netdev - Re: [PATCH net-next v9 0/9] Implement devlink-rate API and extend it

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Tue, 8 Nov 2022 14:34:59 -0800
From:   Jakub Kicinski <kuba@...nel.org>
To:     "Wilczynski, Michal" <michal.wilczynski@...el.com>
Cc:     <netdev@...r.kernel.org>, <alexandr.lobakin@...el.com>,
        <jacob.e.keller@...el.com>, <jesse.brandeburg@...el.com>,
        <przemyslaw.kitszel@...el.com>, <anthony.l.nguyen@...el.com>,
        <ecree.xilinx@...il.com>, <jiri@...nulli.us>
Subject: Re: [PATCH net-next v9 0/9] Implement devlink-rate API and extend
 it

On Tue, 8 Nov 2022 09:36:01 +0100 Wilczynski, Michal wrote:
> > On 11/7/2022 7:31 PM, Jakub Kicinski wrote:  
> >> You can't reply to email and then immediately post a new version :/
> >> How am I supposed to have a conversation with you? Extremely annoying.  
> >
> > I'm sorry if you find this annoying, however I can't see any harm here ?
> > I fixed some legit issues that you've pointed in v9, wrote some
> > documentation and basically said, "I wrote some documentation in
> > the next patchset, is it enough ?". I think it's better to get feedback
> > for smaller commits faster, this way I send the updated patchset
> > quickly.

Perhaps spending some time reading the list would help you understand
what normal development upstream looks like. Posting the N + 1 version
of a patch set and then replying to comments on version N is confusing
because it's impossible to decide where the conversation is taking
place. Should I reply to you on version N or reply to N + 1 even tho
there I can't quote your reply. And will the maintainer who's applying
the patches understand that N + 1 got rejected if the discussion is
happening under the thread of version N?

Perhaps one has to read the list to appreciate the challenges involved.

> >> I'm tossing v10 from patchwork, and v11 better come with the docs :/  
> >
> > I will however create a new devlink-rate.rst file if you insist.
> 
> There is however a mention about rate-object management in
> devlink-port.rst. Would it be okay to extend devlink-por.rstt with new
> attributes tx_priority, tx_weight instead of creating a new
> devlink-rate.rst ?

Sounds good, but please make sure you describe the interaction between
the params and the algorithm. We don't have a SW implementation like we
have for qdiscs here, and we don't want each vendor to be coming up
with their own interpretation of the arguments. So we need solid docs,
and some pseudo code to describe the behavior, perhaps?

Let me ask some extra questions on the doc.