khenaidoo | 59ce9dd | 2019-11-11 13:05:32 -0500 | [diff] [blame] | 1 | // Copyright 2015 The etcd Authors |
| 2 | // |
| 3 | // Licensed under the Apache License, Version 2.0 (the "License"); |
| 4 | // you may not use this file except in compliance with the License. |
| 5 | // You may obtain a copy of the License at |
| 6 | // |
| 7 | // http://www.apache.org/licenses/LICENSE-2.0 |
| 8 | // |
| 9 | // Unless required by applicable law or agreed to in writing, software |
| 10 | // distributed under the License is distributed on an "AS IS" BASIS, |
| 11 | // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 12 | // See the License for the specific language governing permissions and |
| 13 | // limitations under the License. |
| 14 | |
| 15 | package wal |
| 16 | |
| 17 | import ( |
| 18 | "bufio" |
| 19 | "encoding/binary" |
| 20 | "hash" |
| 21 | "io" |
| 22 | "sync" |
| 23 | |
| 24 | "go.etcd.io/etcd/pkg/crc" |
| 25 | "go.etcd.io/etcd/pkg/pbutil" |
| 26 | "go.etcd.io/etcd/raft/raftpb" |
| 27 | "go.etcd.io/etcd/wal/walpb" |
| 28 | ) |
| 29 | |
| 30 | const minSectorSize = 512 |
| 31 | |
| 32 | // frameSizeBytes is frame size in bytes, including record size and padding size. |
| 33 | const frameSizeBytes = 8 |
| 34 | |
| 35 | type decoder struct { |
| 36 | mu sync.Mutex |
| 37 | brs []*bufio.Reader |
| 38 | |
| 39 | // lastValidOff file offset following the last valid decoded record |
| 40 | lastValidOff int64 |
| 41 | crc hash.Hash32 |
| 42 | } |
| 43 | |
| 44 | func newDecoder(r ...io.Reader) *decoder { |
| 45 | readers := make([]*bufio.Reader, len(r)) |
| 46 | for i := range r { |
| 47 | readers[i] = bufio.NewReader(r[i]) |
| 48 | } |
| 49 | return &decoder{ |
| 50 | brs: readers, |
| 51 | crc: crc.New(0, crcTable), |
| 52 | } |
| 53 | } |
| 54 | |
| 55 | func (d *decoder) decode(rec *walpb.Record) error { |
| 56 | rec.Reset() |
| 57 | d.mu.Lock() |
| 58 | defer d.mu.Unlock() |
| 59 | return d.decodeRecord(rec) |
| 60 | } |
| 61 | |
| 62 | func (d *decoder) decodeRecord(rec *walpb.Record) error { |
| 63 | if len(d.brs) == 0 { |
| 64 | return io.EOF |
| 65 | } |
| 66 | |
| 67 | l, err := readInt64(d.brs[0]) |
| 68 | if err == io.EOF || (err == nil && l == 0) { |
| 69 | // hit end of file or preallocated space |
| 70 | d.brs = d.brs[1:] |
| 71 | if len(d.brs) == 0 { |
| 72 | return io.EOF |
| 73 | } |
| 74 | d.lastValidOff = 0 |
| 75 | return d.decodeRecord(rec) |
| 76 | } |
| 77 | if err != nil { |
| 78 | return err |
| 79 | } |
| 80 | |
| 81 | recBytes, padBytes := decodeFrameSize(l) |
| 82 | |
| 83 | data := make([]byte, recBytes+padBytes) |
| 84 | if _, err = io.ReadFull(d.brs[0], data); err != nil { |
| 85 | // ReadFull returns io.EOF only if no bytes were read |
| 86 | // the decoder should treat this as an ErrUnexpectedEOF instead. |
| 87 | if err == io.EOF { |
| 88 | err = io.ErrUnexpectedEOF |
| 89 | } |
| 90 | return err |
| 91 | } |
| 92 | if err := rec.Unmarshal(data[:recBytes]); err != nil { |
| 93 | if d.isTornEntry(data) { |
| 94 | return io.ErrUnexpectedEOF |
| 95 | } |
| 96 | return err |
| 97 | } |
| 98 | |
| 99 | // skip crc checking if the record type is crcType |
| 100 | if rec.Type != crcType { |
| 101 | d.crc.Write(rec.Data) |
| 102 | if err := rec.Validate(d.crc.Sum32()); err != nil { |
| 103 | if d.isTornEntry(data) { |
| 104 | return io.ErrUnexpectedEOF |
| 105 | } |
| 106 | return err |
| 107 | } |
| 108 | } |
| 109 | // record decoded as valid; point last valid offset to end of record |
| 110 | d.lastValidOff += frameSizeBytes + recBytes + padBytes |
| 111 | return nil |
| 112 | } |
| 113 | |
| 114 | func decodeFrameSize(lenField int64) (recBytes int64, padBytes int64) { |
| 115 | // the record size is stored in the lower 56 bits of the 64-bit length |
| 116 | recBytes = int64(uint64(lenField) & ^(uint64(0xff) << 56)) |
| 117 | // non-zero padding is indicated by set MSb / a negative length |
| 118 | if lenField < 0 { |
| 119 | // padding is stored in lower 3 bits of length MSB |
| 120 | padBytes = int64((uint64(lenField) >> 56) & 0x7) |
| 121 | } |
| 122 | return recBytes, padBytes |
| 123 | } |
| 124 | |
| 125 | // isTornEntry determines whether the last entry of the WAL was partially written |
| 126 | // and corrupted because of a torn write. |
| 127 | func (d *decoder) isTornEntry(data []byte) bool { |
| 128 | if len(d.brs) != 1 { |
| 129 | return false |
| 130 | } |
| 131 | |
| 132 | fileOff := d.lastValidOff + frameSizeBytes |
| 133 | curOff := 0 |
| 134 | chunks := [][]byte{} |
| 135 | // split data on sector boundaries |
| 136 | for curOff < len(data) { |
| 137 | chunkLen := int(minSectorSize - (fileOff % minSectorSize)) |
| 138 | if chunkLen > len(data)-curOff { |
| 139 | chunkLen = len(data) - curOff |
| 140 | } |
| 141 | chunks = append(chunks, data[curOff:curOff+chunkLen]) |
| 142 | fileOff += int64(chunkLen) |
| 143 | curOff += chunkLen |
| 144 | } |
| 145 | |
| 146 | // if any data for a sector chunk is all 0, it's a torn write |
| 147 | for _, sect := range chunks { |
| 148 | isZero := true |
| 149 | for _, v := range sect { |
| 150 | if v != 0 { |
| 151 | isZero = false |
| 152 | break |
| 153 | } |
| 154 | } |
| 155 | if isZero { |
| 156 | return true |
| 157 | } |
| 158 | } |
| 159 | return false |
| 160 | } |
| 161 | |
| 162 | func (d *decoder) updateCRC(prevCrc uint32) { |
| 163 | d.crc = crc.New(prevCrc, crcTable) |
| 164 | } |
| 165 | |
| 166 | func (d *decoder) lastCRC() uint32 { |
| 167 | return d.crc.Sum32() |
| 168 | } |
| 169 | |
| 170 | func (d *decoder) lastOffset() int64 { return d.lastValidOff } |
| 171 | |
| 172 | func mustUnmarshalEntry(d []byte) raftpb.Entry { |
| 173 | var e raftpb.Entry |
| 174 | pbutil.MustUnmarshal(&e, d) |
| 175 | return e |
| 176 | } |
| 177 | |
| 178 | func mustUnmarshalState(d []byte) raftpb.HardState { |
| 179 | var s raftpb.HardState |
| 180 | pbutil.MustUnmarshal(&s, d) |
| 181 | return s |
| 182 | } |
| 183 | |
| 184 | func readInt64(r io.Reader) (int64, error) { |
| 185 | var n int64 |
| 186 | err := binary.Read(r, binary.LittleEndian, &n) |
| 187 | return n, err |
| 188 | } |