On Thu, Sep 03, 2020 at 11:05:37AM +0100, Paul Durrant wrote: > From: Paul Durrant <[email protected]> > > The manpage for 'xl' documents that guest co-operation is required for a (non- > forced) block-detach operation and that it may consequently fail. Currently, > however, the implementation of generic device removal means that a time-out > of a block-detach is being automatically re-tried with the force flag set > rather than failing. This patch stops such behaviour.
Won't this break cleanup on domain shutdown if the guest doesn't close the devices itself? I think we need some special-casing on shutdown that keeps the current behavior on that case. > > Signed-off-by: Paul Durrant <[email protected]> > --- > Cc: Ian Jackson <[email protected]> > Cc: Wei Liu <[email protected]> > Cc: Anthony PERARD <[email protected]> > --- > tools/libxl/libxl_device.c | 9 ++++++--- > 1 file changed, 6 insertions(+), 3 deletions(-) > > diff --git a/tools/libxl/libxl_device.c b/tools/libxl/libxl_device.c > index 0381c5d509..d17ca78848 100644 > --- a/tools/libxl/libxl_device.c > +++ b/tools/libxl/libxl_device.c > @@ -1092,7 +1092,8 @@ static void device_backend_callback(libxl__egc *egc, > libxl__ev_devstate *ds, > > if (rc == ERROR_TIMEDOUT && > aodev->action == LIBXL__DEVICE_ACTION_REMOVE && > - !aodev->force) { > + !aodev->force && > + aodev->dev->kind != LIBXL__DEVICE_KIND_VBD) { Doing this differentiation for block only seems weird, we should treat all devices equally. Thanks, Roger.
