Provide root node immediately when add and pin add

License: MIT
Signed-off-by: Michael Avila <davidmichaelavila@gmail.com>

core/builder.go

@@ -5,6 +5,7 @@ import (
 	"crypto/rand"
 	"encoding/base64"
 	"errors"
+	"github.com/ipfs/go-ipfs/provider"
 	"os"
 	"syscall"
 	"time"
@@ -275,6 +276,13 @@ func setupNode(ctx context.Context, n *IpfsNode, cfg *BuildCfg) error {
 	}
 	n.Resolver = resolver.NewBasicResolver(n.DAG)
 
+	// Provider
+	queue, err := provider.NewQueue("provider-v1", ctx, n.Repo.Datastore())
+	if err != nil {
+		return err
+	}
+	n.Provider = provider.NewProvider(ctx, queue, n.Routing)
+
 	if cfg.Online {
 		if err := n.startLateOnlineServices(ctx); err != nil {
 			return err

core/core.go

@@ -14,6 +14,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"github.com/ipfs/go-ipfs/provider"
 	"io"
 	"io/ioutil"
 	"os"
@@ -124,6 +125,7 @@ type IpfsNode struct {
 	Routing      routing.IpfsRouting // the routing system. recommend ipfs-dht
 	Exchange     exchange.Interface  // the block exchange + strategy (bitswap)
 	Namesys      namesys.NameSystem  // the name system, resolves paths to hashes
+	Provider     *provider.Provider  // the value provider system
 	Reprovider   *rp.Reprovider      // the value reprovider system
 	IpnsRepub    *ipnsrp.Republisher
 
@@ -324,6 +326,12 @@ func (n *IpfsNode) startLateOnlineServices(ctx context.Context) error {
 		return err
 	}
 
+	// Provider
+
+	n.Provider.Run()
+
+	// Reprovider
+
 	var keyProvider rp.KeyChanFunc
 
 	switch cfg.Reprovider.Strategy {

core/coreapi/coreapi.go

@@ -20,6 +20,7 @@ import (
 
 	"github.com/ipfs/go-ipfs/core"
 	"github.com/ipfs/go-ipfs/namesys"
+	"github.com/ipfs/go-ipfs/provider"
 	"github.com/ipfs/go-ipfs/pin"
 	"github.com/ipfs/go-ipfs/repo"
 
@@ -66,6 +67,8 @@ type CoreAPI struct {
 	namesys namesys.NameSystem
 	routing routing.IpfsRouting
 
+	provider *provider.Provider
+
 	pubSub *pubsub.PubSub
 
 	checkPublishAllowed func() error
@@ -174,6 +177,8 @@ func (api *CoreAPI) WithOptions(opts ...options.ApiOption) (coreiface.CoreAPI, e
 		exchange:        n.Exchange,
 		routing:         n.Routing,
 
+		provider: n.Provider,
+
 		pubSub: n.PubSub,
 
 		nd:         n,

core/coreapi/pin.go

@@ -32,6 +32,10 @@ func (api *PinAPI) Add(ctx context.Context, p coreiface.Path, opts ...caopts.Pin
 		return fmt.Errorf("pin: %s", err)
 	}
 
+	if err := api.provider.Provide(dagNode.Cid()); err != nil {
+		return err
+	}
+
 	return api.pinning.Flush()
 }
 

core/coreapi/provider.go

+package coreapi
+
+import (
+	cid "github.com/ipfs/go-cid"
+)
+
+type ProviderAPI CoreAPI
+
+func (api *ProviderAPI) Provide(root cid.Cid) error {
+	return api.provider.Provide(root)
+}

core/coreapi/unixfs.go

@@ -129,6 +129,11 @@ func (api *UnixfsAPI) Add(ctx context.Context, files files.Node, opts ...options
 	if err != nil {
 		return nil, err
 	}
+
+	if err := api.provider.Provide(nd.Cid()); err != nil {
+		return nil, err
+	}
+
 	return coreiface.IpfsPath(nd.Cid()), nil
 }
 

provider/provider.go

+// Package provider implements structures and methods to provide blocks,
+// keep track of which blocks are provided, and to allow those blocks to
+// be reprovided.
+package provider
+
+import (
+	"context"
+	"github.com/ipfs/go-cid"
+	logging "github.com/ipfs/go-log"
+	"github.com/libp2p/go-libp2p-routing"
+	"time"
+)
+
+var (
+	log = logging.Logger("provider")
+)
+
+const (
+	provideOutgoingWorkerLimit = 8
+	provideOutgoingTimeout     = 15 * time.Second
+)
+
+// Provider announces blocks to the network, tracks which blocks are
+// being provided, and untracks blocks when they're no longer in the blockstore.
+type Provider struct {
+	ctx context.Context
+	// the CIDs for which provide announcements should be made
+	queue *Queue
+	// used to announce providing to the network
+	contentRouting routing.ContentRouting
+}
+
+func NewProvider(ctx context.Context, queue *Queue, contentRouting routing.ContentRouting) *Provider {
+	return &Provider{
+		ctx:            ctx,
+		queue:          queue,
+		contentRouting: contentRouting,
+	}
+}
+
+// Start workers to handle provide requests.
+func (p *Provider) Run() {
+	p.queue.Run()
+	p.handleAnnouncements()
+}
+
+// Provide the given cid using specified strategy.
+func (p *Provider) Provide(root cid.Cid) error {
+	return p.queue.Enqueue(root)
+}
+
+// Handle all outgoing cids by providing (announcing) them
+func (p *Provider) handleAnnouncements() {
+	for workers := 0; workers < provideOutgoingWorkerLimit; workers++ {
+		go func() {
+			for {
+				select {
+				case <-p.ctx.Done():
+					return
+				case entry := <-p.queue.Dequeue():
+					if err := doProvide(p.ctx, p.contentRouting, entry.cid); err != nil {
+						log.Warningf("Unable to provide entry: %s, %s", entry.cid, err)
+					}
+
+					if err := entry.Complete(); err != nil {
+						log.Warningf("Unable to complete queue entry when providing: %s, %s", entry.cid, err)
+					}
+				}
+			}
+		}()
+	}
+}
+
+// TODO: better document this provide logic
+func doProvide(ctx context.Context, contentRouting routing.ContentRouting, key cid.Cid) error {
+	// announce
+	log.Info("announce - start - ", key)
+	ctx, cancel := context.WithTimeout(ctx, provideOutgoingTimeout)
+	if err := contentRouting.Provide(ctx, key, true); err != nil {
+		log.Warningf("Failed to provide cid: %s", err)
+		// TODO: Maybe put these failures onto a failures queue?
+		cancel()
+		return err
+	}
+	cancel()
+	log.Info("announce - end - ", key)
+	return nil
+}

provider/queue.go

+package provider
+
+import (
+	"context"
+	"errors"
+	"github.com/ipfs/go-cid"
+	ds "github.com/ipfs/go-datastore"
+	"github.com/ipfs/go-datastore/namespace"
+	"github.com/ipfs/go-datastore/query"
+	"math"
+	"strconv"
+	"strings"
+	"sync"
+)
+
+// Entry allows for the durability in the queue. When a cid is dequeued it is
+// not removed from the datastore until you call Complete() on the entry you
+// receive.
+type Entry struct {
+	cid cid.Cid
+	key ds.Key
+	queue *Queue
+}
+
+func (e *Entry) Complete() error {
+	return e.queue.remove(e.key)
+}
+
+// Queue provides a durable, FIFO interface to the datastore for storing cids
+//
+// Durability just means that cids in the process of being provided when a
+// crash or shutdown occurs will still be in the queue when the node is
+// brought back online.
+type Queue struct {
+	// used to differentiate queues in datastore
+	// e.g. provider vs reprovider
+	name string
+
+	ctx context.Context
+
+	tail uint64
+	head uint64
+
+	lock sync.Mutex
+	datastore ds.Datastore
+
+	dequeue chan *Entry
+	notEmpty chan struct{}
+
+	isRunning bool
+}
+
+func NewQueue(name string, ctx context.Context, datastore ds.Datastore) (*Queue, error) {
+	namespaced := namespace.Wrap(datastore, ds.NewKey("/" + name + "/queue/"))
+	head, tail, err := getQueueHeadTail(name, ctx, namespaced)
+	if err != nil {
+		return nil, err
+	}
+	q := &Queue{
+		name: name,
+		ctx: ctx,
+		head: head,
+		tail: tail,
+		lock: sync.Mutex{},
+		datastore: namespaced,
+		dequeue: make(chan *Entry),
+		notEmpty: make(chan struct{}),
+		isRunning: false,
+	}
+	return q, nil
+}
+
+// Put a cid in the queue
+func (q *Queue) Enqueue(cid cid.Cid) error {
+	q.lock.Lock()
+	defer q.lock.Unlock()
+
+	wasEmpty := q.IsEmpty()
+
+	nextKey := q.queueKey(q.tail)
+
+	if err := q.datastore.Put(nextKey, cid.Bytes()); err != nil {
+		return err
+	}
+
+	q.tail++
+
+	if q.isRunning && wasEmpty {
+		select {
+		case q.notEmpty <- struct{}{}:
+		case <-q.ctx.Done():
+		}
+	}
+
+	return nil
+}
+
+// Remove an entry from the queue.
+func (q *Queue) Dequeue() <-chan *Entry {
+	return q.dequeue
+}
+
+func (q *Queue) IsEmpty() bool {
+	return (q.tail - q.head) == 0
+}
+
+func (q *Queue) remove(key ds.Key) error {
+	return q.datastore.Delete(key)
+}
+
+// dequeue items when the dequeue channel is available to
+// be written to
+func (q *Queue) Run() {
+	q.isRunning = true
+	go func() {
+		for {
+			select {
+			case <-q.ctx.Done():
+				return
+			default:
+			}
+			if q.IsEmpty() {
+				select {
+				case <-q.ctx.Done():
+					return
+					// wait for a notEmpty message
+				case <-q.notEmpty:
+				}
+			}
+
+			entry, err := q.next()
+			if err != nil {
+				log.Warningf("Error Dequeue()-ing: %s, %s", entry, err)
+				continue
+			}
+
+			select {
+			case <-q.ctx.Done():
+				return
+			case q.dequeue <- entry:
+			}
+		}
+	}()
+}
+
+// Find the next item in the queue, crawl forward if an entry is not
+// found in the next spot.
+func (q *Queue) next() (*Entry, error) {
+	q.lock.Lock()
+	defer q.lock.Unlock()
+
+	var nextKey ds.Key
+	var value []byte
+	var err error
+	for {
+		if q.head >= q.tail {
+			return nil, errors.New("no more entries in queue")
+		}
+		select {
+		case <-q.ctx.Done():
+			return nil, nil
+		default:
+		}
+		nextKey = q.queueKey(q.head)
+		value, err = q.datastore.Get(nextKey)
+		if err == ds.ErrNotFound {
+			q.head++
+			continue
+		} else if err != nil {
+			return nil, err
+		} else {
+			break
+		}
+	}
+
+	id, err := cid.Parse(value)
+	if err != nil {
+		return nil, err
+	}
+
+	entry := &Entry {
+		cid: id,
+		key: nextKey,
+		queue: q,
+	}
+
+	q.head++
+
+	return entry, nil
+}
+
+func (q *Queue) queueKey(id uint64) ds.Key {
+	return ds.NewKey(strconv.FormatUint(id, 10))
+}
+
+// crawl over the queue entries to find the head and tail
+func getQueueHeadTail(name string, ctx context.Context, datastore ds.Datastore) (uint64, uint64, error) {
+	query := query.Query{}
+	results, err := datastore.Query(query)
+	if err != nil {
+		return 0, 0, err
+	}
+
+	var tail uint64 = 0
+	var head uint64 = math.MaxUint64
+	for entry := range results.Next() {
+		select {
+		case <-ctx.Done():
+			return 0, 0, nil
+		default:
+		}
+		trimmed := strings.TrimPrefix(entry.Key, "/")
+		id, err := strconv.ParseUint(trimmed, 10, 64)
+		if err != nil {
+			return 0, 0, err
+		}
+
+		if id < head {
+			head = id
+		}
+
+		if (id+1) > tail {
+			tail = (id+1)
+		}
+	}
+	if err := results.Close(); err != nil {
+		return 0, 0, err
+	}
+	if head == math.MaxUint64 {
+		head = 0
+	}
+
+	return head, tail, nil
+}
+