EC Audit: N2: storage node panics when another node answers an EC part request with a linker header of huge payload length

Abierto
#4,329 2 comentarios 0 reacciones 0 asignados Ver en GitHub

Los mantenedores suelen responder en 1 día

Nadie ha tomado este issue todavía.

Evaluación

Dificultad
3/5
Tiempo estimado
1-2 días
Aptitud para principiantes
64/100
Tipo de issue
Error
Claridad
Bien especificado
Estado de actividad
Activo
Stack tecnológico
go

Línea de trabajo

Start with the linker allocation sites in pkg/services/object/get/ec.go at the full GET and ranged GET paths, then read InitGetObjectStream in pkg/services/object/get/util.go to understand the unverified header. Add or adapt a regression test in pkg/services/object/get/ using a LINK header with an oversized declared payload. Done means both GET paths reject or safely bound the payload without panicking or allocating based on its declared length.

Escrito por el modelo de indexación a partir del texto del issue.

Descripción

bug hal:proposed neofs-storage

Expected Behavior

A header received from another node cannot make the node allocate an arbitrary amount of memory. A bad reply fails that one request.

Current Behavior

When the reply to an EC part request is an object of type LINK, the node allocates a buffer of the payload length declared in the received header, before any check of that header:

The header comes from InitGetObjectStream, which asks the SDK client to skip verification (get/util.go#L330-L332). The only check before the allocation is that the parent ID field equals the requested ID (get/ec.go#L676-L683). verifyParentHeaderID from #4290 is placed after the linker branch on purpose and does not apply to it.

With PayloadSize = 1<<62 in the reply, make panics with makeslice: len out of range. On the full GET path this happens in an errgroup goroutine; nothing in the node recovers panics, so the process exits. Smaller values that pass make allocate that much memory.

The reply has to come from a node of the container, that is, from a faulty or hostile storage node. In the gRPC handler the full GET first goes through the streaming transport, which reports a LINK first part as ErrLinker and then falls back to this code (get/ec.go#L197-L203).

For non-linker parts the allocation at get/ec.go#L619-L627 is now bounded by the payload size of a parent header that hashes to the requested ID, which is fine for real objects.

Possible Solution

Bound the linker's payload length before allocating: a linker of an object cannot be longer than the network's MaxObjectSize, and a tighter bound follows from the number of children. Reading through io.LimitReader into a growing buffer avoids trusting the declared length at all.

Steps to Reproduce (for bugs)

  1. Save the test below as pkg/services/object/get/ecaudit_test.go.
  2. go test ./pkg/services/object/get/ -run 'TestECAuditN2/ranged' -v passes: it asserts the panic.
  3. go test ./pkg/services/object/get/ -run 'TestECAuditN2/full' -v kills the test binary:
panic: runtime error: makeslice: len out of range
goroutine 14 [running]:
	…/pkg/services/object/get/ec.go:599 +0x228
	…/pkg/services/object/get/ec.go:404 +0xb0
ecaudit_test.go
package getsvc

import (
	"bytes"
	"context"
	"crypto/ecdsa"
	"io"
	"testing"

	iec "github.com/nspcc-dev/neofs-node/internal/ec"
	"github.com/nspcc-dev/neofs-node/internal/testutil"
	cid "github.com/nspcc-dev/neofs-sdk-go/container/id"
	cidtest "github.com/nspcc-dev/neofs-sdk-go/container/id/test"
	neofscryptotest "github.com/nspcc-dev/neofs-sdk-go/crypto/test"
	"github.com/nspcc-dev/neofs-sdk-go/netmap"
	"github.com/nspcc-dev/neofs-sdk-go/object"
	oid "github.com/nspcc-dev/neofs-sdk-go/object/id"
	usertest "github.com/nspcc-dev/neofs-sdk-go/user/test"
	"github.com/stretchr/testify/require"
)

type ecAuditScene struct {
	cnr           cid.ID
	rule          iec.Rule
	parentHdr     object.Object
	parentAddr    oid.Address
	parentPayload []byte
	partObjs      []object.Object
	nodeSvcs      []*Service
	nodeLists     [][]netmap.NodeInfo
	getNodesMap   map[oid.Address]getNodesForObjectValue
	tc            *testECServiceConn
}

func newECAuditScene(t *testing.T, rule iec.Rule, pldLen int) *ecAuditScene {
	s := &ecAuditScene{cnr: cidtest.ID(), rule: rule, parentPayload: testutil.RandByteSlice(pldLen)}
	s.parentHdr.SetContainerID(s.cnr)
	s.parentHdr.SetPayloadSize(uint64(pldLen))
	require.NoError(t, s.parentHdr.CalculateAndSetID())
	s.parentAddr = oid.NewAddress(s.cnr, s.parentHdr.GetID())

	parts, _, err := iec.Encode(rule, s.parentPayload)
	require.NoError(t, err)
	s.partObjs = make([]object.Object, len(parts))
	for i := range parts {
		s.partObjs[i], err = iec.FormObjectForECPart(usertest.User(), s.parentHdr, parts[i], iec.PartInfo{Index: i})
		require.NoError(t, err)
	}

	s.nodeLists, _ = testNodeMatrix(t, []int{len(parts)})
	s.getNodesMap = map[oid.Address]getNodesForObjectValue{
		s.parentAddr: {nodeSets: s.nodeLists, ecRules: []iec.Rule{rule}},
	}
	s.tc = &testECServiceConn{
		mockKeyStorage: mockKeyStorage{privKey: neofscryptotest.ECDSAPrivateKey()},
		parentHdr:      s.parentHdr,
		nodes:          make(map[string]*Service),
	}
	s.nodeSvcs = make([]*Service, len(parts))
	for i := range s.nodeSvcs {
		s.nodeSvcs[i] = New(&mockNeoFSNet{getNodesForObject: s.getNodesMap, localPubKey: s.nodeLists[0][i].PublicKey()})
		s.setPart(i, s.partObjs[i].Payload())
		s.tc.nodes[string(s.nodeLists[0][i].Marshal())] = s.nodeSvcs[i]
	}
	return s
}

// setPart makes node #i serve pld as payload of part #i (nil = part is missing).
func (s *ecAuditScene) setPart(i int, pld []byte) {
	m := map[getECPartKey]getECPartValue{}
	if pld != nil {
		m[getECPartKey{cnr: s.cnr, parent: s.parentHdr.GetID(), pi: iec.PartInfo{Index: i}}] = getECPartValue{
			hdr: *s.partObjs[i].CutPayload(), pld: pld, rdr: io.NopCloser(bytes.NewReader(pld)),
		}
	}
	s.nodeSvcs[i].localObjects = &mockLocalObjects{getECPart: m}
}

func (s *ecAuditScene) service() *Service {
	svc := New(&mockNeoFSNet{getNodesForObject: s.getNodesMap})
	svc.keyStore = s.tc
	svc.conns = s.tc
	return svc
}

// maliciousLinkerConn answers every part request with a header of a LINK
// object declaring a huge payload.
type maliciousLinkerConn struct {
	*testECServiceConn
	parent object.Object
	size   uint64
}

func (x *maliciousLinkerConn) InitGetObjectStream(_ context.Context, _ netmap.NodeInfo, _ ecdsa.PrivateKey,
	_ cid.ID, id oid.ID, _, _ bool, _ *object.Range, _ iec.PartInfo) (object.Object, io.ReadCloser, error) {
	var hdr object.Object
	hdr.SetType(object.TypeLink)
	hdr.SetParentID(id)
	hdr.SetParent(&x.parent)
	hdr.SetPayloadSize(x.size)
	return hdr, io.NopCloser(bytes.NewReader(nil)), nil
}

func TestECAuditN2_PeerControlledAllocation(t *testing.T) {
	ctx := context.Background()
	s := newECAuditScene(t, iec.Rule{DataPartNum: 3, ParityPartNum: 2}, 4096)

	svc := s.service()
	svc.conns = &maliciousLinkerConn{testECServiceConn: s.tc, parent: s.parentHdr, size: 1 << 62}

	t.Run("ranged GET", func(t *testing.T) {
		var w mockObjectWriter
		var prm Prm
		prm.WithAddress(s.parentAddr)
		prm.SetCommonParameters(newCommonParameters(false))
		prm.SetObjectWriter(&w)
		prm.MarkPayloadOnly()
		rng := object.NewRange()
		rng.SetOffset(1)
		rng.SetLength(10)
		prm.SetRange(rng)

		require.PanicsWithError(t, "runtime error: makeslice: len out of range", func() { _ = svc.Get(ctx, prm) })
	})

	t.Run("full GET", func(t *testing.T) {
		var w mockObjectWriter
		var prm Prm
		prm.WithAddress(s.parentAddr)
		prm.SetCommonParameters(newCommonParameters(false))
		prm.SetObjectWriter(&w)

		_ = svc.Get(ctx, prm) // panics inside an errgroup goroutine
	})
}

Context

Found by an audit of the erasure coding code. The audit was static: code reading plus unit-level probe tests in the packages concerned. No cluster was run, so there is no end-to-end reproduction.

Regression

Not known.

Your Environment

  • Version used: 0.58.2 (1f0d51b5d)
  • Server setup and configuration: none, unit tests only
  • Operating System and version (uname -a): Darwin 25.6.0 arm64, Go 1.27.3
Lenguaje dominante
Go
Estrellas
37
Forks
56
Merge medio
1 d 21 h
PR fusionados (30 d)
46

Preparar el entorno

Primeros pasos

  1. Lee el issue completo y luego la guía de contribución del proyecto.
  2. Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
  3. Haz un fork del repositorio y trabaja en una rama.
  4. Abre un pull request que haga referencia al número del issue.

Más de nspcc-dev/neofs-node

Todos los issues de nspcc-dev/neofs-node

Issues similares

Más issues de Go

Recibe los nuevos issues en tu correo

Un resumen breve de issues de GitHub para principiantes.