deb.open-vault/vault/expiration.go

2849 lines
84 KiB
Go

// Copyright (c) HashiCorp, Inc.
// SPDX-License-Identifier: MPL-2.0
package vault
import (
"context"
"encoding/json"
"errors"
"fmt"
"math/rand"
"os"
"path"
"sort"
"strconv"
"strings"
"sync"
"sync/atomic"
"time"
metrics "github.com/armon/go-metrics"
"github.com/hashicorp/errwrap"
log "github.com/hashicorp/go-hclog"
multierror "github.com/hashicorp/go-multierror"
"github.com/hashicorp/go-secure-stdlib/base62"
"github.com/hashicorp/vault/helper/fairshare"
"github.com/hashicorp/vault/helper/locking"
"github.com/hashicorp/vault/helper/metricsutil"
"github.com/hashicorp/vault/helper/namespace"
"github.com/hashicorp/vault/sdk/framework"
"github.com/hashicorp/vault/sdk/helper/consts"
"github.com/hashicorp/vault/sdk/helper/jsonutil"
"github.com/hashicorp/vault/sdk/helper/locksutil"
"github.com/hashicorp/vault/sdk/logical"
"github.com/hashicorp/vault/vault/quotas"
uberAtomic "go.uber.org/atomic"
)
const (
// expirationSubPath is the sub-path used for the expiration manager
// view. This is nested under the system view.
expirationSubPath = "expire/"
// leaseViewPrefix is the prefix used for the ID based lookup of leases.
leaseViewPrefix = "id/"
// tokenViewPrefix is the prefix used for the token based lookup of leases.
tokenViewPrefix = "token/"
// maxRevokeAttempts limits how many revoke attempts are made
maxRevokeAttempts = 6
// revokeRetryBase is a baseline retry time
revokeRetryBase = 10 * time.Second
// maxLeaseDuration is the default maximum lease duration
maxLeaseTTL = 32 * 24 * time.Hour
// defaultLeaseDuration is the default lease duration used when no lease is specified
defaultLeaseTTL = maxLeaseTTL
// maxLeaseThreshold is the maximum lease count before generating log warning
maxLeaseThreshold = 256000
// numExpirationWorkersDefault is the maximum amount of workers working on lease expiration
numExpirationWorkersDefault = 200
// number of workers to use for general purpose testing
numExpirationWorkersTest = 10
fairshareWorkersOverrideVar = "VAULT_LEASE_REVOCATION_WORKERS"
// limit irrevocable error messages to 240 characters to be respectful of
// storage/memory
maxIrrevocableErrorLength = 240
genericIrrevocableErrorMessage = "unknown"
outOfRetriesMessage = "out of retries"
// maximum number of irrevocable leases we return to the irrevocable lease
// list API **without** the `force` flag set
MaxIrrevocableLeasesToReturn = 10000
MaxIrrevocableLeasesWarning = "Command halted because many irrevocable leases were found. To emit the entire list, re-run the command with force set true."
)
type pendingInfo struct {
// A subset of the lease entry, cached in memory
cachedLeaseInfo *leaseEntry
timer *time.Timer
revokesAttempted uint8
loginRole string
}
// ExpirationManager is used by the Core to manage leases. Secrets
// can provide a lease, meaning that they can be renewed or revoked.
// If a secret is not renewed in timely manner, it may be expired, and
// the ExpirationManager will handle doing automatic revocation.
type ExpirationManager struct {
core *Core
router *Router
idView *BarrierView
tokenView *BarrierView
tokenStore *TokenStore
logger log.Logger
// Although the data structure itself is atomic,
// pendingLock should be held to ensure lease modifications
// are atomic (with respect to storage, expiration time,
// and particularly the lease count.)
// The nonexpiring map holds entries for root tokens with
// TTL zero, which we want to count but have no timer associated.
pending sync.Map
nonexpiring sync.Map
leaseCount int
pendingLock locking.RWMutex
// A sync.Lock for every active leaseID
lockPerLease sync.Map
// Track expired leases that have been determined to be irrevocable (without
// manual intervention). We retain a subset of the lease info in memory
irrevocable sync.Map
// Track count for metrics reporting
// This value is protected by pendingLock
irrevocableLeaseCount int
// The uniquePolicies map holds policy sets, so they can
// be deduplicated. It is periodically emptied to prevent
// unbounded growth.
uniquePolicies map[string][]string
emptyUniquePolicies *time.Ticker
tidyLock *int32
restoreMode *int32
restoreModeLock sync.RWMutex
restoreRequestLock sync.RWMutex
restoreLocks []*locksutil.LockEntry
restoreLoaded sync.Map
quitCh chan struct{}
// do not hold coreStateLock in any API handler code - it is already held
coreStateLock locking.RWMutex
quitContext context.Context
leaseCheckCounter *uint32
logLeaseExpirations bool
expireFunc atomic.Pointer[ExpireLeaseStrategy]
// testRegisterAuthFailure, if set to true, triggers an explicit failure on
// RegisterAuth to simulate a partial failure during a token creation
// request. This value should only be set by tests.
testRegisterAuthFailure uberAtomic.Bool
jobManager *fairshare.JobManager
revokeRetryBase time.Duration
}
type ExpireLeaseStrategy func(context.Context, *ExpirationManager, string, *namespace.Namespace)
// revocationJob should only be created through newRevocationJob()
type revocationJob struct {
leaseID string
ns *namespace.Namespace
m *ExpirationManager
nsCtx context.Context
startTime time.Time
}
func newRevocationJob(nsCtx context.Context, leaseID string, ns *namespace.Namespace, m *ExpirationManager) (*revocationJob, error) {
if leaseID == "" {
return nil, fmt.Errorf("cannot have empty lease id")
}
if m == nil {
return nil, fmt.Errorf("cannot have nil expiration manager")
}
if nsCtx == nil {
return nil, fmt.Errorf("cannot have nil namespace context.Context")
}
return &revocationJob{
leaseID: leaseID,
ns: ns,
m: m,
nsCtx: nsCtx,
startTime: time.Now(),
}, nil
}
// errIsUnrecoverable returns true if the logical error is unlikely to resolve
// automatically or with additional retries
func errIsUnrecoverable(err error) bool {
switch {
case errors.Is(err, logical.ErrUnrecoverable),
errors.Is(err, logical.ErrUnsupportedOperation),
errors.Is(err, logical.ErrUnsupportedPath),
errors.Is(err, logical.ErrInvalidRequest):
return true
}
return false
}
func (r *revocationJob) Execute() error {
r.m.core.metricSink.IncrCounterWithLabels([]string{"expire", "lease_expiration"}, 1, []metrics.Label{metricsutil.NamespaceLabel(r.ns)})
r.m.core.metricSink.MeasureSinceWithLabels([]string{"expire", "lease_expiration", "time_in_queue"}, r.startTime, []metrics.Label{metricsutil.NamespaceLabel(r.ns)})
// don't start the timer until the revocation is being executed
revokeCtx, cancel := context.WithTimeout(r.nsCtx, DefaultMaxRequestDuration)
defer cancel()
go func() {
select {
case <-r.m.quitCh:
cancel()
case <-revokeCtx.Done():
}
}()
select {
case <-r.m.quitCh:
r.m.logger.Error("shutting down, not attempting further revocation of lease", "lease_id", r.leaseID)
return nil
case <-r.m.quitContext.Done():
r.m.logger.Error("core context canceled, not attempting further revocation of lease", "lease_id", r.leaseID)
return nil
default:
}
r.m.coreStateLock.RLock()
err := r.m.Revoke(revokeCtx, r.leaseID)
r.m.coreStateLock.RUnlock()
return err
}
func (r *revocationJob) OnFailure(err error) {
r.m.core.metricSink.IncrCounterWithLabels([]string{"expire", "lease_expiration", "error"}, 1, []metrics.Label{metricsutil.NamespaceLabel(r.ns)})
r.m.pendingLock.Lock()
pendingRaw, ok := r.m.pending.Load(r.leaseID)
r.m.pendingLock.Unlock()
if !ok {
r.m.logger.Warn("failed to find lease in pending map for revocation retry", "lease_id", r.leaseID)
return
}
pending := pendingRaw.(pendingInfo)
pending.revokesAttempted++
newTimer := r.revokeExponentialBackoff(pending.revokesAttempted)
if pending.revokesAttempted >= maxRevokeAttempts || errIsUnrecoverable(err) {
reason := "unrecoverable error"
if pending.revokesAttempted >= maxRevokeAttempts {
reason = "lease has consumed all retry attempts"
err = fmt.Errorf("%v: %w", outOfRetriesMessage, err)
}
r.m.logger.Trace("failed to revoke lease, marking lease as irrevocable", "lease_id", r.leaseID, "error", err, "reason", reason)
le, loadErr := r.m.loadEntry(r.nsCtx, r.leaseID)
if loadErr != nil {
r.m.logger.Warn("failed to mark lease as irrevocable - failed to load", "lease_id", r.leaseID, "err", loadErr)
return
}
if le == nil {
r.m.logger.Warn("failed to mark lease as irrevocable - nil lease", "lease_id", r.leaseID)
return
}
r.m.pendingLock.Lock()
r.m.markLeaseIrrevocable(r.nsCtx, le, err)
r.m.pendingLock.Unlock()
return
} else {
r.m.logger.Error("failed to revoke lease", "lease_id", r.leaseID, "error", err,
"attempts", pending.revokesAttempted, "next_attempt", newTimer)
}
pending.timer.Reset(newTimer)
r.m.pendingLock.Lock()
r.m.pending.Store(r.leaseID, pending)
r.m.pendingLock.Unlock()
}
func expireLeaseStrategyFairsharing(ctx context.Context, m *ExpirationManager, leaseID string, ns *namespace.Namespace) {
nsCtx := namespace.ContextWithNamespace(ctx, ns)
mountAccessor := m.getLeaseMountAccessorLocked(ctx, leaseID)
job, err := newRevocationJob(nsCtx, leaseID, ns, m)
if err != nil {
m.logger.Warn("error creating revocation job", "error", err)
return
}
m.jobManager.AddJob(job, mountAccessor)
}
func (r *revocationJob) revokeExponentialBackoff(attempt uint8) time.Duration {
exp := (1 << attempt) * r.m.revokeRetryBase
randomDelta := 0.5 * float64(exp)
// Allow backoff time to be a random value between exp +/- (0.5*exp)
backoffTime := (float64(exp) - randomDelta) + (rand.Float64() * (2 * randomDelta))
return time.Duration(backoffTime)
}
func getNumExpirationWorkers(c *Core, l log.Logger) int {
numWorkers := c.numExpirationWorkers
workerOverride := os.Getenv(fairshareWorkersOverrideVar)
if workerOverride != "" {
i, err := strconv.Atoi(workerOverride)
if err != nil {
l.Warn("vault lease revocation workers override must be an integer", "value", workerOverride)
} else if i < 1 || i > 10000 {
l.Warn("vault lease revocation workers override out of range", "value", i)
} else {
numWorkers = i
}
}
return numWorkers
}
// NewExpirationManager creates a new ExpirationManager that is backed
// using a given view, and uses the provided router for revocation.
func NewExpirationManager(c *Core, view *BarrierView, e ExpireLeaseStrategy, logger log.Logger, detectDeadlocks bool) *ExpirationManager {
managerLogger := logger.Named("job-manager")
jobManager := fairshare.NewJobManager("expire", getNumExpirationWorkers(c, logger), managerLogger, c.metricSink)
jobManager.Start()
c.AddLogger(managerLogger)
exp := &ExpirationManager{
core: c,
router: c.router,
idView: view.SubView(leaseViewPrefix),
tokenView: view.SubView(tokenViewPrefix),
tokenStore: c.tokenStore,
logger: logger,
pending: sync.Map{},
pendingLock: &locking.SyncRWMutex{},
nonexpiring: sync.Map{},
leaseCount: 0,
tidyLock: new(int32),
lockPerLease: sync.Map{},
uniquePolicies: make(map[string][]string),
emptyUniquePolicies: time.NewTicker(7 * 24 * time.Hour),
// new instances of the expiration manager will go immediately into
// restore mode
restoreMode: new(int32),
restoreLocks: locksutil.CreateLocks(),
quitCh: make(chan struct{}),
coreStateLock: c.stateLock,
quitContext: c.activeContext,
leaseCheckCounter: new(uint32),
logLeaseExpirations: os.Getenv("VAULT_SKIP_LOGGING_LEASE_EXPIRATIONS") == "",
jobManager: jobManager,
revokeRetryBase: c.expirationRevokeRetryBase,
}
exp.expireFunc.Store(&e)
if exp.revokeRetryBase == 0 {
exp.revokeRetryBase = revokeRetryBase
}
*exp.restoreMode = 1
if exp.logger == nil {
opts := log.LoggerOptions{Name: "expiration_manager"}
exp.logger = log.New(&opts)
}
if detectDeadlocks {
managerLogger.Debug("enabling deadlock detection")
exp.pendingLock = &locking.DeadlockRWMutex{}
}
go exp.uniquePoliciesGc()
return exp
}
// setupExpiration is invoked after we've loaded the mount table to
// initialize the expiration manager
func (c *Core) setupExpiration(e ExpireLeaseStrategy) error {
c.metricsMutex.Lock()
defer c.metricsMutex.Unlock()
// Create a sub-view
view := c.systemBarrierView.SubView(expirationSubPath)
// Create the manager
expLogger := c.baseLogger.Named("expiration")
c.AddLogger(expLogger)
detectDeadlocks := false
for _, v := range c.detectDeadlocks {
if v == "expiration" {
detectDeadlocks = true
}
}
mgr := NewExpirationManager(c, view, e, expLogger, detectDeadlocks)
c.expiration = mgr
// Link the token store to this
c.tokenStore.SetExpirationManager(mgr)
// Restore the existing state
c.logger.Info("restoring leases")
errorFunc := func() {
c.logger.Error("shutting down")
if err := c.Shutdown(); err != nil {
c.logger.Error("error shutting down core", "error", err)
}
}
go c.expiration.Restore(errorFunc)
quit := c.expiration.quitCh
go func() {
t := time.NewTimer(24 * time.Hour)
for {
select {
case <-quit:
return
case <-t.C:
c.expiration.attemptIrrevocableLeasesRevoke()
t.Reset(24 * time.Hour)
}
}
}()
return nil
}
// stopExpiration is used to stop the expiration manager before
// sealing the Vault.
func (c *Core) stopExpiration() error {
if c.expiration != nil {
if err := c.expiration.Stop(); err != nil {
return err
}
c.metricsMutex.Lock()
defer c.metricsMutex.Unlock()
c.expiration = nil
}
return nil
}
// lockLease takes out a lock for a given lease ID
func (m *ExpirationManager) lockLease(leaseID string) {
locksutil.LockForKey(m.restoreLocks, leaseID).Lock()
}
// unlockLease unlocks a given lease ID
func (m *ExpirationManager) unlockLease(leaseID string) {
locksutil.LockForKey(m.restoreLocks, leaseID).Unlock()
}
// inRestoreMode returns if we are currently in restore mode
func (m *ExpirationManager) inRestoreMode() bool {
return atomic.LoadInt32(m.restoreMode) == 1
}
func (m *ExpirationManager) invalidate(key string) {
switch {
case strings.HasPrefix(key, leaseViewPrefix):
leaseID := strings.TrimPrefix(key, leaseViewPrefix)
ctx := m.quitContext
_, nsID := namespace.SplitIDFromString(leaseID)
leaseNS := namespace.RootNamespace
var err error
if nsID != "" {
leaseNS, err = NamespaceByID(ctx, nsID, m.core)
if err != nil {
m.logger.Error("failed to invalidate lease entry", "error", err)
return
}
}
le, err := m.loadEntryInternal(namespace.ContextWithNamespace(ctx, leaseNS), leaseID, false, false)
if err != nil {
m.logger.Error("failed to invalidate lease entry", "error", err)
return
}
m.pendingLock.Lock()
defer m.pendingLock.Unlock()
info, ok := m.pending.Load(leaseID)
switch {
case ok:
switch {
case le == nil:
// Handle lease deletion
pending := info.(pendingInfo)
pending.timer.Stop()
m.pending.Delete(leaseID)
m.leaseCount--
// Avoid nil pointer dereference. Without cachedLeaseInfo we do not have enough information to
// accurately update quota lease information.
// Note that cachedLeaseInfo should never be nil under normal operation.
if pending.cachedLeaseInfo != nil {
leaseInfo := &quotas.QuotaLeaseInformation{LeaseId: leaseID, Role: pending.cachedLeaseInfo.LoginRole}
if err := m.core.quotasHandleLeases(ctx, quotas.LeaseActionDeleted, []*quotas.QuotaLeaseInformation{leaseInfo}); err != nil {
m.logger.Error("failed to update quota on lease invalidation", "error", err)
return
}
}
default:
// Update the lease in memory
m.updatePendingInternal(le)
}
default:
if le == nil {
// There is no entry in the pending map and the invalidation
// resulted in a nil entry. Therefore we should clean up the
// other maps, and update metrics/quotas if appropriate.
m.nonexpiring.Delete(leaseID)
if info, ok := m.irrevocable.Load(leaseID); ok {
ile := info.(*leaseEntry)
m.irrevocable.Delete(leaseID)
m.irrevocableLeaseCount--
m.leaseCount--
// Note that the leaseEntry should never be nil under normal operation.
if ile != nil {
leaseInfo := &quotas.QuotaLeaseInformation{LeaseId: leaseID, Role: ile.LoginRole}
if err := m.core.quotasHandleLeases(ctx, quotas.LeaseActionDeleted, []*quotas.QuotaLeaseInformation{leaseInfo}); err != nil {
m.logger.Error("failed to update quota on lease invalidation", "error", err)
return
}
}
}
return
}
// Handle lease update (if irrevocable) or creation (if pending)
m.updatePendingInternal(le)
}
}
}
// Tidy cleans up the dangling storage entries for leases. It scans the storage
// view to find all the available leases, checks if the token embedded in it is
// either empty or invalid and in both the cases, it revokes them. It also uses
// a token cache to avoid multiple lookups of the same token ID. It is normally
// not required to use the API that invokes this. This is only intended to
// clean up the corrupt storage due to bugs.
func (m *ExpirationManager) Tidy(ctx context.Context) error {
if m.inRestoreMode() {
return errors.New("cannot run tidy while restoring leases")
}
var tidyErrors *multierror.Error
logger := m.logger.Named("tidy")
m.core.AddLogger(logger)
if !atomic.CompareAndSwapInt32(m.tidyLock, 0, 1) {
logger.Warn("tidy operation on leases is already in progress")
return nil
}
defer atomic.CompareAndSwapInt32(m.tidyLock, 1, 0)
logger.Info("beginning tidy operation on leases")
defer logger.Info("finished tidy operation on leases")
// Create a cache to keep track of looked up tokens
tokenCache := make(map[string]bool)
var countLease, revokedCount, deletedCountInvalidToken, deletedCountEmptyToken int64
tidyFunc := func(leaseID string) {
countLease++
if countLease%500 == 0 {
logger.Info("tidying leases", "progress", countLease)
}
le, err := m.loadEntry(ctx, leaseID)
if err != nil {
tidyErrors = multierror.Append(tidyErrors, fmt.Errorf("failed to load the lease ID %q: %w", leaseID, err))
return
}
if le == nil {
tidyErrors = multierror.Append(tidyErrors, fmt.Errorf("nil entry for lease ID %q: %w", leaseID, err))
return
}
var isValid, ok bool
revokeLease := false
if le.ClientToken == "" {
logger.Debug("revoking lease which has an empty token", "lease_id", leaseID)
revokeLease = true
deletedCountEmptyToken++
goto REVOKE_CHECK
}
isValid, ok = tokenCache[le.ClientToken]
if !ok {
lock := locksutil.LockForKey(m.tokenStore.tokenLocks, le.ClientToken)
lock.RLock()
te, err := m.tokenStore.lookupInternal(ctx, le.ClientToken, false, true)
lock.RUnlock()
if err != nil {
tidyErrors = multierror.Append(tidyErrors, fmt.Errorf("failed to lookup token: %w", err))
return
}
if te == nil {
logger.Debug("revoking lease which holds an invalid token", "lease_id", leaseID)
revokeLease = true
deletedCountInvalidToken++
tokenCache[le.ClientToken] = false
} else {
tokenCache[le.ClientToken] = true
}
goto REVOKE_CHECK
} else {
if isValid {
return
}
logger.Debug("revoking lease which contains an invalid token", "lease_id", leaseID)
revokeLease = true
deletedCountInvalidToken++
goto REVOKE_CHECK
}
REVOKE_CHECK:
if revokeLease {
// Force the revocation and skip going through the token store
// again
leaseLock := m.lockForLeaseID(leaseID)
leaseLock.Lock()
err = m.revokeCommon(ctx, leaseID, true, true)
leaseLock.Unlock()
if err != nil {
tidyErrors = multierror.Append(tidyErrors, fmt.Errorf("failed to revoke an invalid lease with ID %q: %w", leaseID, err))
return
}
revokedCount++
}
}
ns, err := namespace.FromContext(ctx)
if err != nil {
return err
}
leaseView := m.leaseView(ns)
if err := logical.ScanView(m.quitContext, leaseView, tidyFunc); err != nil {
return err
}
logger.Info("number of leases scanned", "count", countLease)
logger.Info("number of leases which had empty tokens", "count", deletedCountEmptyToken)
logger.Info("number of leases which had invalid tokens", "count", deletedCountInvalidToken)
logger.Info("number of leases successfully revoked", "count", revokedCount)
return tidyErrors.ErrorOrNil()
}
// Restore is used to recover the lease states when starting.
// This is used after starting the vault.
func (m *ExpirationManager) Restore(errorFunc func()) (retErr error) {
defer func() {
// Turn off restore mode. We can do this safely without the lock because
// if restore mode finished successfully, restore mode was already
// disabled with the lock. In an error state, this will allow the
// Stop() function to shut everything down.
atomic.StoreInt32(m.restoreMode, 0)
switch {
case retErr == nil:
case strings.Contains(retErr.Error(), context.Canceled.Error()):
// Don't run error func because we lost leadership
m.logger.Warn("context canceled while restoring leases, stopping lease loading")
retErr = nil
case errwrap.Contains(retErr, ErrBarrierSealed.Error()):
// Don't run error func because we're likely already shutting down
m.logger.Warn("barrier sealed while restoring leases, stopping lease loading")
retErr = nil
default:
m.logger.Error("error restoring leases", "error", retErr)
if errorFunc != nil {
errorFunc()
}
}
}()
// Accumulate existing leases
m.logger.Debug("collecting leases")
existing, leaseCount, err := m.collectLeases()
if err != nil {
return err
}
m.logger.Debug("leases collected", "num_existing", leaseCount)
// Make the channels used for the worker pool
type lease struct {
namespace *namespace.Namespace
id string
}
broker := make(chan *lease)
quit := make(chan bool)
// Buffer these channels to prevent deadlocks
errs := make(chan error, leaseCount)
result := make(chan struct{}, leaseCount)
// Use a wait group
wg := &sync.WaitGroup{}
// Create 64 workers to distribute work to
for i := 0; i < consts.ExpirationRestoreWorkerCount; i++ {
wg.Add(1)
go func() {
defer wg.Done()
for {
select {
case lease, ok := <-broker:
// broker has been closed, we are done
if !ok {
return
}
ctx := namespace.ContextWithNamespace(m.quitContext, lease.namespace)
err := m.processRestore(ctx, lease.id)
if err != nil {
errs <- err
continue
}
// Send message that lease is done
result <- struct{}{}
// quit early
case <-quit:
return
case <-m.quitCh:
return
}
}
}()
}
// Distribute the collected keys to the workers in a go routine
wg.Add(1)
go func() {
defer wg.Done()
i := 0
for ns := range existing {
for _, leaseID := range existing[ns] {
i++
if i%500 == 0 {
m.logger.Debug("leases loading", "progress", i)
}
select {
case <-quit:
return
case <-m.quitCh:
return
default:
broker <- &lease{
namespace: ns,
id: leaseID,
}
}
}
}
// Close the broker, causing worker routines to exit
close(broker)
}()
// Ensure all keys on the chan are processed
LOOP:
for i := 0; i < leaseCount; i++ {
select {
case err = <-errs:
// Close all go routines
close(quit)
break LOOP
case <-m.quitCh:
close(quit)
break LOOP
case <-result:
}
}
// Let all go routines finish
wg.Wait()
if err != nil {
return err
}
m.restoreModeLock.Lock()
atomic.StoreInt32(m.restoreMode, 0)
m.restoreLoaded.Range(func(k, v interface{}) bool {
m.restoreLoaded.Delete(k)
return true
})
m.restoreLocks = nil
m.restoreModeLock.Unlock()
m.logger.Info("lease restore complete")
return nil
}
// processRestore takes a lease and restores it in the expiration manager if it has
// not already been seen.
// Once we load the lease, we also update the quotas that are keeping track of those leases
func (m *ExpirationManager) processRestore(ctx context.Context, leaseID string) error {
m.restoreRequestLock.RLock()
defer m.restoreRequestLock.RUnlock()
// Check if the lease has been seen
if _, ok := m.restoreLoaded.Load(leaseID); ok {
return nil
}
m.lockLease(leaseID)
defer m.unlockLease(leaseID)
// Check again with the lease locked
if _, ok := m.restoreLoaded.Load(leaseID); ok {
return nil
}
// Load lease and restore expiration timer
le, err := m.loadEntryInternal(ctx, leaseID, true, false)
if err != nil {
return err
}
// Update quotas with relevant lease information
if le != nil {
leaseInfo := &quotas.QuotaLeaseInformation{LeaseId: le.LeaseID, Role: le.LoginRole}
if err := m.core.quotasHandleLeases(context.Background(), quotas.LeaseActionLoaded, []*quotas.QuotaLeaseInformation{leaseInfo}); err != nil {
// We don't want to fail the start-up due to leases not being able
// to be loaded into the quota manager. When the leases get loaded,
// quota manager will be updated individually too.
m.logger.Error("failed to load lease into the quota sub-system", "LeaseID:", leaseID, "LoginRole", le.LoginRole, "error", err)
}
}
return nil
}
func expireNoop(ctx context.Context, manager *ExpirationManager, s string, n *namespace.Namespace) {}
// Stop is used to prevent further automatic revocations.
// This must be called before sealing the view.
func (m *ExpirationManager) Stop() error {
// Stop all the pending expiration timers
m.logger.Debug("stop triggered")
defer m.logger.Debug("finished stopping")
m.jobManager.Stop()
// Do this before stopping pending timers to avoid potential races with
// expiring timers
close(m.quitCh)
m.pendingLock.Lock()
// We don't want any lease timers that fire to do anything; they can wait
// for the next ExpirationManager to handle them.
newStrategy := ExpireLeaseStrategy(expireNoop)
m.expireFunc.Store(&newStrategy)
oldPending := &m.pending
m.pending, m.nonexpiring, m.irrevocable = sync.Map{}, sync.Map{}, sync.Map{}
m.leaseCount = 0
m.uniquePolicies = make(map[string][]string)
m.irrevocableLeaseCount = 0
m.pendingLock.Unlock()
go oldPending.Range(func(key, value interface{}) bool {
info := value.(pendingInfo)
// We need to stop the timers to prevent memory leaks.
info.timer.Stop()
return true
})
if m.inRestoreMode() {
for {
if !m.inRestoreMode() {
break
}
time.Sleep(10 * time.Millisecond)
}
}
m.emptyUniquePolicies.Stop()
return nil
}
// Revoke is used to revoke a secret named by the given LeaseID
func (m *ExpirationManager) Revoke(ctx context.Context, leaseID string) error {
defer metrics.MeasureSince([]string{"expire", "revoke"}, time.Now())
return m.revokeCommon(ctx, leaseID, false, false)
}
// LazyRevoke is used to queue revocation for a secret named by the given
// LeaseID. If the lease was not found it returns nil; if the lease was found
// it triggers a return of a 202.
func (m *ExpirationManager) LazyRevoke(ctx context.Context, leaseID string) error {
defer metrics.MeasureSince([]string{"expire", "lazy-revoke"}, time.Now())
return m.lazyRevokeInternal(ctx, leaseID)
}
// Mark a lease as expiring immediately
func (m *ExpirationManager) lazyRevokeInternal(ctx context.Context, leaseID string) error {
leaseLock := m.lockForLeaseID(leaseID)
leaseLock.Lock()
defer leaseLock.Unlock()
// Load the entry
le, err := m.loadEntry(ctx, leaseID)
if err != nil {
return err
}
// If there is no entry, nothing to revoke
if le == nil {
return nil
}
le.ExpireTime = time.Now()
if err := m.persistEntry(ctx, le); err != nil {
return err
}
m.updatePending(le)
return nil
}
// should be run on a schedule. something like once a day, maybe once a week
func (m *ExpirationManager) attemptIrrevocableLeasesRevoke() {
m.irrevocable.Range(func(k, v interface{}) bool {
leaseID := k.(string)
le := v.(*leaseEntry)
if le.ExpireTime.Add(time.Hour).Before(time.Now()) {
// if we get an error (or no namespace) note it, but continue attempting
// to revoke other leases
leaseNS, err := m.getNamespaceFromLeaseID(m.core.activeContext, leaseID)
if err != nil {
m.logger.Debug("could not get lease namespace from ID", "error", err)
return true
}
if leaseNS == nil {
m.logger.Debug("could not get lease namespace from ID: nil namespace")
return true
}
ctxWithNS := namespace.ContextWithNamespace(m.core.activeContext, leaseNS)
ctxWithNSAndTimeout, _ := context.WithTimeout(ctxWithNS, time.Minute)
if err := m.revokeCommon(ctxWithNSAndTimeout, leaseID, false, false); err != nil {
// on failure, force some delay to mitigate resource spike while
// this is running. if revocations succeed, we are okay with
// the higher resource consumption.
time.Sleep(10 * time.Millisecond)
}
}
return true
})
}
// revokeCommon does the heavy lifting. If force is true, we ignore a problem
// during revocation and still remove entries/index/lease timers
func (m *ExpirationManager) revokeCommon(ctx context.Context, leaseID string, force, skipToken bool) error {
defer metrics.MeasureSince([]string{"expire", "revoke-common"}, time.Now())
if !skipToken {
// Acquire lock for this lease
// If skipToken is true, then we're either being (1) called via RevokeByToken, so
// probably the lock is already held, and if we re-acquire we get deadlock, or
// (2) called by tidy, in which case the lock is held by the tidy thread.
leaseLock := m.lockForLeaseID(leaseID)
leaseLock.Lock()
defer leaseLock.Unlock()
}
// Load the entry
le, err := m.loadEntry(ctx, leaseID)
if err != nil {
return err
}
// If there is no entry, nothing to revoke
if le == nil {
return nil
}
// Revoke the entry
if !skipToken || le.Auth == nil {
if err := m.revokeEntry(ctx, le); err != nil {
if !force {
return err
}
if m.logger.IsWarn() {
m.logger.Warn("revocation from the backend failed, but in force mode so ignoring", "error", err)
}
}
}
// Delete the entry
if err := m.deleteEntry(ctx, le); err != nil {
return err
}
// Lease has been removed, also remove the in-memory lock.
m.deleteLockForLease(leaseID)
// Delete the secondary index, but only if it's a leased secret (not auth)
if le.Secret != nil {
var indexToken string
// Maintain secondary index by token, except for orphan batch tokens
switch le.ClientTokenType {
case logical.TokenTypeBatch:
te, err := m.tokenStore.lookupBatchTokenInternal(ctx, le.ClientToken)
if err != nil {
return err
}
// If it's a non-orphan batch token, assign the secondary index to its
// parent
indexToken = te.Parent
default:
indexToken = le.ClientToken
}
if indexToken != "" {
if err := m.removeIndexByToken(ctx, le, indexToken); err != nil {
return err
}
}
}
// Clear the expiration handler
m.pendingLock.Lock()
m.removeFromPending(ctx, leaseID, true)
m.nonexpiring.Delete(leaseID)
if _, ok := m.irrevocable.Load(le.LeaseID); ok {
m.irrevocable.Delete(leaseID)
m.irrevocableLeaseCount--
}
m.pendingLock.Unlock()
if m.logger.IsInfo() && !skipToken && m.logLeaseExpirations {
m.logger.Info("revoked lease", "lease_id", leaseID)
}
if m.logger.IsWarn() && !skipToken && le.isIncorrectlyNonExpiring() {
var accessor string
if le.Auth != nil {
accessor = le.Auth.Accessor
}
m.logger.Warn("finished revoking incorrectly non-expiring lease", "leaseID", le.LeaseID, "accessor", accessor)
}
return nil
}
// RevokeForce works similarly to RevokePrefix but continues in the case of a
// revocation error; this is mostly meant for recovery operations
func (m *ExpirationManager) RevokeForce(ctx context.Context, prefix string) error {
defer metrics.MeasureSince([]string{"expire", "revoke-force"}, time.Now())
return m.revokePrefixCommon(ctx, prefix, true, true)
}
// RevokePrefix is used to revoke all secrets with a given prefix.
// The prefix maps to that of the mount table to make this simpler
// to reason about.
func (m *ExpirationManager) RevokePrefix(ctx context.Context, prefix string, sync bool) error {
defer metrics.MeasureSince([]string{"expire", "revoke-prefix"}, time.Now())
return m.revokePrefixCommon(ctx, prefix, false, sync)
}
// RevokeByToken is used to revoke all the secrets issued with a given token.
// This is done by using the secondary index. It also removes the lease entry
// for the token itself. As a result it should *ONLY* ever be called from the
// token store's revokeInternal function.
// (NB: it's called by token tidy as well.)
func (m *ExpirationManager) RevokeByToken(ctx context.Context, te *logical.TokenEntry) error {
defer metrics.MeasureSince([]string{"expire", "revoke-by-token"}, time.Now())
tokenNS, err := NamespaceByID(ctx, te.NamespaceID, m.core)
if err != nil {
return err
}
if tokenNS == nil {
return namespace.ErrNoNamespace
}
tokenCtx := namespace.ContextWithNamespace(ctx, tokenNS)
// Lookup the leases
existing, err := m.lookupLeasesByToken(tokenCtx, te)
if err != nil {
return fmt.Errorf("failed to scan for leases: %w", err)
}
// Revoke all the keys by marking them expired
for _, leaseID := range existing {
err := m.lazyRevokeInternal(ctx, leaseID)
if err != nil {
return err
}
}
// te.Path should never be empty, but we check just in case
if te.Path != "" {
saltCtx := namespace.ContextWithNamespace(ctx, tokenNS)
saltedID, err := m.tokenStore.SaltID(saltCtx, te.ID)
if err != nil {
return err
}
tokenLeaseID := path.Join(te.Path, saltedID)
if tokenNS.ID != namespace.RootNamespaceID {
tokenLeaseID = fmt.Sprintf("%s.%s", tokenLeaseID, tokenNS.ID)
}
// We want to skip the revokeEntry call as that will call back into
// revocation logic in the token store, which is what is running this
// function in the first place -- it'd be a deadlock loop. Since the only
// place that this function is called is revokeSalted in the token store,
// we're already revoking the token, so we just want to clean up the lease.
// This avoids spurious revocations later in the log when the timer runs
// out, and eases up resource usage.
return m.revokeCommon(ctx, tokenLeaseID, false, true)
}
return nil
}
// revoke all leases under `prefix`
// if sync == true, revoke immediately (using a single worker).
// otherwise, mark the lease as expiring `now` and let the expiration manager
// queue it for revocation.
func (m *ExpirationManager) revokePrefixCommon(ctx context.Context, prefix string, force, sync bool) error {
if m.inRestoreMode() {
m.restoreRequestLock.Lock()
defer m.restoreRequestLock.Unlock()
}
// Ensure there is a trailing slash; or, if there is no slash, see if there
// is a matching specific ID
if !strings.HasSuffix(prefix, "/") {
le, err := m.loadEntry(ctx, prefix)
if err == nil && le != nil {
if sync {
if err := m.revokeCommon(ctx, prefix, force, false); err != nil {
return fmt.Errorf("failed to revoke %q: %w", prefix, err)
}
return nil
}
return m.LazyRevoke(ctx, prefix)
}
prefix = prefix + "/"
}
// Accumulate existing leases
ns, err := namespace.FromContext(ctx)
if err != nil {
return err
}
view := m.leaseView(ns)
sub := view.SubView(prefix)
existing, err := logical.CollectKeys(ctx, sub)
if err != nil {
return fmt.Errorf("failed to scan for leases: %w", err)
}
// Revoke all the keys
for idx, suffix := range existing {
leaseID := prefix + suffix
// No need to acquire per-lease lock here, one of these two will do it.
switch {
case sync:
if err := m.revokeCommon(ctx, leaseID, force, false); err != nil {
return fmt.Errorf("failed to revoke %q (%d / %d): %w", leaseID, idx+1, len(existing), err)
}
default:
if err := m.LazyRevoke(ctx, leaseID); err != nil {
return fmt.Errorf("failed to revoke %q (%d / %d): %w", leaseID, idx+1, len(existing), err)
}
}
}
return nil
}
// Renew is used to renew a secret using the given leaseID
// and a renew interval. The increment may be ignored.
func (m *ExpirationManager) Renew(ctx context.Context, leaseID string, increment time.Duration) (*logical.Response, error) {
defer metrics.MeasureSince([]string{"expire", "renew"}, time.Now())
// Acquire lock for this lease
leaseLock := m.lockForLeaseID(leaseID)
leaseLock.Lock()
defer leaseLock.Unlock()
// Load the entry
le, err := m.loadEntry(ctx, leaseID)
if err != nil {
return nil, err
}
// Check if the lease is renewable
if _, err := le.renewable(); err != nil {
return nil, err
}
if le.Secret == nil {
if le.Auth != nil {
return logical.ErrorResponse("tokens cannot be renewed through this endpoint"), nil
}
return logical.ErrorResponse("lease does not correspond to a secret"), nil
}
ns, err := namespace.FromContext(ctx)
if err != nil {
return nil, err
}
if ns.ID != le.namespace.ID {
return nil, errors.New("cannot renew a lease across namespaces")
}
sysViewCtx := namespace.ContextWithNamespace(ctx, le.namespace)
sysView := m.router.MatchingSystemView(sysViewCtx, le.Path)
if sysView == nil {
return nil, fmt.Errorf("unable to retrieve system view from router")
}
// Attempt to renew the entry
resp, err := m.renewEntry(ctx, le, increment)
if err != nil {
return nil, err
}
if resp == nil {
return nil, nil
}
if resp.IsError() {
return &logical.Response{
Data: resp.Data,
}, nil
}
if resp.Secret == nil {
return nil, nil
}
ttl, warnings, err := framework.CalculateTTL(sysView, increment, resp.Secret.TTL, 0, resp.Secret.MaxTTL, 0, le.IssueTime)
if err != nil {
return nil, err
}
for _, warning := range warnings {
resp.AddWarning(warning)
}
resp.Secret.TTL = ttl
// Attach the LeaseID
resp.Secret.LeaseID = leaseID
// Update the lease entry
le.Data = resp.Data
le.Secret = resp.Secret
le.ExpireTime = resp.Secret.ExpirationTime()
le.LastRenewalTime = time.Now()
// If the token it's associated with is a batch token, constrain lease
// times
if le.ClientTokenType == logical.TokenTypeBatch {
te, err := m.tokenStore.Lookup(ctx, le.ClientToken)
if err != nil {
return nil, err
}
if te == nil {
return nil, errors.New("cannot renew lease, no valid associated token")
}
tokenLeaseTimes, err := m.FetchLeaseTimesByToken(ctx, te)
if err != nil {
return nil, err
}
if tokenLeaseTimes == nil {
return nil, errors.New("failed to load batch token expiration time")
}
if le.ExpireTime.After(tokenLeaseTimes.ExpireTime) {
resp.Secret.TTL = tokenLeaseTimes.ExpireTime.Sub(le.LastRenewalTime)
le.ExpireTime = tokenLeaseTimes.ExpireTime
}
}
if err := m.persistEntry(ctx, le); err != nil {
return nil, err
}
// Update the expiration time
m.updatePending(le)
// Return the response
return resp, nil
}
// RenewToken is used to renew a token which does not need to
// invoke a logical backend.
func (m *ExpirationManager) RenewToken(ctx context.Context, req *logical.Request, te *logical.TokenEntry,
increment time.Duration,
) (*logical.Response, error) {
defer metrics.MeasureSince([]string{"expire", "renew-token"}, time.Now())
tokenNS, err := NamespaceByID(ctx, te.NamespaceID, m.core)
if err != nil {
return nil, err
}
if tokenNS == nil {
return nil, namespace.ErrNoNamespace
}
ns, err := namespace.FromContext(ctx)
if err != nil {
return nil, err
}
if ns.ID != tokenNS.ID {
return nil, errors.New("cannot renew a token across namespaces")
}
// Compute the Lease ID
saltedID, err := m.tokenStore.SaltID(ctx, te.ID)
if err != nil {
return nil, err
}
leaseID := path.Join(te.Path, saltedID)
if ns.ID != namespace.RootNamespaceID {
leaseID = fmt.Sprintf("%s.%s", leaseID, ns.ID)
}
// Acquire lock for this lease
leaseLock := m.lockForLeaseID(leaseID)
leaseLock.Lock()
defer leaseLock.Unlock()
// Load the entry
le, err := m.loadEntry(ctx, leaseID)
if err != nil {
return nil, err
}
if le == nil {
return logical.ErrorResponse("invalid lease ID"), logical.ErrInvalidRequest
}
// Check if the lease is renewable. Note that this also checks for a nil
// lease and errors in that case as well.
if _, err := le.renewable(); err != nil {
return logical.ErrorResponse(err.Error()), logical.ErrInvalidRequest
}
// Attempt to renew the auth entry
resp, err := m.renewAuthEntry(ctx, req, le, increment)
if err != nil {
return nil, err
}
if resp == nil {
return nil, nil
}
if resp.IsError() {
return &logical.Response{
Data: resp.Data,
}, nil
}
if resp.Auth == nil {
return nil, nil
}
sysViewCtx := namespace.ContextWithNamespace(ctx, le.namespace)
sysView := m.router.MatchingSystemView(sysViewCtx, le.Path)
if sysView == nil {
return nil, fmt.Errorf("unable to retrieve system view from router")
}
ttl, warnings, err := framework.CalculateTTL(sysView, increment, resp.Auth.TTL, resp.Auth.Period, resp.Auth.MaxTTL, resp.Auth.ExplicitMaxTTL, le.IssueTime)
if err != nil {
return nil, err
}
retResp := &logical.Response{}
for _, warning := range warnings {
retResp.AddWarning(warning)
}
resp.Auth.TTL = ttl
// Attach the ClientToken
resp.Auth.ClientToken = te.ID
// Refresh groups
if resp.Auth.EntityID != "" && m.core.identityStore != nil {
mountAccessor := ""
if resp.Auth.Alias != nil {
mountAccessor = resp.Auth.Alias.MountAccessor
}
validAliases, err := m.core.identityStore.refreshExternalGroupMembershipsByEntityID(ctx, resp.Auth.EntityID, resp.Auth.GroupAliases, mountAccessor)
if err != nil {
return nil, err
}
resp.Auth.GroupAliases = validAliases
}
// Update the lease entry
le.Auth = resp.Auth
le.ExpireTime = resp.Auth.ExpirationTime()
le.LastRenewalTime = time.Now()
if err := m.persistEntry(ctx, le); err != nil {
return nil, err
}
m.updatePending(le)
retResp.Auth = resp.Auth
return retResp, nil
}
// Register is used to take a request and response with an associated
// lease. The secret gets assigned a LeaseID and the management of
// the lease is assumed by the expiration manager.
func (m *ExpirationManager) Register(ctx context.Context, req *logical.Request, resp *logical.Response, loginRole string) (id string, retErr error) {
defer metrics.MeasureSince([]string{"expire", "register"}, time.Now())
te := req.TokenEntry()
if te == nil {
return "", fmt.Errorf("cannot register a lease with an empty client token")
}
// Ignore if there is no leased secret
if resp == nil || resp.Secret == nil {
return "", nil
}
// Validate the secret
if err := resp.Secret.Validate(); err != nil {
return "", err
}
// Create a lease entry
leaseRand, err := base62.Random(TokenLength)
if err != nil {
return "", err
}
ns, err := namespace.FromContext(ctx)
if err != nil {
return "", err
}
leaseID := path.Join(req.Path, leaseRand)
if ns.ID != namespace.RootNamespaceID {
leaseID = fmt.Sprintf("%s.%s", leaseID, ns.ID)
}
le := &leaseEntry{
LeaseID: leaseID,
ClientToken: req.ClientToken,
ClientTokenType: te.Type,
Path: req.Path,
Data: resp.Data,
Secret: resp.Secret,
LoginRole: loginRole,
IssueTime: time.Now(),
ExpireTime: resp.Secret.ExpirationTime(),
namespace: ns,
Version: 1,
}
var indexToken string
// Maintain secondary index by token, except for orphan batch tokens
switch {
case te.Type != logical.TokenTypeBatch:
indexToken = le.ClientToken
case te.Parent != "":
// If it's a non-orphan batch token, assign the secondary index to its
// parent
indexToken = te.Parent
}
defer func() {
// If there is an error we want to rollback as much as possible (note
// that errors here are ignored to do as much cleanup as we can). We
// want to revoke a generated secret (since an error means we may not
// be successfully tracking it), remove indexes, and delete the entry.
if retErr != nil {
revokeCtx := namespace.ContextWithNamespace(m.quitContext, ns)
revResp, err := m.router.Route(revokeCtx, logical.RevokeRequest(req.Path, resp.Secret, resp.Data))
if err != nil {
retErr = multierror.Append(retErr, fmt.Errorf("an additional internal error was encountered revoking the newly-generated secret: %w", err))
} else if revResp != nil && revResp.IsError() {
retErr = multierror.Append(retErr, fmt.Errorf("an additional error was encountered revoking the newly-generated secret: %w", revResp.Error()))
}
if err := m.deleteEntry(ctx, le); err != nil {
retErr = multierror.Append(retErr, fmt.Errorf("an additional error was encountered deleting any lease associated with the newly-generated secret: %w", err))
}
if err := m.removeIndexByToken(ctx, le, indexToken); err != nil {
retErr = multierror.Append(retErr, fmt.Errorf("an additional error was encountered removing lease indexes associated with the newly-generated secret: %w", err))
}
m.deleteLockForLease(leaseID)
}
}()
// If the token is a batch token, we want to constrain the maximum lifetime
// by the token's lifetime
if te.Type == logical.TokenTypeBatch {
tokenLeaseTimes, err := m.FetchLeaseTimesByToken(ctx, te)
if err != nil {
return "", err
}
if tokenLeaseTimes == nil {
return "", errors.New("failed to load batch token expiration time")
}
if le.ExpireTime.After(tokenLeaseTimes.ExpireTime) {
le.ExpireTime = tokenLeaseTimes.ExpireTime
}
}
// Acquire the lock here so persistEntry and updatePending are atomic,
// although it is *very unlikely* that anybody could grab the lease ID
// before this function returns. (They could find it in an index, or
// find it in a list.)
leaseLock := m.lockForLeaseID(leaseID)
leaseLock.Lock()
defer leaseLock.Unlock()
// Encode the entry
if err := m.persistEntry(ctx, le); err != nil {
return "", err
}
if indexToken != "" {
if err := m.createIndexByToken(ctx, le, indexToken); err != nil {
return "", err
}
}
// Setup revocation timer if there is a lease
m.updatePending(le)
// We round here because the clock will have already started
// ticking, so we'll end up always returning 299 instead of 300 or
// 26399 instead of 26400, say, even if it's just a few
// microseconds. This provides a nicer UX.
resp.Secret.TTL = le.ExpireTime.Sub(time.Now()).Round(time.Second)
// Done
return le.LeaseID, nil
}
// RegisterAuth is used to take an Auth response with an associated lease.
// The token does not get a LeaseID, but the lease management is handled by
// the expiration manager.
func (m *ExpirationManager) RegisterAuth(ctx context.Context, te *logical.TokenEntry, auth *logical.Auth, loginRole string) error {
defer metrics.MeasureSince([]string{"expire", "register-auth"}, time.Now())
// Triggers failure of RegisterAuth. This should only be set and triggered
// by tests to simulate partial failure during a token creation request.
if m.testRegisterAuthFailure.Load() {
return fmt.Errorf("failing explicitly on RegisterAuth")
}
authExpirationTime := auth.ExpirationTime()
if te.TTL == 0 && authExpirationTime.IsZero() && (len(te.Policies) != 1 || te.Policies[0] != "root") {
return errors.New("refusing to register a lease for a non-root token with no TTL")
}
if te.Type == logical.TokenTypeBatch {
return errors.New("cannot register a lease for a batch token")
}
if auth.ClientToken == "" {
return errors.New("cannot register an auth lease with an empty token")
}
if strings.Contains(te.Path, "..") {
return consts.ErrPathContainsParentReferences
}
tokenNS, err := NamespaceByID(ctx, te.NamespaceID, m.core)
if err != nil {
return err
}
if tokenNS == nil {
return namespace.ErrNoNamespace
}
saltCtx := namespace.ContextWithNamespace(ctx, tokenNS)
saltedID, err := m.tokenStore.SaltID(saltCtx, auth.ClientToken)
if err != nil {
return err
}
leaseID := path.Join(te.Path, saltedID)
if tokenNS.ID != namespace.RootNamespaceID {
leaseID = fmt.Sprintf("%s.%s", leaseID, tokenNS.ID)
}
// Create a lease entry
le := leaseEntry{
LeaseID: leaseID,
ClientToken: auth.ClientToken,
Auth: auth,
Path: te.Path,
LoginRole: loginRole,
IssueTime: time.Now(),
ExpireTime: authExpirationTime,
namespace: tokenNS,
Version: 1,
}
leaseLock := m.lockForLeaseID(leaseID)
leaseLock.Lock()
defer leaseLock.Unlock()
// Encode the entry
if err := m.persistEntry(ctx, &le); err != nil {
return err
}
// Setup revocation timer
m.updatePending(&le)
if strings.HasPrefix(auth.ClientToken, consts.ServiceTokenPrefix) {
generatedTokenEntry := logical.TokenEntry{Policies: auth.Policies}
tok := m.tokenStore.GenerateSSCTokenID(auth.ClientToken, logical.IndexStateFromContext(ctx), &generatedTokenEntry)
te.ExternalID = tok
}
return nil
}
// FetchLeaseTimesByToken is a helper function to use token values to compute
// the leaseID, rather than pushing that logic back into the token store.
// As a special case, for a batch token it simply returns the information
// encoded on it.
func (m *ExpirationManager) FetchLeaseTimesByToken(ctx context.Context, te *logical.TokenEntry) (*leaseEntry, error) {
defer metrics.MeasureSince([]string{"expire", "fetch-lease-times-by-token"}, time.Now())
if te == nil {
return nil, errors.New("cannot fetch lease times for nil token")
}
if te.Type == logical.TokenTypeBatch {
issueTime := time.Unix(te.CreationTime, 0)
return &leaseEntry{
IssueTime: issueTime,
ExpireTime: issueTime.Add(te.TTL),
ClientTokenType: logical.TokenTypeBatch,
}, nil
}
tokenNS, err := NamespaceByID(ctx, te.NamespaceID, m.core)
if err != nil {
return nil, err
}
if tokenNS == nil {
return nil, namespace.ErrNoNamespace
}
saltCtx := namespace.ContextWithNamespace(ctx, tokenNS)
saltedID, err := m.tokenStore.SaltID(saltCtx, te.ID)
if err != nil {
return nil, err
}
leaseID := path.Join(te.Path, saltedID)
if tokenNS.ID != namespace.RootNamespaceID {
leaseID = fmt.Sprintf("%s.%s", leaseID, tokenNS.ID)
}
return m.FetchLeaseTimes(ctx, leaseID)
}
// FetchLeaseTimes is used to fetch the issue time, expiration time, and last
// renewed time of a lease entry. It returns a leaseEntry itself, but with only
// those values copied over.
func (m *ExpirationManager) FetchLeaseTimes(ctx context.Context, leaseID string) (*leaseEntry, error) {
defer metrics.MeasureSince([]string{"expire", "fetch-lease-times"}, time.Now())
info, ok := m.pending.Load(leaseID)
if ok && info.(pendingInfo).cachedLeaseInfo != nil {
return m.leaseTimesForExport(info.(pendingInfo).cachedLeaseInfo), nil
}
info, ok = m.irrevocable.Load(leaseID)
if ok && info.(*leaseEntry) != nil {
return m.leaseTimesForExport(info.(*leaseEntry)), nil
}
// Load the entry
le, err := m.loadEntryInternal(ctx, leaseID, true, false)
if err != nil {
return nil, err
}
if le == nil {
return nil, nil
}
return m.leaseTimesForExport(le), nil
}
// Returns lease times for outside callers based on the full leaseEntry passed in
func (m *ExpirationManager) leaseTimesForExport(le *leaseEntry) *leaseEntry {
ret := &leaseEntry{
IssueTime: le.IssueTime,
ExpireTime: le.ExpireTime,
LastRenewalTime: le.LastRenewalTime,
}
if le.Secret != nil {
ret.Secret = &logical.Secret{}
ret.Secret.Renewable = le.Secret.Renewable
ret.Secret.TTL = le.Secret.TTL
}
if le.Auth != nil {
ret.Auth = &logical.Auth{}
ret.Auth.Renewable = le.Auth.Renewable
ret.Auth.TTL = le.Auth.TTL
}
return ret
}
// Restricts lease entry stored in pendingInfo to a low-cost subset of the
// information.
func (m *ExpirationManager) inMemoryLeaseInfo(le *leaseEntry) *leaseEntry {
ret := m.leaseTimesForExport(le)
// Need to index:
// namespace -- derived from lease ID
// policies -- stored in Auth object
// auth method -- derived from lease.Path
if le.Auth != nil {
// Ensure that list of policies is not copied more than
// once. This method is called with pendingLock held.
// We could use hashstructure here to generate a key, but that
// seems like it would be substantially slower?
key := strings.Join(le.Auth.Policies, "\n")
uniq, ok := m.uniquePolicies[key]
if ok {
ret.Auth.Policies = uniq
} else {
m.uniquePolicies[key] = le.Auth.Policies
ret.Auth.Policies = le.Auth.Policies
}
ret.Path = le.Path
}
if le.isIrrevocable() {
ret.RevokeErr = le.RevokeErr
}
ret.LoginRole = le.LoginRole
return ret
}
func (m *ExpirationManager) uniquePoliciesGc() {
for {
select {
case <-m.quitCh:
return
case <-m.emptyUniquePolicies.C:
}
// If the maximum lease is a month, and we blow away the unique
// policy cache every week, the pessimal case is 4x larger space
// utilization than keeping the cache indefinitely.
m.pendingLock.Lock()
m.uniquePolicies = make(map[string][]string)
m.pendingLock.Unlock()
}
}
// Placing a lock in pendingMap means that we need to work very hard on reload
// to only create one lock. Instead, we'll create locks on-demand in an atomic fashion.
//
// Acquiring a lock from a leaseEntry is a bad idea because it could change
// between loading and acquiring the lock. So we only provide an ID-based map, and the
// locking discipline should be:
// 1. Lock lease
// 2. Load, or attempt to load, leaseEntry
// 3. Modify leaseEntry and pendingMap (atomic wrt operations on this lease)
// 4. Unlock lease
//
// The lock must be removed from the map when the lease is deleted, or is
// found to not exist in storage. loadEntry does this whenever it returns
// nil, but we should also do it in revokeCommon().
func (m *ExpirationManager) lockForLeaseID(id string) *sync.Mutex {
mutex := &sync.Mutex{}
lock, _ := m.lockPerLease.LoadOrStore(id, mutex)
return lock.(*sync.Mutex)
}
func (m *ExpirationManager) deleteLockForLease(id string) {
m.lockPerLease.Delete(id)
}
// updatePending is used to update a pending invocation for a lease
func (m *ExpirationManager) updatePending(le *leaseEntry) {
m.pendingLock.Lock()
defer m.pendingLock.Unlock()
m.updatePendingInternal(le)
}
// updatePendingInternal is the locked version of updatePending; do not call
// this without a write lock on m.pending
func (m *ExpirationManager) updatePendingInternal(le *leaseEntry) {
// Check for an existing timer
info, leaseInPending := m.pending.Load(le.LeaseID)
var pending pendingInfo
if le.ExpireTime.IsZero() && le.nonexpiringToken() {
// Store this in the nonexpiring map instead of pending.
// There does not appear to be any cases where a token that had
// a nonzero can be can be assigned a zero TTL, but that can be
// handled by the next check
pending.cachedLeaseInfo = m.inMemoryLeaseInfo(le)
m.nonexpiring.Store(le.LeaseID, pending)
// if the timer happened to exist, stop the time and delete it from the
// pending timers.
if leaseInPending {
info.(pendingInfo).timer.Stop()
m.pending.Delete(le.LeaseID)
m.leaseCount--
// Avoid nil pointer dereference. Without cachedLeaseInfo we do not have enough information to
// accurately update quota lease information.
// Note that cachedLeaseInfo should never be nil under normal operation.
if pending.cachedLeaseInfo != nil {
leaseInfo := &quotas.QuotaLeaseInformation{LeaseId: le.LeaseID, Role: le.LoginRole}
if err := m.core.quotasHandleLeases(m.quitContext, quotas.LeaseActionDeleted, []*quotas.QuotaLeaseInformation{leaseInfo}); err != nil {
m.logger.Error("failed to update quota on lease deletion", "error", err)
return
}
}
}
return
}
leaseTotal := le.ExpireTime.Sub(time.Now())
leaseCreated := false
if le.isIrrevocable() {
// It's possible this function is being called to update the in-memory state
// for a lease from pending to irrevocable (we don't support the opposite).
// If this is the case, we need to know if the lease was previously counted
// so that we can maintain correct metric and quota lease counts.
_, leaseInIrrevocable := m.irrevocable.Load(le.LeaseID)
if !(leaseInPending || leaseInIrrevocable) {
leaseCreated = true
}
m.removeFromPending(m.quitContext, le.LeaseID, false)
m.irrevocable.Store(le.LeaseID, m.inMemoryLeaseInfo(le))
// Increment count if the lease was not present in the irrevocable map
// prior to being added to it above
if !leaseInIrrevocable {
m.irrevocableLeaseCount++
}
} else {
// Create entry if it does not exist or reset if it does
if leaseInPending {
pending = info.(pendingInfo)
pending.timer.Reset(leaseTotal)
// No change to lease count in this case
} else {
leaseID, namespace := le.LeaseID, le.namespace
// Extend the timer by the lease total
timer := time.AfterFunc(leaseTotal, func() {
expFn := *m.expireFunc.Load() // Load and deref the pointer
expFn(m.quitContext, m, leaseID, namespace)
})
pending = pendingInfo{
timer: timer,
}
leaseCreated = true
}
pending.loginRole = le.LoginRole
pending.cachedLeaseInfo = m.inMemoryLeaseInfo(le)
m.pending.Store(le.LeaseID, pending)
}
if leaseCreated {
m.leaseCount++
// If we're in restore mode, Vault is still starting. While we may get leases created, it is likely
// 'catching up' on old creates. There will be a core.quotasHandleLeases call to register these leases in
// restore process instead - in other words, !m.inRestoreMode() prevents double counting for quotas.
// We keep m.leaseCount++ out of this check as it is not called in processRestore
if !m.inRestoreMode() {
// Avoid nil pointer dereference. Without cachedLeaseInfo we do not have enough information to
// accurately update quota lease information.
// Note that cachedLeaseInfo should never be nil under normal operation.
if pending.cachedLeaseInfo != nil {
leaseInfo := &quotas.QuotaLeaseInformation{LeaseId: le.LeaseID, Role: le.LoginRole}
if err := m.core.quotasHandleLeases(m.quitContext, quotas.LeaseActionCreated, []*quotas.QuotaLeaseInformation{leaseInfo}); err != nil {
m.logger.Error("failed to update quota on lease creation", "error", err)
return
}
}
}
}
}
// revokeEntry is used to attempt revocation of an internal entry
func (m *ExpirationManager) revokeEntry(ctx context.Context, le *leaseEntry) error {
// Revocation of login tokens is special since we can by-pass the
// backend and directly interact with the token store
if le.Auth != nil {
if le.ClientTokenType == logical.TokenTypeBatch {
return errors.New("batch tokens cannot be revoked")
}
if err := m.tokenStore.revokeTree(ctx, le); err != nil {
return fmt.Errorf("failed to revoke token: %w", err)
}
return nil
}
if le.Secret != nil {
// not sure if this is really valid to have a leaseEntry with a nil Secret
// (if there's a nil Secret, what are you really leasing?), but the tests
// create one, and good to be defensive
le.Secret.IssueTime = le.IssueTime
}
// Make sure we're operating in the right namespace
nsCtx := namespace.ContextWithNamespace(ctx, le.namespace)
// Handle standard revocation via backends
resp, err := m.router.Route(nsCtx, logical.RevokeRequest(le.Path, le.Secret, le.Data))
if err != nil || (resp != nil && resp.IsError()) {
return fmt.Errorf("failed to revoke entry: resp: %#v err: %w", resp, err)
}
return nil
}
// renewEntry is used to attempt renew of an internal entry
func (m *ExpirationManager) renewEntry(ctx context.Context, le *leaseEntry, increment time.Duration) (*logical.Response, error) {
secret := *le.Secret
secret.IssueTime = le.IssueTime
secret.Increment = increment
secret.LeaseID = ""
// Make sure we're operating in the right namespace
nsCtx := namespace.ContextWithNamespace(ctx, le.namespace)
req := logical.RenewRequest(le.Path, &secret, le.Data)
resp, err := m.router.Route(nsCtx, req)
if err != nil || (resp != nil && resp.IsError()) {
return nil, fmt.Errorf("failed to renew entry: resp: %#v err: %w", resp, err)
}
return resp, nil
}
// renewAuthEntry is used to attempt renew of an auth entry. Only the token
// store should get the actual token ID intact.
func (m *ExpirationManager) renewAuthEntry(ctx context.Context, req *logical.Request, le *leaseEntry, increment time.Duration) (*logical.Response, error) {
if le.ClientTokenType == logical.TokenTypeBatch {
return logical.ErrorResponse("batch tokens cannot be renewed"), nil
}
auth := *le.Auth
auth.IssueTime = le.IssueTime
auth.Increment = increment
if strings.HasPrefix(le.Path, "auth/token/") {
auth.ClientToken = le.ClientToken
} else {
auth.ClientToken = ""
}
// Make sure we're operating in the right namespace
nsCtx := namespace.ContextWithNamespace(ctx, le.namespace)
authReq := logical.RenewAuthRequest(le.Path, &auth, nil)
authReq.Connection = req.Connection
resp, err := m.router.Route(nsCtx, authReq)
if err != nil {
return nil, fmt.Errorf("failed to renew entry: %w", err)
}
return resp, nil
}
// loadEntry is used to read a lease entry
func (m *ExpirationManager) loadEntry(ctx context.Context, leaseID string) (*leaseEntry, error) {
// Take out the lease locks after we ensure we are in restore mode
restoreMode := m.inRestoreMode()
if restoreMode {
m.restoreModeLock.RLock()
defer m.restoreModeLock.RUnlock()
restoreMode = m.inRestoreMode()
if restoreMode {
m.lockLease(leaseID)
defer m.unlockLease(leaseID)
}
}
_, nsID := namespace.SplitIDFromString(leaseID)
if nsID != "" {
leaseNS, err := NamespaceByID(ctx, nsID, m.core)
if err != nil {
return nil, err
}
if leaseNS != nil {
ctx = namespace.ContextWithNamespace(ctx, leaseNS)
}
} else {
ctx = namespace.ContextWithNamespace(ctx, namespace.RootNamespace)
}
// If a lease entry is nil, proactively delete the lease lock, in case we
// created one erroneously.
// If there was an error, we don't know whether the lease entry exists or not.
leaseEntry, err := m.loadEntryInternal(ctx, leaseID, restoreMode, true)
if err == nil && leaseEntry == nil {
m.deleteLockForLease(leaseID)
}
return leaseEntry, err
}
// loadEntryInternal is used when you need to load an entry but also need to
// control the lifecycle of the restoreLock
func (m *ExpirationManager) loadEntryInternal(ctx context.Context, leaseID string, restoreMode bool, checkRestored bool) (*leaseEntry, error) {
ns, err := namespace.FromContext(ctx)
if err != nil {
return nil, err
}
view := m.leaseView(ns)
out, err := view.Get(ctx, leaseID)
if err != nil {
return nil, fmt.Errorf("failed to read lease entry %s: %w", leaseID, err)
}
if out == nil {
return nil, nil
}
le, err := decodeLeaseEntry(out.Value)
if err != nil {
return nil, fmt.Errorf("failed to decode lease entry %s: %w", leaseID, err)
}
le.namespace = ns
if restoreMode {
if checkRestored {
// If we have already loaded this lease, we don't need to update on
// load. In the case of renewal and revocation, updatePending will be
// done after making the appropriate modifications to the lease.
if _, ok := m.restoreLoaded.Load(leaseID); ok {
return le, nil
}
}
// Update the cache of restored leases, either synchronously or through
// the lazy loaded restore process
m.restoreLoaded.Store(le.LeaseID, struct{}{})
// Setup revocation timer
m.updatePending(le)
}
return le, nil
}
// persistEntry is used to persist a lease entry
func (m *ExpirationManager) persistEntry(ctx context.Context, le *leaseEntry) error {
// Encode the entry
buf, err := le.encode()
if err != nil {
return fmt.Errorf("failed to encode lease entry: %w", err)
}
// Write out to the view
ent := logical.StorageEntry{
Key: le.LeaseID,
Value: buf,
}
if le.Auth != nil && len(le.Auth.Policies) == 1 && le.Auth.Policies[0] == "root" {
ent.SealWrap = true
}
view := m.leaseView(le.namespace)
if err := view.Put(ctx, &ent); err != nil {
return fmt.Errorf("failed to persist lease entry: %w", err)
}
return nil
}
// deleteEntry is used to delete a lease entry
func (m *ExpirationManager) deleteEntry(ctx context.Context, le *leaseEntry) error {
view := m.leaseView(le.namespace)
if err := view.Delete(ctx, le.LeaseID); err != nil {
return fmt.Errorf("failed to delete lease entry: %w", err)
}
return nil
}
// createIndexByToken creates a secondary index from the token to a lease entry
func (m *ExpirationManager) createIndexByToken(ctx context.Context, le *leaseEntry, token string) error {
tokenNS := namespace.RootNamespace
saltCtx := namespace.ContextWithNamespace(ctx, namespace.RootNamespace)
_, nsID := namespace.SplitIDFromString(token)
if nsID != "" {
var err error
tokenNS, err = NamespaceByID(ctx, nsID, m.core)
if err != nil {
return err
}
if tokenNS != nil {
saltCtx = namespace.ContextWithNamespace(ctx, tokenNS)
}
}
saltedID, err := m.tokenStore.SaltID(saltCtx, token)
if err != nil {
return err
}
leaseSaltedID, err := m.tokenStore.SaltID(saltCtx, le.LeaseID)
if err != nil {
return err
}
ent := logical.StorageEntry{
Key: saltedID + "/" + leaseSaltedID,
Value: []byte(le.LeaseID),
}
tokenView := m.tokenIndexView(tokenNS)
if err := tokenView.Put(ctx, &ent); err != nil {
return fmt.Errorf("failed to persist lease index entry: %w", err)
}
return nil
}
// indexByToken looks up the secondary index from the token to a lease entry
func (m *ExpirationManager) indexByToken(ctx context.Context, le *leaseEntry) (*logical.StorageEntry, error) {
tokenNS := namespace.RootNamespace
saltCtx := namespace.ContextWithNamespace(ctx, tokenNS)
_, nsID := namespace.SplitIDFromString(le.ClientToken)
if nsID != "" {
var err error
tokenNS, err = NamespaceByID(ctx, nsID, m.core)
if err != nil {
return nil, err
}
if tokenNS != nil {
saltCtx = namespace.ContextWithNamespace(ctx, tokenNS)
}
}
saltedID, err := m.tokenStore.SaltID(saltCtx, le.ClientToken)
if err != nil {
return nil, err
}
leaseSaltedID, err := m.tokenStore.SaltID(saltCtx, le.LeaseID)
if err != nil {
return nil, err
}
key := saltedID + "/" + leaseSaltedID
tokenView := m.tokenIndexView(tokenNS)
entry, err := tokenView.Get(ctx, key)
if err != nil {
return nil, fmt.Errorf("failed to look up secondary index entry")
}
return entry, nil
}
// removeIndexByToken removes the secondary index from the token to a lease entry
func (m *ExpirationManager) removeIndexByToken(ctx context.Context, le *leaseEntry, token string) error {
tokenNS := namespace.RootNamespace
saltCtx := namespace.ContextWithNamespace(ctx, namespace.RootNamespace)
_, nsID := namespace.SplitIDFromString(token)
if nsID != "" {
var err error
tokenNS, err = NamespaceByID(ctx, nsID, m.core)
if err != nil {
return err
}
if tokenNS != nil {
saltCtx = namespace.ContextWithNamespace(ctx, tokenNS)
}
// Downgrade logic for old-style (V0) namespace leases that had its
// secondary index live in the root namespace. This reverts to the old
// behavior of looking for the secondary index on these leases in the
// root namespace to be cleaned up properly. We set it here because the
// old behavior used the namespace's token store salt for its saltCtx.
if le.Version < 1 {
tokenNS = namespace.RootNamespace
}
}
saltedID, err := m.tokenStore.SaltID(saltCtx, token)
if err != nil {
return err
}
leaseSaltedID, err := m.tokenStore.SaltID(saltCtx, le.LeaseID)
if err != nil {
return err
}
key := saltedID + "/" + leaseSaltedID
tokenView := m.tokenIndexView(tokenNS)
if err := tokenView.Delete(ctx, key); err != nil {
return fmt.Errorf("failed to delete lease index entry: %w", err)
}
return nil
}
// CreateOrFetchRevocationLeaseByToken is used to create or fetch the matching
// leaseID for a particular token. The lease is set to expire immediately after
// it's created.
func (m *ExpirationManager) CreateOrFetchRevocationLeaseByToken(ctx context.Context, te *logical.TokenEntry) (string, error) {
// Fetch the saltedID of the token and construct the leaseID
tokenNS, err := NamespaceByID(ctx, te.NamespaceID, m.core)
if err != nil {
return "", err
}
if tokenNS == nil {
return "", namespace.ErrNoNamespace
}
saltCtx := namespace.ContextWithNamespace(ctx, tokenNS)
saltedID, err := m.tokenStore.SaltID(saltCtx, te.ID)
if err != nil {
return "", err
}
leaseID := path.Join(te.Path, saltedID)
if tokenNS.ID != namespace.RootNamespaceID {
leaseID = fmt.Sprintf("%s.%s", leaseID, tokenNS.ID)
}
// Load the entry
le, err := m.loadEntry(ctx, leaseID)
if err != nil {
return "", err
}
// If there's no associated leaseEntry for the token, we create one
if le == nil {
// Acquire the lock here so persistEntry and updatePending are atomic,
// although it is *very unlikely* that anybody could grab the lease ID
// before this function returns. (They could find it in an index, or
// find it in a list.)
leaseLock := m.lockForLeaseID(leaseID)
leaseLock.Lock()
defer leaseLock.Unlock()
auth := &logical.Auth{
ClientToken: te.ID,
LeaseOptions: logical.LeaseOptions{
TTL: time.Nanosecond,
},
}
if strings.Contains(te.Path, "..") {
return "", consts.ErrPathContainsParentReferences
}
// Create a lease entry
now := time.Now()
le = &leaseEntry{
LeaseID: leaseID,
ClientToken: auth.ClientToken,
Auth: auth,
Path: te.Path,
IssueTime: now,
ExpireTime: now.Add(time.Nanosecond),
namespace: tokenNS,
Version: 1,
}
// Encode the entry
if err := m.persistEntry(ctx, le); err != nil {
m.deleteLockForLease(leaseID)
return "", err
}
}
return le.LeaseID, nil
}
// lookupLeasesByToken is used to lookup all the leaseID's via the tokenID
func (m *ExpirationManager) lookupLeasesByToken(ctx context.Context, te *logical.TokenEntry) ([]string, error) {
tokenNS, err := NamespaceByID(ctx, te.NamespaceID, m.core)
if err != nil {
return nil, err
}
if tokenNS == nil {
return nil, namespace.ErrNoNamespace
}
saltCtx := namespace.ContextWithNamespace(ctx, tokenNS)
saltedID, err := m.tokenStore.SaltID(saltCtx, te.ID)
if err != nil {
return nil, err
}
tokenView := m.tokenIndexView(tokenNS)
// Scan via the index for sub-leases
prefix := saltedID + "/"
subKeys, err := tokenView.List(ctx, prefix)
if err != nil {
return nil, fmt.Errorf("failed to list leases: %w", err)
}
// Read each index entry
leaseIDs := make([]string, 0, len(subKeys))
for _, sub := range subKeys {
out, err := tokenView.Get(ctx, prefix+sub)
if err != nil {
return nil, fmt.Errorf("failed to read lease index: %w", err)
}
if out == nil {
continue
}
leaseIDs = append(leaseIDs, string(out.Value))
}
// Downgrade logic for old-style (V0) leases entries created by a namespace
// token that lived in the root namespace.
if tokenNS.ID != namespace.RootNamespaceID {
tokenView := m.tokenIndexView(namespace.RootNamespace)
// Scan via the index for sub-leases on the root namespace
prefix := saltedID + "/"
subKeys, err := tokenView.List(ctx, prefix)
if err != nil {
return nil, fmt.Errorf("failed to list leases on root namespace: %w", err)
}
for _, sub := range subKeys {
out, err := tokenView.Get(ctx, prefix+sub)
if err != nil {
return nil, fmt.Errorf("failed to read lease index on root namespace: %w", err)
}
if out == nil {
continue
}
leaseIDs = append(leaseIDs, string(out.Value))
}
}
return leaseIDs, nil
}
// emitMetrics is invoked periodically to emit statistics
func (m *ExpirationManager) emitMetrics() {
// All updates of these values are with the pendingLock held.
m.pendingLock.RLock()
allLeases := m.leaseCount
irrevocableLeases := m.irrevocableLeaseCount
m.pendingLock.RUnlock()
metrics.SetGauge([]string{"expire", "num_leases"}, float32(allLeases))
metrics.SetGauge([]string{"expire", "num_irrevocable_leases"}, float32(irrevocableLeases))
// Check if lease count is greater than the threshold
if allLeases > maxLeaseThreshold {
if atomic.LoadUint32(m.leaseCheckCounter) > 59 {
m.logger.Warn("lease count exceeds warning lease threshold", "have", allLeases, "threshold", maxLeaseThreshold)
atomic.StoreUint32(m.leaseCheckCounter, 0)
} else {
atomic.AddUint32(m.leaseCheckCounter, 1)
}
}
}
func (m *ExpirationManager) leaseAggregationMetrics(ctx context.Context, consts metricsutil.TelemetryConstConfig) ([]metricsutil.GaugeLabelValues, error) {
expiryTimes := make(map[metricsutil.LeaseExpiryLabel]int)
leaseEpsilon := consts.LeaseMetricsEpsilon
nsLabel := consts.LeaseMetricsNameSpaceLabels
rollingWindow := time.Now().Add(time.Duration(consts.NumLeaseMetricsTimeBuckets) * leaseEpsilon)
err := m.walkLeases(func(entryID string, expireTime time.Time) bool {
select {
// Abort and return empty collection if it's taking too much time, nonblocking check.
case <-ctx.Done():
return false
default:
if entryID == "" {
return true
}
_, nsID := namespace.SplitIDFromString(entryID)
if nsID == "" {
nsID = "root" // this is what metricsutil.NamespaceLabel does
}
label := metricsutil.ExpiryBucket(expireTime, leaseEpsilon, rollingWindow, nsID, nsLabel)
if label != nil {
expiryTimes[*label] += 1
}
return true
}
})
if err != nil {
return []metricsutil.GaugeLabelValues{}, suppressRestoreModeError(err)
}
// If collection was cancelled, return an empty array.
select {
case <-ctx.Done():
return []metricsutil.GaugeLabelValues{}, nil
default:
break
}
flattenedResults := make([]metricsutil.GaugeLabelValues, 0, len(expiryTimes))
for bucket, count := range expiryTimes {
if nsLabel {
flattenedResults = append(flattenedResults,
metricsutil.GaugeLabelValues{
Labels: []metrics.Label{{"expiring", bucket.LabelName}, {"namespace", bucket.LabelNS}},
Value: float32(count),
})
} else {
flattenedResults = append(flattenedResults,
metricsutil.GaugeLabelValues{
Labels: []metrics.Label{{"expiring", bucket.LabelName}},
Value: float32(count),
})
}
}
return flattenedResults, nil
}
// Callback function type to walk tokens referenced in the expiration
// manager. Don't want to use leaseEntry here because it's an unexported
// type (though most likely we would only call this from within the "vault" core package.)
type ExpirationWalkFunction = func(leaseID string, auth *logical.Auth, path string) bool
var ErrInRestoreMode = errors.New("expiration manager in restore mode")
// WalkTokens extracts the Auth structure from leases corresponding to tokens.
// Returning false from the walk function terminates the iteration.
func (m *ExpirationManager) WalkTokens(walkFn ExpirationWalkFunction) error {
if m.inRestoreMode() {
return ErrInRestoreMode
}
callback := func(key, value interface{}) bool {
p := value.(pendingInfo)
if p.cachedLeaseInfo == nil {
return true
}
lease := p.cachedLeaseInfo
if lease.Auth != nil {
return walkFn(key.(string), lease.Auth, lease.Path)
}
return true
}
m.pendingLock.RLock()
toWalk := []*sync.Map{&m.pending, &m.nonexpiring}
m.pendingLock.RUnlock()
for _, m := range toWalk {
m.Range(callback)
}
return nil
}
// leaseWalkFunction can only be used by the core package.
type leaseWalkFunction = func(leaseID string, expireTime time.Time) bool
func (m *ExpirationManager) walkLeases(walkFn leaseWalkFunction) error {
if m.inRestoreMode() {
return ErrInRestoreMode
}
callback := func(key, value interface{}) bool {
p := value.(pendingInfo)
if p.cachedLeaseInfo == nil {
return true
}
lease := p.cachedLeaseInfo
expireTime := lease.ExpireTime
return walkFn(key.(string), expireTime)
}
m.pendingLock.RLock()
toWalk := []*sync.Map{&m.pending, &m.nonexpiring}
m.pendingLock.RUnlock()
for _, m := range toWalk {
m.Range(callback)
}
return nil
}
// must be called with m.pendingLock held
// set decrementCounters true to decrement the lease count metric and quota
func (m *ExpirationManager) removeFromPending(ctx context.Context, leaseID string, decrementCounters bool) {
if info, ok := m.pending.Load(leaseID); ok {
pending := info.(pendingInfo)
pending.timer.Stop()
m.pending.Delete(leaseID)
if decrementCounters {
m.leaseCount--
// Avoid nil pointer dereference. Without cachedLeaseInfo we do not have enough information to
// accurately update quota lease information.
// Note that cachedLeaseInfo should never be nil under normal operation.
if pending.cachedLeaseInfo != nil {
leaseInfo := &quotas.QuotaLeaseInformation{LeaseId: leaseID, Role: pending.cachedLeaseInfo.LoginRole}
// Log but do not fail; unit tests (and maybe Tidy on production systems)
if err := m.core.quotasHandleLeases(ctx, quotas.LeaseActionDeleted, []*quotas.QuotaLeaseInformation{leaseInfo}); err != nil {
m.logger.Error("failed to update quota on revocation", "error", err)
}
}
}
}
}
// Marks a pending lease as irrevocable. Because the lease is being moved from
// pending to irrevocable, no total lease count metrics/quotas updates are needed.
// However, irrevocable lease count will need to be incremented
// note: must be called with pending lock held
func (m *ExpirationManager) markLeaseIrrevocable(ctx context.Context, le *leaseEntry, err error) {
if le == nil {
m.logger.Warn("attempted to mark nil lease as irrevocable")
return
}
if le.isIrrevocable() {
m.logger.Info("attempted to re-mark lease as irrevocable", "original_error", le.RevokeErr, "new_error", err.Error())
return
}
var errStr string
if err != nil {
errStr = err.Error()
}
if len(errStr) == 0 {
errStr = genericIrrevocableErrorMessage
}
if len(errStr) > maxIrrevocableErrorLength {
errStr = errStr[:maxIrrevocableErrorLength]
}
le.RevokeErr = errStr
m.persistEntry(ctx, le)
m.irrevocable.Store(le.LeaseID, m.inMemoryLeaseInfo(le))
m.irrevocableLeaseCount++
m.removeFromPending(ctx, le.LeaseID, false)
m.nonexpiring.Delete(le.LeaseID)
}
func (m *ExpirationManager) getNamespaceFromLeaseID(ctx context.Context, leaseID string) (*namespace.Namespace, error) {
_, nsID := namespace.SplitIDFromString(leaseID)
// avoid re-declaring leaseNS and err with scope inside the if
leaseNS := namespace.RootNamespace
var err error
if nsID != "" {
leaseNS, err = NamespaceByID(ctx, nsID, m.core)
if err != nil {
return nil, err
}
}
if leaseNS == nil {
return nil, namespace.ErrNoNamespace
}
return leaseNS, nil
}
func (m *ExpirationManager) getLeaseMountAccessorLocked(ctx context.Context, leaseID string) string {
m.coreStateLock.RLock()
defer m.coreStateLock.RUnlock()
return m.getLeaseMountAccessor(ctx, leaseID)
}
// note: this function must be called with m.coreStateLock held for read
func (m *ExpirationManager) getLeaseMountAccessor(ctx context.Context, leaseID string) string {
mount := m.core.router.MatchingMountEntry(ctx, leaseID)
var mountAccessor string
if mount == nil {
mountAccessor = "mount-accessor-not-found"
} else {
mountAccessor = mount.Accessor
}
return mountAccessor
}
func (m *ExpirationManager) getIrrevocableLeaseCounts(ctx context.Context, includeChildNamespaces bool) (map[string]interface{}, error) {
requestNS, err := namespace.FromContext(ctx)
if err != nil {
m.logger.Error("could not get namespace from context", "error", err)
return nil, err
}
numMatchingLeasesPerMount := make(map[string]int)
numMatchingLeases := 0
m.irrevocable.Range(func(k, v interface{}) bool {
leaseID := k.(string)
leaseNS, err := m.getNamespaceFromLeaseID(ctx, leaseID)
if err != nil {
// We should probably note that an error occured, but continue counting
m.logger.Warn("could not get lease namespace from ID", "error", err)
return true
}
leaseMatches := (leaseNS == requestNS) || (includeChildNamespaces && leaseNS.HasParent(requestNS))
if !leaseMatches {
// the lease doesn't meet our criteria, so keep looking
return true
}
mountAccessor := m.getLeaseMountAccessor(ctx, leaseID)
if _, ok := numMatchingLeasesPerMount[mountAccessor]; !ok {
numMatchingLeasesPerMount[mountAccessor] = 0
}
numMatchingLeases++
numMatchingLeasesPerMount[mountAccessor]++
return true
})
resp := make(map[string]interface{})
resp["lease_count"] = numMatchingLeases
resp["counts"] = numMatchingLeasesPerMount
return resp, nil
}
type leaseResponse struct {
LeaseID string `json:"lease_id"`
MountID string `json:"mount_id"`
ErrMsg string `json:"error"`
expireTime time.Time
}
// returns a warning string, if applicable
// limit specifies how many results to return, and must be >0
// includeAll specifies if all results should be returned, regardless of limit
func (m *ExpirationManager) listIrrevocableLeases(ctx context.Context, includeChildNamespaces, returnAll bool, limit int) (map[string]interface{}, string, error) {
requestNS, err := namespace.FromContext(ctx)
if err != nil {
m.logger.Error("could not get namespace from context", "error", err)
return nil, "", err
}
// map of mount point : lease info
matchingLeases := make([]*leaseResponse, 0)
numMatchingLeases := 0
var warning string
m.irrevocable.Range(func(k, v interface{}) bool {
leaseID := k.(string)
leaseInfo := v.(*leaseEntry)
leaseNS, err := m.getNamespaceFromLeaseID(ctx, leaseID)
if err != nil {
// We probably want to track that an error occured, but continue counting
m.logger.Warn("could not get lease namespace from ID", "error", err)
return true
}
leaseMatches := (leaseNS == requestNS) || (includeChildNamespaces && leaseNS.HasParent(requestNS))
if !leaseMatches {
// the lease doesn't meet our criteria, so keep looking
return true
}
if !returnAll && (numMatchingLeases >= limit) {
m.logger.Warn("hit max irrevocable leases without force flag set")
warning = MaxIrrevocableLeasesWarning
return false
}
mountAccessor := m.getLeaseMountAccessor(ctx, leaseID)
numMatchingLeases++
matchingLeases = append(matchingLeases, &leaseResponse{
LeaseID: leaseID,
MountID: mountAccessor,
ErrMsg: leaseInfo.RevokeErr,
expireTime: leaseInfo.ExpireTime,
})
return true
})
// sort the results for consistent API response. we primarily sort on
// increasing expire time, and break ties with increasing lease id
sort.Slice(matchingLeases, func(i, j int) bool {
if !matchingLeases[i].expireTime.Equal(matchingLeases[j].expireTime) {
return matchingLeases[i].expireTime.Before(matchingLeases[j].expireTime)
}
return matchingLeases[i].LeaseID < matchingLeases[j].LeaseID
})
resp := make(map[string]interface{})
resp["lease_count"] = numMatchingLeases
resp["leases"] = matchingLeases
return resp, warning, nil
}
// leaseEntry is used to structure the values the expiration
// manager stores. This is used to handle renew and revocation.
type leaseEntry struct {
LeaseID string `json:"lease_id"`
ClientToken string `json:"client_token"`
ClientTokenType logical.TokenType `json:"token_type"`
Path string `json:"path"`
Data map[string]interface{} `json:"data"`
Secret *logical.Secret `json:"secret"`
Auth *logical.Auth `json:"auth"`
IssueTime time.Time `json:"issue_time"`
ExpireTime time.Time `json:"expire_time"`
LastRenewalTime time.Time `json:"last_renewal_time"`
// LoginRole is used to indicate which login role (if applicable) this lease
// was created with. This is required to decrement lease count quotas
// based on login roles upon lease expiry.
LoginRole string `json:"login_role"`
// Version is used to track new different versions of leases. V0 (or
// zero-value) had non-root namespaced secondary indexes live in the root
// namespace, and V1 has secondary indexes live in the matching namespace.
Version int `json:"version"`
namespace *namespace.Namespace
// RevokeErr tracks if a lease has failed revocation in a way that is
// unlikely to be automatically resolved. The first time this happens,
// RevokeErr will be set, thus marking this leaseEntry as irrevocable. From
// there, it must be manually removed (force revoked).
RevokeErr string `json:"revokeErr"`
}
// encode is used to JSON encode the lease entry
func (le *leaseEntry) encode() ([]byte, error) {
return json.Marshal(le)
}
func (le *leaseEntry) renewable() (bool, error) {
switch {
// If there is no entry, cannot review to renew
case le == nil:
return false, fmt.Errorf("lease not found")
case le.isIrrevocable():
return false, fmt.Errorf("lease is expired and has failed previous revocation attempts")
case le.ExpireTime.IsZero():
return false, fmt.Errorf("lease is not renewable")
case le.ClientTokenType == logical.TokenTypeBatch:
return false, nil
// Determine if the lease is expired
case le.ExpireTime.Before(time.Now()):
return false, fmt.Errorf("lease expired")
// Determine if the lease is renewable
case le.Secret != nil && !le.Secret.Renewable:
return false, fmt.Errorf("lease is not renewable")
case le.Auth != nil && !le.Auth.Renewable:
return false, fmt.Errorf("lease is not renewable")
}
return true, nil
}
func (le *leaseEntry) ttl() int64 {
return int64(le.ExpireTime.Sub(time.Now().Round(time.Second)).Seconds())
}
func (le *leaseEntry) nonexpiringToken() bool {
if le.Auth == nil {
return false
}
// Note that at this time the only non-expiring tokens are root tokens, this test is more involved as it is trying
// to catch tokens created by the VAULT-1949 non-expiring tokens bug and ensure they become expiring.
return !le.Auth.LeaseEnabled() && len(le.Auth.Policies) == 1 && le.Auth.Policies[0] == "root" && le.namespace != nil &&
le.namespace.ID == namespace.RootNamespaceID
}
// TODO maybe lock RevokeErr once this goes in: https://github.com/hashicorp/vault/pull/11122
func (le *leaseEntry) isIrrevocable() bool {
return le.RevokeErr != ""
}
func (le *leaseEntry) isIncorrectlyNonExpiring() bool {
return le.ExpireTime.IsZero() && !le.nonexpiringToken()
}
// decodeLeaseEntry is used to reverse encode and return a new entry
func decodeLeaseEntry(buf []byte) (*leaseEntry, error) {
out := new(leaseEntry)
return out, jsonutil.DecodeJSON(buf, out)
}
func (e *ExpirationManager) DetectDeadlocks() bool {
if _, ok := e.pendingLock.(*locking.DeadlockRWMutex); ok {
return true
}
return false
}