Re: [PATCH net-next v3 00/23][pull request] Queue configs and large buffer providers

[Mina Almasry <almasrymina@google.com>]

On Wed, Aug 20, 2025 at 6:38 AM Pavel Begunkov <asml.silence@gmail.com> wrote:
>
> On 8/20/25 03:31, Jakub Kicinski wrote:
> > On Mon, 18 Aug 2025 14:57:16 +0100 Pavel Begunkov wrote:
> >> Jakub Kicinski (20):
> >
> > I think we need to revisit how we operate.
> > When we started the ZC work w/ io-uring I suggested a permanent shared
> > branch. That's perhaps an overkill. What I did not expect is that you
> > will not even CC netdev@ on changes to io_uring/zcrx.*
> >
> > I don't mean to assert any sort of ownership of that code, but you're
> > not meeting basic collaboration standards for the kernel. This needs
> > to change first.
>
> You're throwing quite allegations. Basic collaboration standards don't
> include spamming people with unrelated changes via an already busy list.
> I cc'ed netdev on patches that meaningfully change how it interacts
> (incl indirectly) with netdev and/or might be of interest, which is
> beyond of the usual standard expected of a project using infrastructure
> provided by a subsystem. There are pieces that don't touch netdev, like
> how io_uring pins pages, accounts memory, sets up rings, etc. In the
> very same way generic io_uring patches are not normally posted to
> netdev, and netdev patches are not redirected to mm because there
> are kmalloc calls, even though, it's not even the standard used here.
>
> If you have some way you want to work, I'd appreciate a clear
> indication of that, because that message you mentioned was answered
> and I've never heard any objection, or anything else really.
>

We could use tags in the MAINTAINERS file similar to these:

F: include/linux/*fence.h
F: include/linux/dma-buf.h
F: include/linux/dma-resv.h
K: \bdma_(?:buf|fence|resv)\b

We could make sure anything touching io_uring/zcrx. and anything using
netmem_ref/net_iov goes to netdev. I think roughly adding something
like this to general networking entry?

F: io_uring/zcrx.*
K: \bnet(mem_ref|_iov)\b

I had suggested this before but never had time to suggest the actual
changes, and in the back of my mind was a bit weary of spamming the
maintainers, but it seems this is not as much a concern as the patches
not getting to netdev.

-- 
Thanks,
Mina