[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20260127142113.3309fc99@phoenix.local>
Date: Tue, 27 Jan 2026 14:21:13 -0800
From: Stephen Hemminger <stephen@...workplumber.org>
To: Jakub Kicinski <kuba@...nel.org>
Cc: Simon Horman <horms@...nel.org>, "illusion.wang"
<illusion.wang@...ula-matrix.com>, dimon.zhao@...ula-matrix.com,
alvin.wang@...ula-matrix.com, sam.chen@...ula-matrix.com,
netdev@...r.kernel.org, andrew+netdev@...n.ch, corbet@....net,
linux-doc@...r.kernel.org, lorenzo@...nel.org, pabeni@...hat.com,
vadim.fedorenko@...ux.dev, lukas.bulwahn@...hat.com, hawk@...nel.org,
ast@...nel.org, bpf@...r.kernel.org, sdf@...ichev.me, daniel@...earbox.net,
john.fastabend@...il.com, edumazet@...gle.com, open list
<linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v3 net-next 00/15] nbl driver for Nebulamatrix NICs
On Tue, 27 Jan 2026 10:53:04 -0800
Jakub Kicinski <kuba@...nel.org> wrote:
> On Tue, 27 Jan 2026 18:06:49 +0000 Simon Horman wrote:
> > As per Jakub's advice in his review of v1 [*],
> > Please try to create a minimal driver in the order of ~5kLoC.
>
> Perhaps other maintainers will disagree, but for ease of review
> and to avoid the huge reposts perhaps we should consider merging
> something like patches 1 and 3* first, as the initial series.
> And then you can build up the driver from there in reasonable,
> 5kLoC-at-a-time series?
>
> * patch 3 currently includes things which do not seem machine
> generated / like register descriptors like structs added to
> nbl_def_channel.h and nbl_include.h
>
Part of the size problem is that the driver has multiple layers:
resources, channels, etc and lots of code that tries to be common
across OS's. This can lead to bloat. Very few drivers have done
that successfully.
Powered by blists - more mailing lists