open-consul/agent/consul/intention_endpoint.go

368 lines
10 KiB
Go
Raw Normal View History

package consul
import (
2018-02-28 23:54:48 +00:00
"errors"
"fmt"
"time"
"github.com/armon/go-metrics"
"github.com/hashicorp/consul/acl"
"github.com/hashicorp/consul/agent/connect"
"github.com/hashicorp/consul/agent/consul/state"
"github.com/hashicorp/consul/agent/structs"
"github.com/hashicorp/go-memdb"
"github.com/hashicorp/go-uuid"
)
2018-02-28 23:54:48 +00:00
var (
// ErrIntentionNotFound is returned if the intention lookup failed.
ErrIntentionNotFound = errors.New("Intention not found")
)
// Intention manages the Connect intentions.
type Intention struct {
// srv is a pointer back to the server.
srv *Server
}
// Apply creates or updates an intention in the data store.
func (s *Intention) Apply(
args *structs.IntentionRequest,
reply *string) error {
// Forward this request to the primary DC if we're a secondary that's replicating intentions.
if s.srv.intentionReplicationEnabled() {
args.Datacenter = s.srv.config.PrimaryDatacenter
}
if done, err := s.srv.forward("Intention.Apply", args, args, reply); done {
return err
}
defer metrics.MeasureSince([]string{"consul", "intention", "apply"}, time.Now())
defer metrics.MeasureSince([]string{"intention", "apply"}, time.Now())
// Always set a non-nil intention to avoid nil-access below
if args.Intention == nil {
args.Intention = &structs.Intention{}
}
// If no ID is provided, generate a new ID. This must be done prior to
// appending to the Raft log, because the ID is not deterministic. Once
// the entry is in the log, the state update MUST be deterministic or
// the followers will not converge.
if args.Op == structs.IntentionOpCreate {
if args.Intention.ID != "" {
return fmt.Errorf("ID must be empty when creating a new intention")
}
state := s.srv.fsm.State()
for {
var err error
args.Intention.ID, err = uuid.GenerateUUID()
if err != nil {
s.srv.logger.Printf("[ERR] consul.intention: UUID generation failed: %v", err)
return err
}
_, ixn, err := state.IntentionGet(nil, args.Intention.ID)
if err != nil {
s.srv.logger.Printf("[ERR] consul.intention: intention lookup failed: %v", err)
return err
}
if ixn == nil {
break
}
}
// Set the created at
args.Intention.CreatedAt = time.Now().UTC()
}
*reply = args.Intention.ID
// Get the ACL token for the request for the checks below.
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
rule, err := s.srv.ResolveToken(args.Token)
if err != nil {
return err
}
// Perform the ACL check
if prefix, ok := args.Intention.GetACLPrefix(); ok {
if rule != nil && !rule.IntentionWrite(prefix) {
s.srv.logger.Printf("[WARN] consul.intention: Operation on intention '%s' denied due to ACLs", args.Intention.ID)
return acl.ErrPermissionDenied
}
}
// If this is not a create, then we have to verify the ID.
if args.Op != structs.IntentionOpCreate {
state := s.srv.fsm.State()
_, ixn, err := state.IntentionGet(nil, args.Intention.ID)
if err != nil {
return fmt.Errorf("Intention lookup failed: %v", err)
}
if ixn == nil {
return fmt.Errorf("Cannot modify non-existent intention: '%s'", args.Intention.ID)
}
// Perform the ACL check that we have write to the old prefix too,
// which must be true to perform any rename.
if prefix, ok := ixn.GetACLPrefix(); ok {
if rule != nil && !rule.IntentionWrite(prefix) {
s.srv.logger.Printf("[WARN] consul.intention: Operation on intention '%s' denied due to ACLs", args.Intention.ID)
return acl.ErrPermissionDenied
}
}
}
// We always update the updatedat field. This has no effect for deletion.
args.Intention.UpdatedAt = time.Now().UTC()
// Default source type
if args.Intention.SourceType == "" {
args.Intention.SourceType = structs.IntentionSourceConsul
}
// Until we support namespaces, we force all namespaces to be default
if args.Intention.SourceNS == "" {
args.Intention.SourceNS = structs.IntentionDefaultNamespace
}
if args.Intention.DestinationNS == "" {
args.Intention.DestinationNS = structs.IntentionDefaultNamespace
}
// Validate. We do not validate on delete since it is valid to only
// send an ID in that case.
if args.Op != structs.IntentionOpDelete {
// Set the precedence
args.Intention.UpdatePrecedence()
if err := args.Intention.Validate(); err != nil {
return err
}
2018-03-03 17:43:37 +00:00
}
// make sure we set the hash prior to raft application
args.Intention.SetHash(true)
// Commit
resp, err := s.srv.raftApply(structs.IntentionRequestType, args)
if err != nil {
s.srv.logger.Printf("[ERR] consul.intention: Apply failed %v", err)
return err
}
if respErr, ok := resp.(error); ok {
return respErr
}
return nil
}
2018-02-28 18:44:49 +00:00
// Get returns a single intention by ID.
func (s *Intention) Get(
args *structs.IntentionQueryRequest,
reply *structs.IndexedIntentions) error {
// Forward if necessary
if done, err := s.srv.forward("Intention.Get", args, args, reply); done {
return err
}
return s.srv.blockingQuery(
&args.QueryOptions,
&reply.QueryMeta,
func(ws memdb.WatchSet, state *state.Store) error {
index, ixn, err := state.IntentionGet(ws, args.IntentionID)
if err != nil {
return err
}
if ixn == nil {
2018-02-28 23:54:48 +00:00
return ErrIntentionNotFound
2018-02-28 18:44:49 +00:00
}
reply.Index = index
reply.Intentions = structs.Intentions{ixn}
2018-03-04 19:53:52 +00:00
// Filter
if err := s.srv.filterACL(args.Token, reply); err != nil {
return err
}
// If ACLs prevented any responses, error
if len(reply.Intentions) == 0 {
s.srv.logger.Printf("[WARN] consul.intention: Request to get intention '%s' denied due to ACLs", args.IntentionID)
return acl.ErrPermissionDenied
}
2018-02-28 18:44:49 +00:00
return nil
},
)
}
// List returns all the intentions.
func (s *Intention) List(
args *structs.DCSpecificRequest,
reply *structs.IndexedIntentions) error {
// Forward if necessary
if done, err := s.srv.forward("Intention.List", args, args, reply); done {
return err
}
return s.srv.blockingQuery(
&args.QueryOptions, &reply.QueryMeta,
func(ws memdb.WatchSet, state *state.Store) error {
index, ixns, err := state.Intentions(ws)
if err != nil {
return err
}
reply.Index, reply.Intentions = index, ixns
if reply.Intentions == nil {
reply.Intentions = make(structs.Intentions, 0)
}
2018-03-05 02:32:28 +00:00
return s.srv.filterACL(args.Token, reply)
},
)
}
// Match returns the set of intentions that match the given source/destination.
func (s *Intention) Match(
args *structs.IntentionQueryRequest,
reply *structs.IndexedIntentionMatches) error {
// Forward if necessary
if done, err := s.srv.forward("Intention.Match", args, args, reply); done {
return err
}
2018-03-05 02:32:28 +00:00
// Get the ACL token for the request for the checks below.
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
rule, err := s.srv.ResolveToken(args.Token)
2018-03-05 02:32:28 +00:00
if err != nil {
return err
}
if rule != nil {
// We go through each entry and test the destination to check if it
// matches.
for _, entry := range args.Match.Entries {
if prefix := entry.Name; prefix != "" && !rule.IntentionRead(prefix) {
s.srv.logger.Printf("[WARN] consul.intention: Operation on intention prefix '%s' denied due to ACLs", prefix)
return acl.ErrPermissionDenied
}
}
}
return s.srv.blockingQuery(
&args.QueryOptions,
&reply.QueryMeta,
func(ws memdb.WatchSet, state *state.Store) error {
index, matches, err := state.IntentionMatch(ws, args.Match)
if err != nil {
return err
}
reply.Index = index
reply.Matches = matches
return nil
},
)
}
2018-05-11 16:19:22 +00:00
// Check tests a source/destination and returns whether it would be allowed
// or denied based on the current ACL configuration.
//
// Note: Whenever the logic for this method is changed, you should take
// a look at the agent authorize endpoint (agent/agent_endpoint.go) since
// the logic there is similar.
2018-05-11 16:19:22 +00:00
func (s *Intention) Check(
args *structs.IntentionQueryRequest,
2018-05-11 16:19:22 +00:00
reply *structs.IntentionQueryCheckResponse) error {
// Forward maybe
2018-05-11 16:19:22 +00:00
if done, err := s.srv.forward("Intention.Check", args, args, reply); done {
return err
}
// Get the test args, and defensively guard against nil
2018-05-11 16:19:22 +00:00
query := args.Check
if query == nil {
2018-05-11 16:19:22 +00:00
return errors.New("Check must be specified on args")
}
// Build the URI
var uri connect.CertURI
switch query.SourceType {
case structs.IntentionSourceConsul:
uri = &connect.SpiffeIDService{
Namespace: query.SourceNS,
Service: query.SourceName,
}
default:
return fmt.Errorf("unsupported SourceType: %q", query.SourceType)
}
// Get the ACL token for the request for the checks below.
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
rule, err := s.srv.ResolveToken(args.Token)
if err != nil {
return err
}
2018-05-19 04:03:10 +00:00
// Perform the ACL check. For Check we only require ServiceRead and
// NOT IntentionRead because the Check API only returns pass/fail and
// returns no other information about the intentions used.
if prefix, ok := query.GetACLPrefix(); ok {
if rule != nil && !rule.ServiceRead(prefix) {
s.srv.logger.Printf("[WARN] consul.intention: test on intention '%s' denied due to ACLs", prefix)
return acl.ErrPermissionDenied
}
}
// Get the matches for this destination
state := s.srv.fsm.State()
_, matches, err := state.IntentionMatch(nil, &structs.IntentionQueryMatch{
Type: structs.IntentionMatchDestination,
Entries: []structs.IntentionMatchEntry{
structs.IntentionMatchEntry{
Namespace: query.DestinationNS,
Name: query.DestinationName,
},
},
})
if err != nil {
return err
}
if len(matches) != 1 {
// This should never happen since the documented behavior of the
// Match call is that it'll always return exactly the number of results
// as entries passed in. But we guard against misbehavior.
return errors.New("internal error loading matches")
}
2018-05-11 16:19:22 +00:00
// Check the authorization for each match
for _, ixn := range matches[0] {
if auth, ok := uri.Authorize(ixn); ok {
reply.Allowed = auth
return nil
}
}
// No match, we need to determine the default behavior. We do this by
// specifying the anonymous token token, which will get that behavior.
// The default behavior if ACLs are disabled is to allow connections
// to mimic the behavior of Consul itself: everything is allowed if
// ACLs are disabled.
//
// NOTE(mitchellh): This is the same behavior as the agent authorize
// endpoint. If this behavior is incorrect, we should also change it there
// which is much more important.
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
rule, err = s.srv.ResolveToken("")
if err != nil {
return err
}
reply.Allowed = true
if rule != nil {
reply.Allowed = rule.IntentionDefaultAllow()
}
return nil
}