package appId import ( "fmt" "github.com/hashicorp/vault/helper/salt" "github.com/hashicorp/vault/logical" "github.com/hashicorp/vault/logical/framework" ) func Factory(conf *logical.BackendConfig) (logical.Backend, error) { b, err := Backend(conf) if err != nil { return nil, err } return b.Setup(conf) } func Backend(conf *logical.BackendConfig) (*framework.Backend, error) { // Initialize the salt salt, err := salt.NewSalt(conf.View, nil) if err != nil { return nil, err } var b backend b.Salt = salt b.MapAppId = &framework.PolicyMap{ PathMap: framework.PathMap{ Name: "app-id", Salt: salt, Schema: map[string]*framework.FieldSchema{ "display_name": &framework.FieldSchema{ Type: framework.TypeString, Description: "A name to map to this app ID for logs.", }, "value": &framework.FieldSchema{ Type: framework.TypeString, Description: "Policies for the app ID.", }, }, }, DefaultKey: "default", } b.MapUserId = &framework.PathMap{ Name: "user-id", Salt: salt, Schema: map[string]*framework.FieldSchema{ "cidr_block": &framework.FieldSchema{ Type: framework.TypeString, Description: "If not blank, restricts auth by this CIDR block", }, "value": &framework.FieldSchema{ Type: framework.TypeString, Description: "App IDs that this user associates with.", }, }, } b.Backend = &framework.Backend{ Help: backendHelp, PathsSpecial: &logical.Paths{ Unauthenticated: []string{ "login", }, }, Paths: framework.PathAppend([]*framework.Path{ pathLogin(&b), }, b.MapAppId.Paths(), b.MapUserId.Paths(), ), } // Since the salt is new in 0.2, we need to handle this by migrating // any existing keys to use the salt. We can deprecate this eventually, // but for now we want a smooth upgrade experience by automatically // upgrading to use salting. if salt.DidGenerate() { if err := b.upgradeToSalted(conf.View); err != nil { return nil, err } } return b.Backend, nil } type backend struct { *framework.Backend Salt *salt.Salt MapAppId *framework.PolicyMap MapUserId *framework.PathMap } // upgradeToSalted is used to upgrade the non-salted keys prior to // Vault 0.2 to be salted. This is done on mount time and is only // done once. It can be deprecated eventually, but should be around // long enough for all 0.1.x users to upgrade. func (b *backend) upgradeToSalted(view logical.Storage) error { // Create a copy of MapAppId that does not use a Salt nonSaltedAppId := new(framework.PathMap) *nonSaltedAppId = b.MapAppId.PathMap nonSaltedAppId.Salt = nil // Get the list of app-ids keys, err := b.MapAppId.List(view, "") if err != nil { return fmt.Errorf("failed to list app-ids: %v", err) } // Upgrade all the existing keys for _, key := range keys { val, err := nonSaltedAppId.Get(view, key) if err != nil { return fmt.Errorf("failed to read app-id: %v", err) } if err := b.MapAppId.Put(view, key, val); err != nil { return fmt.Errorf("failed to write app-id: %v", err) } if err := nonSaltedAppId.Delete(view, key); err != nil { return fmt.Errorf("failed to delete app-id: %v", err) } } // Create a copy of MapUserId that does not use a Salt nonSaltedUserId := new(framework.PathMap) *nonSaltedUserId = *b.MapUserId nonSaltedUserId.Salt = nil // Get the list of user-ids keys, err = b.MapUserId.List(view, "") if err != nil { return fmt.Errorf("failed to list user-ids: %v", err) } // Upgrade all the existing keys for _, key := range keys { val, err := nonSaltedUserId.Get(view, key) if err != nil { return fmt.Errorf("failed to read user-id: %v", err) } if err := b.MapUserId.Put(view, key, val); err != nil { return fmt.Errorf("failed to write user-id: %v", err) } if err := nonSaltedUserId.Delete(view, key); err != nil { return fmt.Errorf("failed to delete user-id: %v", err) } } return nil } const backendHelp = ` The App ID credential provider is used to perform authentication from within applications or machine by pairing together two hard-to-guess unique pieces of information: a unique app ID, and a unique user ID. The goal of this credential provider is to allow elastic users (dynamic machines, containers, etc.) to authenticate with Vault without having to store passwords outside of Vault. It is a single method of solving the chicken-and-egg problem of setting up Vault access on a machine. With this provider, nobody except the machine itself has access to both pieces of information necessary to authenticate. For example: configuration management will have the app IDs, but the machine itself will detect its user ID based on some unique machine property such as a MAC address (or a hash of it with some salt). An example, real world process for using this provider: 1. Create unique app IDs (UUIDs work well) and map them to policies. (Path: map/app-id/) 2. Store the app IDs within configuration management systems. 3. An out-of-band process run by security operators map unique user IDs to these app IDs. Example: when an instance is launched, a cloud-init system tells security operators a unique ID for this machine. This process can be scripted, but the key is that it is out-of-band and out of reach of configuration management. (Path: map/user-id/) 4. A new server is provisioned. Configuration management configures the app ID, the server itself detects its user ID. With both of these pieces of information, Vault can be accessed according to the policy set by the app ID. More details on this process follow: The app ID is a unique ID that maps to a set of policies. This ID is generated by an operator and configured into the backend. The ID itself is usually a UUID, but any hard-to-guess unique value can be used. After creating app IDs, an operator authorizes a fixed set of user IDs with each app ID. When a valid {app ID, user ID} tuple is given to the "login" path, then the user is authenticated with the configured app ID policies. The user ID can be any value (just like the app ID), however it is generally a value unique to a machine, such as a MAC address or instance ID, or a value hashed from these unique values. (Note that it is also possible to authorize multiple app IDs with each user ID by writing them as comma-separated values to the map/user-id/ path.) `