go-ethereum/swarm/storage/mru/resource.go

1067 lines
37 KiB
Go

// Copyright 2018 The go-ethereum Authors
// This file is part of the go-ethereum library.
//
// The go-ethereum library is free software: you can redistribute it and/or modify
// it under the terms of the GNU Lesser General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// The go-ethereum library 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 Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public License
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
package mru
import (
"bytes"
"context"
"encoding/binary"
"errors"
"fmt"
"math/big"
"path/filepath"
"sync"
"time"
"golang.org/x/net/idna"
"github.com/ethereum/go-ethereum/common"
"github.com/ethereum/go-ethereum/contracts/ens"
"github.com/ethereum/go-ethereum/core/types"
"github.com/ethereum/go-ethereum/crypto"
"github.com/ethereum/go-ethereum/swarm/log"
"github.com/ethereum/go-ethereum/swarm/multihash"
"github.com/ethereum/go-ethereum/swarm/storage"
)
const (
signatureLength = 65
metadataChunkOffsetSize = 18
DbDirName = "resource"
chunkSize = 4096 // temporary until we implement FileStore in the resourcehandler
defaultStoreTimeout = 4000 * time.Millisecond
hasherCount = 8
resourceHash = storage.SHA3Hash
defaultRetrieveTimeout = 100 * time.Millisecond
)
type blockEstimator struct {
Start time.Time
Average time.Duration
}
// TODO: Average must be adjusted when blockchain connection is present and synced
func NewBlockEstimator() *blockEstimator {
sampleDate, _ := time.Parse(time.RFC3339, "2018-05-04T20:35:22Z") // from etherscan.io
sampleBlock := int64(3169691) // from etherscan.io
ropstenStart, _ := time.Parse(time.RFC3339, "2016-11-20T11:48:50Z") // from etherscan.io
ns := sampleDate.Sub(ropstenStart).Nanoseconds()
period := int(ns / sampleBlock)
parsestring := fmt.Sprintf("%dns", int(float64(period)*1.0005)) // increase the blockcount a little, so we don't overshoot the read block height; if we do, we will never find the updates when getting synced data
periodNs, _ := time.ParseDuration(parsestring)
return &blockEstimator{
Start: ropstenStart,
Average: periodNs,
}
}
func (b *blockEstimator) HeaderByNumber(context.Context, string, *big.Int) (*types.Header, error) {
return &types.Header{
Number: big.NewInt(time.Since(b.Start).Nanoseconds() / b.Average.Nanoseconds()),
}, nil
}
type Error struct {
code int
err string
}
func (e *Error) Error() string {
return e.err
}
func (e *Error) Code() int {
return e.code
}
func NewError(code int, s string) error {
if code < 0 || code >= ErrCnt {
panic("no such error code!")
}
r := &Error{
err: s,
}
switch code {
case ErrNotFound, ErrIO, ErrUnauthorized, ErrInvalidValue, ErrDataOverflow, ErrNothingToReturn, ErrInvalidSignature, ErrNotSynced, ErrPeriodDepth, ErrCorruptData:
r.code = code
}
return r
}
type Signature [signatureLength]byte
type LookupParams struct {
Limit bool
Max uint32
}
// Encapsulates an specific resource update. When synced it contains the most recent
// version of the resource update data.
type resource struct {
*bytes.Reader
Multihash bool
name string
nameHash common.Hash
startBlock uint64
lastPeriod uint32
lastKey storage.Address
frequency uint64
version uint32
data []byte
updated time.Time
}
// TODO Expire content after a defined period (to force resync)
func (r *resource) isSynced() bool {
return !r.updated.IsZero()
}
func (r *resource) NameHash() common.Hash {
return r.nameHash
}
func (r *resource) Size(chan bool) (int64, error) {
if !r.isSynced() {
return 0, NewError(ErrNotSynced, "Not synced")
}
return int64(len(r.data)), nil
}
func (r *resource) Name() string {
return r.name
}
func (r *resource) UnmarshalBinary(data []byte) error {
r.startBlock = binary.LittleEndian.Uint64(data[:8])
r.frequency = binary.LittleEndian.Uint64(data[8:16])
r.name = string(data[16:])
return nil
}
func (r *resource) MarshalBinary() ([]byte, error) {
b := make([]byte, 16+len(r.name))
binary.LittleEndian.PutUint64(b, r.startBlock)
binary.LittleEndian.PutUint64(b[8:], r.frequency)
copy(b[16:], []byte(r.name))
return b, nil
}
type headerGetter interface {
HeaderByNumber(context.Context, string, *big.Int) (*types.Header, error)
}
type ownerValidator interface {
ValidateOwner(name string, address common.Address) (bool, error)
}
// Mutable resource is an entity which allows updates to a resource
// without resorting to ENS on each update.
// The update scheme is built on swarm chunks with chunk keys following
// a predictable, versionable pattern.
//
// Updates are defined to be periodic in nature, where periods are
// expressed in terms of number of blocks.
//
// The root entry of a mutable resource is tied to a unique identifier,
// typically - but not necessarily - an ens name. The identifier must be
// an valid IDNA string. It also contains the block number
// when the resource update was first registered, and
// the block frequency with which the resource will be updated, both of
// which are stored as little-endian uint64 values in the database (for a
// total of 16 bytes). It also contains the unique identifier.
// It is stored in a separate content-addressed chunk (call it the metadata chunk),
// with the following layout:
//
// (0x0000|startblock|frequency|identifier)
//
// (The two first zero-value bytes are used for disambiguation by the chunk validator,
// and update chunk will always have a value > 0 there.)
//
// The root entry tells the requester from when the mutable resource was
// first added (block number) and in which block number to look for the
// actual updates. Thus, a resource update for identifier "føø.bar"
// starting at block 4200 with frequency 42 will have updates on block 4242,
// 4284, 4326 and so on.
//
// Actual data updates are also made in the form of swarm chunks. The keys
// of the updates are the hash of a concatenation of properties as follows:
//
// sha256(period|version|namehash)
//
// The period is (currentblock - startblock) / frequency
//
// Using our previous example, this means that a period 3 will have 4326 as
// the block number.
//
// If more than one update is made to the same block number, incremental
// version numbers are used successively.
//
// A lookup agent need only know the identifier name in order to get the versions
//
// the resourcedata is:
// headerlength|period|version|identifier|data
//
// if a validator is active, the chunk data is:
// resourcedata|sign(resourcedata)
// otherwise, the chunk data is the same as the resourcedata
//
// headerlength is a 16 bit value containing the byte length of period|version|name
//
// TODO: Include modtime in chunk data + signature
type Handler struct {
chunkStore *storage.NetStore
HashSize int
signer Signer
headerGetter headerGetter
ownerValidator ownerValidator
resources map[string]*resource
hashPool sync.Pool
resourceLock sync.RWMutex
storeTimeout time.Duration
queryMaxPeriods *LookupParams
}
type HandlerParams struct {
QueryMaxPeriods *LookupParams
Signer Signer
HeaderGetter headerGetter
OwnerValidator ownerValidator
}
// Create or open resource update chunk store
func NewHandler(params *HandlerParams) (*Handler, error) {
if params.QueryMaxPeriods == nil {
params.QueryMaxPeriods = &LookupParams{
Limit: false,
}
}
rh := &Handler{
headerGetter: params.HeaderGetter,
ownerValidator: params.OwnerValidator,
resources: make(map[string]*resource),
storeTimeout: defaultStoreTimeout,
signer: params.Signer,
hashPool: sync.Pool{
New: func() interface{} {
return storage.MakeHashFunc(resourceHash)()
},
},
queryMaxPeriods: params.QueryMaxPeriods,
}
for i := 0; i < hasherCount; i++ {
hashfunc := storage.MakeHashFunc(resourceHash)()
if rh.HashSize == 0 {
rh.HashSize = hashfunc.Size()
}
rh.hashPool.Put(hashfunc)
}
return rh, nil
}
// SetStore sets the store backend for resource updates
func (h *Handler) SetStore(store *storage.NetStore) {
h.chunkStore = store
}
// Validate is a chunk validation method (matches ChunkValidatorFunc signature)
//
// If resource update, owner is checked against ENS record of resource name inferred from chunk data
// If parsed signature is nil, validates automatically
// If not resource update, it validates are root chunk if length is metadataChunkOffsetSize and first two bytes are 0
func (h *Handler) Validate(addr storage.Address, data []byte) bool {
signature, period, version, name, parseddata, _, err := h.parseUpdate(data)
if err != nil {
log.Warn(err.Error())
if len(data) > metadataChunkOffsetSize { // identifier comes after this byte range, and must be at least one byte
if bytes.Equal(data[:2], []byte{0, 0}) {
return true
}
}
log.Error("Invalid resource chunk")
return false
} else if signature == nil {
return bytes.Equal(h.resourceHash(period, version, ens.EnsNode(name)), addr)
}
digest := h.keyDataHash(addr, parseddata)
addrSig, err := getAddressFromDataSig(digest, *signature)
if err != nil {
log.Error("Invalid signature on resource chunk")
return false
}
ok, _ := h.checkAccess(name, addrSig)
return ok
}
// If no ens client is supplied, resource updates are not validated
func (h *Handler) IsValidated() bool {
return h.ownerValidator != nil
}
// Create the resource update digest used in signatures
func (h *Handler) keyDataHash(addr storage.Address, data []byte) common.Hash {
hasher := h.hashPool.Get().(storage.SwarmHash)
defer h.hashPool.Put(hasher)
hasher.Reset()
hasher.Write(addr[:])
hasher.Write(data)
return common.BytesToHash(hasher.Sum(nil))
}
// Checks if current address matches owner address of ENS
func (h *Handler) checkAccess(name string, address common.Address) (bool, error) {
if h.ownerValidator == nil {
return true, nil
}
return h.ownerValidator.ValidateOwner(name, address)
}
// get data from current resource
func (h *Handler) GetContent(name string) (storage.Address, []byte, error) {
rsrc := h.get(name)
if rsrc == nil || !rsrc.isSynced() {
return nil, nil, NewError(ErrNotFound, " does not exist or is not synced")
}
return rsrc.lastKey, rsrc.data, nil
}
// Gets the period of the current data loaded in the resource
func (h *Handler) GetLastPeriod(nameHash string) (uint32, error) {
rsrc := h.get(nameHash)
if rsrc == nil {
return 0, NewError(ErrNotFound, " does not exist")
} else if !rsrc.isSynced() {
return 0, NewError(ErrNotSynced, " is not synced")
}
return rsrc.lastPeriod, nil
}
// Gets the version of the current data loaded in the resource
func (h *Handler) GetVersion(nameHash string) (uint32, error) {
rsrc := h.get(nameHash)
if rsrc == nil {
return 0, NewError(ErrNotFound, " does not exist")
} else if !rsrc.isSynced() {
return 0, NewError(ErrNotSynced, " is not synced")
}
return rsrc.version, nil
}
// \TODO should be hashsize * branches from the chosen chunker, implement with FileStore
func (h *Handler) chunkSize() int64 {
return chunkSize
}
// Creates a new root entry for a mutable resource identified by `name` with the specified `frequency`.
//
// The signature data should match the hash of the idna-converted name by the validator's namehash function, NOT the raw name bytes.
//
// The start block of the resource update will be the actual current block height of the connected network.
func (h *Handler) New(ctx context.Context, name string, frequency uint64) (storage.Address, *resource, error) {
// frequency 0 is invalid
if frequency == 0 {
return nil, nil, NewError(ErrInvalidValue, "Frequency cannot be 0")
}
// make sure name only contains ascii values
if !isSafeName(name) {
return nil, nil, NewError(ErrInvalidValue, fmt.Sprintf("Invalid name: '%s'", name))
}
nameHash := ens.EnsNode(name)
// if the signer function is set, validate that the key of the signer has access to modify this ENS name
if h.signer != nil {
signature, err := h.signer.Sign(nameHash)
if err != nil {
return nil, nil, NewError(ErrInvalidSignature, fmt.Sprintf("Sign fail: %v", err))
}
addr, err := getAddressFromDataSig(nameHash, signature)
if err != nil {
return nil, nil, NewError(ErrInvalidSignature, fmt.Sprintf("Retrieve address from signature fail: %v", err))
}
ok, err := h.checkAccess(name, addr)
if err != nil {
return nil, nil, err
} else if !ok {
return nil, nil, NewError(ErrUnauthorized, fmt.Sprintf("Not owner of '%s'", name))
}
}
// get our blockheight at this time
currentblock, err := h.getBlock(ctx, name)
if err != nil {
return nil, nil, err
}
chunk := h.newMetaChunk(name, currentblock, frequency)
h.chunkStore.Put(chunk)
log.Debug("new resource", "name", name, "key", nameHash, "startBlock", currentblock, "frequency", frequency)
// create the internal index for the resource and populate it with the data of the first version
rsrc := &resource{
startBlock: currentblock,
frequency: frequency,
name: name,
nameHash: nameHash,
updated: time.Now(),
}
h.set(nameHash.Hex(), rsrc)
return chunk.Addr, rsrc, nil
}
func (h *Handler) newMetaChunk(name string, startBlock uint64, frequency uint64) *storage.Chunk {
// the metadata chunk points to data of first blockheight + update frequency
// from this we know from what blockheight we should look for updates, and how often
// it also contains the name of the resource, so we know what resource we are working with
data := make([]byte, metadataChunkOffsetSize+len(name))
// root block has first two bytes both set to 0, which distinguishes from update bytes
val := make([]byte, 8)
binary.LittleEndian.PutUint64(val, startBlock)
copy(data[2:10], val)
binary.LittleEndian.PutUint64(val, frequency)
copy(data[10:18], val)
copy(data[18:], []byte(name))
// the key of the metadata chunk is content-addressed
// if it wasn't we couldn't replace it later
// resolving this relationship is left up to external agents (for example ENS)
hasher := h.hashPool.Get().(storage.SwarmHash)
hasher.Reset()
hasher.Write(data)
key := hasher.Sum(nil)
h.hashPool.Put(hasher)
// make the chunk and send it to swarm
chunk := storage.NewChunk(key, nil)
chunk.SData = make([]byte, metadataChunkOffsetSize+len(name))
copy(chunk.SData, data)
return chunk
}
// Searches and retrieves the specific version of the resource update identified by `name`
// at the specific block height
//
// If refresh is set to true, the resource data will be reloaded from the resource update
// metadata chunk.
// It is the callers responsibility to make sure that this chunk exists (if the resource
// update root data was retrieved externally, it typically doesn't)
func (h *Handler) LookupVersionByName(ctx context.Context, name string, period uint32, version uint32, refresh bool, maxLookup *LookupParams) (*resource, error) {
return h.LookupVersion(ctx, ens.EnsNode(name), period, version, refresh, maxLookup)
}
func (h *Handler) LookupVersion(ctx context.Context, nameHash common.Hash, period uint32, version uint32, refresh bool, maxLookup *LookupParams) (*resource, error) {
rsrc := h.get(nameHash.Hex())
if rsrc == nil {
return nil, NewError(ErrNothingToReturn, "resource not loaded")
}
return h.lookup(rsrc, period, version, refresh, maxLookup)
}
// Retrieves the latest version of the resource update identified by `name`
// at the specified block height
//
// If an update is found, version numbers are iterated until failure, and the last
// successfully retrieved version is copied to the corresponding resources map entry
// and returned.
//
// See also (*Handler).LookupVersion
func (h *Handler) LookupHistoricalByName(ctx context.Context, name string, period uint32, refresh bool, maxLookup *LookupParams) (*resource, error) {
return h.LookupHistorical(ctx, ens.EnsNode(name), period, refresh, maxLookup)
}
func (h *Handler) LookupHistorical(ctx context.Context, nameHash common.Hash, period uint32, refresh bool, maxLookup *LookupParams) (*resource, error) {
rsrc := h.get(nameHash.Hex())
if rsrc == nil {
return nil, NewError(ErrNothingToReturn, "resource not loaded")
}
return h.lookup(rsrc, period, 0, refresh, maxLookup)
}
// Retrieves the latest version of the resource update identified by `name`
// at the next update block height
//
// It starts at the next period after the current block height, and upon failure
// tries the corresponding keys of each previous period until one is found
// (or startBlock is reached, in which case there are no updates).
//
// Version iteration is done as in (*Handler).LookupHistorical
//
// See also (*Handler).LookupHistorical
func (h *Handler) LookupLatestByName(ctx context.Context, name string, refresh bool, maxLookup *LookupParams) (*resource, error) {
return h.LookupLatest(ctx, ens.EnsNode(name), refresh, maxLookup)
}
func (h *Handler) LookupLatest(ctx context.Context, nameHash common.Hash, refresh bool, maxLookup *LookupParams) (*resource, error) {
// get our blockheight at this time and the next block of the update period
rsrc := h.get(nameHash.Hex())
if rsrc == nil {
return nil, NewError(ErrNothingToReturn, "resource not loaded")
}
currentblock, err := h.getBlock(ctx, rsrc.name)
if err != nil {
return nil, err
}
nextperiod, err := getNextPeriod(rsrc.startBlock, currentblock, rsrc.frequency)
if err != nil {
return nil, err
}
return h.lookup(rsrc, nextperiod, 0, refresh, maxLookup)
}
// Returns the resource before the one currently loaded in the resource index
//
// This is useful where resource updates are used incrementally in contrast to
// merely replacing content.
//
// Requires a synced resource object
func (h *Handler) LookupPreviousByName(ctx context.Context, name string, maxLookup *LookupParams) (*resource, error) {
return h.LookupPrevious(ctx, ens.EnsNode(name), maxLookup)
}
func (h *Handler) LookupPrevious(ctx context.Context, nameHash common.Hash, maxLookup *LookupParams) (*resource, error) {
rsrc := h.get(nameHash.Hex())
if rsrc == nil {
return nil, NewError(ErrNothingToReturn, "resource not loaded")
}
if !rsrc.isSynced() {
return nil, NewError(ErrNotSynced, "LookupPrevious requires synced resource.")
} else if rsrc.lastPeriod == 0 {
return nil, NewError(ErrNothingToReturn, " not found")
}
if rsrc.version > 1 {
rsrc.version--
} else if rsrc.lastPeriod == 1 {
return nil, NewError(ErrNothingToReturn, "Current update is the oldest")
} else {
rsrc.version = 0
rsrc.lastPeriod--
}
return h.lookup(rsrc, rsrc.lastPeriod, rsrc.version, false, maxLookup)
}
// base code for public lookup methods
func (h *Handler) lookup(rsrc *resource, period uint32, version uint32, refresh bool, maxLookup *LookupParams) (*resource, error) {
// we can't look for anything without a store
if h.chunkStore == nil {
return nil, NewError(ErrInit, "Call Handler.SetStore() before performing lookups")
}
// period 0 does not exist
if period == 0 {
return nil, NewError(ErrInvalidValue, "period must be >0")
}
// start from the last possible block period, and iterate previous ones until we find a match
// if we hit startBlock we're out of options
var specificversion bool
if version > 0 {
specificversion = true
} else {
version = 1
}
var hops uint32
if maxLookup == nil {
maxLookup = h.queryMaxPeriods
}
log.Trace("resource lookup", "period", period, "version", version, "limit", maxLookup.Limit, "max", maxLookup.Max)
for period > 0 {
if maxLookup.Limit && hops > maxLookup.Max {
return nil, NewError(ErrPeriodDepth, fmt.Sprintf("Lookup exceeded max period hops (%d)", maxLookup.Max))
}
key := h.resourceHash(period, version, rsrc.nameHash)
chunk, err := h.chunkStore.GetWithTimeout(key, defaultRetrieveTimeout)
if err == nil {
if specificversion {
return h.updateIndex(rsrc, chunk)
}
// check if we have versions > 1. If a version fails, the previous version is used and returned.
log.Trace("rsrc update version 1 found, checking for version updates", "period", period, "key", key)
for {
newversion := version + 1
key := h.resourceHash(period, newversion, rsrc.nameHash)
newchunk, err := h.chunkStore.GetWithTimeout(key, defaultRetrieveTimeout)
if err != nil {
return h.updateIndex(rsrc, chunk)
}
chunk = newchunk
version = newversion
log.Trace("version update found, checking next", "version", version, "period", period, "key", key)
}
}
log.Trace("rsrc update not found, checking previous period", "period", period, "key", key)
period--
hops++
}
return nil, NewError(ErrNotFound, "no updates found")
}
// Retrieves a resource metadata chunk and creates/updates the index entry for it
// with the resulting metadata
func (h *Handler) Load(addr storage.Address) (*resource, error) {
chunk, err := h.chunkStore.GetWithTimeout(addr, defaultRetrieveTimeout)
if err != nil {
return nil, NewError(ErrNotFound, err.Error())
}
// minimum sanity check for chunk data (an update chunk first two bytes is headerlength uint16, and cannot be 0)
// \TODO this is not enough to make sure the data isn't bogus. A normal content addressed chunk could still satisfy these criteria
if !bytes.Equal(chunk.SData[:2], []byte{0x0, 0x0}) {
return nil, NewError(ErrCorruptData, fmt.Sprintf("Chunk is not a resource metadata chunk"))
} else if len(chunk.SData) <= metadataChunkOffsetSize {
return nil, NewError(ErrNothingToReturn, fmt.Sprintf("Invalid chunk length %d, should be minimum %d", len(chunk.SData), metadataChunkOffsetSize+1))
}
// create the index entry
rsrc := &resource{}
rsrc.UnmarshalBinary(chunk.SData[2:])
rsrc.nameHash = ens.EnsNode(rsrc.name)
h.set(rsrc.nameHash.Hex(), rsrc)
log.Trace("resource index load", "rootkey", addr, "name", rsrc.name, "namehash", rsrc.nameHash, "startblock", rsrc.startBlock, "frequency", rsrc.frequency)
return rsrc, nil
}
// update mutable resource index map with specified content
func (h *Handler) updateIndex(rsrc *resource, chunk *storage.Chunk) (*resource, error) {
// retrieve metadata from chunk data and check that it matches this mutable resource
signature, period, version, name, data, multihash, err := h.parseUpdate(chunk.SData)
if rsrc.name != name {
return nil, NewError(ErrNothingToReturn, fmt.Sprintf("Update belongs to '%s', but have '%s'", name, rsrc.name))
}
log.Trace("resource index update", "name", rsrc.name, "namehash", rsrc.nameHash, "updatekey", chunk.Addr, "period", period, "version", version)
// check signature (if signer algorithm is present)
// \TODO maybe this check is redundant if also checked upon retrieval of chunk
if signature != nil {
digest := h.keyDataHash(chunk.Addr, data)
_, err = getAddressFromDataSig(digest, *signature)
if err != nil {
return nil, NewError(ErrUnauthorized, fmt.Sprintf("Invalid signature: %v", err))
}
}
// update our rsrcs entry map
rsrc.lastKey = chunk.Addr
rsrc.lastPeriod = period
rsrc.version = version
rsrc.updated = time.Now()
rsrc.data = make([]byte, len(data))
rsrc.Multihash = multihash
rsrc.Reader = bytes.NewReader(rsrc.data)
copy(rsrc.data, data)
log.Debug(" synced", "name", rsrc.name, "key", chunk.Addr, "period", rsrc.lastPeriod, "version", rsrc.version)
h.set(rsrc.nameHash.Hex(), rsrc)
return rsrc, nil
}
// retrieve update metadata from chunk data
// mirrors newUpdateChunk()
func (h *Handler) parseUpdate(chunkdata []byte) (*Signature, uint32, uint32, string, []byte, bool, error) {
// absolute minimum an update chunk can contain:
// 14 = header + one byte of name + one byte of data
if len(chunkdata) < 14 {
return nil, 0, 0, "", nil, false, NewError(ErrNothingToReturn, "chunk less than 13 bytes cannot be a resource update chunk")
}
cursor := 0
headerlength := binary.LittleEndian.Uint16(chunkdata[cursor : cursor+2])
cursor += 2
datalength := binary.LittleEndian.Uint16(chunkdata[cursor : cursor+2])
cursor += 2
var exclsignlength int
// we need extra magic if it's a multihash, since we used datalength 0 in header as an indicator of multihash content
// retrieve the second varint and set this as the data length
// TODO: merge with isMultihash code
if datalength == 0 {
uvarintbuf := bytes.NewBuffer(chunkdata[headerlength+4:])
r, err := binary.ReadUvarint(uvarintbuf)
if err != nil {
errstr := fmt.Sprintf("corrupt multihash, hash id varint could not be read: %v", err)
log.Warn(errstr)
return nil, 0, 0, "", nil, false, NewError(ErrCorruptData, errstr)
}
r, err = binary.ReadUvarint(uvarintbuf)
if err != nil {
errstr := fmt.Sprintf("corrupt multihash, hash length field could not be read: %v", err)
log.Warn(errstr)
return nil, 0, 0, "", nil, false, NewError(ErrCorruptData, errstr)
}
exclsignlength = int(headerlength + uint16(r))
} else {
exclsignlength = int(headerlength + datalength + 4)
}
// the total length excluding signature is headerlength and datalength fields plus the length of the header and the data given in these fields
exclsignlength = int(headerlength + datalength + 4)
if exclsignlength > len(chunkdata) || exclsignlength < 14 {
return nil, 0, 0, "", nil, false, NewError(ErrNothingToReturn, fmt.Sprintf("Reported headerlength %d + datalength %d longer than actual chunk data length %d", headerlength, exclsignlength, len(chunkdata)))
} else if exclsignlength < 14 {
return nil, 0, 0, "", nil, false, NewError(ErrNothingToReturn, fmt.Sprintf("Reported headerlength %d + datalength %d is smaller than minimum valid resource chunk length %d", headerlength, datalength, 14))
}
// at this point we can be satisfied that the data integrity is ok
var period uint32
var version uint32
var name string
var data []byte
period = binary.LittleEndian.Uint32(chunkdata[cursor : cursor+4])
cursor += 4
version = binary.LittleEndian.Uint32(chunkdata[cursor : cursor+4])
cursor += 4
namelength := int(headerlength) - cursor + 4
if l := len(chunkdata); l < cursor+namelength {
return nil, 0, 0, "", nil, false, NewError(ErrNothingToReturn, fmt.Sprintf("chunk less than %v bytes is too short to read the name", l))
}
name = string(chunkdata[cursor : cursor+namelength])
cursor += namelength
// if multihash content is indicated we check the validity of the multihash
// \TODO the check above for multihash probably is sufficient also for this case (or can be with a small adjustment) and if so this code should be removed
var intdatalength int
var ismultihash bool
if datalength == 0 {
var intheaderlength int
var err error
intdatalength, intheaderlength, err = multihash.GetMultihashLength(chunkdata[cursor:])
if err != nil {
log.Error("multihash parse error", "err", err)
return nil, 0, 0, "", nil, false, err
}
intdatalength += intheaderlength
multihashboundary := cursor + intdatalength
if len(chunkdata) != multihashboundary && len(chunkdata) < multihashboundary+signatureLength {
log.Debug("multihash error", "chunkdatalen", len(chunkdata), "multihashboundary", multihashboundary)
return nil, 0, 0, "", nil, false, errors.New("Corrupt multihash data")
}
ismultihash = true
} else {
intdatalength = int(datalength)
}
data = make([]byte, intdatalength)
copy(data, chunkdata[cursor:cursor+intdatalength])
// omit signatures if we have no validator
var signature *Signature
cursor += intdatalength
if h.signer != nil {
sigdata := chunkdata[cursor : cursor+signatureLength]
if len(sigdata) > 0 {
signature = &Signature{}
copy(signature[:], sigdata)
}
}
return signature, period, version, name, data, ismultihash, nil
}
// Adds an actual data update
//
// Uses the data currently loaded in the resources map entry.
// It is the caller's responsibility to make sure that this data is not stale.
//
// A resource update cannot span chunks, and thus has max length 4096
func (h *Handler) UpdateMultihash(ctx context.Context, name string, data []byte) (storage.Address, error) {
// \TODO perhaps this check should be in newUpdateChunk()
if _, _, err := multihash.GetMultihashLength(data); err != nil {
return nil, NewError(ErrNothingToReturn, err.Error())
}
return h.update(ctx, name, data, true)
}
func (h *Handler) Update(ctx context.Context, name string, data []byte) (storage.Address, error) {
return h.update(ctx, name, data, false)
}
// create and commit an update
func (h *Handler) update(ctx context.Context, name string, data []byte, multihash bool) (storage.Address, error) {
// zero-length updates are bogus
if len(data) == 0 {
return nil, NewError(ErrInvalidValue, "I refuse to waste swarm space for updates with empty values, amigo (data length is 0)")
}
// we can't update anything without a store
if h.chunkStore == nil {
return nil, NewError(ErrInit, "Call Handler.SetStore() before updating")
}
// signature length is 0 if we are not using them
var signaturelength int
if h.signer != nil {
signaturelength = signatureLength
}
// get the cached information
nameHash := ens.EnsNode(name)
nameHashHex := nameHash.Hex()
rsrc := h.get(nameHashHex)
if rsrc == nil {
return nil, NewError(ErrNotFound, fmt.Sprintf(" object '%s' not in index", name))
} else if !rsrc.isSynced() {
return nil, NewError(ErrNotSynced, " object not in sync")
}
// an update can be only one chunk long; data length less header and signature data
// 12 = length of header and data length fields (2xuint16) plus period and frequency value fields (2xuint32)
datalimit := h.chunkSize() - int64(signaturelength-len(name)-12)
if int64(len(data)) > datalimit {
return nil, NewError(ErrDataOverflow, fmt.Sprintf("Data overflow: %d / %d bytes", len(data), datalimit))
}
// get our blockheight at this time and the next block of the update period
currentblock, err := h.getBlock(ctx, name)
if err != nil {
return nil, NewError(ErrIO, fmt.Sprintf("Could not get block height: %v", err))
}
nextperiod, err := getNextPeriod(rsrc.startBlock, currentblock, rsrc.frequency)
if err != nil {
return nil, err
}
// if we already have an update for this block then increment version
// resource object MUST be in sync for version to be correct, but we checked this earlier in the method already
var version uint32
if h.hasUpdate(nameHashHex, nextperiod) {
version = rsrc.version
}
version++
// calculate the chunk key
key := h.resourceHash(nextperiod, version, rsrc.nameHash)
// if we have a signing function, sign the update
// \TODO this code should probably be consolidated with corresponding code in New()
var signature *Signature
if h.signer != nil {
// sign the data hash with the key
digest := h.keyDataHash(key, data)
sig, err := h.signer.Sign(digest)
if err != nil {
return nil, NewError(ErrInvalidSignature, fmt.Sprintf("Sign fail: %v", err))
}
signature = &sig
// get the address of the signer (which also checks that it's a valid signature)
addr, err := getAddressFromDataSig(digest, *signature)
if err != nil {
return nil, NewError(ErrInvalidSignature, fmt.Sprintf("Invalid data/signature: %v", err))
}
if h.signer != nil {
// check if the signer has access to update
ok, err := h.checkAccess(name, addr)
if err != nil {
return nil, NewError(ErrIO, fmt.Sprintf("Access check fail: %v", err))
} else if !ok {
return nil, NewError(ErrUnauthorized, fmt.Sprintf("Address %x does not have access to update %s", addr, name))
}
}
}
// a datalength field set to 0 means the content is a multihash
var datalength int
if !multihash {
datalength = len(data)
}
chunk := newUpdateChunk(key, signature, nextperiod, version, name, data, datalength)
// send the chunk
h.chunkStore.Put(chunk)
log.Trace("resource update", "name", name, "key", key, "currentblock", currentblock, "lastperiod", nextperiod, "version", version, "data", chunk.SData, "multihash", multihash)
// update our resources map entry and return the new key
rsrc.lastPeriod = nextperiod
rsrc.version = version
rsrc.data = make([]byte, len(data))
copy(rsrc.data, data)
return key, nil
}
// Closes the datastore.
// Always call this at shutdown to avoid data corruption.
func (h *Handler) Close() {
h.chunkStore.Close()
}
// gets the current block height
func (h *Handler) getBlock(ctx context.Context, name string) (uint64, error) {
blockheader, err := h.headerGetter.HeaderByNumber(ctx, name, nil)
if err != nil {
return 0, err
}
return blockheader.Number.Uint64(), nil
}
// Calculate the period index (aka major version number) from a given block number
func (h *Handler) BlockToPeriod(name string, blocknumber uint64) (uint32, error) {
return getNextPeriod(h.resources[name].startBlock, blocknumber, h.resources[name].frequency)
}
// Calculate the block number from a given period index (aka major version number)
func (h *Handler) PeriodToBlock(name string, period uint32) uint64 {
return h.resources[name].startBlock + (uint64(period) * h.resources[name].frequency)
}
// Retrieves the resource index value for the given nameHash
func (h *Handler) get(nameHash string) *resource {
h.resourceLock.RLock()
defer h.resourceLock.RUnlock()
rsrc := h.resources[nameHash]
return rsrc
}
// Sets the resource index value for the given nameHash
func (h *Handler) set(nameHash string, rsrc *resource) {
h.resourceLock.Lock()
defer h.resourceLock.Unlock()
h.resources[nameHash] = rsrc
}
// used for chunk keys
func (h *Handler) resourceHash(period uint32, version uint32, namehash common.Hash) storage.Address {
// format is: hash(period|version|namehash)
hasher := h.hashPool.Get().(storage.SwarmHash)
defer h.hashPool.Put(hasher)
hasher.Reset()
b := make([]byte, 4)
binary.LittleEndian.PutUint32(b, period)
hasher.Write(b)
binary.LittleEndian.PutUint32(b, version)
hasher.Write(b)
hasher.Write(namehash[:])
return hasher.Sum(nil)
}
// Checks if we already have an update on this resource, according to the value in the current state of the resource index
func (h *Handler) hasUpdate(nameHash string, period uint32) bool {
return h.resources[nameHash].lastPeriod == period
}
func getAddressFromDataSig(datahash common.Hash, signature Signature) (common.Address, error) {
pub, err := crypto.SigToPub(datahash.Bytes(), signature[:])
if err != nil {
return common.Address{}, err
}
return crypto.PubkeyToAddress(*pub), nil
}
// create an update chunk
func newUpdateChunk(addr storage.Address, signature *Signature, period uint32, version uint32, name string, data []byte, datalength int) *storage.Chunk {
// no signatures if no validator
var signaturelength int
if signature != nil {
signaturelength = signatureLength
}
// prepend version and period to allow reverse lookups
headerlength := len(name) + 4 + 4
actualdatalength := len(data)
chunk := storage.NewChunk(addr, nil)
chunk.SData = make([]byte, 4+signaturelength+headerlength+actualdatalength) // initial 4 are uint16 length descriptors for headerlength and datalength
// data header length does NOT include the header length prefix bytes themselves
cursor := 0
binary.LittleEndian.PutUint16(chunk.SData[cursor:], uint16(headerlength))
cursor += 2
// data length
binary.LittleEndian.PutUint16(chunk.SData[cursor:], uint16(datalength))
cursor += 2
// header = period + version + name
binary.LittleEndian.PutUint32(chunk.SData[cursor:], period)
cursor += 4
binary.LittleEndian.PutUint32(chunk.SData[cursor:], version)
cursor += 4
namebytes := []byte(name)
copy(chunk.SData[cursor:], namebytes)
cursor += len(namebytes)
// add the data
copy(chunk.SData[cursor:], data)
// if signature is present it's the last item in the chunk data
if signature != nil {
cursor += actualdatalength
copy(chunk.SData[cursor:], signature[:])
}
chunk.Size = int64(len(chunk.SData))
return chunk
}
// Helper function to calculate the next update period number from the current block, start block and frequency
func getNextPeriod(start uint64, current uint64, frequency uint64) (uint32, error) {
if current < start {
return 0, NewError(ErrInvalidValue, fmt.Sprintf("given current block value %d < start block %d", current, start))
}
blockdiff := current - start
period := blockdiff / frequency
return uint32(period + 1), nil
}
// ToSafeName is a helper function to create an valid idna of a given resource update name
func ToSafeName(name string) (string, error) {
return idna.ToASCII(name)
}
// check that name identifiers contain valid bytes
// Strings created using ToSafeName() should satisfy this check
func isSafeName(name string) bool {
if name == "" {
return false
}
validname, err := idna.ToASCII(name)
if err != nil {
return false
}
return validname == name
}
func NewTestHandler(datadir string, params *HandlerParams) (*Handler, error) {
path := filepath.Join(datadir, DbDirName)
rh, err := NewHandler(params)
if err != nil {
return nil, fmt.Errorf("resource handler create fail: %v", err)
}
localstoreparams := storage.NewDefaultLocalStoreParams()
localstoreparams.Init(path)
localStore, err := storage.NewLocalStore(localstoreparams, nil)
if err != nil {
return nil, fmt.Errorf("localstore create fail, path %s: %v", path, err)
}
localStore.Validators = append(localStore.Validators, storage.NewContentAddressValidator(storage.MakeHashFunc(resourceHash)))
localStore.Validators = append(localStore.Validators, rh)
netStore := storage.NewNetStore(localStore, nil)
rh.SetStore(netStore)
return rh, nil
}