Add basic deadlock detection for container start/remove

We can easily tell if we're going to deadlock by comparing lock
IDs before actually taking the lock. Add a few checks for this in
common places where deadlocks might occur.

This does not yet cover pod operations, where detection is more
difficult (and costly) due to the number of locks being involved
being higher than 2.

Also, add some error wrapping on the Podman side, so we can tell
people to use `system renumber` when it occurs.

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

5 files changed, 14 insertions, 3 deletions

diff --git a/libpod/container_internal.go b/libpod/container_internal.go
index 11f9721dc..ff43bfc8f 100644
--- a/libpod/container_internal.go
+++ b/libpod/container_internal.go
@@ -1401,6 +1401,9 @@ func (c *Container) mountNamedVolume(v *ContainerNamedVolume, mountpoint string)
 		return nil, errors.Wrapf(err, "error retrieving named volume %s for container %s", v.Name, c.ID())
 	}
 
+	if vol.config.LockID == c.config.LockID {
+		return nil, errors.Wrapf(define.ErrWillDeadlock, "container %s and volume %s share lock ID %d", c.ID(), vol.Name(), c.config.LockID)
+	}
 	vol.lock.Lock()
 	defer vol.lock.Unlock()
 	if vol.needsMount() {
diff --git a/libpod/define/errors.go b/libpod/define/errors.go
index 523062866..b79cf08dc 100644
--- a/libpod/define/errors.go
+++ b/libpod/define/errors.go
@@ -61,6 +61,11 @@ var (
 	// the user.
 	ErrDetach = utils.ErrDetach
 
+	// ErrWillDeadlock indicates that the requested operation will cause a
+	// deadlock. This is usually caused by upgrade issues, and is resolved
+	// by renumbering the locks.
+	ErrWillDeadlock = errors.New("deadlock due to lock mismatch")
+
 	// ErrNoCgroups indicates that the container does not have its own
 	// CGroup.
 	ErrNoCgroups = errors.New("this container does not have a cgroup")
diff --git a/libpod/runtime_ctr.go b/libpod/runtime_ctr.go
index 3ad09f27c..39284026c 100644
--- a/libpod/runtime_ctr.go
+++ b/libpod/runtime_ctr.go
@@ -412,6 +412,9 @@ func (r *Runtime) removeContainer(ctx context.Context, c *Container, force bool,
 		}
 
 		// Lock the pod while we're removing container
+		if pod.config.LockID == c.config.LockID {
+			return errors.Wrapf(define.ErrWillDeadlock, "container %s and pod %s share lock ID %d", c.ID(), pod.ID(), c.config.LockID)
+		}
 		pod.lock.Lock()
 		defer pod.lock.Unlock()
 		if err := pod.updatePod(); err != nil {
diff --git a/libpod/runtime_volume.go b/libpod/runtime_volume.go
index 835dccf9c..efc3c5bd9 100644
--- a/libpod/runtime_volume.go
+++ b/libpod/runtime_volume.go
@@ -36,9 +36,6 @@ func (r *Runtime) RemoveVolume(ctx context.Context, v *Volume, force bool) error
 		}
 	}
 
-	v.lock.Lock()
-	defer v.lock.Unlock()
-
 	return r.removeVolume(ctx, v, force)
 }
 
diff --git a/libpod/runtime_volume_linux.go b/libpod/runtime_volume_linux.go
index 037cf4cc2..e9cfda9d4 100644
--- a/libpod/runtime_volume_linux.go
+++ b/libpod/runtime_volume_linux.go
@@ -124,6 +124,9 @@ func (r *Runtime) removeVolume(ctx context.Context, v *Volume, force bool) error
 		return define.ErrVolumeRemoved
 	}
 
+	v.lock.Lock()
+	defer v.lock.Unlock()
+
 	// Update volume status to pick up a potential removal from state
 	if err := v.update(); err != nil {
 		return err