Re: [PATCH net-next v3 01/11] devlink: retain error in struct devlink_fmsg

[Przemek Kitszel]

On 10/19/23 15:00, Simon Horman wrote:
> On Wed, Oct 18, 2023 at 10:26:37PM +0200, Przemek Kitszel wrote:
> > Retain error value in struct devlink_fmsg, to relieve drivers from
> > checking it after each call.
> > Note that fmsg is an in-memory builder/buffer of formatted message,
> > so it's not the case that half baked message was sent somewhere.
> >
> > We could find following scheme in multiple drivers:
> >    err = devlink_fmsg_obj_nest_start(fmsg);
> >    if (err)
> >    	return err;
> >    err = devlink_fmsg_string_pair_put(fmsg, "src", src);
> >    if (err)
> >    	return err;
> >    err = devlink_fmsg_something(fmsg, foo, bar);
> >    if (err)
> > 	return err;
> >    // and so on...
> >    err = devlink_fmsg_obj_nest_end(fmsg);
> >
> > With retaining error API that translates to:
> >    devlink_fmsg_obj_nest_start(fmsg);
> >    devlink_fmsg_string_pair_put(fmsg, "src", src);
> >    devlink_fmsg_something(fmsg, foo, bar);
> >    // and so on...
> >    devlink_fmsg_obj_nest_end(fmsg);
> >
> > What means we check error just when is time to send.
> >
> > Possible error scenarios are developer error (API misuse) and memory
> > exhaustion, both cases are good candidates to choose readability
> > over fastest possible exit.
> >
> > Note that this patch keeps returning errors, to allow per-driver conversion
> > to the new API, but those are not needed at this point already.
> >
> > This commit itself is an illustration of benefits for the dev-user,
> > more of it will be in separate commits of the series.
> >
> > Reviewed-by: Jesse Brandeburg <jesse.brandeburg@xxxxxxxxx>
> > Reviewed-by: Jiri Pirko <jiri@xxxxxxxxxx>
> > Signed-off-by: Przemek Kitszel <przemyslaw.kitszel@xxxxxxxxx>
>
> ...
>
> > @@ -1027,14 +934,12 @@ int devlink_fmsg_binary_pair_put(struct devlink_fmsg *fmsg, const char *name,
>
> Hi Przemek,
>
> The line before this hunk is:
>
> 		err = devlink_fmsg_binary_put(fmsg, value + offset, data_size);
>
> And, as of this patch, the implementation of
> devlink_fmsg_binary_pair_nest_start() looks like this:
>
> int devlink_fmsg_binary_put(struct devlink_fmsg *fmsg, const void *value,
>                              u16 value_len)
> {
>          if (!fmsg->putting_binary)
>                  return -EINVAL;
>
>          return devlink_fmsg_put_value(fmsg, value, value_len, NLA_BINARY);
> }
>
> Which may return an error, if the if condition is met, without setting
> fmsg->err.
>
> >   		if (err)
> >   			break;
> >   		/* Exit from loop with a break (instead of
> > -		 * return) to make sure putting_binary is turned off in
> > -		 * devlink_fmsg_binary_pair_nest_end
> > +		 * return) to make sure putting_binary is turned off
> >   		 */
> >   	}
> >   
> > -	end_err = devlink_fmsg_binary_pair_nest_end(fmsg);
> > -	if (end_err)
> > -		err = end_err;
>
> Prior to this patch, the value of err from the loop above was preserved,
> unless devlink_fmsg_binary_pair_nest_end generated an error.
>
> > +	err = devlink_fmsg_binary_pair_nest_end(fmsg);
>
> But now it looks like this is only the case if fmsg->err corresponds to err
> when the loop was exited.
>
> Or in other words, the err returned by devlink_fmsg_binary_put()
> is not propagated to the caller if !fmsg->putting_binary.
>
> If so, is this intentional?

In scope of devlink_fmsg_binary_pair_put() all it's fine, and perhaps
that lead me into thinking that it's fine in *general*, which is not,
as devlink_fmsg_binary_put() is exported (so it's return value should
be preserved).

On the other note, it's called in two places (one of which was just
covered, and the second place is mlx5e_health_rsc_fmsg_binary(), which
really just mimics what we have here. So this is also fine in practice.

Last patch of the series fixes it for general case, so barring
pathological cherry-picks or reverts we are safe.
Do you want a v4 with that fixed?

Other thing is that this function could be changed into some internal 
helper without export.

> > +	fmsg->putting_binary = false;
> >   
> >   	return err;
> >   }
> > --
> > 2.38.1