Hi Hangbin, Martin,
On 15/02/2023 04:43, Hangbin Liu wrote:
> On Tue, Feb 14, 2023 at 11:22:58AM -0800, Martin KaFai Lau wrote:
>> On 2/13/23 7:11 PM, Hangbin Liu wrote:
>>> On Mon, Feb 13, 2023 at 05:28:19PM +0100, Matthieu Baerts wrote:
>>>> But again, I'm not totally against that, I'm just saying that if these
>>>> tests are executed in dedicated netns, this modification is not needed
>>>> when using a vanilla kernel ;-)
>>>>
>>>> Except if I misunderstood and these tests are not executed in dedicated
>>>> netns?
>>>
>>> I tried on my test machine, it looks the test is executed in init netns.
>>
>> The new test is needed to run under its own netns whenever possible. The
>> existing mptcp test should be changed to run in its own netns also. Then
>> changing any pernet sysctl (eg. mptcp.enabled) is a less concern to other
>> tests. You can take a look at how other tests doing it (eg. decap_sanity.c).
>
> Thanks for the reference. I will update the patch to make mptcp run in it's
> own netns.
Thank you both for your replies!
Yes, that would be good to have this test running in a dedicated NS.
Then mptcp.enabled can be forced using write_sysctl() or SYS("sysctl (...)".
Cheers,
Matt
--
Tessares | Belgium | Hybrid Access Solutions
www.tessares.net
