Remove resize race condition

Since podman-remote resize requests can come in at random times, this
generates a real potential for race conditions. We should only be
attempting to resize TTY on running containers, but the containers can
go from running to stopped at any time, and returning an error to the
caller is just causing noice.

This change will basically ignore requests to resize terminals if the
container is not running and return the caller to success.  All other
callers will still return failure.

Fixes: https://github.com/containers/podman/issues/9831

Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>

7 files changed, 44 insertions, 13 deletions

diff --git a/pkg/api/handlers/compat/resize.go b/pkg/api/handlers/compat/resize.go
index 1bf7ad460..23ed33a22 100644
--- a/pkg/api/handlers/compat/resize.go
+++ b/pkg/api/handlers/compat/resize.go
@@ -19,8 +19,9 @@ func ResizeTTY(w http.ResponseWriter, r *http.Request) {
 
 	// /containers/{id}/resize
 	query := struct {
-		Height uint16 `schema:"h"`
-		Width  uint16 `schema:"w"`
+		Height           uint16 `schema:"h"`
+		Width            uint16 `schema:"w"`
+		IgnoreNotRunning bool   `schema:"running"`
 	}{
 		// override any golang type defaults
 	}
@@ -48,14 +49,17 @@ func ResizeTTY(w http.ResponseWriter, r *http.Request) {
 		if state, err := ctnr.State(); err != nil {
 			utils.InternalServerError(w, errors.Wrapf(err, "cannot obtain container state"))
 			return
-		} else if state != define.ContainerStateRunning {
+		} else if state != define.ContainerStateRunning && !query.IgnoreNotRunning {
 			utils.Error(w, "Container not running", http.StatusConflict,
 				fmt.Errorf("container %q in wrong state %q", name, state.String()))
 			return
 		}
+		// If container is not running, ignore since this can be a race condition, and is expected
 		if err := ctnr.AttachResize(sz); err != nil {
-			utils.InternalServerError(w, errors.Wrapf(err, "cannot resize container"))
-			return
+			if errors.Cause(err) != define.ErrCtrStateInvalid || !query.IgnoreNotRunning {
+				utils.InternalServerError(w, errors.Wrapf(err, "cannot resize container"))
+				return
+			}
 		}
 		// This is not a 204, even though we write nothing, for compatibility
 		// reasons.
@@ -70,14 +74,16 @@ func ResizeTTY(w http.ResponseWriter, r *http.Request) {
 		if state, err := ctnr.State(); err != nil {
 			utils.InternalServerError(w, errors.Wrapf(err, "cannot obtain session container state"))
 			return
-		} else if state != define.ContainerStateRunning {
+		} else if state != define.ContainerStateRunning && !query.IgnoreNotRunning {
 			utils.Error(w, "Container not running", http.StatusConflict,
 				fmt.Errorf("container %q in wrong state %q", name, state.String()))
 			return
 		}
 		if err := ctnr.ExecResize(name, sz); err != nil {
-			utils.InternalServerError(w, errors.Wrapf(err, "cannot resize session"))
-			return
+			if errors.Cause(err) != define.ErrCtrStateInvalid || !query.IgnoreNotRunning {
+				utils.InternalServerError(w, errors.Wrapf(err, "cannot resize session"))
+				return
+			}
 		}
 		// This is not a 204, even though we write nothing, for compatibility
 		// reasons.
diff --git a/pkg/api/server/register_containers.go b/pkg/api/server/register_containers.go
index 31196aa9e..b379d52ce 100644
--- a/pkg/api/server/register_containers.go
+++ b/pkg/api/server/register_containers.go
@@ -587,6 +587,11 @@ func (s *APIServer) registerContainersHandlers(r *mux.Router) error {
 	//    type: integer
 	//    required: false
 	//    description: Width to set for the terminal, in characters
+	//  - in: query
+	//    name: running
+	//    type: boolean
+	//    required: false
+	//    description: Ignore containers not running errors
 	// produces:
 	// - application/json
 	// responses:
diff --git a/pkg/api/server/register_exec.go b/pkg/api/server/register_exec.go
index 0f8c827c8..de437ab1a 100644
--- a/pkg/api/server/register_exec.go
+++ b/pkg/api/server/register_exec.go
@@ -136,6 +136,11 @@ func (s *APIServer) registerExecHandlers(r *mux.Router) error {
 	//    name: w
 	//    type: integer
 	//    description: Width of the TTY session in characters
+	//  - in: query
+	//    name: running
+	//    type: boolean
+	//    required: false
+	//    description: Ignore containers not running errors
 	// produces:
 	// - application/json
 	// responses:
diff --git a/pkg/bindings/containers/attach.go b/pkg/bindings/containers/attach.go
index ecae22a1b..fd8a7011d 100644
--- a/pkg/bindings/containers/attach.go
+++ b/pkg/bindings/containers/attach.go
@@ -307,6 +307,7 @@ func resizeTTY(ctx context.Context, endpoint string, height *int, width *int) er
 	if width != nil {
 		params.Set("w", strconv.Itoa(*width))
 	}
+	params.Set("running", "true")
 	rsp, err := conn.DoRequest(nil, http.MethodPost, endpoint, params, nil)
 	if err != nil {
 		return err
diff --git a/pkg/bindings/containers/types.go b/pkg/bindings/containers/types.go
index 2d0e65bb4..f63e35bf1 100644
--- a/pkg/bindings/containers/types.go
+++ b/pkg/bindings/containers/types.go
@@ -210,8 +210,9 @@ type RenameOptions struct {
 // ResizeTTYOptions are optional options for resizing
 // container TTYs
 type ResizeTTYOptions struct {
-	Height *int
-	Width  *int
+	Height  *int
+	Width   *int
+	Running *bool
 }
 
 //go:generate go run ../generator/generator.go ResizeExecTTYOptions
diff --git a/pkg/bindings/containers/types_resizetty_options.go b/pkg/bindings/containers/types_resizetty_options.go
index 68527b330..94946692f 100644
--- a/pkg/bindings/containers/types_resizetty_options.go
+++ b/pkg/bindings/containers/types_resizetty_options.go
@@ -51,3 +51,19 @@ func (o *ResizeTTYOptions) GetWidth() int {
 	}
 	return *o.Width
 }
+
+// WithRunning
+func (o *ResizeTTYOptions) WithRunning(value bool) *ResizeTTYOptions {
+	v := &value
+	o.Running = v
+	return o
+}
+
+// GetRunning
+func (o *ResizeTTYOptions) GetRunning() bool {
+	var running bool
+	if o.Running == nil {
+		return running
+	}
+	return *o.Running
+}
diff --git a/test/system/450-interactive.bats b/test/system/450-interactive.bats
index d047b9f25..a9bf52ee8 100644
--- a/test/system/450-interactive.bats
+++ b/test/system/450-interactive.bats
@@ -57,9 +57,6 @@ function teardown() {
 
     # ...and make sure stty under podman reads that.
     # FIXME: 'sleep 1' is needed for podman-remote; without it, there's
-    # a race condition resulting in the following warning:
-    #   WARN[0000] failed to resize TTY: container "xx" in wrong state "stopped"
-    # (also "created")
     run_podman run -it --name mystty $IMAGE sh -c 'sleep 1;stty size' <$PODMAN_TEST_PTY
     is "$output" "$rows $cols" "stty under podman reads the correct dimensions"
 }