github.com/cuiweixie/go-ethereum@v1.8.2-0.20180303084001-66cd41af1e38/swarm/storage/types.go (about) 1 // Copyright 2016 The go-ethereum Authors 2 // This file is part of the go-ethereum library. 3 // 4 // The go-ethereum library is free software: you can redistribute it and/or modify 5 // it under the terms of the GNU Lesser General Public License as published by 6 // the Free Software Foundation, either version 3 of the License, or 7 // (at your option) any later version. 8 // 9 // The go-ethereum library is distributed in the hope that it will be useful, 10 // but WITHOUT ANY WARRANTY; without even the implied warranty of 11 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 12 // GNU Lesser General Public License for more details. 13 // 14 // You should have received a copy of the GNU Lesser General Public License 15 // along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>. 16 17 package storage 18 19 import ( 20 "bytes" 21 "crypto" 22 "fmt" 23 "hash" 24 "io" 25 "sync" 26 27 "github.com/ethereum/go-ethereum/bmt" 28 "github.com/ethereum/go-ethereum/common" 29 "github.com/ethereum/go-ethereum/crypto/sha3" 30 ) 31 32 type Hasher func() hash.Hash 33 type SwarmHasher func() SwarmHash 34 35 // Peer is the recorded as Source on the chunk 36 // should probably not be here? but network should wrap chunk object 37 type Peer interface{} 38 39 type Key []byte 40 41 func (x Key) Size() uint { 42 return uint(len(x)) 43 } 44 45 func (x Key) isEqual(y Key) bool { 46 return bytes.Equal(x, y) 47 } 48 49 func (h Key) bits(i, j uint) uint { 50 ii := i >> 3 51 jj := i & 7 52 if ii >= h.Size() { 53 return 0 54 } 55 56 if jj+j <= 8 { 57 return uint((h[ii] >> jj) & ((1 << j) - 1)) 58 } 59 60 res := uint(h[ii] >> jj) 61 jj = 8 - jj 62 j -= jj 63 for j != 0 { 64 ii++ 65 if j < 8 { 66 res += uint(h[ii]&((1<<j)-1)) << jj 67 return res 68 } 69 res += uint(h[ii]) << jj 70 jj += 8 71 j -= 8 72 } 73 return res 74 } 75 76 func IsZeroKey(key Key) bool { 77 return len(key) == 0 || bytes.Equal(key, ZeroKey) 78 } 79 80 var ZeroKey = Key(common.Hash{}.Bytes()) 81 82 func MakeHashFunc(hash string) SwarmHasher { 83 switch hash { 84 case "SHA256": 85 return func() SwarmHash { return &HashWithLength{crypto.SHA256.New()} } 86 case "SHA3": 87 return func() SwarmHash { return &HashWithLength{sha3.NewKeccak256()} } 88 case "BMT": 89 return func() SwarmHash { 90 hasher := sha3.NewKeccak256 91 pool := bmt.NewTreePool(hasher, bmt.DefaultSegmentCount, bmt.DefaultPoolSize) 92 return bmt.New(pool) 93 } 94 } 95 return nil 96 } 97 98 func (key Key) Hex() string { 99 return fmt.Sprintf("%064x", []byte(key[:])) 100 } 101 102 func (key Key) Log() string { 103 if len(key[:]) < 4 { 104 return fmt.Sprintf("%x", []byte(key[:])) 105 } 106 return fmt.Sprintf("%08x", []byte(key[:4])) 107 } 108 109 func (key Key) String() string { 110 return fmt.Sprintf("%064x", []byte(key)[:]) 111 } 112 113 func (key Key) MarshalJSON() (out []byte, err error) { 114 return []byte(`"` + key.String() + `"`), nil 115 } 116 117 func (key *Key) UnmarshalJSON(value []byte) error { 118 s := string(value) 119 *key = make([]byte, 32) 120 h := common.Hex2Bytes(s[1 : len(s)-1]) 121 copy(*key, h) 122 return nil 123 } 124 125 // each chunk when first requested opens a record associated with the request 126 // next time a request for the same chunk arrives, this record is updated 127 // this request status keeps track of the request ID-s as well as the requesting 128 // peers and has a channel that is closed when the chunk is retrieved. Multiple 129 // local callers can wait on this channel (or combined with a timeout, block with a 130 // select). 131 type RequestStatus struct { 132 Key Key 133 Source Peer 134 C chan bool 135 Requesters map[uint64][]interface{} 136 } 137 138 func newRequestStatus(key Key) *RequestStatus { 139 return &RequestStatus{ 140 Key: key, 141 Requesters: make(map[uint64][]interface{}), 142 C: make(chan bool), 143 } 144 } 145 146 // Chunk also serves as a request object passed to ChunkStores 147 // in case it is a retrieval request, Data is nil and Size is 0 148 // Note that Size is not the size of the data chunk, which is Data.Size() 149 // but the size of the subtree encoded in the chunk 150 // 0 if request, to be supplied by the dpa 151 type Chunk struct { 152 Key Key // always 153 SData []byte // nil if request, to be supplied by dpa 154 Size int64 // size of the data covered by the subtree encoded in this chunk 155 Source Peer // peer 156 C chan bool // to signal data delivery by the dpa 157 Req *RequestStatus // request Status needed by netStore 158 wg *sync.WaitGroup // wg to synchronize 159 dbStored chan bool // never remove a chunk from memStore before it is written to dbStore 160 } 161 162 func NewChunk(key Key, rs *RequestStatus) *Chunk { 163 return &Chunk{Key: key, Req: rs} 164 } 165 166 /* 167 The ChunkStore interface is implemented by : 168 169 - MemStore: a memory cache 170 - DbStore: local disk/db store 171 - LocalStore: a combination (sequence of) memStore and dbStore 172 - NetStore: cloud storage abstraction layer 173 - DPA: local requests for swarm storage and retrieval 174 */ 175 type ChunkStore interface { 176 Put(*Chunk) // effectively there is no error even if there is an error 177 Get(Key) (*Chunk, error) 178 Close() 179 } 180 181 /* 182 Chunker is the interface to a component that is responsible for disassembling and assembling larger data and indended to be the dependency of a DPA storage system with fixed maximum chunksize. 183 184 It relies on the underlying chunking model. 185 186 When calling Split, the caller provides a channel (chan *Chunk) on which it receives chunks to store. The DPA delegates to storage layers (implementing ChunkStore interface). 187 188 Split returns an error channel, which the caller can monitor. 189 After getting notified that all the data has been split (the error channel is closed), the caller can safely read or save the root key. Optionally it times out if not all chunks get stored or not the entire stream of data has been processed. By inspecting the errc channel the caller can check if any explicit errors (typically IO read/write failures) occurred during splitting. 190 191 When calling Join with a root key, the caller gets returned a seekable lazy reader. The caller again provides a channel on which the caller receives placeholder chunks with missing data. The DPA is supposed to forward this to the chunk stores and notify the chunker if the data has been delivered (i.e. retrieved from memory cache, disk-persisted db or cloud based swarm delivery). As the seekable reader is used, the chunker then puts these together the relevant parts on demand. 192 */ 193 type Splitter interface { 194 /* 195 When splitting, data is given as a SectionReader, and the key is a hashSize long byte slice (Key), the root hash of the entire content will fill this once processing finishes. 196 New chunks to store are coming to caller via the chunk storage channel, which the caller provides. 197 wg is a Waitgroup (can be nil) that can be used to block until the local storage finishes 198 The caller gets returned an error channel, if an error is encountered during splitting, it is fed to errC error channel. 199 A closed error signals process completion at which point the key can be considered final if there were no errors. 200 */ 201 Split(io.Reader, int64, chan *Chunk, *sync.WaitGroup, *sync.WaitGroup) (Key, error) 202 203 /* This is the first step in making files mutable (not chunks).. 204 Append allows adding more data chunks to the end of the already existsing file. 205 The key for the root chunk is supplied to load the respective tree. 206 Rest of the parameters behave like Split. 207 */ 208 Append(Key, io.Reader, chan *Chunk, *sync.WaitGroup, *sync.WaitGroup) (Key, error) 209 } 210 211 type Joiner interface { 212 /* 213 Join reconstructs original content based on a root key. 214 When joining, the caller gets returned a Lazy SectionReader, which is 215 seekable and implements on-demand fetching of chunks as and where it is read. 216 New chunks to retrieve are coming to caller via the Chunk channel, which the caller provides. 217 If an error is encountered during joining, it appears as a reader error. 218 The SectionReader. 219 As a result, partial reads from a document are possible even if other parts 220 are corrupt or lost. 221 The chunks are not meant to be validated by the chunker when joining. This 222 is because it is left to the DPA to decide which sources are trusted. 223 */ 224 Join(key Key, chunkC chan *Chunk) LazySectionReader 225 } 226 227 type Chunker interface { 228 Joiner 229 Splitter 230 // returns the key length 231 // KeySize() int64 232 } 233 234 // Size, Seek, Read, ReadAt 235 type LazySectionReader interface { 236 Size(chan bool) (int64, error) 237 io.Seeker 238 io.Reader 239 io.ReaderAt 240 } 241 242 type LazyTestSectionReader struct { 243 *io.SectionReader 244 } 245 246 func (self *LazyTestSectionReader) Size(chan bool) (int64, error) { 247 return self.SectionReader.Size(), nil 248 }