[Review] 12455 SO_REUSEPORT

[Review] 12455 SO_REUSEPORT

[Carlos Neira]

[-- Attachment #1: Type: text/plain, Size: 446 bytes --]

Hi,

I've integrated the work done by araragihokuto@outlook.com into the current illumos-gate.
 Over the past few days, I've tested these changes by running illumos-gate and illumos-omnios,  serving DNS queries using Unbound compiled with SO_REUSEPORT support.
Things so far have been working smoothly, and the tests provided in the patch passed.
Let me know which other tests or changes are required to proceed into a RTI.


Bests!


[-- Attachment #2: Type: text/html, Size: 683 bytes --]

Re: [developer] [Review] 12455 SO_REUSEPORT

[Bill Sommerfeld]

On 5/3/24 09:28, Carlos Neira wrote:
> Hi,
> 
> I've integrated the work done by araragihokuto@outlook.com 
> <mailto:araragihokuto@outlook.com> into the current illumos-gate.
> Over the past few days, I've tested these changes by running 
> illumos-gate and illumos-omnios,  serving DNS queries using Unbound 
> compiled with SO_REUSEPORT support.
> Things so far have been working smoothly, and the tests provided in the 
> patch passed.
> Let me know which other tests or changes are required to proceed into a RTI.

Is this the change you pushed for review at:

  https://code.illumos.org/c/illumos-gate/+/3457 ?

I'm somewhat familiar with how SO_REUSEPORT is used on Linux; I'll take 
a look.

					- Bill

Re: [developer] [Review] 12455 SO_REUSEPORT

[Carlos Neira]

[-- Attachment #1: Type: text/plain, Size: 26 bytes --]

Thank you very much, Bill.

[-- Attachment #2: Type: text/html, Size: 78 bytes --]

Re: [developer] [Review] 12455 SO_REUSEPORT

[Patrick Mooney]

Hi Carlos,

I'd like to raise a few things relating to how this change is evaluated.  Hokuto's work is based on the (TCP-only) SO_REUSEPORT implementation added to SmartOS (primarily for LX).  I've noted the related (SmartOS) tickets in #12455.  It would be useful to document how this change meets those expectations, and where it may choose to diverge.  Considering the challenge noted in OS-5613, solved by relying on TCP squeues for exclusion, we need to carefully evaluate how this change is navigating that same challenge while being general to IP.


A few other notes:
1. Have you run pbchk on what you've posted for CR?  It appears there are at least several cases of trailing whitespace which should trip checks there.
2. If things move forward with this change, correcting it to include proper attribution for the work it is based on would be prudent.


Thanks
--
Patrick Mooney

Re: [developer] [Review] 12455 SO_REUSEPORT

[Carlos Neira]

[-- Attachment #1: Type: text/plain, Size: 649 bytes --]

Hi Patrick,

Thank you for your feedback.
 I'll take a look at tickets #12455 and OS-5613. I was testing in illumos-gate so I was not aware of LX having SO_REUSEPORT.
1. I'll be more diligent in running pbcheck, seems I always miss running it.


2. I'm sorry I'll add you pmooney@pfmooney.com  and  Hokuto Araragi  (araragihokuto@outlook.com) in the authored-by string. Would that be correct or I'm missing someone else? I checked on the OS-5613 and related OS-* tickets and if I'm not mistaken all the work was done by you, for the LX SO_REUSEPORT.





I'm sorry for the late response.





Bests  


[-- Attachment #2: Type: text/html, Size: 2905 bytes --]