netdev - Re: [RFC net-next 00/22] net: per-queue rx-buf-len configuration

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]

Message-ID: <20250425165531.7926490d@kernel.org>
Date: Fri, 25 Apr 2025 16:55:31 -0700
From: Jakub Kicinski <kuba@...nel.org>
To: Mina Almasry <almasrymina@...gle.com>
Cc: davem@...emloft.net, netdev@...r.kernel.org, edumazet@...gle.com,
 pabeni@...hat.com, andrew+netdev@...n.ch, horms@...nel.org,
 donald.hunter@...il.com, sdf@...ichev.me, dw@...idwei.uk,
 asml.silence@...il.com, ap420073@...il.com, jdamato@...tly.com,
 dtatulea@...dia.com, michael.chan@...adcom.com
Subject: Re: [RFC net-next 00/22] net: per-queue rx-buf-len configuration

On Wed, 23 Apr 2025 13:02:52 -0700 Mina Almasry wrote:
> On Mon, Apr 21, 2025 at 3:28 PM Jakub Kicinski <kuba@...nel.org> wrote:
> > Add support for per-queue rx-buf-len configuration.
> >
> > I'm sending this as RFC because I'd like to ponder the uAPI side
> > a little longer but it's good enough for people to work on
> > the memory provider side and support in other drivers.
> >  
> 
> May be silly question, but I assume opting into this is optional for
> queue API drivers? Or do you need GVE to implement dependencies of
> this very soon otherwise it's blocking your work?

Completely optional, I think it has to be.

> I think it needs to be the former. Memory providers will have wildly
> differing restrictions in regards to size. I think already the dmabuf
> mp can allocate any byte size net_iov. I think the io_uring mp can
> allocate any multiple of PAGE_SIZE net_iov. MPs communicating their
> restrictions over a uniform interface with the driver seems difficult
> to define. Better for the driver to ask the pp/mp what it wants, and
> the mp can complain if it doesn't support it.
> 
> Also this mirrors what we do today with page_pool_params.order arg
> IIRC. You probably want to piggy back off that or rework it.

Yup. I added the rx-buf-len-max size reporting because I suspect
applications will need to discover the NIC capabilities.