On Fri, Jul 19, 2019 at 10:55:49AM -0600, David Ahern wrote:
> Hi:
>
> On 7/18/19 10:17 PM, Hangbin Liu wrote:
> > Hi David,
> >
> > Before commit 18a8021a7be3 ("net/ipv4: Plumb support for filtering route
> > dumps"), when we dump a non-exist table, ip cmd exits silently.
> >
> > # ip -4 route list table 1
> > # echo $?
> > 0
> >
> > After commit 18a8021a7be3 ("net/ipv4: Plumb support for filtering route
> > dumps"). When we dump a non-exist table, as we returned -ENOENT, ip route
> > shows:
> >
> > # ip -4 route show table 1
> > Error: ipv4: FIB table does not exist.
> > Dump terminated
> > # echo $?
> > 2
> >
> > For me it looks make sense to return -ENOENT if we do not have the route
> > table. But this changes the userspace behavior. Do you think if we need to
> > keep backward compatible or just let it do as it is right now?
> >
>
> It is not change in userspace behavior; ip opted into the strict
> checking. The impact is to 'ip' users.
>
> A couple of people have asked about this, and I am curious as to why
> people run a route dump for a table that does not exist and do not like
> being told that it does not exist.
Hi David,
Thanks for the reply. We have some route function tests and the new behavior
break the test. I just want to make sure this is expected so we can change our
tests to match the new behavior.
Cheers
Hangbin
