On Thu, 2023-01-26 at 09:28 +0200, Leon Romanovsky wrote:
> On Wed, Jan 25, 2023 at 11:02:26AM -0800, Jakub Kicinski wrote:
> > On Tue, 24 Jan 2023 13:54:57 +0200 Leon Romanovsky wrote:
> > > - err = dev->xfrmdev_ops->xdo_dev_policy_add(xp);
> > > + err = dev->xfrmdev_ops->xdo_dev_policy_add(xp, extack);
> > > if (err) {
> > > xdo->dev = NULL;
> > > xdo->real_dev = NULL;
> > > xdo->type = XFRM_DEV_OFFLOAD_UNSPECIFIED;
> > > xdo->dir = 0;
> > > netdev_put(dev, &xdo->dev_tracker);
> > > - NL_SET_ERR_MSG(extack, "Device failed to offload this policy");
> >
> > In a handful of places we do:
> >
> > if (!extack->msg)
> > NL_SET_ERR_MSG(extack, "Device failed to offload this policy");
> >
> > in case the device did not provide the extack.
> > Dunno if it's worth doing here.
>
> Honestly, I followed devlink.c which didn't do that, but looked again
> and found that devlink can potentially overwrite messages :)
>
> For example in this case:
> 997 err = ops->port_fn_state_get(port, &state, &opstate, extack);
> 998 if (err) {
> 999 if (err == -EOPNOTSUPP)
> 1000 return 0;
> 1001 return err;
> 1002 }
> 1003 if (!devlink_port_fn_state_valid(state)) {
> 1004 WARN_ON_ONCE(1);
> 1005 NL_SET_ERR_MSG_MOD(extack, "Invalid state read from driver");
> 1006 return -EINVAL;
> 1007 }
>
>
> So what do you think about the following change, so we can leave
> NL_SET_ERR_MSG_MOD() in devlink and xfrm intact?
>
> diff --git a/include/linux/netlink.h b/include/linux/netlink.h
> index 38f6334f408c..d6f3a958e30b 100644
> --- a/include/linux/netlink.h
> +++ b/include/linux/netlink.h
> @@ -101,7 +101,7 @@ struct netlink_ext_ack {
> \
> do_trace_netlink_extack(__msg); \
> \
> - if (__extack) \
> + if (__extack && !__extack->msg) \
> __extack->_msg = __msg; \
> } while (0)
>
> @@ -111,7 +111,7 @@ struct netlink_ext_ack {
> #define NL_SET_ERR_MSG_FMT(extack, fmt, args...) do { \
> struct netlink_ext_ack *__extack = (extack); \
> \
> - if (!__extack) \
> + if (!__extack || __extack->msg) \
> break; \
> if (snprintf(__extack->_msg_buf, NETLINK_MAX_FMTMSG_LEN, \
> "%s" fmt "%s", "", ##args, "") >= \
>
I think it makes sense. With the above patch 3/10 should be updated to
preserve the 'catch-all' error message, I guess.
Let's see what Jakub thinks ;)
Cheers,
Paolo
