Re: [patch net-next 09/10] net: sch: prio: Add offload ability for grafting a child

[Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>]

On Fri, Mar 02, 2018 at 09:21:56AM -0500, Alexander Aring wrote:
> Hi,
> 
> On Fri, Mar 2, 2018 at 3:37 AM, Jiri Pirko <jiri@resnulli.us> wrote:
> > Fri, Mar 02, 2018 at 04:48:40AM CET, kubakici@wp.pl wrote:
> >>On Thu, 1 Mar 2018 22:38:50 -0500, Alexander Aring wrote:
> >>> I guess to make extack working, you need to return an errno if failed.
> >>
> >>AFAIK extack is printed as a warning if operation did not fail.
> >
> > Yes, I checked this and it is printed as warning.
> 
> ouch, so far I know extack it allows only one messages delivered back
> to the user space.
> 
> If we introduce a warning in the successful path here, it could be
> that in the callpath (after "successful" part of this callback), that
> somebody else want to add a warning and overwrites actually your
> warning (even, he is not aware that this warning was set - okay I
> suppose you can do another check on NULL and set a warning, that
> somebody overwrites a warning :-D).
> 
> If extack messages get's append and is some kind of for_each_nested
> string in netlink -> we have no problem, but I guess this not how it's
> working. :-/

IOW I guess what we are looking for here is a way to use extack to
track more than an error/warning message, but to be a bit more
complete error reporting tool.

The case here is pretty much like the case with tc flower offloading,
to issue a message when it couldn't offload while it wasn't fatal for
the rule (in case SKIP_SW wasn't specified). The reason for why the
offloading couldn't happen could have been a temporary one and such
log of a warning is important for troubleshooting.

  Marcelo