Re: [PATCH net] netdevsim: Restore per-network namespace accounting for fib entries

[David Ahern <dsahern@gmail.com>]

On 8/7/19 12:27 AM, Jiri Pirko wrote:
> Wed, Aug 07, 2019 at 12:32:14AM CEST, jakub.kicinski@netronome.com wrote:
>> On Tue,  6 Aug 2019 12:15:17 -0700, David Ahern wrote:
>>> From: David Ahern <dsahern@gmail.com>
>>>
>>> Prior to the commit in the fixes tag, the resource controller in netdevsim
>>> tracked fib entries and rules per network namespace. Restore that behavior.
>>>
>>> Fixes: 5fc494225c1e ("netdevsim: create devlink instance per netdevsim instance")
>>> Signed-off-by: David Ahern <dsahern@gmail.com>
>>
>> Thanks.
>>
>> Let's see what Jiri says, but to me this patch seems to indeed restore
>> the original per-namespace accounting when the more natural way forward
>> may perhaps be to make nsim only count the fib entries where
> 
> I think that:
> 1) netdevsim is a glorified dummy device for testing kernel api, not for
>    configuring per-namespace resource limitation.
> 2) If the conclusion os to use devlink instead of cgroups for resourse
>    limitations, it should be done in a separate code, not in netdevsim.
> 
> I would definitelly want to wait what is the result of discussion around 2)
> first. But one way or another netdevsim code should not do this, I would
> like to cutout the fib limitations from it instead, just to expose the
> setup resource limits through debugfs like the rest of the configuration
> of netdevsim.
> 
> 

This is the most incredulous response. You break code you don't
understand and argue that it should be left broken in older releases and
ripped out in later ones rather than fixed back to its intent.

Again, the devlink resource controller was added by me specifically to
test the ability to fail in-kernel fib notifiers. When I add the nexthop
in-kernel notifier, netdevsim again provides a simple means of testing it.

It satisfies all of the conditions and intents of netdevsim - test
kernel APIs without hardware.