review scoring

review scoring

[Jakub Kicinski]

Hi!

Alexei just sent an announcement to the BPF list describing
a new system of requiring review participation:
https://lore.kernel.org/all/CAADnVQ+TKKptnNB25V3=bcdybh5G6c2DyW2sYtXvyRaVnPN8MA@mail.gmail.com/
I'm very curious to see how this goes, it definitely feels like
a logical step.

On the netdev side I've been using a patch review queue which
takes into account 
 - the reviewer score of the author
 - the reviewer socre of their employer (using the same logic 
   I use for the periodic "developer statistics" email),
 - quality metrics like "is checkpatch clean"
 - whether someone already reviewed the patches on the list
   (reviews from the same company do not count)

I've been meaning to send an email to let folks know that I'm doing
this. Most large companies are getting a 2 day delay from this scoring
system. To be clear, I'm not intentionally waiting 2 days before
looking at code (at the cost of my sanity). It's just that with our
patch volume if something gets a 2 day delay there's plenty of patches
that will jump earlier in the queue. And so far advancements in AI
failed to make the days any longer..

the queue: https://netdev.bots.linux.dev/suie.html?delegate=netdev

Re: review scoring

[Dawid Osuchowski]

On 2026-03-04 3:01 AM, Jakub Kicinski wrote:
> Hi!
> 
> Alexei just sent an announcement to the BPF list describing
> a new system of requiring review participation:
> https://lore.kernel.org/all/CAADnVQ+TKKptnNB25V3=bcdybh5G6c2DyW2sYtXvyRaVnPN8MA@mail.gmail.com/
> I'm very curious to see how this goes, it definitely feels like
> a logical step.
> 
> On the netdev side I've been using a patch review queue which
> takes into account
>   - the reviewer score of the author
>   - the reviewer socre of their employer (using the same logic
>     I use for the periodic "developer statistics" email),
>   - quality metrics like "is checkpatch clean"
>   - whether someone already reviewed the patches on the list
>     (reviews from the same company do not count)
> 
> I've been meaning to send an email to let folks know that I'm doing
> this. Most large companies are getting a 2 day delay from this scoring
> system. To be clear, I'm not intentionally waiting 2 days before
> looking at code (at the cost of my sanity). It's just that with our
> patch volume if something gets a 2 day delay there's plenty of patches
...
> that will jump earlier in the queue. And so far advancements in AI
> failed to make the days any longer..

Hey Kuba,

I don't think that's possible, but it hopefully AI could make our 
working days shorter by offloading some of our work or making it more 
efficient ;) I just hope we will circumvent / prevent AI slop from 
taking over.

> the queue: https://netdev.bots.linux.dev/suie.html?delegate=netdev

In all seriousness, thanks for sharing this and maybe this will help get 
some execs and other higher-ups across the broad corporate Linux kernel 
world to give more time to engineers to perform reviews for other 
company / community patches as well.

Best regards,
Dawid