Re: [Patch net 08/12] selftests/tc-testing: Add a test case for CODEL with HTB parent

[Cong Wang <xiyou.wangcong@gmail.com>]

On Wed, Apr 02, 2025 at 10:23:22PM -0300, Victor Nogueira wrote:
> On 02/04/2025 21:40, Cong Wang wrote:
> > On Sun, Mar 30, 2025 at 4:32 PM Cong Wang <xiyou.wangcong@gmail.com> wrote:
> > > 
> > > On Sun, Mar 30, 2025 at 06:05:06PM -0300, Victor Nogueira wrote:
> > > > On 28/03/2025 17:35, Cong Wang wrote:
> > > > > On Sun, Mar 23, 2025 at 07:48:39PM -0300, Victor Nogueira wrote:
> > > > > > On 20/03/2025 20:25, Cong Wang wrote:
> > > > > > > Add a test case for CODEL with HTB parent to verify packet drop
> > > > > > > behavior when the queue becomes empty. This helps ensure proper
> > > > > > > notification mechanisms between qdiscs.
> > > > > > > 
> > > > > > > Note this is best-effort, it is hard to play with those parameters
> > > > > > > perfectly to always trigger ->qlen_notify().
> > > > > > > 
> > > > > > > Cc: Pedro Tammela <pctammela@mojatatu.com>
> > > > > > > Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
> > > > > > 
> > > > > > Cong, can you double check this test?
> > > > > > I ran all of your other tests and they all succeeded, however
> > > > > > this one specifically is always failing:
> > > > > 
> > > > > Interesting, I thought I completely fixed this before posting after several
> > > > > rounds of playing with the parameters. I will double check it, maybe it
> > > > > just becomes less reproducible.
> > > > 
> > > > I see.
> > > > I experimented with it a bit today and found out that changing the ping
> > > > command to:
> > > > 
> > > > ping -c 2 -i 0 -s 1500 -I $DUMMY 10.10.10.1 > /dev/null || true
> > > > 
> > > > makes the test pass consistently (at least in my environment).
> > > > So essentially just changing the "-s" option to 1500.
> > > > 
> > > > If you could, please try it out as well.
> > > > Maybe I just got lucky.
> > > 
> > > Sure, I will change it to 1500.
> > 
> > Hmm, it failed on my side:
> > 
> > Test a4bd: Test CODEL with HTB parent - force packet drop with empty queue
> > [...]
> 
> I see, this test seems to be very environment sensitive.
> I think it will be better if we do what you suggested earlier.
> It's not fair to stall this set because of this single test.
> Can you resend your patches excluding this test?

Sure, will do. It sounds like a good plan.

> I can try and figure this one out later so that we can
> unblock.

I really appreciate this.

Thanks!