[dpdk-dev] 【BUG REPORT】l3fwd-power can not exit by ctrl+c

[Burakov, Anatoly]

On 22-May-20 9:17 AM, oulijun wrote:
> 
> 
> 在 2020/5/20 15:22, David Marchand 写道:
>> Hello,
>>
>> On Wed, May 20, 2020 at 5:18 AM oulijun <oulijun at huawei.com> wrote:
>>>      I am using 20.05-rc2 version to test based on HNS3 NIC hardware, 
>>> and
>>> found that after starting l3fwd-power,
>>>
>>> using ctrl+c cannot force quit. But I revert the patch(33666b4 service:
>>> fix crash on exit) and it is ok.
>>
>> We had a fix in rc1 that is supposed to fix this.
>> https://git.dpdk.org/dpdk/commit?id=613ce6691c0d5ac0f99d7995f1e8e4ac86643882 
>>
>>
>> Copying Anatoly and David H. too.
>>
> Hi, David Marchand
>     I am only merged the patch(examples/l3fwd-power: implement proper 
> shutdown) based on 20.05-rc0 and test.

Why are you testing 20.05-rc0 when there is a later version that already 
has all of this fixed? What's the point of creating additional 
difficulties for yourself, and, more importantly, reporting bugs to the 
community that are already fixed while also not testing the version that 
has those fixes? What are you trying to do that requires you to use rc0 
instead of rc2 or later?

We know rc0 was buggy. We had one patch (linked to by David above) in 
rc1 that fixed it. We had another one in rc2 (linked by myself in 
another email: 
https://git.dpdk.org/dpdk/commit/?id=f4d1e19c293dc95073614f630ea729cf0bfb57b7) 
which completed the fix. Both of these patches are present in rc2 and 
later. So why not try and test rc2 or laster, instead of cherry-picking 
patches on top of rc0?

  the bug is not resolved.
> Can you provide me with complete solution?  I want to understand the 
> cause of the whole problem
> 

The cause of the problem is described in the commit messages for the 
patches. The one David linked to was because l3fwd was exiting 
incorrectly and was relying on particulars of implementation of 
rte_exit() which exited the application straight away, without waiting 
for threads to finish. The one i linked to fixed an additional issue 
related to waiting for interrupts - there was no timeout, so the thread 
never woke up if there was no traffic, and thus rte_exit() couldn't 
finish as well.

All of the above was fixed by rc2, and later versions have these fixes 
as well. So, why are you insisting on using an outdated version for testing?

-- 
Thanks,
Anatoly