pod: ps does not race with rm

the "pod ps" command first retrieves the list of all pods, then
iterates over the list to inspect each pod.  This introduce a race
since a pod could be deleted in the meanwhile by another process.

Solve it by ignoring the define.ErrNoSuchPod error.

Closes: https://github.com/containers/podman/issues/14736

Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>

pkg/domain/infra/abi/pods.go

@@ -483,6 +483,9 @@ func (ic *ContainerEngine) PodPs(ctx context.Context, options entities.PodPSOpti
 	for _, p := range pds {
 		r, err := ic.listPodReportFromPod(p)
 		if err != nil {
+			if errors.Is(err, define.ErrNoSuchPod) || errors.Is(err, define.ErrNoSuchCtr) {
+				continue
+			}
 			return nil, err
 		}
 		reports = append(reports, r)

test/system/200-pod.bats

@@ -495,7 +495,22 @@ spec:
     local actual2=$(< /sys/fs/cgroup/$path2/cpu.max)
     is "$actual2" "500000 100000" "resource limits set properly"
     run_podman --cgroup-manager=cgroupfs pod rm $name2
+}
 
+@test "podman pod ps doesn't race with pod rm" {
+    # create a few pods
+    for i in {0..10}; do
+        run_podman pod create
+    done
+
+    # and delete them
+    $PODMAN pod rm -a &
+
+    # pod ps should not fail while pods are deleted
+    run_podman pod ps -q
+
+    # wait for pod rm -a
+    wait
 }
 
 # vim: filetype=sh