlockLen*int(k))
for i := 0; i < int(k); i++ {
decodedData = append(decodedData, encodedDataBlocks[i]...)
}
return decodedData[:dataLen], nil
}
-174
View File
@@ -1,174 +0,0 @@
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package erasure
// #include <stdlib.h>
// #include "ec.h"
// #include "ec_minio_common.h"
import "C"
import (
"errors"
"sync"
"unsafe"
)
// Block alignment
const (
SIMDAlign = 32
)
// Params is a configuration set for building an encoder. It is created using ValidateParams().
type Params struct {
K uint8
M uint8
}
// Erasure is an object used to encode and decode data.
type Erasure struct {
params *Params
encodeMatrix, encodeTbls *C.uchar
decodeMatrix, decodeTbls *C.uchar
decodeIndex *C.uint32_t
mutex *sync.Mutex
}
// ValidateParams creates an Params object.
//
// k and m represent the matrix size, which corresponds to the protection level
// technique is the matrix type. Valid inputs are Cauchy (recommended) or Vandermonde.
//
func ValidateParams(k, m uint8) (*Params, error) {
if k < 1 {
return nil, errors.New("k cannot be zero")
}
if m < 1 {
return nil, errors.New("m cannot be zero")
}
if k+m > 255 {
return nil, errors.New("(k + m) cannot be bigger than Galois field GF(2^8) - 1")
}
return &Params{
K: k,
M: m,
}, nil
}
// NewErasure creates an encoder object with a given set of parameters.
func NewErasure(ep *Params) *Erasure {
var k = C.int(ep.K)
var m = C.int(ep.M)
var encodeMatrix *C.uchar
var encodeTbls *C.uchar
C.minio_init_encoder(k, m, &encodeMatrix, &encodeTbls)
return &Erasure{
params: ep,
encodeMatrix: encodeMatrix,
encodeTbls: encodeTbls,
decodeMatrix: nil,
decodeTbls: nil,
decodeIndex: nil,
mutex: new(sync.Mutex),
}
}
// GetEncodedBlocksLen - total length of all encoded blocks
func GetEncodedBlocksLen(inputLen int, k, m uint8) (outputLen int) {
outputLen = GetEncodedBlockLen(inputLen, k) * int(k+m)
return outputLen
}
// GetEncodedBlockLen - length per block of encoded blocks
func GetEncodedBlockLen(inputLen int, k uint8) (encodedOutputLen int) {
alignment := int(k) * SIMDAlign
remainder := inputLen % alignment
paddedInputLen := inputLen
if remainder != 0 {
paddedInputLen = inputLen + (alignment - remainder)
}
encodedOutputLen = paddedInputLen / int(k)
return encodedOutputLen
}
// Encode erasure codes a block of data in "k" data blocks and "m" parity blocks.
// Output is [k+m][]blocks of data and parity slices.
func (e *Erasure) Encode(inputData []byte) (encodedBlocks [][]byte, err error) {
e.mutex.Lock()
defer e.mutex.Unlock()
k := int(e.params.K) // "k" data blocks
m := int(e.params.M) // "m" parity blocks
n := k + m // "n" total encoded blocks
// Length of a single encoded chunk.
// Total number of encoded chunks = "k" data + "m" parity blocks
encodedBlockLen := GetEncodedBlockLen(len(inputData), uint8(k))
// Length of total number of "k" data chunks
encodedDataBlocksLen := encodedBlockLen * k
// Length of extra padding required for the data blocks.
encodedDataBlocksPadLen := encodedDataBlocksLen - len(inputData)
// Extend inputData buffer to accommodate coded data blocks if necesssary
if encodedDataBlocksPadLen > 0 {
padding := make([]byte, encodedDataBlocksPadLen)
// Expand with new padded blocks to the byte array
inputData = append(inputData, padding...)
}
// Extend inputData buffer to accommodate coded parity blocks
{ // Local Scope
encodedParityBlocksLen := encodedBlockLen * m
parityBlocks := make([]byte, encodedParityBlocksLen)
inputData = append(inputData, parityBlocks...)
}
// Allocate memory to the "encoded blocks" return buffer
encodedBlocks = make([][]byte, n) // Return buffer
// Neccessary to bridge Go to the C world. C requires 2D arry of pointers to
// byte array. "encodedBlocks" is a 2D slice.
pointersToEncodedBlock := make([]*byte, n) // Pointers to encoded blocks.
// Copy data block slices to encoded block buffer
for i := 0; i < k; i++ {
encodedBlocks[i] = inputData[i*encodedBlockLen : (i+1)*encodedBlockLen]
pointersToEncodedBlock[i] = &encodedBlocks[i][0]
}
// Copy erasure block slices to encoded block buffer
for i := k; i < n; i++ {
encodedBlocks[i] = make([]byte, encodedBlockLen)
pointersToEncodedBlock[i] = &encodedBlocks[i][0]
}
// Erasure code the data into K data blocks and M parity
// blocks. Only the parity blocks are filled. Data blocks remain
// intact.
C.ec_encode_data(C.int(encodedBlockLen), C.int(k), C.int(m), e.encodeTbls,
(**C.uchar)(unsafe.Pointer(&pointersToEncodedBlock[:k][0])), // Pointers to data blocks
(**C.uchar)(unsafe.Pointer(&pointersToEncodedBlock[k:][0]))) // Pointers to parity blocks
return encodedBlocks, nil
}
-80
View File
@@ -1,80 +0,0 @@
/*
* Minio Cloud Storage, (C) 2014 Minio, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package erasure
import (
"bytes"
"testing"
. "gopkg.in/check.v1"
)
type MySuite struct{}
var _ = Suite(&MySuite{})
func Test(t *testing.T) { TestingT(t) }
const (
k = 10
m = 5
)
func corruptChunks(chunks [][]byte, errorIndex []int) [][]byte {
for _, err := range errorIndex {
chunks[err] = nil
}
return chunks
}
func (s *MySuite) TestEncodeDecodeFailure(c *C) {
ep, err := ValidateParams(k, m)
c.Assert(err, IsNil)
data := []byte("Lorem Ipsum is simply dummy text of the printing and typesetting industry. Lorem Ipsum has been the industry's standard dummy text ever since the 1500s, when an unknown printer took a galley of type and scrambled it to make a type specimen book. It has survived not only five centuries, but also the leap into electronic typesetting, remaining essentially unchanged. It was popularised in the 1960s with the release of Letraset sheets containing Lorem Ipsum passages, and more recently with desktop publishing software like Aldus PageMaker including versions of Lorem Ipsum.")
e := NewErasure(ep)
chunks, err := e.Encode(data)
c.Assert(err, IsNil)
errorIndex := []int{0, 3, 5, 9, 11, 13}
chunks = corruptChunks(chunks, errorIndex)
_, err = e.Decode(chunks, len(data))
c.Assert(err, Not(IsNil))
}
func (s *MySuite) TestEncodeDecodeSuccess(c *C) {
ep, err := ValidateParams(k, m)
c.Assert(err, IsNil)
data := []byte("Lorem Ipsum is simply dummy text of the printing and typesetting industry. Lorem Ipsum has been the industry's standard dummy text ever since the 1500s, when an unknown printer took a galley of type and scrambled it to make a type specimen book. It has survived not only five centuries, but also the leap into electronic typesetting, remaining essentially unchanged. It was popularised in the 1960s with the release of Letraset sheets containing Lorem Ipsum passages, and more recently with desktop publishing software like Aldus PageMaker including versions of Lorem Ipsum.")
e := NewErasure(ep)
chunks, err := e.Encode(data)
c.Assert(err, IsNil)
errorIndex := []int{0, 3, 5, 9, 13}
chunks = corruptChunks(chunks, errorIndex)
recoveredData, err := e.Decode(chunks, len(data))
c.Assert(err, IsNil)
if !bytes.Equal(data, recoveredData) {
c.Fatalf("Recovered data mismatches with original data")
}
}
-43
View File
@@ -1,43 +0,0 @@
// !build amd64
package erasure
//go:generate yasm -f macho64 --prefix=_ ec_multibinary.asm -o ec_multibinary.syso
//go:generate yasm -f macho64 --prefix=_ gf_2vect_mad_avx2.asm -o gf_2vect_mad_avx2.syso
//go:generate yasm -f macho64 --prefix=_ gf_2vect_mad_avx.asm -o gf_2vect_mad_avx.syso
//go:generate yasm -f macho64 --prefix=_ gf_2vect_mad_sse.asm -o gf_2vect_mad_sse.syso
//go:generate yasm -f macho64 --prefix=_ gf_3vect_mad_avx2.asm -o gf_3vect_mad_avx2.syso
//go:generate yasm -f macho64 --prefix=_ gf_3vect_mad_avx.asm -o gf_3vect_mad_avx.syso
//go:generate yasm -f macho64 --prefix=_ gf_3vect_mad_sse.asm -o gf_3vect_mad_sse.syso
//go:generate yasm -f macho64 --prefix=_ gf_4vect_mad_avx2.asm -o gf_4vect_mad_avx2.syso
//go:generate yasm -f macho64 --prefix=_ gf_4vect_mad_avx.asm -o gf_4vect_mad_avx.syso
//go:generate yasm -f macho64 --prefix=_ gf_4vect_mad_sse.asm -o gf_4vect_mad_sse.syso
//go:generate yasm -f macho64 --prefix=_ gf_5vect_mad_avx2.asm -o gf_5vect_mad_avx2.syso
//go:generate yasm -f macho64 --prefix=_ gf_5vect_mad_avx.asm -o gf_5vect_mad_avx.syso
//go:generate yasm -f macho64 --prefix=_ gf_5vect_mad_sse.asm -o gf_5vect_mad_sse.syso
//go:generate yasm -f macho64 --prefix=_ gf_6vect_mad_avx2.asm -o gf_6vect_mad_avx2.syso
//go:generate yasm -f macho64 --prefix=_ gf_6vect_mad_avx.asm -o gf_6vect_mad_avx.syso
//go:generate yasm -f macho64 --prefix=_ gf_6vect_mad_sse.asm -o gf_6vect_mad_sse.syso
//go:generate yasm -f macho64 --prefix=_ gf_vect_mad_avx2.asm -o gf_vect_mad_avx2.syso
//go:generate yasm -f macho64 --prefix=_ gf_vect_mad_avx.asm -o gf_vect_mad_avx.syso
//go:generate yasm -f macho64 --prefix=_ gf_vect_mad_sse.asm -o gf_vect_mad_sse.syso
//go:generate yasm -f macho64 --prefix=_ gf_2vect_dot_prod_avx2.asm -o gf_2vect_dot_prod_avx2.syso
//go:generate yasm -f macho64 --prefix=_ gf_2vect_dot_prod_avx.asm -o gf_2vect_dot_prod_avx.syso
//go:generate yasm -f macho64 --prefix=_ gf_2vect_dot_prod_sse.asm -o gf_2vect_dot_prod_sse.syso
//go:generate yasm -f macho64 --prefix=_ gf_3vect_dot_prod_avx2.asm -o gf_3vect_dot_prod_avx2.syso
//go:generate yasm -f macho64 --prefix=_ gf_3vect_dot_prod_avx.asm -o gf_3vect_dot_prod_avx.syso
//go:generate yasm -f macho64 --prefix=_ gf_3vect_dot_prod_sse.asm -o gf_3vect_dot_prod_sse.syso
//go:generate yasm -f macho64 --prefix=_ gf_4vect_dot_prod_avx2.asm -o gf_4vect_dot_prod_avx2.syso
//go:generate yasm -f macho64 --prefix=_ gf_4vect_dot_prod_avx.asm -o gf_4vect_dot_prod_avx.syso
//go:generate yasm -f macho64 --prefix=_ gf_4vect_dot_prod_sse.asm -o gf_4vect_dot_prod_sse.syso
//go:generate yasm -f macho64 --prefix=_ gf_5vect_dot_prod_avx2.asm -o gf_5vect_dot_prod_avx2.syso
//go:generate yasm -f macho64 --prefix=_ gf_5vect_dot_prod_avx.asm -o gf_5vect_dot_prod_avx.syso
//go:generate yasm -f macho64 --prefix=_ gf_5vect_dot_prod_sse.asm -o gf_5vect_dot_prod_sse.syso
//go:generate yasm -f macho64 --prefix=_ gf_6vect_dot_prod_avx2.asm -o gf_6vect_dot_prod_avx2.syso
//go:generate yasm -f macho64 --prefix=_ gf_6vect_dot_prod_avx.asm -o gf_6vect_dot_prod_avx.syso
//go:generate yasm -f macho64 --prefix=_ gf_6vect_dot_prod_sse.asm -o gf_6vect_dot_prod_sse.syso
//go:generate yasm -f macho64 --prefix=_ gf_vect_dot_prod_avx2.asm -o gf_vect_dot_prod_avx2.syso
//go:generate yasm -f macho64 --prefix=_ gf_vect_dot_prod_avx.asm -o gf_vect_dot_prod_avx.syso
//go:generate yasm -f macho64 --prefix=_ gf_vect_dot_prod_sse.asm -o gf_vect_dot_prod_sse.syso
//go:generate yasm -f macho64 --prefix=_ gf_vect_mul_avx.asm -o gf_vect_mul_avx.syso
//go:generate yasm -f macho64 --prefix=_ gf_vect_mul_sse.asm -o gf_vect_mul_sse.syso
-43
View File
@@ -1,43 +0,0 @@
// !build amd64
package erasure
//go:generate yasm -f elf64 ec_multibinary.asm -o ec_multibinary.syso
//go:generate yasm -f elf64 gf_2vect_mad_avx2.asm -o gf_2vect_mad_avx2.syso
//go:generate yasm -f elf64 gf_2vect_mad_avx.asm -o gf_2vect_mad_avx.syso
//go:generate yasm -f elf64 gf_2vect_mad_sse.asm -o gf_2vect_mad_sse.syso
//go:generate yasm -f elf64 gf_3vect_mad_avx2.asm -o gf_3vect_mad_avx2.syso
//go:generate yasm -f elf64 gf_3vect_mad_avx.asm -o gf_3vect_mad_avx.syso
//go:generate yasm -f elf64 gf_3vect_mad_sse.asm -o gf_3vect_mad_sse.syso
//go:generate yasm -f elf64 gf_4vect_mad_avx2.asm -o gf_4vect_mad_avx2.syso
//go:generate yasm -f elf64 gf_4vect_mad_avx.asm -o gf_4vect_mad_avx.syso
//go:generate yasm -f elf64 gf_4vect_mad_sse.asm -o gf_4vect_mad_sse.syso
//go:generate yasm -f elf64 gf_5vect_mad_avx2.asm -o gf_5vect_mad_avx2.syso
//go:generate yasm -f elf64 gf_5vect_mad_avx.asm -o gf_5vect_mad_avx.syso
//go:generate yasm -f elf64 gf_5vect_mad_sse.asm -o gf_5vect_mad_sse.syso
//go:generate yasm -f elf64 gf_6vect_mad_avx2.asm -o gf_6vect_mad_avx2.syso
//go:generate yasm -f elf64 gf_6vect_mad_avx.asm -o gf_6vect_mad_avx.syso
//go:generate yasm -f elf64 gf_6vect_mad_sse.asm -o gf_6vect_mad_sse.syso
//go:generate yasm -f elf64 gf_vect_mad_avx2.asm -o gf_vect_mad_avx2.syso
//go:generate yasm -f elf64 gf_vect_mad_avx.asm -o gf_vect_mad_avx.syso
//go:generate yasm -f elf64 gf_vect_mad_sse.asm -o gf_vect_mad_sse.syso
//go:generate yasm -f elf64 gf_2vect_dot_prod_avx2.asm -o gf_2vect_dot_prod_avx2.syso
//go:generate yasm -f elf64 gf_2vect_dot_prod_avx.asm -o gf_2vect_dot_prod_avx.syso
//go:generate yasm -f elf64 gf_2vect_dot_prod_sse.asm -o gf_2vect_dot_prod_sse.syso
//go:generate yasm -f elf64 gf_3vect_dot_prod_avx2.asm -o gf_3vect_dot_prod_avx2.syso
//go:generate yasm -f elf64 gf_3vect_dot_prod_avx.asm -o gf_3vect_dot_prod_avx.syso
//go:generate yasm -f elf64 gf_3vect_dot_prod_sse.asm -o gf_3vect_dot_prod_sse.syso
//go:generate yasm -f elf64 gf_4vect_dot_prod_avx2.asm -o gf_4vect_dot_prod_avx2.syso
//go:generate yasm -f elf64 gf_4vect_dot_prod_avx.asm -o gf_4vect_dot_prod_avx.syso
//go:generate yasm -f elf64 gf_4vect_dot_prod_sse.asm -o gf_4vect_dot_prod_sse.syso
//go:generate yasm -f elf64 gf_5vect_dot_prod_avx2.asm -o gf_5vect_dot_prod_avx2.syso
//go:generate yasm -f elf64 gf_5vect_dot_prod_avx.asm -o gf_5vect_dot_prod_avx.syso
//go:generate yasm -f elf64 gf_5vect_dot_prod_sse.asm -o gf_5vect_dot_prod_sse.syso
//go:generate yasm -f elf64 gf_6vect_dot_prod_avx2.asm -o gf_6vect_dot_prod_avx2.syso
//go:generate yasm -f elf64 gf_6vect_dot_prod_avx.asm -o gf_6vect_dot_prod_avx.syso
//go:generate yasm -f elf64 gf_6vect_dot_prod_sse.asm -o gf_6vect_dot_prod_sse.syso
//go:generate yasm -f elf64 gf_vect_dot_prod_avx2.asm -o gf_vect_dot_prod_avx2.syso
//go:generate yasm -f elf64 gf_vect_dot_prod_avx.asm -o gf_vect_dot_prod_avx.syso
//go:generate yasm -f elf64 gf_vect_dot_prod_sse.asm -o gf_vect_dot_prod_sse.syso
//go:generate yasm -f elf64 gf_vect_mul_avx.asm -o gf_vect_mul_avx.syso
//go:generate yasm -f elf64 gf_vect_mul_sse.asm -o gf_vect_mul_sse.syso
-43
View File
@@ -1,43 +0,0 @@
// !build amd64
package erasure
//go:generate yasm -f win64 ec_multibinary.asm -o ec_multibinary.syso
//go:generate yasm -f win64 gf_2vect_mad_avx2.asm -o gf_2vect_mad_avx2.syso
//go:generate yasm -f win64 gf_2vect_mad_avx.asm -o gf_2vect_mad_avx.syso
//go:generate yasm -f win64 gf_2vect_mad_sse.asm -o gf_2vect_mad_sse.syso
//go:generate yasm -f win64 gf_3vect_mad_avx2.asm -o gf_3vect_mad_avx2.syso
//go:generate yasm -f win64 gf_3vect_mad_avx.asm -o gf_3vect_mad_avx.syso
//go:generate yasm -f win64 gf_3vect_mad_sse.asm -o gf_3vect_mad_sse.syso
//go:generate yasm -f win64 gf_4vect_mad_avx2.asm -o gf_4vect_mad_avx2.syso
//go:generate yasm -f win64 gf_4vect_mad_avx.asm -o gf_4vect_mad_avx.syso
//go:generate yasm -f win64 gf_4vect_mad_sse.asm -o gf_4vect_mad_sse.syso
//go:generate yasm -f win64 gf_5vect_mad_avx2.asm -o gf_5vect_mad_avx2.syso
//go:generate yasm -f win64 gf_5vect_mad_avx.asm -o gf_5vect_mad_avx.syso
//go:generate yasm -f win64 gf_5vect_mad_sse.asm -o gf_5vect_mad_sse.syso
//go:generate yasm -f win64 gf_6vect_mad_avx2.asm -o gf_6vect_mad_avx2.syso
//go:generate yasm -f win64 gf_6vect_mad_avx.asm -o gf_6vect_mad_avx.syso
//go:generate yasm -f win64 gf_6vect_mad_sse.asm -o gf_6vect_mad_sse.syso
//go:generate yasm -f win64 gf_vect_mad_avx2.asm -o gf_vect_mad_avx2.syso
//go:generate yasm -f win64 gf_vect_mad_avx.asm -o gf_vect_mad_avx.syso
//go:generate yasm -f win64 gf_vect_mad_sse.asm -o gf_vect_mad_sse.syso
//go:generate yasm -f win64 gf_2vect_dot_prod_avx2.asm -o gf_2vect_dot_prod_avx2.syso
//go:generate yasm -f win64 gf_2vect_dot_prod_avx.asm -o gf_2vect_dot_prod_avx.syso
//go:generate yasm -f win64 gf_2vect_dot_prod_sse.asm -o gf_2vect_dot_prod_sse.syso
//go:generate yasm -f win64 gf_3vect_dot_prod_avx2.asm -o gf_3vect_dot_prod_avx2.syso
//go:generate yasm -f win64 gf_3vect_dot_prod_avx.asm -o gf_3vect_dot_prod_avx.syso
//go:generate yasm -f win64 gf_3vect_dot_prod_sse.asm -o gf_3vect_dot_prod_sse.syso
//go:generate yasm -f win64 gf_4vect_dot_prod_avx2.asm -o gf_4vect_dot_prod_avx2.syso
//go:generate yasm -f win64 gf_4vect_dot_prod_avx.asm -o gf_4vect_dot_prod_avx.syso
//go:generate yasm -f win64 gf_4vect_dot_prod_sse.asm -o gf_4vect_dot_prod_sse.syso
//go:generate yasm -f win64 gf_5vect_dot_prod_avx2.asm -o gf_5vect_dot_prod_avx2.syso
//go:generate yasm -f win64 gf_5vect_dot_prod_avx.asm -o gf_5vect_dot_prod_avx.syso
//go:generate yasm -f win64 gf_5vect_dot_prod_sse.asm -o gf_5vect_dot_prod_sse.syso
//go:generate yasm -f win64 gf_6vect_dot_prod_avx2.asm -o gf_6vect_dot_prod_avx2.syso
//go:generate yasm -f win64 gf_6vect_dot_prod_avx.asm -o gf_6vect_dot_prod_avx.syso
//go:generate yasm -f win64 gf_6vect_dot_prod_sse.asm -o gf_6vect_dot_prod_sse.syso
//go:generate yasm -f win64 gf_vect_dot_prod_avx2.asm -o gf_vect_dot_prod_avx2.syso
//go:generate yasm -f win64 gf_vect_dot_prod_avx.asm -o gf_vect_dot_prod_avx.syso
//go:generate yasm -f win64 gf_vect_dot_prod_sse.asm -o gf_vect_dot_prod_sse.syso
//go:generate yasm -f win64 gf_vect_mul_avx.asm -o gf_vect_mul_avx.syso
//go:generate yasm -f win64 gf_vect_mul_sse.asm -o gf_vect_mul_sse.syso
-374
View File
@@ -1,374 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; 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 Intel Corporation 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "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 THE COPYRIGHT
; OWNER OR CONTRIBUTORS 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.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_2vect_dot_prod_avx(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r9
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
%endmacro
%macro FUNC_RESTORE 0
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define stack_size 3*16 + 3*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
save_xmm128 xmm6, 0*16
save_xmm128 xmm7, 1*16
save_xmm128 xmm8, 2*16
save_reg r12, 3*16 + 0*8
save_reg r13, 3*16 + 1*8
save_reg r14, 3*16 + 2*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp + 0*16]
vmovdqa xmm7, [rsp + 1*16]
vmovdqa xmm8, [rsp + 2*16]
mov r12, [rsp + 3*16 + 0*8]
mov r13, [rsp + 3*16 + 1*8]
mov r14, [rsp + 3*16 + 2*8]
add rsp, stack_size
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf32
;;;================== High Address;
;;; arg4
;;; arg3
;;; arg2
;;; arg1
;;; arg0
;;; return
;;;<================= esp of caller
;;; ebp
;;;<================= ebp = esp
;;; var0
;;; esi
;;; edi
;;; ebx
;;;<================= esp of callee
;;;
;;;================== Low Address;
%define PS 4
%define LOG_PS 2
%define func(x) x:
%define arg(x) [ebp + PS*2 + PS*x]
%define var(x) [ebp - PS - PS*x]
%define trans ecx
%define trans2 esi
%define arg0 trans ;trans and trans2 are for the variables in stack
%define arg0_m arg(0)
%define arg1 ebx
%define arg2 arg2_m
%define arg2_m arg(2)
%define arg3 trans
%define arg3_m arg(3)
%define arg4 trans
%define arg4_m arg(4)
%define tmp edx
%define tmp2 edi
%define tmp3 trans2
%define tmp4 trans2
%define tmp4_m var(0)
%define return eax
%macro SLDR 2 ;; stack load/restore
mov %1, %2
%endmacro
%define SSTR SLDR
%macro FUNC_SAVE 0
push ebp
mov ebp, esp
sub esp, PS*1 ;1 local variable
push esi
push edi
push ebx
mov arg1, arg(1)
%endmacro
%macro FUNC_RESTORE 0
pop ebx
pop edi
pop esi
add esp, PS*1 ;1 local variable
pop ebp
%endmacro
%endif ; output formats
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest1 arg4
%define vec_i tmp2
%define ptr tmp3
%define dest2 tmp4
%define pos return
%ifidn PS,4 ;32-bit code
%define len_m arg0_m
%define src_m arg3_m
%define dest1_m arg4_m
%define dest2_m tmp4_m
%endif
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
%ifidn PS,8 ; 64-bit code
default rel
[bits 64]
%endif
section .text
%ifidn PS,8 ;64-bit code
%define xmask0f xmm8
%define xgft1_lo xmm7
%define xgft1_hi xmm6
%define xgft2_lo xmm5
%define xgft2_hi xmm4
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%else ;32-bit code
%define xmask0f xmm4
%define xgft1_lo xmm7
%define xgft1_hi xmm6
%define xgft2_lo xgft1_lo
%define xgft2_hi xgft1_hi
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%endif
align 16
global gf_2vect_dot_prod_avx:function
func(gf_2vect_dot_prod_avx)
FUNC_SAVE
SLDR len, len_m
sub len, 16
SSTR len_m, len
jl .return_fail
xor pos, pos
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
SLDR dest1, dest1_m
mov dest2, [dest1+PS]
SSTR dest2_m, dest2
mov dest1, [dest1]
SSTR dest1_m, dest1
.loop16
vpxor xp1, xp1
vpxor xp2, xp2
mov tmp, mul_array
xor vec_i, vec_i
.next_vect
SLDR src, src_m
mov ptr, [src+vec_i]
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
vmovdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, ..., Ax{f0}
%ifidn PS,8 ; 64-bit code
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
vmovdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
add tmp, 32
add vec_i, PS
%endif
XLDR x0, [ptr+pos] ;Get next source vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp1, xgft1_hi ;xp1 += partial
%ifidn PS,4 ; 32-bit code
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
vmovdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
add tmp, 32
add vec_i, PS
%endif
vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_lo ;GF add high and low partials
vpxor xp2, xgft2_hi ;xp2 += partial
cmp vec_i, vec
jl .next_vect
SLDR dest1, dest1_m
SLDR dest2, dest2_m
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
SLDR len, len_m
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop16 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion gf_2vect_dot_prod_avx, 02, 04, 0191
-391
View File
@@ -1,391 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; 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 Intel Corporation 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "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 THE COPYRIGHT
; OWNER OR CONTRIBUTORS 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.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_2vect_dot_prod_avx2(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r9
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
%endmacro
%macro FUNC_RESTORE 0
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r9
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
%endmacro
%macro FUNC_RESTORE 0
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define stack_size 3*16 + 3*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
vmovdqa [rsp + 0*16], xmm6
vmovdqa [rsp + 1*16], xmm7
vmovdqa [rsp + 2*16], xmm8
save_reg r12, 3*16 + 0*8
save_reg r13, 3*16 + 1*8
save_reg r14, 3*16 + 2*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp + 0*16]
vmovdqa xmm7, [rsp + 1*16]
vmovdqa xmm8, [rsp + 2*16]
mov r12, [rsp + 3*16 + 0*8]
mov r13, [rsp + 3*16 + 1*8]
mov r14, [rsp + 3*16 + 2*8]
add rsp, stack_size
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf32
;;;================== High Address;
;;; arg4
;;; arg3
;;; arg2
;;; arg1
;;; arg0
;;; return
;;;<================= esp of caller
;;; ebp
;;;<================= ebp = esp
;;; var0
;;; esi
;;; edi
;;; ebx
;;;<================= esp of callee
;;;
;;;================== Low Address;
%define PS 4
%define LOG_PS 2
%define func(x) x:
%define arg(x) [ebp + PS*2 + PS*x]
%define var(x) [ebp - PS - PS*x]
%define trans ecx
%define trans2 esi
%define arg0 trans ;trans and trans2 are for the variables in stack
%define arg0_m arg(0)
%define arg1 ebx
%define arg2 arg2_m
%define arg2_m arg(2)
%define arg3 trans
%define arg3_m arg(3)
%define arg4 trans
%define arg4_m arg(4)
%define tmp edx
%define tmp.w edx
%define tmp.b dl
%define tmp2 edi
%define tmp3 trans2
%define tmp4 trans2
%define tmp4_m var(0)
%define return eax
%macro SLDR 2 ;stack load/restore
mov %1, %2
%endmacro
%define SSTR SLDR
%macro FUNC_SAVE 0
push ebp
mov ebp, esp
sub esp, PS*1 ;1 local variable
push esi
push edi
push ebx
mov arg1, arg(1)
%endmacro
%macro FUNC_RESTORE 0
pop ebx
pop edi
pop esi
add esp, PS*1 ;1 local variable
pop ebp
%endmacro
%endif ; output formats
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest1 arg4
%define vec_i tmp2
%define ptr tmp3
%define dest2 tmp4
%define pos return
%ifidn PS,4 ;32-bit code
%define len_m arg0_m
%define src_m arg3_m
%define dest1_m arg4_m
%define dest2_m tmp4_m
%endif
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
%ifidn PS,8 ;64-bit code
default rel
[bits 64]
%endif
section .text
%ifidn PS,8 ;64-bit code
%define xmask0f ymm8
%define xmask0fx xmm8
%define xgft1_lo ymm7
%define xgft1_hi ymm6
%define xgft2_lo ymm5
%define xgft2_hi ymm4
%define x0 ymm0
%define xtmpa ymm1
%define xp1 ymm2
%define xp2 ymm3
%else ;32-bit code
%define xmask0f ymm7
%define xmask0fx xmm7
%define xgft1_lo ymm5
%define xgft1_hi ymm4
%define xgft2_lo xgft1_lo
%define xgft2_hi xgft1_hi
%define x0 ymm0
%define xtmpa ymm1
%define xp1 ymm2
%define xp2 ymm3
%endif
align 16
global gf_2vect_dot_prod_avx2:function
func(gf_2vect_dot_prod_avx2)
FUNC_SAVE
SLDR len, len_m
sub len, 32
SSTR len_m, len
jl .return_fail
xor pos, pos
mov tmp.b, 0x0f
vpinsrb xmask0fx, xmask0fx, tmp.w, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
SLDR dest1, dest1_m
mov dest2, [dest1+PS]
SSTR dest2_m, dest2
mov dest1, [dest1]
SSTR dest1_m, dest1
.loop32
vpxor xp1, xp1
vpxor xp2, xp2
mov tmp, mul_array
xor vec_i, vec_i
.next_vect
SLDR src, src_m
mov ptr, [src+vec_i]
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
; " Ax{00}, Ax{10}, ..., Ax{f0}
vperm2i128 xgft1_hi, xgft1_lo, xgft1_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft1_lo, xgft1_lo, xgft1_lo, 0x00 ; swapped to lo | lo
%ifidn PS,8 ; 64-bit code
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
; " Bx{00}, Bx{10}, ..., Bx{f0}
vperm2i128 xgft2_hi, xgft2_lo, xgft2_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft2_lo, xgft2_lo, xgft2_lo, 0x00 ; swapped to lo | lo
XLDR x0, [ptr+pos] ;Get next source vector
add tmp, 32
add vec_i, PS
%else
XLDR x0, [ptr+pos] ;Get next source vector
%endif
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp1, xgft1_hi ;xp1 += partial
%ifidn PS,4 ; 32-bit code
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
; " Bx{00}, Bx{10}, ..., Bx{f0}
vperm2i128 xgft2_hi, xgft2_lo, xgft2_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft2_lo, xgft2_lo, xgft2_lo, 0x00 ; swapped to lo | lo
add tmp, 32
add vec_i, PS
%endif
vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_lo ;GF add high and low partials
vpxor xp2, xgft2_hi ;xp2 += partial
cmp vec_i, vec
jl .next_vect
SLDR dest1, dest1_m
SLDR dest2, dest2_m
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
SLDR len, len_m
add pos, 32 ;Loop on 32 bytes at a time
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop32 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion gf_2vect_dot_prod_avx2, 04, 04, 0196
-376
View File
@@ -1,376 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; 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 Intel Corporation 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "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 THE COPYRIGHT
; OWNER OR CONTRIBUTORS 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.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_2vect_dot_prod_sse(len, vec, *g_tbls, **buffs, **dests);
;;;
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define tmp3 r9
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
%endmacro
%macro FUNC_RESTORE 0
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define tmp r11
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define return rax
%macro SLDR 2
%endmacro
%define SSTR SLDR
%define PS 8
%define LOG_PS 3
%define stack_size 3*16 + 3*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
save_xmm128 xmm6, 0*16
save_xmm128 xmm7, 1*16
save_xmm128 xmm8, 2*16
save_reg r12, 3*16 + 0*8
save_reg r13, 3*16 + 1*8
save_reg r14, 3*16 + 2*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp + 0*16]
movdqa xmm7, [rsp + 1*16]
movdqa xmm8, [rsp + 2*16]
mov r12, [rsp + 3*16 + 0*8]
mov r13, [rsp + 3*16 + 1*8]
mov r14, [rsp + 3*16 + 2*8]
add rsp, stack_size
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf32
;;;================== High Address;
;;; arg4
;;; arg3
;;; arg2
;;; arg1
;;; arg0
;;; return
;;;<================= esp of caller
;;; ebp
;;;<================= ebp = esp
;;; var0
;;; esi
;;; edi
;;; ebx
;;;<================= esp of callee
;;;
;;;================== Low Address;
%define PS 4
%define LOG_PS 2
%define func(x) x:
%define arg(x) [ebp + PS*2 + PS*x]
%define var(x) [ebp - PS - PS*x]
%define trans ecx
%define trans2 esi
%define arg0 trans ;trans and trans2 are for the variables in stack
%define arg0_m arg(0)
%define arg1 ebx
%define arg2 arg2_m
%define arg2_m arg(2)
%define arg3 trans
%define arg3_m arg(3)
%define arg4 trans
%define arg4_m arg(4)
%define tmp edx
%define tmp2 edi
%define tmp3 trans2
%define tmp4 trans2
%define tmp4_m var(0)
%define return eax
%macro SLDR 2 ;; stack load/restore
mov %1, %2
%endmacro
%define SSTR SLDR
%macro FUNC_SAVE 0
push ebp
mov ebp, esp
sub esp, PS*1 ;1 local variable
push esi
push edi
push ebx
mov arg1, arg(1)
%endmacro
%macro FUNC_RESTORE 0
pop ebx
pop edi
pop esi
add esp, PS*1 ;1 local variable
pop ebp
%endmacro
%endif ; output formats
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest1 arg4
%define vec_i tmp2
%define ptr tmp3
%define dest2 tmp4
%define pos return
%ifidn PS,4 ;32-bit code
%define len_m arg0_m
%define src_m arg3_m
%define dest1_m arg4_m
%define dest2_m tmp4_m
%endif
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR movdqu
%define XSTR movdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR movdqa
%define XSTR movdqa
%else
%define XLDR movntdqa
%define XSTR movntdq
%endif
%endif
%ifidn PS,8 ;64-bit code
default rel
[bits 64]
%endif
section .text
%ifidn PS,8 ;64-bit code
%define xmask0f xmm8
%define xgft1_lo xmm7
%define xgft1_hi xmm6
%define xgft2_lo xmm5
%define xgft2_hi xmm4
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%else ;32-bit code
%define xmask0f xmm4
%define xgft1_lo xmm7
%define xgft1_hi xmm6
%define xgft2_lo xgft1_lo
%define xgft2_hi xgft1_hi
%define x0 xmm0
%define xtmpa xmm1
%define xp1 xmm2
%define xp2 xmm3
%endif
align 16
global gf_2vect_dot_prod_sse:function
func(gf_2vect_dot_prod_sse)
FUNC_SAVE
SLDR len, len_m
sub len, 16
SSTR len_m, len
jl .return_fail
xor pos, pos
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
SLDR dest1, dest1_m
mov dest2, [dest1+PS]
SSTR dest2_m, dest2
mov dest1, [dest1]
SSTR dest1_m, dest1
.loop16
pxor xp1, xp1
pxor xp2, xp2
mov tmp, mul_array
xor vec_i, vec_i
.next_vect
SLDR src, src_m
mov ptr, [src+vec_i]
movdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
movdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, ..., Ax{f0}
%ifidn PS,8 ;64-bit code
movdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
movdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
add tmp, 32
add vec_i, PS
%endif
XLDR x0, [ptr+pos] ;Get next source vector
movdqa xtmpa, x0 ;Keep unshifted copy of src
psraw x0, 4 ;Shift to put high nibble into bits 4-0
pand x0, xmask0f ;Mask high src nibble in bits 4-0
pand xtmpa, xmask0f ;Mask low src nibble in bits 4-0
pshufb xgft1_hi, x0 ;Lookup mul table of high nibble
pshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft1_hi, xgft1_lo ;GF add high and low partials
pxor xp1, xgft1_hi ;xp1 += partial
%ifidn PS,4 ;32-bit code
movdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
movdqu xgft2_hi, [tmp+vec*(32/PS)+16] ; " Bx{00}, Bx{10}, ..., Bx{f0}
add tmp, 32
add vec_i, PS
%endif
pshufb xgft2_hi, x0 ;Lookup mul table of high nibble
pshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
pxor xgft2_hi, xgft2_lo ;GF add high and low partials
pxor xp2, xgft2_hi ;xp2 += partial
cmp vec_i, vec
jl .next_vect
SLDR dest1, dest1_m
SLDR dest2, dest2_m
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
SLDR len, len_m
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop16 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion gf_2vect_dot_prod_sse, 00, 03, 0062
-258
View File
@@ -1,258 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; 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 Intel Corporation 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "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 THE COPYRIGHT
; OWNER OR CONTRIBUTORS 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.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_2vect_mad_avx(len, vec, vec_i, mul_array, src, dest);
;;;
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define tmp2 r10
%define return rax
%define return.w eax
%define stack_size 16*9 + 3*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
save_reg r12, 9*16 + 0*8
save_reg r15, 9*16 + 1*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
mov r12, [rsp + 9*16 + 0*8]
mov r15, [rsp + 9*16 + 1*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp2 r10
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
;;; gf_2vect_mad_avx(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 tmp2
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f xmm14
%define xgft1_lo xmm13
%define xgft1_hi xmm12
%define xgft2_lo xmm11
%define xgft2_hi xmm10
%define x0 xmm0
%define xtmpa xmm1
%define xtmph1 xmm2
%define xtmpl1 xmm3
%define xtmph2 xmm4
%define xtmpl2 xmm5
%define xd1 xmm6
%define xd2 xmm7
%define xtmpd1 xmm8
%define xtmpd2 xmm9
align 16
global gf_2vect_mad_avx:function
func(gf_2vect_mad_avx)
FUNC_SAVE
sub len, 16
jl .return_fail
xor pos, pos
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
sal vec_i, 5 ;Multiply by 32
sal vec, 5
lea tmp, [mul_array + vec_i]
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, Ax{02}, ...
vmovdqu xgft1_hi, [tmp+16] ; " Ax{00}, Ax{10}, Ax{20}, ... , Ax{f0}
vmovdqu xgft2_lo, [tmp+vec] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
vmovdqu xgft2_hi, [tmp+vec+16] ; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
mov dest2, [dest1+PS]
mov dest1, [dest1]
XLDR xtmpd1, [dest1+len] ;backup the last 16 bytes in dest
XLDR xtmpd2, [dest2+len] ;backup the last 16 bytes in dest
.loop16
XLDR xd1, [dest1+pos] ;Get next dest vector
XLDR xd2, [dest2+pos] ;Get next dest vector
.loop16_overlap:
XLDR x0, [src+pos] ;Get next source vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xtmph1, xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl1, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl1 ;GF add high and low partials
vpxor xd1, xd1, xtmph1 ;xd1 += partial
vpshufb xtmph2, xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl2, xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xtmpl2 ;GF add high and low partials
vpxor xd2, xd2, xtmph2 ;xd2 += partial
XSTR [dest1+pos], xd1
XSTR [dest2+pos], xd2
add pos, 16 ;Loop on 16 bytes at a time
cmp pos, len
jle .loop16
lea tmp, [len + 16]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
vmovdqa xd1, xtmpd1 ;Restore xd1
vmovdqa xd2, xtmpd2 ;Restore xd2
jmp .loop16_overlap ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 16
mask0f: ddq 0x0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f0f
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion gf_2vect_mad_avx, 02, 00, 0204
-273
View File
@@ -1,273 +0,0 @@
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 Intel Corporation All rights reserved.
;
; 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 Intel Corporation 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 THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "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 THE COPYRIGHT
; OWNER OR CONTRIBUTORS 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.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_2vect_mad_avx2(len, vec, vec_i, mul_array, src, dest);
;;;
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12
%define arg5 r15
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define return rax
%define return.w eax
%define stack_size 16*9 + 3*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
vmovdqa [rsp+16*0],xmm6
vmovdqa [rsp+16*1],xmm7
vmovdqa [rsp+16*2],xmm8
vmovdqa [rsp+16*3],xmm9
vmovdqa [rsp+16*4],xmm10
vmovdqa [rsp+16*5],xmm11
vmovdqa [rsp+16*6],xmm12
vmovdqa [rsp+16*7],xmm13
vmovdqa [rsp+16*8],xmm14
save_reg r12, 9*16 + 0*8
save_reg r15, 9*16 + 1*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp+16*0]
vmovdqa xmm7, [rsp+16*1]
vmovdqa xmm8, [rsp+16*2]
vmovdqa xmm9, [rsp+16*3]
vmovdqa xmm10, [rsp+16*4]
vmovdqa xmm11, [rsp+16*5]
vmovdqa xmm12, [rsp+16*6]
vmovdqa xmm13, [rsp+16*7]
vmovdqa xmm14, [rsp+16*8]
mov r12, [rsp + 9*16 + 0*8]
mov r15, [rsp + 9*16 + 1*8]
add rsp, stack_size
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
;;; gf_2vect_mad_avx2(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 tmp2
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f ymm14
%define xmask0fx xmm14
%define xgft1_lo ymm13
%define xgft1_hi ymm12
%define xgft2_lo ymm11
%define xgft2_hi ymm10
%define x0 ymm0
%define xtmpa ymm1
%define xtmph1 ymm2
%define xtmpl1 ymm3
%define xtmph2 ymm4
%define xtmpl2 ymm5
%define xd1 ymm6
%define xd2 ymm7
%define xtmpd1 ymm8
%define xtmpd2 ymm9
align 16
global gf_2vect_mad_avx2:function
func(gf_2vect_mad_avx2)
FUNC_SAVE
sub len, 32
jl .return_fail
xor pos, pos
mov tmp.b, 0x0f
vpinsrb xmask0fx, xmask0fx, tmp.w, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
sal vec_i, 5 ;Multiply by 32
sal vec, 5
lea tmp, [mul_array + vec_i]
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
; " Ax{00}, Ax{10}, ..., Ax{f0}
vmovdqu xgft2_lo, [tmp+vec] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
; " Bx{00}, Bx{10}, ..., Bx{f0}
vperm2i128 xgft1_hi, xgft1_lo, xgft1_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft1_lo, xgft1_lo, xgft1_lo, 0x00 ; swapped to lo | lo
vperm2i128 xgft2_hi, xgft2_lo, xgft2_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft2_lo, xgft2_lo, xgft2_lo, 0x00 ; swapped to lo | lo
mov dest2, [dest1+PS] ; reuse mul_array
mov dest1, [dest1]
XLDR xtmpd1, [dest1+len] ;backup the last 16 bytes in dest
XLDR xtmpd2, [dest2+len] ;backup the last 16 bytes in dest
.loop32
XLDR xd1, [dest1+pos] ;Get next dest vector
XLDR xd2, [dest2+pos] ;Get next dest vector
.loop32_overlap:
XLDR x0, [src+pos] ;Get next source vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xtmph1, xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl1, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl1 ;GF add high and low partials
vpxor xd1, xd1, xtmph1 ;xd1 += partial
vpshufb xtmph2, xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl2, xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xtmpl2 ;GF add high and low partials
vpxor xd2, xd2, xtmph2 ;xd2 += partial
XSTR [dest1+pos], xd1
XSTR [dest2+pos], xd2
add pos, 32 ;Loop on 32 bytes at a time
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-32
vmovdqa xd1, xtmpd1 ;Restore xd1
vmovdqa xd2, xtmpd2 ;Restore xd2
jmp .loop32_overlap ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion gf_2vect_mad_avx2, 04, 00, 0205

Some files were not shown because too many files have changed in this diff Show More