Move RenumberLocks into runtime init

We can't do renumbering after init - we need to open a
potentially invalid locks file (too many/too few locks), and then
potentially delete the old locks and make new ones.

We need to be in init to bypass the checks that would otherwise
make this impossible.

This leaves us with two choices: make RenumberLocks a separate
entrypoint from NewRuntime, duplicating a lot of configuration
load code (we need to know where the locks live, how many there
are, etc) - or modify NewRuntime to allow renumbering during it.
Previous experience says the first is not really a viable option
and produces massive code bloat, so the second it is.

Signed-off-by: Matthew Heon <matthew.heon@pm.me>

3 files changed, 37 insertions, 29 deletions

diff --git a/libpod/options.go b/libpod/options.go
index 06737776b..4a3dd582d 100644
--- a/libpod/options.go
+++ b/libpod/options.go
@@ -394,6 +394,26 @@ func WithDefaultInfraCommand(cmd string) RuntimeOption {
 	}
 }
 
+// WithRenumber instructs libpod to perform a lock renumbering instead of a
+// normal init.
+// When this is specified, no valid runtime will be returned by NewRuntime.
+// Instead, libpod will reinitialize lock numbers on all pods and containers,
+// shut down the runtime, and return.
+// Renumber is intended to be used from a dedicated entrypoint, where it will
+// handle a changed maximum number of locks and return, with the program
+// exiting after that.
+func WithRenumber() RuntimeOption {
+	return func(rt *Runtime) error {
+		if rt.valid {
+			return ErrRuntimeFinalized
+		}
+
+		rt.doRenumber = true
+
+		return nil
+	}
+}
+
 // Container Creation Options
 
 // WithShmDir sets the directory that should be mounted on /dev/shm.
diff --git a/libpod/runtime.go b/libpod/runtime.go
index 4f5d1e292..bc7c061c4 100644
--- a/libpod/runtime.go
+++ b/libpod/runtime.go
@@ -79,7 +79,8 @@ type RuntimeOption func(*Runtime) error
 
 // Runtime is the core libpod runtime
 type Runtime struct {
-	config          *RuntimeConfig
+	config *RuntimeConfig
+
 	state           State
 	store           storage.Store
 	storageService  *storageService
@@ -88,12 +89,15 @@ type Runtime struct {
 	netPlugin       ocicni.CNIPlugin
 	ociRuntimePath  OCIRuntimePath
 	conmonPath      string
-	valid           bool
-	lock            sync.RWMutex
 	imageRuntime    *image.Runtime
 	firewallBackend firewall.FirewallBackend
 	lockManager     lock.Manager
 	configuredFrom  *runtimeConfiguredFrom
+
+	doRenumber bool
+
+	valid bool
+	lock  sync.RWMutex
 }
 
 // OCIRuntimePath contains information about an OCI runtime.
@@ -753,6 +757,13 @@ func makeRuntime(runtime *Runtime) (err error) {
 			aliveLock.Unlock()
 		}
 	}()
+	// If we're renumbering locks, do it now.
+	// It breaks out of normal runtime init, and will not return a valid
+	// runtime.
+	if runtime.doRenumber {
+		return runtime.renumberLocks()
+	}
+
 	_, err = os.Stat(runtimeAliveFile)
 	if err != nil {
 		// If the file doesn't exist, we need to refresh the state
diff --git a/libpod/runtime_renumber.go b/libpod/runtime_renumber.go
index d5279934e..04abc84d1 100644
--- a/libpod/runtime_renumber.go
+++ b/libpod/runtime_renumber.go
@@ -1,14 +1,11 @@
 package libpod
 
 import (
-	"path/filepath"
-
-	"github.com/containers/storage"
 	"github.com/pkg/errors"
 )
 
-// RenumberLocks reassigns lock numbers for all containers, pods, and volumes in
-// the state.
+// renumberLocks reassigns lock numbers for all containers and pods in the
+// state.
 // It renders the runtime it is called on, and all container/pod/volume structs
 // from that runtime, unusable, and requires that a new runtime be initialized
 // after it is called.
@@ -18,24 +15,7 @@ import (
 // lock as read, renumber attempting to take a write lock?
 // The alternative is some sort of session tracking, and I don't know how
 // reliable that can be.
-func (r *Runtime) RenumberLocks() error {
-	r.lock.Lock()
-	locked := true
-	defer func() {
-		if locked {
-			r.lock.Unlock()
-		}
-	}()
-
-	runtimeAliveLock := filepath.Join(r.config.TmpDir, "alive.lck")
-	aliveLock, err := storage.GetLockfile(runtimeAliveLock)
-	if err != nil {
-		return errors.Wrapf(err, "error acquiring runtime init lock")
-	}
-	aliveLock.Lock()
-	// It's OK to defer until Shutdown() has run, so no need to check locked
-	defer aliveLock.Unlock()
-
+func (r *Runtime) renumberLocks() error {
 	// Start off by deallocating all locks
 	if err := r.lockManager.FreeAllLocks(); err != nil {
 		return err
@@ -76,8 +56,5 @@ func (r *Runtime) RenumberLocks() error {
 		}
 	}
 
-	r.lock.Unlock()
-	locked = false
-
 	return r.Shutdown(false)
 }