netdev - Re: [RFC net-next 0/2] igc: Link IRQs and queues to NAPIs

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

Message-ID: <Zwa5zdhtxlqJxIj7@LQ3V64L9R2>
Date: Wed, 9 Oct 2024 10:13:49 -0700
From: Joe Damato <jdamato@...tly.com>
To: Vinicius Costa Gomes <vinicius.gomes@...el.com>
Cc: netdev@...r.kernel.org, "David S. Miller" <davem@...emloft.net>,
	Eric Dumazet <edumazet@...gle.com>,
	"moderated list:INTEL ETHERNET DRIVERS" <intel-wired-lan@...ts.osuosl.org>,
	Jakub Kicinski <kuba@...nel.org>,
	open list <linux-kernel@...r.kernel.org>,
	Paolo Abeni <pabeni@...hat.com>,
	Przemek Kitszel <przemyslaw.kitszel@...el.com>,
	Tony Nguyen <anthony.l.nguyen@...el.com>
Subject: Re: [RFC net-next 0/2] igc: Link IRQs and queues to NAPIs

On Mon, Oct 07, 2024 at 04:03:00PM -0700, Vinicius Costa Gomes wrote:
> Joe Damato <jdamato@...tly.com> writes:
> 
> > Greetings:
> >
> > This is an RFC to get feedback before submitting an actual series and
> > because I have a question for igc maintainers, see below.
> >
> > This series addss support for netdev-genl to igc so that userland apps
> > can query IRQ, queue, and NAPI instance relationships. This is useful
> > because developers who have igc NICs (for example, in their Intel NUCs)
> > who are working on epoll-based busy polling apps and using
> > SO_INCOMING_NAPI_ID, need access to this API to map NAPI IDs back to
> > queues.
> >
> > See the commit messages of each patch for example output I got on my igc
> > hardware.
> >
> > My question for maintainers:
> >
> > In patch 2, the linking should be avoided for XDP queues. Is there a way
> > to test that somehow in the driver? I looked around a bit, but didn't
> > notice anything. Sorry if I'm missing something obvious.
> >
> 
> From a quick look, it seems that you could "unlink" the XDP queues in
> igc_xdp_enable_pool() and (re-)link them in igc_xdp_disable_poll().

That approach seems reasonable to me, but I am not an igc expert by
any means :)

I checked and it seems that igc_xdp_enable_pool and
igc_xdp_disable_poll are only called while RTNL is held, which is
good because netif_queue_set_napi uses ASSERT_RTNL.
 
> Or just the existence of the flag IGC_RING_FLAG_AF_XDP_ZC in the rings
> associated with the queue is enough?

I didn't notice that flag, thanks for pointing that out.

It might be better to go the link/unlink route as you described
above, though.

> I still have to take a better look at your work to help more, sorry.

No worries, thanks for taking a look.

I'll implement what you suggested above and send another RFC.