Update reaper for multipe subscribers

Depends on https://github.com/containerd/go-runc/pull/24

The is currently a race with the reaper where you could miss some exit
events from processes.

The problem before and why the reaper was so complex was because
processes could fork, getting a pid, and then fail on an execve before
we would have time to register the process with the reaper.  This could
cause pids to fill up in a map as a way to reduce the race.

This changes makes the reaper handle multiple subscribers so that the
caller can handle locking, for when they want to wait for a specific
pid, without affecting other callers using the reaper code.

Exit events are broadcast to multiple subscribers, in the case, the runc
commands and container pids that we get from a pid-file.  Locking while
the entire container stats no longs affects runc commands where you want
to call `runc create` and wait until that has been completed.

Signed-off-by: Michael Crosby <crosbymichael@gmail.com>

vendor/github.com/containerd/go-runc/monitor.go

@@ -3,10 +3,17 @@ package runc
 import (
 	"os/exec"
 	"syscall"
+	"time"
 )
 
 var Monitor ProcessMonitor = &defaultMonitor{}
 
+type Exit struct {
+	Timestamp time.Time
+	Pid       int
+	Status    int
+}
+
 // ProcessMonitor is an interface for process monitoring
 //
 // It allows daemons using go-runc to have a SIGCHLD handler
@@ -18,8 +25,8 @@ type ProcessMonitor interface {
 	Output(*exec.Cmd) ([]byte, error)
 	CombinedOutput(*exec.Cmd) ([]byte, error)
 	Run(*exec.Cmd) error
-	Start(*exec.Cmd) error
-	Wait(*exec.Cmd) (int, error)
+	Start(*exec.Cmd) (chan Exit, error)
+	Wait(*exec.Cmd, chan Exit) (int, error)
 }
 
 type defaultMonitor struct {
@@ -37,18 +44,32 @@ func (m *defaultMonitor) Run(c *exec.Cmd) error {
 	return c.Run()
 }
 
-func (m *defaultMonitor) Start(c *exec.Cmd) error {
-	return c.Start()
-}
-
-func (m *defaultMonitor) Wait(c *exec.Cmd) (int, error) {
-	if err := c.Wait(); err != nil {
-		if exitErr, ok := err.(*exec.ExitError); ok {
-			if status, ok := exitErr.Sys().(syscall.WaitStatus); ok {
-				return status.ExitStatus(), nil
+func (m *defaultMonitor) Start(c *exec.Cmd) (chan Exit, error) {
+	if err := c.Start(); err != nil {
+		return nil, err
+	}
+	ec := make(chan Exit, 1)
+	go func() {
+		var status int
+		if err := c.Wait(); err != nil {
+			status = 255
+			if exitErr, ok := err.(*exec.ExitError); ok {
+				if ws, ok := exitErr.Sys().(syscall.WaitStatus); ok {
+					status = ws.ExitStatus()
+				}
 			}
 		}
-		return -1, err
-	}
-	return 0, nil
+		ec <- Exit{
+			Timestamp: time.Now(),
+			Pid:       c.Process.Pid,
+			Status:    status,
+		}
+		close(ec)
+	}()
+	return ec, nil
+}
+
+func (m *defaultMonitor) Wait(c *exec.Cmd, ec chan Exit) (int, error) {
+	e := <-ec
+	return e.Status, nil
 }

vendor/github.com/containerd/go-runc/runc.go

@@ -41,7 +41,7 @@ type Runc struct {
 	PdeathSignal  syscall.Signal
 	Setpgid       bool
 	Criu          string
-	SystemdCgroup string
+	SystemdCgroup bool
 }
 
 // List returns all containers created inside the provided runc root directory
@@ -134,7 +134,8 @@ func (r *Runc) Create(context context.Context, id, bundle string, opts *CreateOp
 		}
 		return nil
 	}
-	if err := Monitor.Start(cmd); err != nil {
+	ec, err := Monitor.Start(cmd)
+	if err != nil {
 		return err
 	}
 	if opts != nil && opts.IO != nil {
@@ -144,7 +145,7 @@ func (r *Runc) Create(context context.Context, id, bundle string, opts *CreateOp
 			}
 		}
 	}
-	_, err := Monitor.Wait(cmd)
+	_, err = Monitor.Wait(cmd, ec)
 	return err
 }
 
@@ -209,7 +210,8 @@ func (r *Runc) Exec(context context.Context, id string, spec specs.Process, opts
 		}
 		return nil
 	}
-	if err := Monitor.Start(cmd); err != nil {
+	ec, err := Monitor.Start(cmd)
+	if err != nil {
 		return err
 	}
 	if opts != nil && opts.IO != nil {
@@ -219,7 +221,7 @@ func (r *Runc) Exec(context context.Context, id string, spec specs.Process, opts
 			}
 		}
 	}
-	_, err = Monitor.Wait(cmd)
+	_, err = Monitor.Wait(cmd, ec)
 	return err
 }
 
@@ -238,10 +240,11 @@ func (r *Runc) Run(context context.Context, id, bundle string, opts *CreateOpts)
 	if opts != nil {
 		opts.Set(cmd)
 	}
-	if err := Monitor.Start(cmd); err != nil {
+	ec, err := Monitor.Start(cmd)
+	if err != nil {
 		return -1, err
 	}
-	return Monitor.Wait(cmd)
+	return Monitor.Wait(cmd, ec)
 }
 
 type DeleteOpts struct {
@@ -294,13 +297,14 @@ func (r *Runc) Stats(context context.Context, id string) (*Stats, error) {
 	if err != nil {
 		return nil, err
 	}
-	defer func() {
-		rd.Close()
-		Monitor.Wait(cmd)
-	}()
-	if err := Monitor.Start(cmd); err != nil {
+	ec, err := Monitor.Start(cmd)
+	if err != nil {
 		return nil, err
 	}
+	defer func() {
+		rd.Close()
+		Monitor.Wait(cmd, ec)
+	}()
 	var e Event
 	if err := json.NewDecoder(rd).Decode(&e); err != nil {
 		return nil, err
@@ -315,7 +319,8 @@ func (r *Runc) Events(context context.Context, id string, interval time.Duration
 	if err != nil {
 		return nil, err
 	}
-	if err := Monitor.Start(cmd); err != nil {
+	ec, err := Monitor.Start(cmd)
+	if err != nil {
 		rd.Close()
 		return nil, err
 	}
@@ -327,7 +332,7 @@ func (r *Runc) Events(context context.Context, id string, interval time.Duration
 		defer func() {
 			close(c)
 			rd.Close()
-			Monitor.Wait(cmd)
+			Monitor.Wait(cmd, ec)
 		}()
 		for {
 			var e Event
@@ -505,7 +510,8 @@ func (r *Runc) Restore(context context.Context, id, bundle string, opts *Restore
 	if opts != nil {
 		opts.Set(cmd)
 	}
-	if err := Monitor.Start(cmd); err != nil {
+	ec, err := Monitor.Start(cmd)
+	if err != nil {
 		return -1, err
 	}
 	if opts != nil && opts.IO != nil {
@@ -515,7 +521,7 @@ func (r *Runc) Restore(context context.Context, id, bundle string, opts *Restore
 			}
 		}
 	}
-	return Monitor.Wait(cmd)
+	return Monitor.Wait(cmd, ec)
 }
 
 // Update updates the current container with the provided resource spec
@@ -596,8 +602,8 @@ func (r *Runc) args() (out []string) {
 	if r.Criu != "" {
 		out = append(out, "--criu", r.Criu)
 	}
-	if r.SystemdCgroup != "" {
-		out = append(out, "--systemd-cgroup", r.SystemdCgroup)
+	if r.SystemdCgroup {
+		out = append(out, "--systemd-cgroup")
 	}
 	return out
 }