open-vault/vault/mount.go
Christopher Swenson b136a7ecd8
Add plugin version to GRPC interface (#17088)
Add plugin version to GRPC interface

Added a version interface in the sdk/logical so that it can be shared between all plugin types, and then wired it up to RunningVersion in the mounts, auth list, and database systems.

I've tested that this works with auth, database, and secrets plugin types, with the following logic to populate RunningVersion:

If a plugin has a PluginVersion() method implemented, then that is used
If not, and the plugin is built into the Vault binary, then the go.mod version is used
Otherwise, the it will be the empty string.
My apologies for the length of this PR.

* Placeholder backend should be external

We use a placeholder backend (previously a framework.Backend) before a
GRPC plugin is lazy-loaded. This makes us later think the plugin is a
builtin plugin.

So we added a `placeholderBackend` type that overrides the
`IsExternal()` method so that later we know that the plugin is external,
and don't give it a default builtin version.
2022-09-15 16:37:59 -07:00

1799 lines
58 KiB
Go

package vault
import (
"context"
"errors"
"fmt"
"os"
"sort"
"strings"
"sync"
"time"
"github.com/armon/go-metrics"
"github.com/hashicorp/go-secure-stdlib/strutil"
"github.com/hashicorp/go-uuid"
"github.com/hashicorp/vault/builtin/plugin"
"github.com/hashicorp/vault/helper/metricsutil"
"github.com/hashicorp/vault/helper/namespace"
"github.com/hashicorp/vault/helper/versions"
"github.com/hashicorp/vault/sdk/helper/consts"
"github.com/hashicorp/vault/sdk/helper/jsonutil"
"github.com/hashicorp/vault/sdk/logical"
"github.com/mitchellh/copystructure"
)
const (
// coreMountConfigPath is used to store the mount configuration.
// Mounts are protected within the Vault itself, which means they
// can only be viewed or modified after an unseal.
coreMountConfigPath = "core/mounts"
// coreLocalMountConfigPath is used to store mount configuration for local
// (non-replicated) mounts
coreLocalMountConfigPath = "core/local-mounts"
// backendBarrierPrefix is the prefix to the UUID used in the
// barrier view for the backends.
backendBarrierPrefix = "logical/"
// systemBarrierPrefix is the prefix used for the
// system logical backend.
systemBarrierPrefix = "sys/"
// mountTableType is the value we expect to find for the mount table and
// corresponding entries
mountTableType = "mounts"
)
// ListingVisibilityType represents the types for listing visibility
type ListingVisibilityType string
const (
// ListingVisibilityDefault is the default value for listing visibility
ListingVisibilityDefault ListingVisibilityType = ""
// ListingVisibilityHidden is the hidden type for listing visibility
ListingVisibilityHidden ListingVisibilityType = "hidden"
// ListingVisibilityUnauth is the unauth type for listing visibility
ListingVisibilityUnauth ListingVisibilityType = "unauth"
systemMountPath = "sys/"
identityMountPath = "identity/"
cubbyholeMountPath = "cubbyhole/"
systemMountType = "system"
identityMountType = "identity"
cubbyholeMountType = "cubbyhole"
pluginMountType = "plugin"
mountTypeNSCubbyhole = "ns_cubbyhole"
MountTableUpdateStorage = true
MountTableNoUpdateStorage = false
)
var (
// loadMountsFailed if loadMounts encounters an error
errLoadMountsFailed = errors.New("failed to setup mount table")
// protectedMounts cannot be remounted
protectedMounts = []string{
"audit/",
"auth/",
systemMountPath,
cubbyholeMountPath,
identityMountPath,
}
untunableMounts = []string{
cubbyholeMountPath,
systemMountPath,
"audit/",
identityMountPath,
}
// singletonMounts can only exist in one location and are
// loaded by default. These are types, not paths.
singletonMounts = []string{
cubbyholeMountType,
systemMountType,
"token",
identityMountType,
}
// mountAliases maps old backend names to new backend names, allowing us
// to move/rename backends but maintain backwards compatibility
mountAliases = map[string]string{"generic": "kv"}
)
func (c *Core) generateMountAccessor(entryType string) (string, error) {
var accessor string
for {
randBytes, err := uuid.GenerateRandomBytes(4)
if err != nil {
return "", err
}
accessor = fmt.Sprintf("%s_%s", entryType, fmt.Sprintf("%08x", randBytes[0:4]))
if entry := c.router.MatchingMountByAccessor(accessor); entry == nil {
break
}
}
return accessor, nil
}
// MountTable is used to represent the internal mount table
type MountTable struct {
Type string `json:"type"`
Entries []*MountEntry `json:"entries"`
}
type MountMigrationStatus int
const (
MigrationInProgressStatus MountMigrationStatus = iota
MigrationSuccessStatus
MigrationFailureStatus
)
func (m MountMigrationStatus) String() string {
switch m {
case MigrationInProgressStatus:
return "in-progress"
case MigrationSuccessStatus:
return "success"
case MigrationFailureStatus:
return "failure"
}
return "unknown"
}
type MountMigrationInfo struct {
SourceMount string `json:"source_mount"`
TargetMount string `json:"target_mount"`
MigrationStatus string `json:"status"`
}
// tableMetrics is responsible for setting gauge metrics for
// mount table storage sizes (in bytes) and mount table num
// entries. It does this via setGaugeWithLabels. It then
// saves these metrics in a cache for regular reporting in
// a loop, via AddGaugeLoopMetric.
// Note that the reported storage sizes are pre-encryption
// sizes. Currently barrier uses aes-gcm for encryption, which
// preserves plaintext size, adding a constant of 30 bytes of
// padding, which is negligable and subject to change, and thus
// not accounted for.
func (c *Core) tableMetrics(entryCount int, isLocal bool, isAuth bool, compressedTable []byte) {
if c.metricsHelper == nil {
// do nothing if metrics are not initialized
return
}
typeAuthLabelMap := map[bool]metrics.Label{
true: {Name: "type", Value: "auth"},
false: {Name: "type", Value: "logical"},
}
typeLocalLabelMap := map[bool]metrics.Label{
true: {Name: "local", Value: "true"},
false: {Name: "local", Value: "false"},
}
c.metricSink.SetGaugeWithLabels(metricsutil.LogicalTableSizeName,
float32(entryCount), []metrics.Label{
typeAuthLabelMap[isAuth],
typeLocalLabelMap[isLocal],
})
c.metricsHelper.AddGaugeLoopMetric(metricsutil.LogicalTableSizeName,
float32(entryCount), []metrics.Label{
typeAuthLabelMap[isAuth],
typeLocalLabelMap[isLocal],
})
c.metricSink.SetGaugeWithLabels(metricsutil.PhysicalTableSizeName,
float32(len(compressedTable)), []metrics.Label{
typeAuthLabelMap[isAuth],
typeLocalLabelMap[isLocal],
})
c.metricsHelper.AddGaugeLoopMetric(metricsutil.PhysicalTableSizeName,
float32(len(compressedTable)), []metrics.Label{
typeAuthLabelMap[isAuth],
typeLocalLabelMap[isLocal],
})
}
// shallowClone returns a copy of the mount table that
// keeps the MountEntry locations, so as not to invalidate
// other locations holding pointers. Care needs to be taken
// if modifying entries rather than modifying the table itself
func (t *MountTable) shallowClone() *MountTable {
mt := &MountTable{
Type: t.Type,
Entries: make([]*MountEntry, len(t.Entries)),
}
for i, e := range t.Entries {
mt.Entries[i] = e
}
return mt
}
// setTaint is used to set the taint on given entry Accepts either the mount
// entry's path or namespace + path, i.e. <ns-path>/secret/ or <ns-path>/token/
func (t *MountTable) setTaint(nsID, path string, tainted bool, mountState string) (*MountEntry, error) {
n := len(t.Entries)
for i := 0; i < n; i++ {
if entry := t.Entries[i]; entry.Path == path && entry.Namespace().ID == nsID {
t.Entries[i].Tainted = tainted
t.Entries[i].MountState = mountState
return t.Entries[i], nil
}
}
return nil, nil
}
// remove is used to remove a given path entry; returns the entry that was
// removed
func (t *MountTable) remove(ctx context.Context, path string) (*MountEntry, error) {
n := len(t.Entries)
ns, err := namespace.FromContext(ctx)
if err != nil {
return nil, err
}
for i := 0; i < n; i++ {
if entry := t.Entries[i]; entry.Path == path && entry.Namespace().ID == ns.ID {
t.Entries[i], t.Entries[n-1] = t.Entries[n-1], nil
t.Entries = t.Entries[:n-1]
return entry, nil
}
}
return nil, nil
}
func (t *MountTable) find(ctx context.Context, path string) (*MountEntry, error) {
n := len(t.Entries)
ns, err := namespace.FromContext(ctx)
if err != nil {
return nil, err
}
for i := 0; i < n; i++ {
if entry := t.Entries[i]; entry.Path == path && entry.Namespace().ID == ns.ID {
return entry, nil
}
}
return nil, nil
}
func (t *MountTable) findByBackendUUID(ctx context.Context, backendUUID string) (*MountEntry, error) {
n := len(t.Entries)
ns, err := namespace.FromContext(ctx)
if err != nil {
return nil, err
}
for i := 0; i < n; i++ {
if entry := t.Entries[i]; entry.BackendAwareUUID == backendUUID && entry.Namespace().ID == ns.ID {
return entry, nil
}
}
return nil, nil
}
// sortEntriesByPath sorts the entries in the table by path and returns the
// table; this is useful for tests
func (t *MountTable) sortEntriesByPath() *MountTable {
sort.Slice(t.Entries, func(i, j int) bool {
return t.Entries[i].Path < t.Entries[j].Path
})
return t
}
// sortEntriesByPath sorts the entries in the table by path and returns the
// table; this is useful for tests
func (t *MountTable) sortEntriesByPathDepth() *MountTable {
sort.Slice(t.Entries, func(i, j int) bool {
return len(strings.Split(t.Entries[i].Namespace().Path+t.Entries[i].Path, "/")) < len(strings.Split(t.Entries[j].Namespace().Path+t.Entries[j].Path, "/"))
})
return t
}
const mountStateUnmounting = "unmounting"
// MountEntry is used to represent a mount table entry
type MountEntry struct {
Table string `json:"table"` // The table it belongs to
Path string `json:"path"` // Mount Path
Type string `json:"type"` // Logical backend Type. NB: This is the plugin name, e.g. my-vault-plugin, NOT plugin type (e.g. auth).
Description string `json:"description"` // User-provided description
UUID string `json:"uuid"` // Barrier view UUID
BackendAwareUUID string `json:"backend_aware_uuid"` // UUID that can be used by the backend as a helper when a consistent value is needed outside of storage.
Accessor string `json:"accessor"` // Unique but more human-friendly ID. Does not change, not used for any sensitive things (like as a salt, which the UUID sometimes is).
Config MountConfig `json:"config"` // Configuration related to this mount (but not backend-derived)
Options map[string]string `json:"options"` // Backend options
Local bool `json:"local"` // Local mounts are not replicated or affected by replication
SealWrap bool `json:"seal_wrap"` // Whether to wrap CSPs
ExternalEntropyAccess bool `json:"external_entropy_access,omitempty"` // Whether to allow external entropy source access
Tainted bool `json:"tainted,omitempty"` // Set as a Write-Ahead flag for unmount/remount
MountState string `json:"mount_state,omitempty"` // The current mount state. The only non-empty mount state right now is "unmounting"
NamespaceID string `json:"namespace_id"`
// namespace contains the populated namespace
namespace *namespace.Namespace
// synthesizedConfigCache is used to cache configuration values. These
// particular values are cached since we want to get them at a point-in-time
// without separately managing their locks individually. See SyncCache() for
// the specific values that are being cached.
synthesizedConfigCache sync.Map
// version info
Version string `json:"version,omitempty"` // The semantic version of the mounted plugin, e.g. v1.2.3.
Sha string `json:"sha,omitempty"` // The SHA256 sum of the plugin binary.
RunningVersion string `json:"running_version,omitempty"` // The semantic version of the mounted plugin as reported by the plugin.
RunningSha string `json:"running_sha,omitempty"`
}
// MountConfig is used to hold settable options
type MountConfig struct {
DefaultLeaseTTL time.Duration `json:"default_lease_ttl,omitempty" structs:"default_lease_ttl" mapstructure:"default_lease_ttl"` // Override for global default
MaxLeaseTTL time.Duration `json:"max_lease_ttl,omitempty" structs:"max_lease_ttl" mapstructure:"max_lease_ttl"` // Override for global default
ForceNoCache bool `json:"force_no_cache,omitempty" structs:"force_no_cache" mapstructure:"force_no_cache"` // Override for global default
AuditNonHMACRequestKeys []string `json:"audit_non_hmac_request_keys,omitempty" structs:"audit_non_hmac_request_keys" mapstructure:"audit_non_hmac_request_keys"`
AuditNonHMACResponseKeys []string `json:"audit_non_hmac_response_keys,omitempty" structs:"audit_non_hmac_response_keys" mapstructure:"audit_non_hmac_response_keys"`
ListingVisibility ListingVisibilityType `json:"listing_visibility,omitempty" structs:"listing_visibility" mapstructure:"listing_visibility"`
PassthroughRequestHeaders []string `json:"passthrough_request_headers,omitempty" structs:"passthrough_request_headers" mapstructure:"passthrough_request_headers"`
AllowedResponseHeaders []string `json:"allowed_response_headers,omitempty" structs:"allowed_response_headers" mapstructure:"allowed_response_headers"`
TokenType logical.TokenType `json:"token_type,omitempty" structs:"token_type" mapstructure:"token_type"`
AllowedManagedKeys []string `json:"allowed_managed_keys,omitempty" mapstructure:"allowed_managed_keys"`
// PluginName is the name of the plugin registered in the catalog.
//
// Deprecated: MountEntry.Type should be used instead for Vault 1.0.0 and beyond.
PluginName string `json:"plugin_name,omitempty" structs:"plugin_name,omitempty" mapstructure:"plugin_name"`
}
// APIMountConfig is an embedded struct of api.MountConfigInput
type APIMountConfig struct {
DefaultLeaseTTL string `json:"default_lease_ttl" structs:"default_lease_ttl" mapstructure:"default_lease_ttl"`
MaxLeaseTTL string `json:"max_lease_ttl" structs:"max_lease_ttl" mapstructure:"max_lease_ttl"`
ForceNoCache bool `json:"force_no_cache" structs:"force_no_cache" mapstructure:"force_no_cache"`
AuditNonHMACRequestKeys []string `json:"audit_non_hmac_request_keys,omitempty" structs:"audit_non_hmac_request_keys" mapstructure:"audit_non_hmac_request_keys"`
AuditNonHMACResponseKeys []string `json:"audit_non_hmac_response_keys,omitempty" structs:"audit_non_hmac_response_keys" mapstructure:"audit_non_hmac_response_keys"`
ListingVisibility ListingVisibilityType `json:"listing_visibility,omitempty" structs:"listing_visibility" mapstructure:"listing_visibility"`
PassthroughRequestHeaders []string `json:"passthrough_request_headers,omitempty" structs:"passthrough_request_headers" mapstructure:"passthrough_request_headers"`
AllowedResponseHeaders []string `json:"allowed_response_headers,omitempty" structs:"allowed_response_headers" mapstructure:"allowed_response_headers"`
TokenType string `json:"token_type" structs:"token_type" mapstructure:"token_type"`
AllowedManagedKeys []string `json:"allowed_managed_keys,omitempty" mapstructure:"allowed_managed_keys"`
// PluginName is the name of the plugin registered in the catalog.
//
// Deprecated: MountEntry.Type should be used instead for Vault 1.0.0 and beyond.
PluginName string `json:"plugin_name,omitempty" structs:"plugin_name,omitempty" mapstructure:"plugin_name"`
}
// Clone returns a deep copy of the mount entry
func (e *MountEntry) Clone() (*MountEntry, error) {
cp, err := copystructure.Copy(e)
if err != nil {
return nil, err
}
return cp.(*MountEntry), nil
}
// Namespace returns the namespace for the mount entry
func (e *MountEntry) Namespace() *namespace.Namespace {
return e.namespace
}
// APIPath returns the full API Path for the given mount entry
func (e *MountEntry) APIPath() string {
path := e.Path
if e.Table == credentialTableType {
path = credentialRoutePrefix + path
}
return e.namespace.Path + path
}
// APIPathNoNamespace returns the API Path without the namespace for the given mount entry
func (e *MountEntry) APIPathNoNamespace() string {
path := e.Path
if e.Table == credentialTableType {
path = credentialRoutePrefix + path
}
return path
}
// SyncCache syncs tunable configuration values to the cache. In the case of
// cached values, they should be retrieved via synthesizedConfigCache.Load()
// instead of accessing them directly through MountConfig.
func (e *MountEntry) SyncCache() {
if len(e.Config.AuditNonHMACRequestKeys) == 0 {
e.synthesizedConfigCache.Delete("audit_non_hmac_request_keys")
} else {
e.synthesizedConfigCache.Store("audit_non_hmac_request_keys", e.Config.AuditNonHMACRequestKeys)
}
if len(e.Config.AuditNonHMACResponseKeys) == 0 {
e.synthesizedConfigCache.Delete("audit_non_hmac_response_keys")
} else {
e.synthesizedConfigCache.Store("audit_non_hmac_response_keys", e.Config.AuditNonHMACResponseKeys)
}
if len(e.Config.PassthroughRequestHeaders) == 0 {
e.synthesizedConfigCache.Delete("passthrough_request_headers")
} else {
e.synthesizedConfigCache.Store("passthrough_request_headers", e.Config.PassthroughRequestHeaders)
}
if len(e.Config.AllowedResponseHeaders) == 0 {
e.synthesizedConfigCache.Delete("allowed_response_headers")
} else {
e.synthesizedConfigCache.Store("allowed_response_headers", e.Config.AllowedResponseHeaders)
}
if len(e.Config.AllowedManagedKeys) == 0 {
e.synthesizedConfigCache.Delete("allowed_managed_keys")
} else {
e.synthesizedConfigCache.Store("allowed_managed_keys", e.Config.AllowedManagedKeys)
}
}
func (c *Core) decodeMountTable(ctx context.Context, raw []byte) (*MountTable, error) {
// Decode into mount table
mountTable := new(MountTable)
if err := jsonutil.DecodeJSON(raw, mountTable); err != nil {
return nil, err
}
// Populate the namespace in memory
var mountEntries []*MountEntry
for _, entry := range mountTable.Entries {
if entry.NamespaceID == "" {
entry.NamespaceID = namespace.RootNamespaceID
}
ns, err := NamespaceByID(ctx, entry.NamespaceID, c)
if err != nil {
return nil, err
}
if ns == nil {
c.logger.Error("namespace on mount entry not found", "namespace_id", entry.NamespaceID, "mount_path", entry.Path, "mount_description", entry.Description)
continue
}
// Immediately shutdown the core if deprecated mounts are detected and VAULT_ALLOW_PENDING_REMOVAL_MOUNTS is unset
if _, err := c.handleDeprecatedMountEntry(ctx, entry, consts.PluginTypeUnknown); err != nil {
c.logger.Error("shutting down core", "error", err)
c.Shutdown()
}
entry.namespace = ns
mountEntries = append(mountEntries, entry)
}
return &MountTable{
Type: mountTable.Type,
Entries: mountEntries,
}, nil
}
// Mount is used to mount a new backend to the mount table.
func (c *Core) mount(ctx context.Context, entry *MountEntry) error {
// Ensure we end the path in a slash
if !strings.HasSuffix(entry.Path, "/") {
entry.Path += "/"
}
// Prevent protected paths from being mounted
for _, p := range protectedMounts {
if strings.HasPrefix(entry.Path, p) && entry.namespace == nil {
return logical.CodedError(403, fmt.Sprintf("cannot mount %q", entry.Path))
}
}
// Do not allow more than one instance of a singleton mount
for _, p := range singletonMounts {
if entry.Type == p {
return logical.CodedError(403, fmt.Sprintf("mount type of %q is not mountable", entry.Type))
}
}
// Mount internally
if err := c.mountInternal(ctx, entry, MountTableUpdateStorage); err != nil {
return err
}
// Re-evaluate filtered paths
if err := runFilteredPathsEvaluation(ctx, c); err != nil {
c.logger.Error("failed to evaluate filtered paths", "error", err)
// We failed to evaluate filtered paths so we are undoing the mount operation
if unmountInternalErr := c.unmountInternal(ctx, entry.Path, MountTableUpdateStorage); unmountInternalErr != nil {
c.logger.Error("failed to unmount", "error", unmountInternalErr)
}
return err
}
return nil
}
func (c *Core) mountInternal(ctx context.Context, entry *MountEntry, updateStorage bool) error {
c.mountsLock.Lock()
defer c.mountsLock.Unlock()
ns, err := namespace.FromContext(ctx)
if err != nil {
return err
}
if err := verifyNamespace(c, ns, entry); err != nil {
return err
}
entry.NamespaceID = ns.ID
entry.namespace = ns
// Ensure the cache is populated, don't need the result
NamespaceByID(ctx, ns.ID, c)
// Basic check for matching names
for _, ent := range c.mounts.Entries {
if ns.ID == ent.NamespaceID {
switch {
// Existing is oauth/github/ new is oauth/ or
// existing is oauth/ and new is oauth/github/
case strings.HasPrefix(ent.Path, entry.Path):
fallthrough
case strings.HasPrefix(entry.Path, ent.Path):
return logical.CodedError(409, fmt.Sprintf("path is already in use at %s", ent.Path))
}
}
}
// Verify there are no conflicting mounts in the router
if match := c.router.MountConflict(ctx, entry.Path); match != "" {
return logical.CodedError(409, fmt.Sprintf("existing mount at %s", match))
}
// Generate a new UUID and view
if entry.UUID == "" {
entryUUID, err := uuid.GenerateUUID()
if err != nil {
return err
}
entry.UUID = entryUUID
}
if entry.BackendAwareUUID == "" {
bUUID, err := uuid.GenerateUUID()
if err != nil {
return err
}
entry.BackendAwareUUID = bUUID
}
if entry.Accessor == "" {
accessor, err := c.generateMountAccessor(entry.Type)
if err != nil {
return err
}
entry.Accessor = accessor
}
// Sync values to the cache
entry.SyncCache()
viewPath := entry.ViewPath()
view := NewBarrierView(c.barrier, viewPath)
// Singleton mounts cannot be filtered manually on a per-secondary basis
// from replication.
if strutil.StrListContains(singletonMounts, entry.Type) {
addFilterablePath(c, viewPath)
}
nilMount, err := preprocessMount(c, entry, view)
if err != nil {
return err
}
origReadOnlyErr := view.getReadOnlyErr()
// Mark the view as read-only until the mounting is complete and
// ensure that it is reset after. This ensures that there will be no
// writes during the construction of the backend.
view.setReadOnlyErr(logical.ErrSetupReadOnly)
// We defer this because we're already up and running so we don't need to
// time it for after postUnseal
defer view.setReadOnlyErr(origReadOnlyErr)
var backend logical.Backend
sysView := c.mountEntrySysView(entry)
backend, err = c.newLogicalBackend(ctx, entry, sysView, view)
if err != nil {
return err
}
if backend == nil {
return fmt.Errorf("nil backend of type %q returned from creation function", entry.Type)
}
// Check for the correct backend type
backendType := backend.Type()
if backendType != logical.TypeLogical {
if entry.Type != "kv" && entry.Type != "system" && entry.Type != "cubbyhole" {
return fmt.Errorf(`unknown backend type: "%s"`, entry.Type)
}
}
// update the entry running version with the backend's reported version
if versioner, ok := backend.(logical.PluginVersioner); ok {
entry.RunningVersion = versioner.PluginVersion().Version
}
if entry.RunningVersion == "" {
// don't set the running version to a builtin if it is running as an external plugin
if externaler, ok := backend.(logical.Externaler); !ok || !externaler.IsExternal() {
entry.RunningVersion = versions.GetBuiltinVersion(consts.PluginTypeSecrets, entry.Type)
}
}
addPathCheckers(c, entry, backend, viewPath)
c.setCoreBackend(entry, backend, view)
// If the mount is filtered or we are on a DR secondary we don't want to
// keep the actual backend running, so we clean it up and set it to nil
// so the router does not have a pointer to the object.
if nilMount {
backend.Cleanup(ctx)
backend = nil
}
newTable := c.mounts.shallowClone()
newTable.Entries = append(newTable.Entries, entry)
if updateStorage {
if err := c.persistMounts(ctx, newTable, &entry.Local); err != nil {
c.logger.Error("failed to update mount table", "error", err)
if err == logical.ErrReadOnly && c.perfStandby {
return err
}
return logical.CodedError(500, "failed to update mount table")
}
}
c.mounts = newTable
if err := c.router.Mount(backend, entry.Path, entry, view); err != nil {
return err
}
if !nilMount {
// restore the original readOnlyErr, so we can write to the view in
// Initialize() if necessary
view.setReadOnlyErr(origReadOnlyErr)
// initialize, using the core's active context.
err := backend.Initialize(c.activeContext, &logical.InitializationRequest{Storage: view})
if err != nil {
return err
}
}
if c.logger.IsInfo() {
c.logger.Info("successful mount", "namespace", entry.Namespace().Path, "path", entry.Path, "type", entry.Type)
}
return nil
}
// builtinTypeFromMountEntry attempts to find a builtin PluginType associated
// with the specified MountEntry. Returns consts.PluginTypeUnknown if not found.
func (c *Core) builtinTypeFromMountEntry(ctx context.Context, entry *MountEntry) consts.PluginType {
if c.builtinRegistry == nil || entry == nil {
return consts.PluginTypeUnknown
}
builtinPluginType := func(name string, pluginType consts.PluginType) (consts.PluginType, bool) {
plugin, err := c.pluginCatalog.Get(ctx, name, pluginType, "")
if err == nil && plugin != nil && plugin.Builtin {
return plugin.Type, true
}
return consts.PluginTypeUnknown, false
}
// auth plugins have their own dedicated mount table
if pluginType, err := consts.ParsePluginType(entry.Table); err == nil {
if builtinType, ok := builtinPluginType(entry.Type, pluginType); ok {
return builtinType
}
}
// Check for possible matches
var builtinTypes []consts.PluginType
for _, pluginType := range [...]consts.PluginType{consts.PluginTypeSecrets, consts.PluginTypeDatabase} {
if builtinType, ok := builtinPluginType(entry.Type, pluginType); ok {
builtinTypes = append(builtinTypes, builtinType)
}
}
if len(builtinTypes) == 1 {
return builtinTypes[0]
}
return consts.PluginTypeUnknown
}
// Unmount is used to unmount a path. The boolean indicates whether the mount
// was found.
func (c *Core) unmount(ctx context.Context, path string) error {
// Ensure we end the path in a slash
if !strings.HasSuffix(path, "/") {
path += "/"
}
// Prevent protected paths from being unmounted
for _, p := range protectedMounts {
if strings.HasPrefix(path, p) {
return fmt.Errorf("cannot unmount %q", path)
}
}
// Unmount mount internally
if err := c.unmountInternal(ctx, path, MountTableUpdateStorage); err != nil {
return err
}
// Re-evaluate filtered paths
if err := runFilteredPathsEvaluation(ctx, c); err != nil {
// Even we failed to evaluate filtered paths, the unmount operation was still successful
c.logger.Error("failed to evaluate filtered paths", "error", err)
}
return nil
}
func (c *Core) unmountInternal(ctx context.Context, path string, updateStorage bool) error {
ns, err := namespace.FromContext(ctx)
if err != nil {
return err
}
// Verify exact match of the route
match := c.router.MatchingMount(ctx, path)
if match == "" || ns.Path+path != match {
return fmt.Errorf("no matching mount")
}
// Get the view for this backend
view := c.router.MatchingStorageByAPIPath(ctx, path)
// Get the backend/mount entry for this path, used to remove ignored
// replication prefixes
backend := c.router.MatchingBackend(ctx, path)
entry := c.router.MatchingMountEntry(ctx, path)
// Mark the entry as tainted
if err := c.taintMountEntry(ctx, ns.ID, path, updateStorage, true); err != nil {
c.logger.Error("failed to taint mount entry for path being unmounted", "error", err, "path", path)
return err
}
// Taint the router path to prevent routing. Note that in-flight requests
// are uncertain, right now.
if err := c.router.Taint(ctx, path); err != nil {
return err
}
rCtx := namespace.ContextWithNamespace(c.activeContext, ns)
if backend != nil && c.rollback != nil {
// Invoke the rollback manager a final time
if err := c.rollback.Rollback(rCtx, path); err != nil {
return err
}
}
if backend != nil && c.expiration != nil && updateStorage {
// Revoke all the dynamic keys
if err := c.expiration.RevokePrefix(rCtx, path, true); err != nil {
return err
}
}
if backend != nil {
// Call cleanup function if it exists
backend.Cleanup(ctx)
}
viewPath := entry.ViewPath()
switch {
case !updateStorage:
// Don't attempt to clear data, replication will handle this
case c.IsDRSecondary():
// If we are a dr secondary we want to clear the view, but the provided
// view is marked as read only. We use the barrier here to get around
// it.
if err := logical.ClearViewWithLogging(ctx, NewBarrierView(c.barrier, viewPath), c.logger.Named("secrets.deletion").With("namespace", ns.ID, "path", path)); err != nil {
c.logger.Error("failed to clear view for path being unmounted", "error", err, "path", path)
return err
}
case entry.Local, !c.ReplicationState().HasState(consts.ReplicationPerformanceSecondary):
// Have writable storage, remove the whole thing
if err := logical.ClearViewWithLogging(ctx, view, c.logger.Named("secrets.deletion").With("namespace", ns.ID, "path", path)); err != nil {
c.logger.Error("failed to clear view for path being unmounted", "error", err, "path", path)
return err
}
case !entry.Local && c.ReplicationState().HasState(consts.ReplicationPerformanceSecondary):
if err := clearIgnoredPaths(ctx, c, backend, viewPath); err != nil {
return err
}
}
// Remove the mount table entry
if err := c.removeMountEntry(ctx, path, updateStorage); err != nil {
c.logger.Error("failed to remove mount entry for path being unmounted", "error", err, "path", path)
return err
}
// Unmount the backend entirely
if err := c.router.Unmount(ctx, path); err != nil {
return err
}
removePathCheckers(c, entry, viewPath)
if c.quotaManager != nil {
if err := c.quotaManager.HandleBackendDisabling(ctx, ns.Path, path); err != nil {
c.logger.Error("failed to update quotas after disabling mount", "path", path, "error", err)
return err
}
}
if c.logger.IsInfo() {
c.logger.Info("successfully unmounted", "path", path, "namespace", ns.Path)
}
return nil
}
// removeMountEntry is used to remove an entry from the mount table
func (c *Core) removeMountEntry(ctx context.Context, path string, updateStorage bool) error {
c.mountsLock.Lock()
defer c.mountsLock.Unlock()
// Remove the entry from the mount table
newTable := c.mounts.shallowClone()
entry, err := newTable.remove(ctx, path)
if err != nil {
return err
}
if entry == nil {
c.logger.Error("nil entry found removing entry in mounts table", "path", path)
return logical.CodedError(500, "failed to remove entry in mounts table")
}
// When unmounting all entries the JSON code will load back up from storage
// as a nil slice, which kills tests...just set it nil explicitly
if len(newTable.Entries) == 0 {
newTable.Entries = nil
}
if updateStorage {
// Update the mount table
if err := c.persistMounts(ctx, newTable, &entry.Local); err != nil {
c.logger.Error("failed to remove entry from mounts table", "error", err)
return logical.CodedError(500, "failed to remove entry from mounts table")
}
}
c.mounts = newTable
return nil
}
// taintMountEntry is used to mark an entry in the mount table as tainted
func (c *Core) taintMountEntry(ctx context.Context, nsID, mountPath string, updateStorage, unmounting bool) error {
c.mountsLock.Lock()
defer c.mountsLock.Unlock()
mountState := ""
if unmounting {
mountState = mountStateUnmounting
}
// As modifying the taint of an entry affects shallow clones,
// we simply use the original
entry, err := c.mounts.setTaint(nsID, mountPath, true, mountState)
if err != nil {
return err
}
if entry == nil {
c.logger.Error("nil entry found tainting entry in mounts table", "path", mountPath)
return logical.CodedError(500, "failed to taint entry in mounts table")
}
if updateStorage {
// Update the mount table
if err := c.persistMounts(ctx, c.mounts, &entry.Local); err != nil {
if err == logical.ErrReadOnly && c.perfStandby {
return err
}
c.logger.Error("failed to taint entry in mounts table", "error", err)
return logical.CodedError(500, "failed to taint entry in mounts table")
}
}
return nil
}
// handleDeprecatedMountEntry handles the Deprecation Status of the specified
// mount entry's builtin engine as follows:
//
// * Supported - do nothing
// * Deprecated - log a warning about builtin deprecation
// * PendingRemoval - log an error about builtin deprecation and return an error
// if VAULT_ALLOW_PENDING_REMOVAL_MOUNTS is unset
// * Removed - log an error about builtin deprecation and return an error
func (c *Core) handleDeprecatedMountEntry(ctx context.Context, entry *MountEntry, pluginType consts.PluginType) (*logical.Response, error) {
if c.builtinRegistry == nil || entry == nil {
return nil, nil
}
// Allow type to be determined from mount entry when not otherwise specified
if pluginType == consts.PluginTypeUnknown {
pluginType = c.builtinTypeFromMountEntry(ctx, entry)
}
// Handle aliases
t := entry.Type
if alias, ok := mountAliases[t]; ok {
t = alias
}
status, ok := c.builtinRegistry.DeprecationStatus(t, pluginType)
if ok {
resp := &logical.Response{}
// Deprecation sublogger with some identifying information
dl := c.logger.With("name", t, "type", pluginType, "status", status, "path", entry.Path)
errDeprecatedMount := fmt.Errorf("mount entry associated with %s builtin", status)
switch status {
case consts.Deprecated:
dl.Warn(errDeprecatedMount.Error())
resp.AddWarning(errDeprecatedMount.Error())
return resp, nil
case consts.PendingRemoval:
dl.Error(errDeprecatedMount.Error())
if allow := os.Getenv(consts.VaultAllowPendingRemovalMountsEnv); allow == "" {
return nil, fmt.Errorf("could not mount %q: %w", t, errDeprecatedMount)
}
resp.AddWarning(errDeprecatedMount.Error())
c.Logger().Info("mount allowed by environment variable", "env", consts.VaultAllowPendingRemovalMountsEnv)
return resp, nil
case consts.Removed:
return nil, fmt.Errorf("could not mount %s: %w", t, errDeprecatedMount)
}
}
return nil, nil
}
// remountForceInternal takes a copy of the mount entry for the path and fully unmounts
// and remounts the backend to pick up any changes, such as filtered paths.
// Should be only used for internal usage.
func (c *Core) remountForceInternal(ctx context.Context, path string, updateStorage bool) error {
me := c.router.MatchingMountEntry(ctx, path)
if me == nil {
return fmt.Errorf("cannot find mount for path %q", path)
}
me, err := me.Clone()
if err != nil {
return err
}
if err := c.unmountInternal(ctx, path, updateStorage); err != nil {
return err
}
// Mount internally
if err := c.mountInternal(ctx, me, updateStorage); err != nil {
return err
}
// Re-evaluate filtered paths
if err := runFilteredPathsEvaluation(ctx, c); err != nil {
c.logger.Error("failed to evaluate filtered paths", "error", err)
return err
}
return nil
}
func (c *Core) remountSecretsEngineCurrentNamespace(ctx context.Context, src, dst string, updateStorage bool) error {
ns, err := namespace.FromContext(ctx)
if err != nil {
return err
}
srcPathDetails := c.splitNamespaceAndMountFromPath(ns.Path, src)
dstPathDetails := c.splitNamespaceAndMountFromPath(ns.Path, dst)
return c.remountSecretsEngine(ctx, srcPathDetails, dstPathDetails, updateStorage)
}
// remountSecretsEngine is used to remount a path at a new mount point.
func (c *Core) remountSecretsEngine(ctx context.Context, src, dst namespace.MountPathDetails, updateStorage bool) error {
ns, err := namespace.FromContext(ctx)
if err != nil {
return err
}
// Prevent protected paths from being remounted, or target mounts being in protected paths
for _, p := range protectedMounts {
if strings.HasPrefix(src.MountPath, p) {
return fmt.Errorf("cannot remount %q", src.MountPath)
}
if strings.HasPrefix(dst.MountPath, p) {
return fmt.Errorf("cannot remount to destination %+v", dst)
}
}
srcRelativePath := src.GetRelativePath(ns)
dstRelativePath := dst.GetRelativePath(ns)
// Verify exact match of the route
srcMatch := c.router.MatchingMountEntry(ctx, srcRelativePath)
if srcMatch == nil {
return fmt.Errorf("no matching mount at %q", src.Namespace.Path+src.MountPath)
}
if match := c.router.MountConflict(ctx, dstRelativePath); match != "" {
return fmt.Errorf("path in use at %q", match)
}
// Mark the entry as tainted
if err := c.taintMountEntry(ctx, src.Namespace.ID, src.MountPath, updateStorage, false); err != nil {
return err
}
// Taint the router path to prevent routing
if err := c.router.Taint(ctx, srcRelativePath); err != nil {
return err
}
if !c.IsDRSecondary() {
// Invoke the rollback manager a final time
rCtx := namespace.ContextWithNamespace(c.activeContext, ns)
if c.rollback != nil && c.router.MatchingBackend(ctx, srcRelativePath) != nil {
if err := c.rollback.Rollback(rCtx, srcRelativePath); err != nil {
return err
}
}
revokeCtx := namespace.ContextWithNamespace(ctx, src.Namespace)
// Revoke all the dynamic keys
if err := c.expiration.RevokePrefix(revokeCtx, src.MountPath, true); err != nil {
return err
}
}
c.mountsLock.Lock()
if match := c.router.MountConflict(ctx, dstRelativePath); match != "" {
c.mountsLock.Unlock()
return fmt.Errorf("path in use at %q", match)
}
srcMatch.Tainted = false
srcMatch.NamespaceID = dst.Namespace.ID
srcMatch.namespace = dst.Namespace
srcPath := srcMatch.Path
srcMatch.Path = dst.MountPath
// Update the mount table
if err := c.persistMounts(ctx, c.mounts, &srcMatch.Local); err != nil {
srcMatch.Path = srcPath
srcMatch.Tainted = true
c.mountsLock.Unlock()
if err == logical.ErrReadOnly && c.perfStandby {
return err
}
return fmt.Errorf("failed to update mount table with error %+v", err)
}
// Remount the backend
if err := c.router.Remount(ctx, srcRelativePath, dstRelativePath); err != nil {
c.mountsLock.Unlock()
return err
}
c.mountsLock.Unlock()
// Un-taint the path
if err := c.router.Untaint(ctx, dstRelativePath); err != nil {
return err
}
return nil
}
// From an input path that has a relative namespace heirarchy followed by a mount point, return the full
// namespace of the mount point, along with the mount point without the namespace related prefix.
// For example, in a heirarchy ns1/ns2/ns3/secret-mount, when currNs is ns1 and path is ns2/ns3/secret-mount,
// this returns the namespace object for ns1/ns2/ns3/, and the string "secret-mount"
func (c *Core) splitNamespaceAndMountFromPath(currNs, path string) namespace.MountPathDetails {
fullPath := currNs + path
fullNs := c.namespaceByPath(fullPath)
mountPath := strings.TrimPrefix(fullPath, fullNs.Path)
return namespace.MountPathDetails{
Namespace: fullNs,
MountPath: sanitizePath(mountPath),
}
}
// loadMounts is invoked as part of postUnseal to load the mount table
func (c *Core) loadMounts(ctx context.Context) error {
// Load the existing mount table
raw, err := c.barrier.Get(ctx, coreMountConfigPath)
if err != nil {
c.logger.Error("failed to read mount table", "error", err)
return errLoadMountsFailed
}
rawLocal, err := c.barrier.Get(ctx, coreLocalMountConfigPath)
if err != nil {
c.logger.Error("failed to read local mount table", "error", err)
return errLoadMountsFailed
}
c.mountsLock.Lock()
defer c.mountsLock.Unlock()
if raw != nil {
// Check if the persisted value has canary in the beginning. If
// yes, decompress the table and then JSON decode it. If not,
// simply JSON decode it.
mountTable, err := c.decodeMountTable(ctx, raw.Value)
if err != nil {
c.logger.Error("failed to decompress and/or decode the mount table", "error", err)
return err
}
c.tableMetrics(len(mountTable.Entries), false, false, raw.Value)
c.mounts = mountTable
}
var needPersist bool
if c.mounts == nil {
c.logger.Info("no mounts; adding default mount table")
c.mounts = c.defaultMountTable()
needPersist = true
}
if rawLocal != nil {
localMountTable, err := c.decodeMountTable(ctx, rawLocal.Value)
if err != nil {
c.logger.Error("failed to decompress and/or decode the local mount table", "error", err)
return err
}
if localMountTable != nil && len(localMountTable.Entries) > 0 {
c.tableMetrics(len(localMountTable.Entries), true, false, rawLocal.Value)
c.mounts.Entries = append(c.mounts.Entries, localMountTable.Entries...)
}
}
// If this node is a performance standby we do not want to attempt to
// upgrade the mount table, this will be the active node's responsibility.
if !c.perfStandby {
err := c.runMountUpdates(ctx, needPersist)
if err != nil {
c.logger.Error("failed to run mount table upgrades", "error", err)
return err
}
}
for _, entry := range c.mounts.Entries {
if entry.NamespaceID == "" {
entry.NamespaceID = namespace.RootNamespaceID
}
ns, err := NamespaceByID(ctx, entry.NamespaceID, c)
if err != nil {
return err
}
if ns == nil {
return namespace.ErrNoNamespace
}
entry.namespace = ns
// Sync values to the cache
entry.SyncCache()
}
return nil
}
// Note that this is only designed to work with singletons, as it checks by
// type only.
func (c *Core) runMountUpdates(ctx context.Context, needPersist bool) error {
// Upgrade to typed mount table
if c.mounts.Type == "" {
c.mounts.Type = mountTableType
needPersist = true
}
for _, requiredMount := range c.requiredMountTable().Entries {
foundRequired := false
for _, coreMount := range c.mounts.Entries {
if coreMount.Type == requiredMount.Type {
foundRequired = true
coreMount.Config = requiredMount.Config
break
}
}
// In a replication scenario we will let sync invalidation take
// care of creating a new required mount that doesn't exist yet.
// This should only happen in the upgrade case where a new one is
// introduced on the primary; otherwise initial bootstrapping will
// ensure this comes over. If we upgrade first, we simply don't
// create the mount, so we won't conflict when we sync. If this is
// local (e.g. cubbyhole) we do still add it.
if !foundRequired && (!c.ReplicationState().HasState(consts.ReplicationPerformanceSecondary) || requiredMount.Local) {
c.mounts.Entries = append(c.mounts.Entries, requiredMount)
needPersist = true
}
}
// Upgrade to table-scoped entries
for _, entry := range c.mounts.Entries {
if !c.PR1103disabled && entry.Type == mountTypeNSCubbyhole && !entry.Local && !c.ReplicationState().HasState(consts.ReplicationPerformanceSecondary|consts.ReplicationDRSecondary) {
entry.Local = true
needPersist = true
}
if entry.Type == cubbyholeMountType && !entry.Local {
entry.Local = true
needPersist = true
}
if entry.Table == "" {
entry.Table = c.mounts.Type
needPersist = true
}
if entry.Accessor == "" {
accessor, err := c.generateMountAccessor(entry.Type)
if err != nil {
return err
}
entry.Accessor = accessor
needPersist = true
}
if entry.BackendAwareUUID == "" {
bUUID, err := uuid.GenerateUUID()
if err != nil {
return err
}
entry.BackendAwareUUID = bUUID
needPersist = true
}
if entry.NamespaceID == "" {
entry.NamespaceID = namespace.RootNamespaceID
needPersist = true
}
}
// Done if we have restored the mount table and we don't need
// to persist
if !needPersist {
return nil
}
// Persist both mount tables
if err := c.persistMounts(ctx, c.mounts, nil); err != nil {
c.logger.Error("failed to persist mount table", "error", err)
return errLoadMountsFailed
}
return nil
}
// persistMounts is used to persist the mount table after modification
func (c *Core) persistMounts(ctx context.Context, table *MountTable, local *bool) error {
if table.Type != mountTableType {
c.logger.Error("given table to persist has wrong type", "actual_type", table.Type, "expected_type", mountTableType)
return fmt.Errorf("invalid table type given, not persisting")
}
for _, entry := range table.Entries {
if entry.Table != table.Type {
c.logger.Error("given entry to persist in mount table has wrong table value", "path", entry.Path, "entry_table_type", entry.Table, "actual_type", table.Type)
return fmt.Errorf("invalid mount entry found, not persisting")
}
}
nonLocalMounts := &MountTable{
Type: mountTableType,
}
localMounts := &MountTable{
Type: mountTableType,
}
for _, entry := range table.Entries {
if entry.Local {
localMounts.Entries = append(localMounts.Entries, entry)
} else {
nonLocalMounts.Entries = append(nonLocalMounts.Entries, entry)
}
}
writeTable := func(mt *MountTable, path string) ([]byte, error) {
// Encode the mount table into JSON and compress it (lzw).
compressedBytes, err := jsonutil.EncodeJSONAndCompress(mt, nil)
if err != nil {
c.logger.Error("failed to encode or compress mount table", "error", err)
return nil, err
}
// Create an entry
entry := &logical.StorageEntry{
Key: path,
Value: compressedBytes,
}
// Write to the physical backend
if err := c.barrier.Put(ctx, entry); err != nil {
c.logger.Error("failed to persist mount table", "error", err)
return nil, err
}
return compressedBytes, nil
}
var err error
var compressedBytes []byte
switch {
case local == nil:
// Write non-local mounts
compressedBytes, err := writeTable(nonLocalMounts, coreMountConfigPath)
if err != nil {
return err
}
c.tableMetrics(len(nonLocalMounts.Entries), false, false, compressedBytes)
// Write local mounts
compressedBytes, err = writeTable(localMounts, coreLocalMountConfigPath)
if err != nil {
return err
}
c.tableMetrics(len(localMounts.Entries), true, false, compressedBytes)
case *local:
// Write local mounts
compressedBytes, err = writeTable(localMounts, coreLocalMountConfigPath)
if err != nil {
return err
}
c.tableMetrics(len(localMounts.Entries), true, false, compressedBytes)
default:
// Write non-local mounts
compressedBytes, err = writeTable(nonLocalMounts, coreMountConfigPath)
if err != nil {
return err
}
c.tableMetrics(len(nonLocalMounts.Entries), false, false, compressedBytes)
}
return nil
}
// setupMounts is invoked after we've loaded the mount table to
// initialize the logical backends and setup the router
func (c *Core) setupMounts(ctx context.Context) error {
c.mountsLock.Lock()
defer c.mountsLock.Unlock()
for _, entry := range c.mounts.sortEntriesByPathDepth().Entries {
// Initialize the backend, special casing for system
barrierPath := entry.ViewPath()
// Create a barrier view using the UUID
view := NewBarrierView(c.barrier, barrierPath)
// Singleton mounts cannot be filtered manually on a per-secondary basis
// from replication
if strutil.StrListContains(singletonMounts, entry.Type) {
addFilterablePath(c, barrierPath)
}
// Determining the replicated state of the mount
nilMount, err := preprocessMount(c, entry, view)
if err != nil {
return err
}
origReadOnlyErr := view.getReadOnlyErr()
// Mark the view as read-only until the mounting is complete and
// ensure that it is reset after. This ensures that there will be no
// writes during the construction of the backend.
view.setReadOnlyErr(logical.ErrSetupReadOnly)
if strutil.StrListContains(singletonMounts, entry.Type) {
defer view.setReadOnlyErr(origReadOnlyErr)
} else {
c.postUnsealFuncs = append(c.postUnsealFuncs, func() {
view.setReadOnlyErr(origReadOnlyErr)
})
}
var backend logical.Backend
// Create the new backend
sysView := c.mountEntrySysView(entry)
backend, err = c.newLogicalBackend(ctx, entry, sysView, view)
if err != nil {
c.logger.Error("failed to create mount entry", "path", entry.Path, "error", err)
if !c.builtinRegistry.Contains(entry.Type, consts.PluginTypeSecrets) {
// If we encounter an error instantiating the backend due to an error,
// skip backend initialization but register the entry to the mount table
// to preserve storage and path.
c.logger.Warn("skipping plugin-based mount entry", "path", entry.Path)
goto ROUTER_MOUNT
}
return errLoadMountsFailed
}
if backend == nil {
return fmt.Errorf("created mount entry of type %q is nil", entry.Type)
}
{
// Check for the correct backend type
backendType := backend.Type()
if backendType != logical.TypeLogical {
if entry.Type != "kv" && entry.Type != "system" && entry.Type != "cubbyhole" {
return fmt.Errorf(`unknown backend type: "%s"`, entry.Type)
}
}
addPathCheckers(c, entry, backend, barrierPath)
c.setCoreBackend(entry, backend, view)
}
// If the mount is filtered or we are on a DR secondary we don't want to
// keep the actual backend running, so we clean it up and set it to nil
// so the router does not have a pointer to the object.
if nilMount {
backend.Cleanup(ctx)
backend = nil
}
ROUTER_MOUNT:
// Mount the backend
err = c.router.Mount(backend, entry.Path, entry, view)
if err != nil {
c.logger.Error("failed to mount entry", "path", entry.Path, "error", err)
return errLoadMountsFailed
}
// Initialize
if !nilMount {
// Bind locally
localEntry := entry
c.postUnsealFuncs = append(c.postUnsealFuncs, func() {
if backend == nil {
c.logger.Error("skipping initialization on nil backend", "path", localEntry.Path)
return
}
err := backend.Initialize(ctx, &logical.InitializationRequest{Storage: view})
if err != nil {
c.logger.Error("failed to initialize mount entry", "path", localEntry.Path, "error", err)
}
})
}
if c.logger.IsInfo() {
c.logger.Info("successfully mounted backend", "type", entry.Type, "path", entry.Path)
}
// Ensure the path is tainted if set in the mount table
if entry.Tainted {
c.router.Taint(ctx, entry.Path)
}
// Ensure the cache is populated, don't need the result
NamespaceByID(ctx, entry.NamespaceID, c)
}
return nil
}
// unloadMounts is used before we seal the vault to reset the mounts to
// their unloaded state, calling Cleanup if defined. This is reversed by load and setup mounts.
func (c *Core) unloadMounts(ctx context.Context) error {
c.mountsLock.Lock()
defer c.mountsLock.Unlock()
if c.mounts != nil {
mountTable := c.mounts.shallowClone()
for _, e := range mountTable.Entries {
backend := c.router.MatchingBackend(namespace.ContextWithNamespace(ctx, e.namespace), e.Path)
if backend != nil {
backend.Cleanup(ctx)
}
viewPath := e.ViewPath()
removePathCheckers(c, e, viewPath)
}
}
c.mounts = nil
c.router.reset()
c.systemBarrierView = nil
return nil
}
// newLogicalBackend is used to create and configure a new logical backend by name
func (c *Core) newLogicalBackend(ctx context.Context, entry *MountEntry, sysView logical.SystemView, view logical.Storage) (logical.Backend, error) {
t := entry.Type
if alias, ok := mountAliases[t]; ok {
t = alias
}
f, ok := c.logicalBackends[t]
if !ok {
plug, err := c.pluginCatalog.Get(ctx, t, consts.PluginTypeSecrets, entry.Version)
if err != nil {
return nil, err
}
if plug == nil {
errContext := t
if entry.Version != "" {
errContext += fmt.Sprintf(", version=%s", entry.Version)
}
return nil, fmt.Errorf("%w: %s", ErrPluginNotFound, errContext)
}
f = plugin.Factory
if !plug.Builtin {
f = wrapFactoryCheckPerms(c, plugin.Factory)
}
}
// Set up conf to pass in plugin_name
conf := make(map[string]string)
for k, v := range entry.Options {
conf[k] = v
}
switch {
case entry.Type == "plugin":
conf["plugin_name"] = entry.Config.PluginName
default:
conf["plugin_name"] = t
}
conf["plugin_type"] = consts.PluginTypeSecrets.String()
conf["plugin_version"] = entry.Version
backendLogger := c.baseLogger.Named(fmt.Sprintf("secrets.%s.%s", t, entry.Accessor))
c.AddLogger(backendLogger)
config := &logical.BackendConfig{
StorageView: view,
Logger: backendLogger,
Config: conf,
System: sysView,
BackendUUID: entry.BackendAwareUUID,
}
ctx = context.WithValue(ctx, "core_number", c.coreNumber)
b, err := f(ctx, config)
if err != nil {
return nil, err
}
if b == nil {
return nil, fmt.Errorf("nil backend of type %q returned from factory", t)
}
addLicenseCallback(c, b)
return b, nil
}
// defaultMountTable creates a default mount table
func (c *Core) defaultMountTable() *MountTable {
table := &MountTable{
Type: mountTableType,
}
table.Entries = append(table.Entries, c.requiredMountTable().Entries...)
if os.Getenv("VAULT_INTERACTIVE_DEMO_SERVER") != "" {
mountUUID, err := uuid.GenerateUUID()
if err != nil {
panic(fmt.Sprintf("could not create default secret mount UUID: %v", err))
}
mountAccessor, err := c.generateMountAccessor("kv")
if err != nil {
panic(fmt.Sprintf("could not generate default secret mount accessor: %v", err))
}
bUUID, err := uuid.GenerateUUID()
if err != nil {
panic(fmt.Sprintf("could not create default secret mount backend UUID: %v", err))
}
kvMount := &MountEntry{
Table: mountTableType,
Path: "secret/",
Type: "kv",
Description: "key/value secret storage",
UUID: mountUUID,
Accessor: mountAccessor,
BackendAwareUUID: bUUID,
Options: map[string]string{
"version": "2",
},
RunningVersion: versions.GetBuiltinVersion(consts.PluginTypeSecrets, "kv"),
}
table.Entries = append(table.Entries, kvMount)
}
return table
}
// requiredMountTable() creates a mount table with entries required
// to be available
func (c *Core) requiredMountTable() *MountTable {
table := &MountTable{
Type: mountTableType,
}
cubbyholeUUID, err := uuid.GenerateUUID()
if err != nil {
panic(fmt.Sprintf("could not create cubbyhole UUID: %v", err))
}
cubbyholeAccessor, err := c.generateMountAccessor("cubbyhole")
if err != nil {
panic(fmt.Sprintf("could not generate cubbyhole accessor: %v", err))
}
cubbyholeBackendUUID, err := uuid.GenerateUUID()
if err != nil {
panic(fmt.Sprintf("could not create cubbyhole backend UUID: %v", err))
}
cubbyholeMount := &MountEntry{
Table: mountTableType,
Path: cubbyholeMountPath,
Type: cubbyholeMountType,
Description: "per-token private secret storage",
UUID: cubbyholeUUID,
Accessor: cubbyholeAccessor,
Local: true,
BackendAwareUUID: cubbyholeBackendUUID,
RunningVersion: versions.GetBuiltinVersion(consts.PluginTypeSecrets, "cubbyhole"),
}
sysUUID, err := uuid.GenerateUUID()
if err != nil {
panic(fmt.Sprintf("could not create sys UUID: %v", err))
}
sysAccessor, err := c.generateMountAccessor("system")
if err != nil {
panic(fmt.Sprintf("could not generate sys accessor: %v", err))
}
sysBackendUUID, err := uuid.GenerateUUID()
if err != nil {
panic(fmt.Sprintf("could not create sys backend UUID: %v", err))
}
sysMount := &MountEntry{
Table: mountTableType,
Path: "sys/",
Type: systemMountType,
Description: "system endpoints used for control, policy and debugging",
UUID: sysUUID,
Accessor: sysAccessor,
BackendAwareUUID: sysBackendUUID,
SealWrap: true, // Enable SealWrap since SystemBackend utilizes SealWrapStorage, see factory in addExtraLogicalBackends().
Config: MountConfig{
PassthroughRequestHeaders: []string{"Accept"},
},
RunningVersion: versions.DefaultBuiltinVersion,
}
identityUUID, err := uuid.GenerateUUID()
if err != nil {
panic(fmt.Sprintf("could not create identity mount entry UUID: %v", err))
}
identityAccessor, err := c.generateMountAccessor("identity")
if err != nil {
panic(fmt.Sprintf("could not generate identity accessor: %v", err))
}
identityBackendUUID, err := uuid.GenerateUUID()
if err != nil {
panic(fmt.Sprintf("could not create identity backend UUID: %v", err))
}
identityMount := &MountEntry{
Table: mountTableType,
Path: "identity/",
Type: "identity",
Description: "identity store",
UUID: identityUUID,
Accessor: identityAccessor,
BackendAwareUUID: identityBackendUUID,
Config: MountConfig{
PassthroughRequestHeaders: []string{"Authorization"},
},
RunningVersion: versions.DefaultBuiltinVersion,
}
table.Entries = append(table.Entries, cubbyholeMount)
table.Entries = append(table.Entries, sysMount)
table.Entries = append(table.Entries, identityMount)
return table
}
// This function returns tables that are singletons. The main usage of this is
// for replication, so we can send over mount info (especially, UUIDs of
// mounts, which are used for salts) for mounts that may not be able to be
// handled normally. After saving these values on the secondary, we let normal
// sync invalidation do its thing. Because of its use for replication, we
// exclude local mounts.
func (c *Core) singletonMountTables() (mounts, auth *MountTable) {
mounts = &MountTable{}
auth = &MountTable{}
c.mountsLock.RLock()
for _, entry := range c.mounts.Entries {
if strutil.StrListContains(singletonMounts, entry.Type) && !entry.Local && entry.Namespace().ID == namespace.RootNamespaceID {
mounts.Entries = append(mounts.Entries, entry)
}
}
c.mountsLock.RUnlock()
c.authLock.RLock()
for _, entry := range c.auth.Entries {
if strutil.StrListContains(singletonMounts, entry.Type) && !entry.Local && entry.Namespace().ID == namespace.RootNamespaceID {
auth.Entries = append(auth.Entries, entry)
}
}
c.authLock.RUnlock()
return
}
func (c *Core) setCoreBackend(entry *MountEntry, backend logical.Backend, view *BarrierView) {
switch entry.Type {
case systemMountType:
c.systemBackend = backend.(*SystemBackend)
c.systemBarrierView = view
case cubbyholeMountType:
ch := backend.(*CubbyholeBackend)
ch.saltUUID = entry.UUID
ch.storageView = view
c.cubbyholeBackend = ch
case identityMountType:
c.identityStore = backend.(*IdentityStore)
}
}
func (c *Core) createMigrationStatus(from, to namespace.MountPathDetails) (string, error) {
migrationID, err := uuid.GenerateUUID()
if err != nil {
return "", fmt.Errorf("error generating uuid for mount move invocation: %w", err)
}
migrationInfo := MountMigrationInfo{
SourceMount: from.Namespace.Path + from.MountPath,
TargetMount: to.Namespace.Path + to.MountPath,
MigrationStatus: MigrationInProgressStatus.String(),
}
c.mountMigrationTracker.Store(migrationID, migrationInfo)
return migrationID, nil
}
func (c *Core) setMigrationStatus(migrationID string, migrationStatus MountMigrationStatus) error {
migrationInfoRaw, ok := c.mountMigrationTracker.Load(migrationID)
if !ok {
return fmt.Errorf("Migration Tracker entry missing for ID %s", migrationID)
}
migrationInfo := migrationInfoRaw.(MountMigrationInfo)
migrationInfo.MigrationStatus = migrationStatus.String()
c.mountMigrationTracker.Store(migrationID, migrationInfo)
return nil
}
func (c *Core) readMigrationStatus(migrationID string) *MountMigrationInfo {
migrationInfoRaw, ok := c.mountMigrationTracker.Load(migrationID)
if !ok {
return nil
}
migrationInfo := migrationInfoRaw.(MountMigrationInfo)
return &migrationInfo
}