Merge pull request ethereum#214 from ethersphere/swarm-mutableresourc…

…es-multihash

Infer multihash resource update content when datalength header is 0

swarm/api/http/server.go

@@ -415,22 +415,16 @@ func (s *Server) translateResourceError(w http.ResponseWriter, r *Request, supEr
 	}
 	switch code {
 	case storage.ErrInvalidValue:
-		//s.BadRequest(w, r, defaultErr.Error())
 		return http.StatusBadRequest, defaultErr
 	case storage.ErrNotFound, storage.ErrNotSynced, storage.ErrNothingToReturn:
-		//s.NotFound(w, r, defaultErr)
 		return http.StatusNotFound, defaultErr
 	case storage.ErrUnauthorized, storage.ErrInvalidSignature:
-		//ShowError(w, &r.Request, defaultErr.Error(), http.StatusUnauthorized)
 		return http.StatusUnauthorized, defaultErr
 	case storage.ErrDataOverflow:
-		//ShowError(w, &r.Request, defaultErr.Error(), http.StatusRequestEntityTooLarge)
 		return http.StatusRequestEntityTooLarge, defaultErr
 	}
 
 	return http.StatusInternalServerError, defaultErr
-
-	//s.Error(w, r, defaultErr)
 }
 
 // HandleGet handles a GET request to

swarm/storage/resource.go

@@ -3,6 +3,7 @@ package storage
 import (
 	"context"
 	"encoding/binary"
+	"errors"
 	"fmt"
 	"math/big"
 	"path/filepath"
@@ -44,8 +45,11 @@ func NewResourceError(code int, s string) error {
 		panic("no such error code!")
 	}
 	r := &ResourceError{
-		err:  s,
-		code: code,
+		err: s,
+	}
+	switch code {
+	case ErrNotFound, ErrIO, ErrUnauthorized, ErrInvalidValue, ErrDataOverflow, ErrNothingToReturn, ErrInvalidSignature, ErrNotSynced:
+		r.code = code
 	}
 	return r
 }
@@ -513,7 +517,17 @@ func (self *ResourceHandler) parseUpdate(chunkdata []byte) (*Signature, uint32,
 	namelength := int(headerlength) - cursor + 4
 	name = string(chunkdata[cursor : cursor+namelength])
 	cursor += namelength
-	intdatalength := int(datalength)
+	var intdatalength int
+	if datalength == 0 {
+		intdatalength = isMultihash(chunkdata[cursor:])
+		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, errors.New("Corrupt multihash data")
+		}
+	} else {
+		intdatalength = int(datalength)
+	}
 	data = make([]byte, intdatalength)
 	copy(data, chunkdata[cursor:cursor+intdatalength])
 
@@ -534,7 +548,19 @@ func (self *ResourceHandler) parseUpdate(chunkdata []byte) (*Signature, uint32,
 // 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 (self *ResourceHandler) UpdateMultihash(ctx context.Context, name string, data []byte) (Key, error) {
+	if isMultihash(data) == 0 {
+		return nil, NewResourceError(ErrNothingToReturn, "Invalid multihash")
+	}
+	return self.update(ctx, name, data, true)
+}
+
 func (self *ResourceHandler) Update(ctx context.Context, name string, data []byte) (Key, error) {
+	return self.update(ctx, name, data, false)
+}
+
+func (self *ResourceHandler) update(ctx context.Context, name string, data []byte, multihash bool) (Key, error) {
 
 	var signaturelength int
 	if self.validator != nil {
@@ -599,7 +625,11 @@ func (self *ResourceHandler) Update(ctx context.Context, name string, data []byt
 		}
 	}
 
-	chunk := newUpdateChunk(key, signature, nextperiod, version, name, data)
+	var datalength int
+	if !multihash {
+		datalength = len(data)
+	}
+	chunk := newUpdateChunk(key, signature, nextperiod, version, name, data, datalength)
 
 	// send the chunk
 	self.Put(chunk)
@@ -684,7 +714,7 @@ func getAddressFromDataSig(datahash common.Hash, signature Signature) (common.Ad
 }
 
 // create an update chunk
-func newUpdateChunk(key Key, signature *Signature, period uint32, version uint32, name string, data []byte) *Chunk {
+func newUpdateChunk(key Key, signature *Signature, period uint32, version uint32, name string, data []byte, datalength int) *Chunk {
 
 	// no signatures if no validator
 	var signaturelength int
@@ -695,11 +725,9 @@ func newUpdateChunk(key Key, signature *Signature, period uint32, version uint32
 	// prepend version and period to allow reverse lookups
 	headerlength := len(name) + 4 + 4
 
-	// also prepend datalength
-	datalength := len(data)
-
+	actualdatalength := len(data)
 	chunk := NewChunk(key, nil)
-	chunk.SData = make([]byte, 4+signaturelength+headerlength+datalength) // initial 4 are uint16 length descriptors for headerlength and datalength
+	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
@@ -726,7 +754,7 @@ func newUpdateChunk(key Key, signature *Signature, period uint32, version uint32
 
 	// if signature is present it's the last item in the chunk data
 	if signature != nil {
-		cursor += datalength
+		cursor += actualdatalength
 		copy(chunk.SData[cursor:], signature[:])
 	}
 
@@ -812,6 +840,31 @@ func (self *ResourceHandler) keyDataHash(key Key, data []byte) common.Hash {
 	return common.BytesToHash(hasher.Sum(nil))
 }
 
+// if first byte is the start of a multihash this function will try to parse it
+// if successful it returns the length of multihash data, 0 otherwise
+func isMultihash(data []byte) int {
+	cursor := 0
+	_, c := binary.Uvarint(data)
+	if c <= 0 {
+		log.Warn("Corrupt multihash data, hashtype is unreadable")
+		return 0
+	}
+	cursor += c
+	hashlength, c := binary.Uvarint(data[cursor:])
+	if c <= 0 {
+		log.Warn("Corrupt multihash data, hashlength is unreadable")
+		return 0
+	}
+	cursor += c
+	// we cheekily assume hashlength < maxint
+	inthashlength := int(hashlength)
+	if len(data[cursor:]) < inthashlength {
+		log.Warn("Corrupt multihash data, hash does not align with data boundary")
+		return 0
+	}
+	return cursor + inthashlength
+}
+
 // TODO: this should not be exposed, but swarm/testutil/http.go needs it
 func NewTestResourceHandler(datadir string, ethClient ethApi, validator ResourceValidator) (*ResourceHandler, error) {
 	path := filepath.Join(datadir, DbDirName)

swarm/storage/resource_test.go

@@ -15,6 +15,8 @@ import (
 	"testing"
 	"time"
 
+	"github.com/multiformats/go-multihash"
+
 	"github.com/ethereum/go-ethereum/accounts/abi/bind"
 	"github.com/ethereum/go-ethereum/accounts/abi/bind/backends"
 	"github.com/ethereum/go-ethereum/common"
@@ -107,7 +109,7 @@ func TestResourceReverse(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	chunk := newUpdateChunk(key, &sig, period, version, safeName, data)
+	chunk := newUpdateChunk(key, &sig, period, version, safeName, data, len(data))
 
 	// check that we can recover the owner account from the update chunk's signature
 	checksig, checkperiod, checkversion, checkname, checkdata, err := rh.parseUpdate(chunk.SData)
@@ -262,6 +264,133 @@ func TestResourceHandler(t *testing.T) {
 
 }
 
+func TestResourceMultihash(t *testing.T) {
+
+	// signer containing private key
+	signer, err := newTestSigner()
+	if err != nil {
+		t.Fatal(err)
+	}
+	validator := newTestValidator(signer.signContent)
+
+	// make fake backend, set up rpc and create resourcehandler
+	backend := &fakeBackend{
+		blocknumber: int64(startBlock),
+	}
+
+	// set up rpc and create resourcehandler
+	rh, datadir, _, teardownTest, err := setupTest(backend, nil)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer teardownTest()
+
+	// create a new resource
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	_, err = rh.NewResource(ctx, safeName, resourceFrequency)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// we're naïvely assuming keccak256 for swarm hashes
+	// if it ever changes this test should also change
+	swarmhashbytes := rh.nameHash("foo")
+	swarmhashmulti, err := multihash.Encode(swarmhashbytes.Bytes(), multihash.KECCAK_256)
+	if err != nil {
+		t.Fatal(err)
+	}
+	swarmhashkey, err := rh.UpdateMultihash(ctx, safeName, swarmhashmulti)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	sha1bytes := make([]byte, multihash.DefaultLengths[multihash.SHA1])
+	sha1multi, err := multihash.Encode(sha1bytes, multihash.SHA1)
+	if err != nil {
+		t.Fatal(err)
+	}
+	sha1key, err := rh.UpdateMultihash(ctx, safeName, sha1multi)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// invalid multihashes
+	_, err = rh.UpdateMultihash(ctx, safeName, swarmhashmulti[1:])
+	if err == nil {
+		t.Fatalf("Expected update to fail with first byte skipped")
+	}
+	_, err = rh.UpdateMultihash(ctx, safeName, swarmhashmulti[:len(swarmhashmulti)-2])
+	if err == nil {
+		t.Fatalf("Expected update to fail with last byte skipped")
+	}
+
+	data, err := getUpdateDirect(rh, swarmhashkey)
+	if err != nil {
+		t.Fatal(err)
+	}
+	swarmhashdecode, err := multihash.Decode(data)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !bytes.Equal(swarmhashdecode.Digest, swarmhashbytes.Bytes()) {
+		t.Fatalf("Decoded SHA1 hash '%x' does not match original hash '%x'", swarmhashdecode.Digest, swarmhashbytes.Bytes())
+	}
+	data, err = getUpdateDirect(rh, sha1key)
+	if err != nil {
+		t.Fatal(err)
+	}
+	sha1decode, err := multihash.Decode(data)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !bytes.Equal(sha1decode.Digest, sha1bytes) {
+		t.Fatalf("Decoded SHA1 hash '%x' does not match original hash '%x'", sha1decode.Digest, sha1bytes)
+	}
+	rh.Close()
+
+	// test with signed data
+	rh2, err := NewTestResourceHandler(datadir, rh.ethClient, validator)
+	if err != nil {
+		t.Fatal(err)
+	}
+	_, err = rh2.NewResource(ctx, safeName, resourceFrequency)
+	if err != nil {
+		t.Fatal(err)
+	}
+	swarmhashsignedkey, err := rh2.UpdateMultihash(ctx, safeName, swarmhashmulti)
+	if err != nil {
+		t.Fatal(err)
+	}
+	sha1signedkey, err := rh2.UpdateMultihash(ctx, safeName, sha1multi)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	data, err = getUpdateDirect(rh2, swarmhashsignedkey)
+	if err != nil {
+		t.Fatal(err)
+	}
+	swarmhashdecode, err = multihash.Decode(data)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !bytes.Equal(swarmhashdecode.Digest, swarmhashbytes.Bytes()) {
+		t.Fatalf("Decoded SHA1 hash '%x' does not match original hash '%x'", swarmhashdecode.Digest, swarmhashbytes.Bytes())
+	}
+	data, err = getUpdateDirect(rh2, sha1signedkey)
+	if err != nil {
+		t.Fatal(err)
+	}
+	sha1decode, err = multihash.Decode(data)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !bytes.Equal(sha1decode.Digest, sha1bytes) {
+		t.Fatalf("Decoded SHA1 hash '%x' does not match original hash '%x'", sha1decode.Digest, sha1bytes)
+	}
+}
+
 // create ENS enabled resource update, with and without valid owner
 func TestResourceENSOwner(t *testing.T) {
 
@@ -444,3 +573,15 @@ func (self *testValidator) checkAccess(name string, address common.Address) (boo
 func (self *testValidator) nameHash(name string) common.Hash {
 	return self.hashFunc(name)
 }
+
+func getUpdateDirect(rh *ResourceHandler, key Key) ([]byte, error) {
+	chunk, err := rh.ChunkStore.(*resourceChunkStore).localStore.(*LocalStore).memStore.Get(key)
+	if err != nil {
+		return nil, err
+	}
+	_, _, _, _, data, err := rh.parseUpdate(chunk.SData)
+	if err != nil {
+		return nil, err
+	}
+	return data, nil
+}