On Mon, Apr 15, 2024 at 11:10:50AM +0200, Jiri Pirko wrote:
> Mon, Apr 15, 2024 at 10:39:39AM CEST, michal.swiatkow...@linux.intel.com
> wrote:
> >On Fri, Apr 12, 2024 at 09:12:18AM +0200, Jiri Pirko wrote:
> >> Fri, Apr 12, 2024 at 08:30:49AM CEST, michal.swiatkow...@linux.intel.com
> >> wrote:
> >> >From: Piotr Raczynski <piotr.raczyn...@intel.com>
>
> [...]
>
> >> >+static int
> >> >+ice_devlink_port_fn_state_get(struct devlink_port *port,
> >> >+ enum devlink_port_fn_state *state,
> >> >+ enum devlink_port_fn_opstate *opstate,
> >> >+ struct netlink_ext_ack *extack)
> >> >+{
> >> >+ struct ice_dynamic_port *dyn_port;
> >> >+
> >> >+ dyn_port = ice_devlink_port_to_dyn(port);
> >> >+
> >> >+ if (dyn_port->active) {
> >> >+ *state = DEVLINK_PORT_FN_STATE_ACTIVE;
> >> >+ *opstate = DEVLINK_PORT_FN_OPSTATE_ATTACHED;
> >>
> >> Interesting. This means that you don't distinguish between admin state
> >> and operational state. Meaning, when user does activate, you atomically
> >> achive the hw attachment and it is ready to go before activation cmd
> >> returns, correct? I'm just making sure I understand the code.
> >>
> >
> >I am setting the dyn_port->active after the activation heppens, so it is
> >true, when active is set it is ready to go.
> >
> >Do you mean that dyn_port->active should be set even before the activation is
> >finished? I mean when user only call devlink to active the port?
>
> The devlink instance lock is taken the whole time, isn't it?
>
I don't take PF devlink lock here. Only subfunction devlink lock is
taken during the initialization of subfunction.
> >
> >>
> >> >+ } else {
> >> >+ *state = DEVLINK_PORT_FN_STATE_INACTIVE;
> >> >+ *opstate = DEVLINK_PORT_FN_OPSTATE_DETACHED;
> >> >+ }
> >> >+
> >> >+ return 0;
> >> >+}
> >> >+
>
> [...]
