etcd/etcdserver/apply_auth.go

245 lines
6.4 KiB
Go
Raw Normal View History

// Copyright 2016 The etcd Authors
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package etcdserver
import (
"context"
"sync"
"go.etcd.io/etcd/auth"
pb "go.etcd.io/etcd/etcdserver/etcdserverpb"
"go.etcd.io/etcd/lease"
"go.etcd.io/etcd/mvcc"
2019-10-02 01:38:52 +03:00
"go.etcd.io/etcd/pkg/traceutil"
)
type authApplierV3 struct {
applierV3
as auth.AuthStore
lessor lease.Lessor
// mu serializes Apply so that user isn't corrupted and so that
// serialized requests don't leak data from TOCTOU errors
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
mu sync.Mutex
authInfo auth.AuthInfo
}
func newAuthApplierV3(as auth.AuthStore, base applierV3, lessor lease.Lessor) *authApplierV3 {
return &authApplierV3{applierV3: base, as: as, lessor: lessor}
}
func (aa *authApplierV3) Apply(r *pb.InternalRaftRequest) *applyResult {
aa.mu.Lock()
defer aa.mu.Unlock()
if r.Header != nil {
// backward-compatible with pre-3.0 releases when internalRaftRequest
// does not have header field
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
aa.authInfo.Username = r.Header.Username
aa.authInfo.Revision = r.Header.AuthRevision
}
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
if needAdminPermission(r) {
if err := aa.as.IsAdminPermitted(&aa.authInfo); err != nil {
aa.authInfo.Username = ""
aa.authInfo.Revision = 0
return &applyResult{err: err}
}
}
ret := aa.applierV3.Apply(r)
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
aa.authInfo.Username = ""
aa.authInfo.Revision = 0
return ret
}
2019-10-02 01:38:52 +03:00
func (aa *authApplierV3) Put(txn mvcc.TxnWrite, r *pb.PutRequest) (*pb.PutResponse, *traceutil.Trace, error) {
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
if err := aa.as.IsPutPermitted(&aa.authInfo, r.Key); err != nil {
2019-10-02 01:38:52 +03:00
return nil, nil, err
}
if err := aa.checkLeasePuts(lease.LeaseID(r.Lease)); err != nil {
// The specified lease is already attached with a key that cannot
// be written by this user. It means the user cannot revoke the
// lease so attaching the lease to the newly written key should
// be forbidden.
2019-10-02 01:38:52 +03:00
return nil, nil, err
}
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
if r.PrevKv {
err := aa.as.IsRangePermitted(&aa.authInfo, r.Key, nil)
if err != nil {
2019-10-02 01:38:52 +03:00
return nil, nil, err
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
}
2016-07-06 03:08:09 +03:00
}
2017-01-05 12:13:47 +03:00
return aa.applierV3.Put(txn, r)
}
func (aa *authApplierV3) Range(ctx context.Context, txn mvcc.TxnRead, r *pb.RangeRequest) (*pb.RangeResponse, error) {
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
if err := aa.as.IsRangePermitted(&aa.authInfo, r.Key, r.RangeEnd); err != nil {
return nil, err
}
return aa.applierV3.Range(ctx, txn, r)
}
2017-01-05 12:13:47 +03:00
func (aa *authApplierV3) DeleteRange(txn mvcc.TxnWrite, r *pb.DeleteRangeRequest) (*pb.DeleteRangeResponse, error) {
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
if err := aa.as.IsDeleteRangePermitted(&aa.authInfo, r.Key, r.RangeEnd); err != nil {
return nil, err
}
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
if r.PrevKv {
err := aa.as.IsRangePermitted(&aa.authInfo, r.Key, r.RangeEnd)
if err != nil {
return nil, err
}
2016-07-01 07:52:57 +03:00
}
2017-01-05 12:13:47 +03:00
return aa.applierV3.DeleteRange(txn, r)
}
func checkTxnReqsPermission(as auth.AuthStore, ai *auth.AuthInfo, reqs []*pb.RequestOp) error {
for _, requ := range reqs {
switch tv := requ.Request.(type) {
case *pb.RequestOp_RequestRange:
if tv.RequestRange == nil {
continue
}
if err := as.IsRangePermitted(ai, tv.RequestRange.Key, tv.RequestRange.RangeEnd); err != nil {
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
return err
}
case *pb.RequestOp_RequestPut:
if tv.RequestPut == nil {
continue
}
if err := as.IsPutPermitted(ai, tv.RequestPut.Key); err != nil {
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
return err
}
case *pb.RequestOp_RequestDeleteRange:
if tv.RequestDeleteRange == nil {
continue
}
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
if tv.RequestDeleteRange.PrevKv {
err := as.IsRangePermitted(ai, tv.RequestDeleteRange.Key, tv.RequestDeleteRange.RangeEnd)
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
if err != nil {
return err
}
2016-07-01 07:52:57 +03:00
}
err := as.IsDeleteRangePermitted(ai, tv.RequestDeleteRange.Key, tv.RequestDeleteRange.RangeEnd)
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
if err != nil {
return err
}
}
}
auth, etcdserver: introduce revision of authStore for avoiding TOCTOU problem This commit introduces revision of authStore. The revision number represents a version of authStore that is incremented by updating auth related information. The revision is required for avoiding TOCTOU problems. Currently there are two types of the TOCTOU problems in v3 auth. The first one is in ordinal linearizable requests with a sequence like below (): 1. Request from client CA is processed in follower FA. FA looks up the username (let it U) for the request from a token of the request. At this time, the request is authorized correctly. 2. Another request from client CB is processed in follower FB. CB is for changing U's password. 3. FB forwards the request from CB to the leader before FA. Now U's password is updated and the request from CA should be rejected. 4. However, the request from CA is processed by the leader because authentication is already done in FA. For avoiding the above sequence, this commit lets etcdserverpb.RequestHeader have a member revision. The member is initialized during authentication by followers and checked in a leader. If the revision in RequestHeader is lower than the leader's authStore revision, it means a sequence like above happened. In such a case, the state machine returns auth.ErrAuthRevisionObsolete. The error code lets nodes retry their requests. The second one, a case of serializable range and txn, is more subtle. Because these requests are processed in follower directly. The TOCTOU problem can be caused by a sequence like below: 1. Serializable request from client CA is processed in follower FA. At first, FA looks up the username (let it U) and its permission before actual access to KV. 2. Another request from client CB is processed in follower FB and forwarded to the leader. The cluster including FA now commits a log entry of the request from CB. Assume the request changed the permission or password of U. 3. Now the serializable request from CA is accessing to KV. Even if the access is allowed at the point of 1, now it can be invalid because of the change introduced in 2. For avoiding the above sequence, this commit lets the functions of serializable requests (EtcdServer.Range() and EtcdServer.Txn()) compare the revision in the request header with the latest revision of authStore after the actual access. If the saved revision is lower than the latest one, it means the permission can be changed. Although it would introduce false positives (e.g. changing other user's password), it prevents the TOCTOU problem. This idea is an implementation of Anthony's comment: https://github.com/coreos/etcd/pull/5739#issuecomment-228128254
2016-06-23 12:31:12 +03:00
return nil
}
func checkTxnAuth(as auth.AuthStore, ai *auth.AuthInfo, rt *pb.TxnRequest) error {
for _, c := range rt.Compare {
if err := as.IsRangePermitted(ai, c.Key, c.RangeEnd); err != nil {
return err
}
}
if err := checkTxnReqsPermission(as, ai, rt.Success); err != nil {
return err
}
return checkTxnReqsPermission(as, ai, rt.Failure)
}
func (aa *authApplierV3) Txn(rt *pb.TxnRequest) (*pb.TxnResponse, error) {
if err := checkTxnAuth(aa.as, &aa.authInfo, rt); err != nil {
return nil, err
}
return aa.applierV3.Txn(rt)
}
func (aa *authApplierV3) LeaseRevoke(lc *pb.LeaseRevokeRequest) (*pb.LeaseRevokeResponse, error) {
if err := aa.checkLeasePuts(lease.LeaseID(lc.ID)); err != nil {
return nil, err
}
return aa.applierV3.LeaseRevoke(lc)
}
func (aa *authApplierV3) checkLeasePuts(leaseID lease.LeaseID) error {
lease := aa.lessor.Lookup(leaseID)
if lease != nil {
for _, key := range lease.Keys() {
if err := aa.as.IsPutPermitted(&aa.authInfo, []byte(key)); err != nil {
return err
}
}
}
return nil
}
func (aa *authApplierV3) UserGet(r *pb.AuthUserGetRequest) (*pb.AuthUserGetResponse, error) {
err := aa.as.IsAdminPermitted(&aa.authInfo)
if err != nil && r.Name != aa.authInfo.Username {
aa.authInfo.Username = ""
aa.authInfo.Revision = 0
return &pb.AuthUserGetResponse{}, err
}
return aa.applierV3.UserGet(r)
}
func (aa *authApplierV3) RoleGet(r *pb.AuthRoleGetRequest) (*pb.AuthRoleGetResponse, error) {
err := aa.as.IsAdminPermitted(&aa.authInfo)
if err != nil && !aa.as.HasRole(aa.authInfo.Username, r.Role) {
aa.authInfo.Username = ""
aa.authInfo.Revision = 0
return &pb.AuthRoleGetResponse{}, err
}
return aa.applierV3.RoleGet(r)
}
func needAdminPermission(r *pb.InternalRaftRequest) bool {
switch {
case r.AuthEnable != nil:
return true
case r.AuthDisable != nil:
return true
case r.AuthUserAdd != nil:
return true
case r.AuthUserDelete != nil:
return true
case r.AuthUserChangePassword != nil:
return true
case r.AuthUserGrantRole != nil:
return true
case r.AuthUserRevokeRole != nil:
return true
case r.AuthRoleAdd != nil:
return true
case r.AuthRoleGrantPermission != nil:
return true
case r.AuthRoleRevokePermission != nil:
return true
case r.AuthRoleDelete != nil:
return true
case r.AuthUserList != nil:
return true
case r.AuthRoleList != nil:
return true
default:
return false
}
}