open-consul/agent/acl.go

337 lines
10 KiB
Go
Raw Normal View History

package agent
import (
"fmt"
2020-11-17 22:10:21 +00:00
"github.com/hashicorp/serf/serf"
"github.com/hashicorp/consul/acl"
"github.com/hashicorp/consul/agent/structs"
)
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
// resolveToken is the primary interface used by ACL-checkers in the agent
// endpoints, which is the one place where we do some ACL enforcement on
// clients. Some of the enforcement is normative (e.g. self and monitor)
// and some is informative (e.g. catalog and health).
func (a *Agent) resolveToken(id string) (acl.Authorizer, error) {
return a.resolveTokenAndDefaultMeta(id, nil, nil)
}
// resolveTokenAndDefaultMeta is used to resolve an ACL token secret to an
// acl.Authorizer and to default any enterprise specific metadata for the request.
// The defaulted metadata is then used to fill in an acl.AuthorizationContext.
func (a *Agent) resolveTokenAndDefaultMeta(id string, entMeta *structs.EnterpriseMeta, authzContext *acl.AuthorizerContext) (acl.Authorizer, error) {
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
// ACLs are disabled
if !a.config.ACLsEnabled {
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
return nil, nil
}
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
if acl.RootAuthorizer(id) != nil {
return nil, acl.ErrRootDenied
}
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
if a.tokens.IsAgentMasterToken(id) {
return a.aclMasterAuthorizer, nil
}
return a.delegate.ResolveTokenAndDefaultMeta(id, entMeta, authzContext)
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
}
// aclAccessorID is used to convert an ACLToken's secretID to its accessorID for non-
// critical purposes, such as logging. Therefore we interpret all errors as empty-string
// so we can safely log it without handling non-critical errors at the usage site.
func (a *Agent) aclAccessorID(secretID string) string {
2020-11-17 22:10:21 +00:00
ident, err := a.delegate.ResolveTokenToIdentity(secretID)
if acl.IsErrNotFound(err) {
return ""
}
if err != nil {
a.logger.Debug("non-critical error resolving acl token accessor for logging", "error", err)
return ""
}
if ident == nil {
return ""
}
return ident.ID()
}
func initializeACLs(nodeName string) (acl.Authorizer, error) {
// Build a policy for the agent master token.
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
// The builtin agent master policy allows reading any node information
// and allows writes to the agent with the node name of the running agent
// only. This used to allow a prefix match on agent names but that seems
// entirely unnecessary so it is now using an exact match.
policy := &acl.Policy{
PolicyRules: acl.PolicyRules{
Agents: []*acl.AgentRule{
{
Node: nodeName,
Policy: acl.PolicyWrite,
},
},
NodePrefixes: []*acl.NodeRule{
{
Name: "",
Policy: acl.PolicyRead,
},
},
},
}
return acl.NewPolicyAuthorizerWithDefaults(acl.DenyAll(), []*acl.Policy{policy}, nil)
}
// vetServiceRegister makes sure the service registration action is allowed by
// the given token.
func (a *Agent) vetServiceRegister(token string, service *structs.NodeService) error {
// Resolve the token and bail if ACLs aren't enabled.
authz, err := a.resolveToken(token)
if err != nil {
return err
}
return a.vetServiceRegisterWithAuthorizer(authz, service)
}
func (a *Agent) vetServiceRegisterWithAuthorizer(authz acl.Authorizer, service *structs.NodeService) error {
if authz == nil {
return nil
}
var authzContext acl.AuthorizerContext
service.FillAuthzContext(&authzContext)
// Vet the service itself.
if authz.ServiceWrite(service.Service, &authzContext) != acl.Allow {
serviceName := service.CompoundServiceName()
return acl.PermissionDenied("Missing service:write on %s", serviceName.String())
}
// Vet any service that might be getting overwritten.
if existing := a.State.Service(service.CompoundServiceID()); existing != nil {
existing.FillAuthzContext(&authzContext)
if authz.ServiceWrite(existing.Service, &authzContext) != acl.Allow {
serviceName := service.CompoundServiceName()
return acl.PermissionDenied("Missing service:write on %s", serviceName.String())
}
}
// If the service is a proxy, ensure that it has write on the destination too
// since it can be discovered as an instance of that service.
if service.Kind == structs.ServiceKindConnectProxy {
service.FillAuthzContext(&authzContext)
if authz.ServiceWrite(service.Proxy.DestinationServiceName, &authzContext) != acl.Allow {
return acl.PermissionDenied("Missing service:write on %s", service.Proxy.DestinationServiceName)
}
}
return nil
}
// vetServiceUpdate makes sure the service update action is allowed by the given
// token.
func (a *Agent) vetServiceUpdate(token string, serviceID structs.ServiceID) error {
// Resolve the token and bail if ACLs aren't enabled.
authz, err := a.resolveToken(token)
if err != nil {
return err
}
return a.vetServiceUpdateWithAuthorizer(authz, serviceID)
}
func (a *Agent) vetServiceUpdateWithAuthorizer(authz acl.Authorizer, serviceID structs.ServiceID) error {
if authz == nil {
return nil
}
var authzContext acl.AuthorizerContext
// Vet any changes based on the existing services's info.
if existing := a.State.Service(serviceID); existing != nil {
existing.FillAuthzContext(&authzContext)
if authz.ServiceWrite(existing.Service, &authzContext) != acl.Allow {
serviceName := existing.CompoundServiceName()
return acl.PermissionDenied("Missing service:write on %s", serviceName.String())
}
} else {
return fmt.Errorf("Unknown service %q", serviceID)
}
return nil
}
// vetCheckRegister makes sure the check registration action is allowed by the
// given token.
func (a *Agent) vetCheckRegister(token string, check *structs.HealthCheck) error {
// Resolve the token and bail if ACLs aren't enabled.
authz, err := a.resolveToken(token)
if err != nil {
return err
}
return a.vetCheckRegisterWithAuthorizer(authz, check)
}
func (a *Agent) vetCheckRegisterWithAuthorizer(authz acl.Authorizer, check *structs.HealthCheck) error {
if authz == nil {
return nil
}
var authzContext acl.AuthorizerContext
check.FillAuthzContext(&authzContext)
// Vet the check itself.
if len(check.ServiceName) > 0 {
if authz.ServiceWrite(check.ServiceName, &authzContext) != acl.Allow {
return acl.PermissionDenied("Missing service:write on %v", structs.ServiceIDString(check.ServiceName, &check.EnterpriseMeta))
}
} else {
if authz.NodeWrite(a.config.NodeName, &authzContext) != acl.Allow {
return acl.PermissionDenied("Missing node:write on %s", a.config.NodeName)
}
}
// Vet any check that might be getting overwritten.
if existing := a.State.Check(check.CompoundCheckID()); existing != nil {
if len(existing.ServiceName) > 0 {
if authz.ServiceWrite(existing.ServiceName, &authzContext) != acl.Allow {
return acl.PermissionDenied("Missing service:write on %s", structs.ServiceIDString(existing.ServiceName, &existing.EnterpriseMeta))
}
} else {
if authz.NodeWrite(a.config.NodeName, &authzContext) != acl.Allow {
return acl.PermissionDenied("Missing node:write on %s", a.config.NodeName)
}
}
}
return nil
}
// vetCheckUpdate makes sure that a check update is allowed by the given token.
func (a *Agent) vetCheckUpdate(token string, checkID structs.CheckID) error {
// Resolve the token and bail if ACLs aren't enabled.
authz, err := a.resolveToken(token)
if err != nil {
return err
}
return a.vetCheckUpdateWithAuthorizer(authz, checkID)
}
func (a *Agent) vetCheckUpdateWithAuthorizer(authz acl.Authorizer, checkID structs.CheckID) error {
if authz == nil {
return nil
}
var authzContext acl.AuthorizerContext
checkID.FillAuthzContext(&authzContext)
// Vet any changes based on the existing check's info.
if existing := a.State.Check(checkID); existing != nil {
if len(existing.ServiceName) > 0 {
if authz.ServiceWrite(existing.ServiceName, &authzContext) != acl.Allow {
return acl.PermissionDenied("Missing service:write on %s", existing.ServiceName)
}
} else {
if authz.NodeWrite(a.config.NodeName, &authzContext) != acl.Allow {
return acl.PermissionDenied("Missing node:write on %s", a.config.NodeName)
}
}
} else {
return fmt.Errorf("Unknown check %q", checkID.String())
}
return nil
}
// filterMembers redacts members that the token doesn't have access to.
func (a *Agent) filterMembers(token string, members *[]serf.Member) error {
// Resolve the token and bail if ACLs aren't enabled.
rule, err := a.resolveToken(token)
if err != nil {
return err
}
if rule == nil {
return nil
}
var authzContext acl.AuthorizerContext
structs.DefaultEnterpriseMeta().FillAuthzContext(&authzContext)
// Filter out members based on the node policy.
m := *members
for i := 0; i < len(m); i++ {
node := m[i].Name
if rule.NodeRead(node, &authzContext) == acl.Allow {
continue
}
accessorID := a.aclAccessorID(token)
a.logger.Debug("dropping node from result due to ACLs", "node", node, "accessorID", accessorID)
m = append(m[:i], m[i+1:]...)
i--
}
*members = m
return nil
}
// filterServices redacts services that the token doesn't have access to.
func (a *Agent) filterServices(token string, services *map[structs.ServiceID]*structs.NodeService) error {
// Resolve the token and bail if ACLs aren't enabled.
authz, err := a.resolveToken(token)
if err != nil {
return err
}
return a.filterServicesWithAuthorizer(authz, services)
}
func (a *Agent) filterServicesWithAuthorizer(authz acl.Authorizer, services *map[structs.ServiceID]*structs.NodeService) error {
if authz == nil {
return nil
}
var authzContext acl.AuthorizerContext
// Filter out services based on the service policy.
for id, service := range *services {
service.FillAuthzContext(&authzContext)
if authz.ServiceRead(service.Service, &authzContext) == acl.Allow {
continue
}
a.logger.Debug("dropping service from result due to ACLs", "service", id.String())
delete(*services, id)
}
return nil
}
// filterChecks redacts checks that the token doesn't have access to.
func (a *Agent) filterChecks(token string, checks *map[structs.CheckID]*structs.HealthCheck) error {
// Resolve the token and bail if ACLs aren't enabled.
authz, err := a.resolveToken(token)
if err != nil {
return err
}
return a.filterChecksWithAuthorizer(authz, checks)
}
func (a *Agent) filterChecksWithAuthorizer(authz acl.Authorizer, checks *map[structs.CheckID]*structs.HealthCheck) error {
if authz == nil {
return nil
}
var authzContext acl.AuthorizerContext
// Filter out checks based on the node or service policy.
for id, check := range *checks {
if len(check.ServiceName) > 0 {
check.FillAuthzContext(&authzContext)
if authz.ServiceRead(check.ServiceName, &authzContext) == acl.Allow {
continue
}
} else {
structs.DefaultEnterpriseMeta().FillAuthzContext(&authzContext)
if authz.NodeRead(a.config.NodeName, &authzContext) == acl.Allow {
continue
}
}
a.logger.Debug("dropping check from result due to ACLs", "check", id.String())
delete(*checks, id)
}
return nil
}