mirror of https://github.com/minio/minio.git
Merge pull request #355 from abperiasamy/janitor_erasure
This commit is contained in:
commit
7d25490a2e
|
@ -24,7 +24,7 @@ import (
|
|||
)
|
||||
|
||||
// Integer to Int conversion
|
||||
func int2cInt(src_err_list []int) *C.int32_t {
|
||||
func int2CInt(src_err_list []int) *C.int32_t {
|
||||
var sizeErrInt = int(unsafe.Sizeof(src_err_list[0]))
|
||||
switch sizeInt {
|
||||
case sizeErrInt:
|
||||
|
|
|
@ -32,47 +32,47 @@ import (
|
|||
//
|
||||
// Decoded data is exactly similar in length and content as the original data.
|
||||
func (e *Encoder) Decode(chunks [][]byte, length int) ([]byte, error) {
|
||||
var decode_matrix *C.uint8_t
|
||||
var decode_tbls *C.uint8_t
|
||||
var decode_index *C.uint32_t
|
||||
var decodeMatrix *C.uint8_t
|
||||
var decodeTbls *C.uint8_t
|
||||
var decodeIndex *C.uint32_t
|
||||
var source, target **C.uint8_t
|
||||
|
||||
k := e.params.K
|
||||
m := e.params.M
|
||||
k := int(e.params.K)
|
||||
m := int(e.params.M)
|
||||
n := k + m
|
||||
if len(chunks) != int(n) {
|
||||
if len(chunks) != n {
|
||||
return nil, errors.New(fmt.Sprintf("chunks length must be %d", n))
|
||||
}
|
||||
chunk_size := GetEncodedChunkLen(length, uint8(k))
|
||||
chunkLen := GetEncodedBlockLen(length, uint8(k))
|
||||
|
||||
error_index := make([]int, n+1)
|
||||
var err_count int = 0
|
||||
errorIndex := make([]int, n+1)
|
||||
var errCount int = 0
|
||||
|
||||
for i := range chunks {
|
||||
// Check of chunks are really null
|
||||
if chunks[i] == nil || len(chunks[i]) == 0 {
|
||||
error_index[err_count] = i
|
||||
err_count++
|
||||
errorIndex[errCount] = i
|
||||
errCount++
|
||||
}
|
||||
}
|
||||
error_index[err_count] = -1
|
||||
err_count++
|
||||
errorIndex[errCount] = -1
|
||||
errCount++
|
||||
|
||||
// Too many missing chunks, cannot be more than parity `m`
|
||||
if err_count-1 > int(n-k) {
|
||||
if errCount-1 > int(n-k) {
|
||||
return nil, errors.New("too many erasures requested, can't decode")
|
||||
}
|
||||
|
||||
error_index_ptr := int2cInt(error_index[:err_count])
|
||||
errorIndex_ptr := int2CInt(errorIndex[:errCount])
|
||||
|
||||
for i := range chunks {
|
||||
if chunks[i] == nil || len(chunks[i]) == 0 {
|
||||
chunks[i] = make([]byte, chunk_size)
|
||||
chunks[i] = make([]byte, chunkLen)
|
||||
}
|
||||
}
|
||||
|
||||
C.minio_init_decoder(error_index_ptr, C.int(k), C.int(n), C.int(err_count-1),
|
||||
e.encode_matrix, &decode_matrix, &decode_tbls, &decode_index)
|
||||
C.minio_init_decoder(errorIndex_ptr, C.int(k), C.int(n), C.int(errCount-1),
|
||||
e.encodeMatrix, &decodeMatrix, &decodeTbls, &decodeIndex)
|
||||
|
||||
pointers := make([]*byte, n)
|
||||
for i := range chunks {
|
||||
|
@ -81,24 +81,24 @@ func (e *Encoder) Decode(chunks [][]byte, length int) ([]byte, error) {
|
|||
|
||||
data := (**C.uint8_t)(unsafe.Pointer(&pointers[0]))
|
||||
|
||||
ret := C.minio_get_source_target(C.int(err_count-1), C.int(k), C.int(m), error_index_ptr,
|
||||
decode_index, data, &source, &target)
|
||||
ret := C.minio_get_source_target(C.int(errCount-1), C.int(k), C.int(m), errorIndex_ptr,
|
||||
decodeIndex, data, &source, &target)
|
||||
|
||||
if int(ret) == -1 {
|
||||
return nil, errors.New("Decoding source target failed")
|
||||
}
|
||||
|
||||
C.ec_encode_data(C.int(chunk_size), C.int(k), C.int(err_count-1), decode_tbls,
|
||||
C.ec_encode_data(C.int(chunkLen), C.int(k), C.int(errCount-1), decodeTbls,
|
||||
source, target)
|
||||
|
||||
recovered_output := make([]byte, 0, chunk_size*int(k))
|
||||
recoveredOutput := make([]byte, 0, chunkLen*int(k))
|
||||
for i := 0; i < int(k); i++ {
|
||||
recovered_output = append(recovered_output, chunks[i]...)
|
||||
recoveredOutput = append(recoveredOutput, chunks[i]...)
|
||||
}
|
||||
|
||||
// TODO cache this if necessary
|
||||
e.decode_matrix = decode_matrix
|
||||
e.decode_tbls = decode_tbls
|
||||
e.decodeMatrix = decodeMatrix
|
||||
e.decodeTbls = decodeTbls
|
||||
|
||||
return recovered_output[:length], nil
|
||||
return recoveredOutput[:length], nil
|
||||
}
|
||||
|
|
|
@ -26,7 +26,7 @@ import (
|
|||
"unsafe"
|
||||
)
|
||||
|
||||
type Technique int
|
||||
type Technique uint8
|
||||
|
||||
const (
|
||||
Vandermonde Technique = iota
|
||||
|
@ -52,10 +52,10 @@ type EncoderParams struct {
|
|||
// Encoder is an object used to encode and decode data.
|
||||
type Encoder struct {
|
||||
params *EncoderParams
|
||||
encode_matrix,
|
||||
encode_tbls,
|
||||
decode_matrix,
|
||||
decode_tbls *C.uint8_t
|
||||
encodeMatrix,
|
||||
encodeTbls,
|
||||
decodeMatrix,
|
||||
decodeTbls *C.uint8_t
|
||||
}
|
||||
|
||||
// ParseEncoderParams creates an EncoderParams object.
|
||||
|
@ -94,30 +94,30 @@ func ParseEncoderParams(k, m uint8, technique Technique) (*EncoderParams, error)
|
|||
|
||||
// NewEncoder creates an encoder object with a given set of parameters.
|
||||
func NewEncoder(ep *EncoderParams) *Encoder {
|
||||
var encode_matrix *C.uint8_t
|
||||
var encode_tbls *C.uint8_t
|
||||
var k = C.int(ep.K)
|
||||
var m = C.int(ep.M)
|
||||
|
||||
k := C.int(ep.K)
|
||||
m := C.int(ep.M)
|
||||
var encodeMatrix *C.uint8_t
|
||||
var encodeTbls *C.uint8_t
|
||||
|
||||
C.minio_init_encoder(C.int(ep.Technique), k, m, &encode_matrix,
|
||||
&encode_tbls)
|
||||
C.minio_init_encoder(C.int(ep.Technique), k, m, &encodeMatrix,
|
||||
&encodeTbls)
|
||||
|
||||
return &Encoder{
|
||||
params: ep,
|
||||
encode_matrix: encode_matrix,
|
||||
encode_tbls: encode_tbls,
|
||||
decode_matrix: nil,
|
||||
decode_tbls: nil,
|
||||
params: ep,
|
||||
encodeMatrix: encodeMatrix,
|
||||
encodeTbls: encodeTbls,
|
||||
decodeMatrix: nil,
|
||||
decodeTbls: nil,
|
||||
}
|
||||
}
|
||||
|
||||
func GetEncodedLen(inputLen int, k, m uint8) (outputLen int) {
|
||||
outputLen = GetEncodedChunkLen(inputLen, k) * int(k+m)
|
||||
func GetEncodedBlocksLen(inputLen int, k, m uint8) (outputLen int) {
|
||||
outputLen = GetEncodedBlockLen(inputLen, k) * int(k+m)
|
||||
return outputLen
|
||||
}
|
||||
|
||||
func GetEncodedChunkLen(inputLen int, k uint8) (outputChunkLen int) {
|
||||
func GetEncodedBlockLen(inputLen int, k uint8) (encodedOutputLen int) {
|
||||
alignment := int(k) * SIMDAlign
|
||||
remainder := inputLen % alignment
|
||||
|
||||
|
@ -125,55 +125,66 @@ func GetEncodedChunkLen(inputLen int, k uint8) (outputChunkLen int) {
|
|||
if remainder != 0 {
|
||||
paddedInputLen = inputLen + (alignment - remainder)
|
||||
}
|
||||
outputChunkLen = paddedInputLen / int(k)
|
||||
return outputChunkLen
|
||||
encodedOutputLen = paddedInputLen / int(k)
|
||||
return encodedOutputLen
|
||||
}
|
||||
|
||||
// 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.
|
||||
func (e *Encoder) Encode(input []byte) ([][]byte, error) {
|
||||
inputLen := len(input)
|
||||
k := C.int(e.params.K)
|
||||
m := C.int(e.params.M)
|
||||
n := k + m
|
||||
// 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 *Encoder) Encode(inputData []byte) (encodedBlocks [][]byte, err error) {
|
||||
k := int(e.params.K) // "k" data blocks
|
||||
m := int(e.params.M) // "m" parity blocks
|
||||
n := k + m // "n" total encoded blocks
|
||||
|
||||
chunkLen := GetEncodedChunkLen(inputLen, e.params.K)
|
||||
encodedDataLen := chunkLen * int(k)
|
||||
paddedDataLen := int(encodedDataLen) - inputLen
|
||||
// Length of a single encoded chunk.
|
||||
// Total number of encoded chunks = "k" data + "m" parity blocks
|
||||
encodedBlockLen := GetEncodedBlockLen(len(inputData), uint8(k))
|
||||
|
||||
if paddedDataLen > 0 {
|
||||
s := make([]byte, paddedDataLen)
|
||||
// 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
|
||||
input = append(input, s...)
|
||||
inputData = append(inputData, padding...)
|
||||
}
|
||||
|
||||
encodedParityLen := chunkLen * int(e.params.M)
|
||||
c := make([]byte, encodedParityLen)
|
||||
input = append(input, c...)
|
||||
|
||||
// encodedOutLen := encodedDataLen + encodedParityLen
|
||||
|
||||
// Allocate chunks
|
||||
chunks := make([][]byte, k+m)
|
||||
pointers := make([]*byte, k+m)
|
||||
|
||||
var i int
|
||||
// Add data blocks to chunks
|
||||
for i = 0; i < int(k); i++ {
|
||||
chunks[i] = input[i*chunkLen : (i+1)*chunkLen]
|
||||
pointers[i] = &chunks[i][0]
|
||||
// Extend inputData buffer to accommodate coded parity blocks
|
||||
if true { // create a temporary scope to trigger garbage collect
|
||||
encodedParityBlocksLen := encodedBlockLen * m
|
||||
parityBlocks := make([]byte, encodedParityBlocksLen)
|
||||
inputData = append(inputData, parityBlocks...)
|
||||
}
|
||||
|
||||
for i = int(k); i < int(n); i++ {
|
||||
chunks[i] = make([]byte, chunkLen)
|
||||
pointers[i] = &chunks[i][0]
|
||||
// Allocate memory to the "encoded blocks" return buffer
|
||||
encodedBlocks = make([][]byte, n) // Return buffer
|
||||
|
||||
// Nessary 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]
|
||||
}
|
||||
|
||||
data := (**C.uint8_t)(unsafe.Pointer(&pointers[:k][0]))
|
||||
coding := (**C.uint8_t)(unsafe.Pointer(&pointers[k:][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]
|
||||
}
|
||||
|
||||
C.ec_encode_data(C.int(chunkLen), k, m, e.encode_tbls, data,
|
||||
coding)
|
||||
return chunks, nil
|
||||
// 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.uint8_t)(unsafe.Pointer(&pointersToEncodedBlock[:k][0])), // Pointers to data blocks
|
||||
(**C.uint8_t)(unsafe.Pointer(&pointersToEncodedBlock[k:][0]))) // Pointers to parity blocks
|
||||
|
||||
return encodedBlocks, nil
|
||||
}
|
||||
|
|
|
@ -75,7 +75,9 @@ func erasureReader(readers []io.ReadCloser, donutMetadata map[string]string, wri
|
|||
if blockSize < totalLeft {
|
||||
curBlockSize = blockSize
|
||||
}
|
||||
curChunkSize := erasure.GetEncodedChunkLen(curBlockSize, uint8(k))
|
||||
|
||||
curChunkSize := erasure.GetEncodedBlockLen(curBlockSize, uint8(k))
|
||||
|
||||
encodedBytes := make([][]byte, 16)
|
||||
for i, reader := range readers {
|
||||
var bytesBuffer bytes.Buffer
|
||||
|
|
Loading…
Reference in New Issue