go-ethereum/swarm/storage/netstore.go

298 lines
10 KiB
Go
Raw Normal View History

// Copyright 2016 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 storage
import (
"context"
"encoding/hex"
"fmt"
"sync"
"sync/atomic"
"time"
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
2018-09-25 01:59:00 +03:00
"github.com/ethereum/go-ethereum/p2p/enode"
2018-06-20 15:06:27 +03:00
"github.com/ethereum/go-ethereum/swarm/log"
lru "github.com/hashicorp/golang-lru"
)
type (
NewNetFetcherFunc func(ctx context.Context, addr Address, peers *sync.Map) NetFetcher
2018-06-20 15:06:27 +03:00
)
type NetFetcher interface {
Request(ctx context.Context, hopCount uint8)
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
2018-09-25 01:59:00 +03:00
Offer(ctx context.Context, source *enode.ID)
}
// NetStore is an extension of local storage
// it implements the ChunkStore interface
// on request it initiates remote cloud retrieval using a fetcher
// fetchers are unique to a chunk and are stored in fetchers LRU memory cache
// fetchFuncFactory is a factory object to create a fetch function for a specific chunk address
type NetStore struct {
mu sync.Mutex
store SyncChunkStore
fetchers *lru.Cache
NewNetFetcherFunc NewNetFetcherFunc
closeC chan struct{}
}
var fetcherTimeout = 2 * time.Minute // timeout to cancel the fetcher even if requests are coming in
// NewNetStore creates a new NetStore object using the given local store. newFetchFunc is a
// constructor function that can create a fetch function for a specific chunk address.
func NewNetStore(store SyncChunkStore, nnf NewNetFetcherFunc) (*NetStore, error) {
fetchers, err := lru.New(defaultChunkRequestsCacheCapacity)
if err != nil {
return nil, err
}
return &NetStore{
store: store,
fetchers: fetchers,
NewNetFetcherFunc: nnf,
closeC: make(chan struct{}),
}, nil
}
// Put stores a chunk in localstore, and delivers to all requestor peers using the fetcher stored in
// the fetchers cache
func (n *NetStore) Put(ctx context.Context, ch Chunk) error {
n.mu.Lock()
defer n.mu.Unlock()
// put to the chunk to the store, there should be no error
err := n.store.Put(ctx, ch)
if err != nil {
return err
}
2018-06-20 15:06:27 +03:00
// if chunk is now put in the store, check if there was an active fetcher and call deliver on it
// (this delivers the chunk to requestors via the fetcher)
if f := n.getFetcher(ch.Address()); f != nil {
f.deliver(ctx, ch)
}
return nil
}
// Get retrieves the chunk from the NetStore DPA synchronously.
// It calls NetStore.get, and if the chunk is not in local Storage
// it calls fetch with the request, which blocks until the chunk
// arrived or context is done
func (n *NetStore) Get(rctx context.Context, ref Address) (Chunk, error) {
chunk, fetch, err := n.get(rctx, ref)
if err != nil {
return nil, err
}
if chunk != nil {
return chunk, nil
}
return fetch(rctx)
}
func (n *NetStore) BinIndex(po uint8) uint64 {
return n.store.BinIndex(po)
2018-06-20 15:06:27 +03:00
}
func (n *NetStore) Iterator(from uint64, to uint64, po uint8, f func(Address, uint64) bool) error {
return n.store.Iterator(from, to, po, f)
}
// FetchFunc returns nil if the store contains the given address. Otherwise it returns a wait function,
// which returns after the chunk is available or the context is done
func (n *NetStore) FetchFunc(ctx context.Context, ref Address) func(context.Context) error {
chunk, fetch, _ := n.get(ctx, ref)
if chunk != nil {
return nil
}
return func(ctx context.Context) error {
_, err := fetch(ctx)
return err
}
}
// Close chunk store
func (n *NetStore) Close() {
close(n.closeC)
n.store.Close()
// TODO: loop through fetchers to cancel them
}
// get attempts at retrieving the chunk from LocalStore
// If it is not found then using getOrCreateFetcher:
// 1. Either there is already a fetcher to retrieve it
// 2. A new fetcher is created and saved in the fetchers cache
// From here on, all Get will hit on this fetcher until the chunk is delivered
// or all fetcher contexts are done.
// It returns a chunk, a fetcher function and an error
// If chunk is nil, the returned fetch function needs to be called with a context to return the chunk.
func (n *NetStore) get(ctx context.Context, ref Address) (Chunk, func(context.Context) (Chunk, error), error) {
n.mu.Lock()
defer n.mu.Unlock()
2018-06-20 15:06:27 +03:00
chunk, err := n.store.Get(ctx, ref)
if err != nil {
if err != ErrChunkNotFound {
log.Debug("Received error from LocalStore other than ErrNotFound", "err", err)
2018-06-20 15:06:27 +03:00
}
// The chunk is not available in the LocalStore, let's get the fetcher for it, or create a new one
// if it doesn't exist yet
f := n.getOrCreateFetcher(ref)
// If the caller needs the chunk, it has to use the returned fetch function to get it
return nil, f.Fetch, nil
}
return chunk, nil, nil
}
// getOrCreateFetcher attempts at retrieving an existing fetchers
// if none exists, creates one and saves it in the fetchers cache
// caller must hold the lock
func (n *NetStore) getOrCreateFetcher(ref Address) *fetcher {
if f := n.getFetcher(ref); f != nil {
return f
}
2018-06-20 15:06:27 +03:00
// no fetcher for the given address, we have to create a new one
key := hex.EncodeToString(ref)
// create the context during which fetching is kept alive
ctx, cancel := context.WithTimeout(context.Background(), fetcherTimeout)
// destroy is called when all requests finish
destroy := func() {
// remove fetcher from fetchers
n.fetchers.Remove(key)
// stop fetcher by cancelling context called when
// all requests cancelled/timedout or chunk is delivered
cancel()
}
// peers always stores all the peers which have an active request for the chunk. It is shared
// between fetcher and the NewFetchFunc function. It is needed by the NewFetchFunc because
// the peers which requested the chunk should not be requested to deliver it.
peers := &sync.Map{}
2018-06-20 15:06:27 +03:00
fetcher := newFetcher(ref, n.NewNetFetcherFunc(ctx, ref, peers), destroy, peers, n.closeC)
n.fetchers.Add(key, fetcher)
return fetcher
}
// getFetcher retrieves the fetcher for the given address from the fetchers cache if it exists,
// otherwise it returns nil
func (n *NetStore) getFetcher(ref Address) *fetcher {
key := hex.EncodeToString(ref)
f, ok := n.fetchers.Get(key)
if ok {
return f.(*fetcher)
2018-06-20 15:06:27 +03:00
}
return nil
}
// RequestsCacheLen returns the current number of outgoing requests stored in the cache
func (n *NetStore) RequestsCacheLen() int {
return n.fetchers.Len()
2018-06-20 15:06:27 +03:00
}
// One fetcher object is responsible to fetch one chunk for one address, and keep track of all the
// peers who have requested it and did not receive it yet.
type fetcher struct {
addr Address // address of chunk
chunk Chunk // fetcher can set the chunk on the fetcher
deliveredC chan struct{} // chan signalling chunk delivery to requests
cancelledC chan struct{} // chan signalling the fetcher has been cancelled (removed from fetchers in NetStore)
netFetcher NetFetcher // remote fetch function to be called with a request source taken from the context
cancel func() // cleanup function for the remote fetcher to call when all upstream contexts are called
peers *sync.Map // the peers which asked for the chunk
requestCnt int32 // number of requests on this chunk. If all the requests are done (delivered or context is done) the cancel function is called
deliverOnce *sync.Once // guarantees that we only close deliveredC once
}
// newFetcher creates a new fetcher object for the fiven addr. fetch is the function which actually
// does the retrieval (in non-test cases this is coming from the network package). cancel function is
// called either
// 1. when the chunk has been fetched all peers have been either notified or their context has been done
// 2. the chunk has not been fetched but all context from all the requests has been done
// The peers map stores all the peers which have requested chunk.
func newFetcher(addr Address, nf NetFetcher, cancel func(), peers *sync.Map, closeC chan struct{}) *fetcher {
cancelOnce := &sync.Once{} // cancel should only be called once
return &fetcher{
addr: addr,
deliveredC: make(chan struct{}),
deliverOnce: &sync.Once{},
cancelledC: closeC,
netFetcher: nf,
cancel: func() {
cancelOnce.Do(func() {
cancel()
})
},
peers: peers,
}
}
// Fetch fetches the chunk synchronously, it is called by NetStore.Get is the chunk is not available
// locally.
func (f *fetcher) Fetch(rctx context.Context) (Chunk, error) {
atomic.AddInt32(&f.requestCnt, 1)
defer func() {
// if all the requests are done the fetcher can be cancelled
if atomic.AddInt32(&f.requestCnt, -1) == 0 {
f.cancel()
}
}()
// The peer asking for the chunk. Store in the shared peers map, but delete after the request
// has been delivered
peer := rctx.Value("peer")
if peer != nil {
f.peers.Store(peer, time.Now())
defer f.peers.Delete(peer)
}
// If there is a source in the context then it is an offer, otherwise a request
sourceIF := rctx.Value("source")
hopCount, _ := rctx.Value("hopcount").(uint8)
if sourceIF != nil {
all: new p2p node representation (#17643) Package p2p/enode provides a generalized representation of p2p nodes which can contain arbitrary information in key/value pairs. It is also the new home for the node database. The "v4" identity scheme is also moved here from p2p/enr to remove the dependency on Ethereum crypto from that package. Record signature handling is changed significantly. The identity scheme registry is removed and acceptable schemes must be passed to any method that needs identity. This means records must now be validated explicitly after decoding. The enode API is designed to make signature handling easy and safe: most APIs around the codebase work with enode.Node, which is a wrapper around a valid record. Going from enr.Record to enode.Node requires a valid signature. * p2p/discover: port to p2p/enode This ports the discovery code to the new node representation in p2p/enode. The wire protocol is unchanged, this can be considered a refactoring change. The Kademlia table can now deal with nodes using an arbitrary identity scheme. This requires a few incompatible API changes: - Table.Lookup is not available anymore. It used to take a public key as argument because v4 protocol requires one. Its replacement is LookupRandom. - Table.Resolve takes *enode.Node instead of NodeID. This is also for v4 protocol compatibility because nodes cannot be looked up by ID alone. - Types Node and NodeID are gone. Further commits in the series will be fixes all over the the codebase to deal with those removals. * p2p: port to p2p/enode and discovery changes This adapts package p2p to the changes in p2p/discover. All uses of discover.Node and discover.NodeID are replaced by their equivalents from p2p/enode. New API is added to retrieve the enode.Node instance of a peer. The behavior of Server.Self with discovery disabled is improved. It now tries much harder to report a working IP address, falling back to 127.0.0.1 if no suitable address can be determined through other means. These changes were needed for tests of other packages later in the series. * p2p/simulations, p2p/testing: port to p2p/enode No surprises here, mostly replacements of discover.Node, discover.NodeID with their new equivalents. The 'interesting' API changes are: - testing.ProtocolSession tracks complete nodes, not just their IDs. - adapters.NodeConfig has a new method to create a complete node. These changes were needed to make swarm tests work. Note that the NodeID change makes the code incompatible with old simulation snapshots. * whisper/whisperv5, whisper/whisperv6: port to p2p/enode This port was easy because whisper uses []byte for node IDs and URL strings in the API. * eth: port to p2p/enode Again, easy to port because eth uses strings for node IDs and doesn't care about node information in any way. * les: port to p2p/enode Apart from replacing discover.NodeID with enode.ID, most changes are in the server pool code. It now deals with complete nodes instead of (Pubkey, IP, Port) triples. The database format is unchanged for now, but we should probably change it to use the node database later. * node: port to p2p/enode This change simply replaces discover.Node and discover.NodeID with their new equivalents. * swarm/network: port to p2p/enode Swarm has its own node address representation, BzzAddr, containing both an overlay address (the hash of a secp256k1 public key) and an underlay address (enode:// URL). There are no changes to the BzzAddr format in this commit, but certain operations such as creating a BzzAddr from a node ID are now impossible because node IDs aren't public keys anymore. Most swarm-related changes in the series remove uses of NewAddrFromNodeID, replacing it with NewAddr which takes a complete node as argument. ToOverlayAddr is removed because we can just use the node ID directly.
2018-09-25 01:59:00 +03:00
var source enode.ID
if err := source.UnmarshalText([]byte(sourceIF.(string))); err != nil {
return nil, err
}
f.netFetcher.Offer(rctx, &source)
} else {
f.netFetcher.Request(rctx, hopCount)
}
// wait until either the chunk is delivered or the context is done
select {
case <-rctx.Done():
return nil, rctx.Err()
case <-f.deliveredC:
return f.chunk, nil
case <-f.cancelledC:
return nil, fmt.Errorf("fetcher cancelled")
}
}
// deliver is called by NetStore.Put to notify all pending requests
func (f *fetcher) deliver(ctx context.Context, ch Chunk) {
f.deliverOnce.Do(func() {
f.chunk = ch
// closing the deliveredC channel will terminate ongoing requests
close(f.deliveredC)
})
2018-06-20 15:06:27 +03:00
}