Re: [PATCH net-next v2] ipv4: Flush the FIB once per dev nexthop removal

[David Ahern <dsahern@kernel.org>]

On 5/6/26 11:53 AM, Cosmin Ratiu wrote:
> On Wed, 2026-05-06 at 10:26 -0600, David Ahern wrote:
>> On 5/6/26 7:01 AM, Ido Schimmel wrote:
>>> ... it would have been easier to review if split into
>>> multiple patches (not saying you should do it). Something like:
>>>
>>> 1. Change the various nexthop remove functions to return an
>>> indication if
>>> flushing is required, but keep doing the flushing in
>>> __remove_nexthop_fib(). Referring to these functions:
>>>
>>> remove_nexthop()
>>> __remove_nexthop()
>>> __remove_nexthop_fib()
>>> remove_nexthop_from_groups()
>>> remove_nh_grp_entry()
>>>
>>> 2. Act upon the flushing indication in the various callers of
>>> remove_nexthop() and remove the flushing from
>>> __remove_nexthop_fib().
>>>
>>> 3. Add __must_check annotations.
>>>
>>
>> +1. Always send the smallest patches possible to evolve the code.
>> Make
>> it easy for reviewers - and yourself should you introduce an intended
>> side effect.
> 
> I didn't split it as the whole thing is tightly coupled across multiple
> functions,

The authors of this code are telling you made a choice, and we are
asking for something easier to review. I spent more time than I should
have to reviewing this patch and trying to recall why I did the code the
way it is.