netdev - Re: [PATCH v4 net-next 07/12] ipv6: add IFLA_GRO_IPV6_MAX

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: <CANn89iL7bjhVTtbhLGPJv=_srr_epVBe-ZroSDkWjFafxTat3Q@mail.gmail.com>
Date:   Fri, 6 May 2022 15:08:27 -0700
From:   Eric Dumazet <edumazet@...gle.com>
To:     Alexander Duyck <alexander.duyck@...il.com>
Cc:     Eric Dumazet <eric.dumazet@...il.com>,
        "David S . Miller" <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>,
        Paolo Abeni <pabeni@...hat.com>,
        netdev <netdev@...r.kernel.org>, Coco Li <lixiaoyan@...gle.com>
Subject: Re: [PATCH v4 net-next 07/12] ipv6: add IFLA_GRO_IPV6_MAX_SIZE

On Fri, May 6, 2022 at 3:01 PM Alexander Duyck
<alexander.duyck@...il.com> wrote:
>
> On Fri, May 6, 2022 at 2:22 PM Eric Dumazet <edumazet@...gle.com> wrote:
> >
> > On Fri, May 6, 2022 at 2:06 PM Alexander H Duyck
> > <alexander.duyck@...il.com> wrote:
> > >
> > > On Fri, 2022-05-06 at 08:30 -0700, Eric Dumazet wrote:
> > > > From: Coco Li <lixiaoyan@...gle.com>
> > > >
> > > > Enable GRO to have IPv6 specific limit for max packet size.
> > > >
> > > > This patch introduces new dev->gro_ipv6_max_size
> > > > that is modifiable through ip link.
> > > >
> > > > ip link set dev eth0 gro_ipv6_max_size 185000
> > > >
> > > > Note that this value is only considered if bigger than
> > > > gro_max_size, and for non encapsulated TCP/ipv6 packets.
> > > >
> > > > Signed-off-by: Coco Li <lixiaoyan@...gle.com>
> > > > Signed-off-by: Eric Dumazet <edumazet@...gle.com>
> > >
> > > This is another spot where it doesn't make much sense to me to add yet
> > > another control. Instead it would make much more sense to simply remove
> > > the cap from the existing control and simply add a check that caps the
> > > non-IPv6 protocols at GRO_MAX_SIZE.
> >
> > Can you please send a diff on top of our patch series ?
>
> I would rather not as it would essentially just be a revert of the two
> problematic patches since what I am suggesting is significantly
> smaller.
>
> > It is kind of hard to see what you want, and _why_ you want this.
> >
> > Note that GRO_MAX_SIZE has been replaced by dev->gro_max_size last year.
>
> I am using GRO_MAX_SIZE as a legacy value for everything that is not
> IPv6. If it would help you could go back and take a look at Jakub's
> patch series and see what he did with TSO_LEGACY_MAX_SIZE.

Yes, I was the one suggesting this TSO_LEGACY_MAX_SIZE.

> You could
> think of my use here as GRO_LEGACY_MAX_SIZE. What I am doing is
> capping all the non-ipv6/tcp flows at the default maximum limit for
> legacy setups.
>
> > Yes, yet another control, but some people want more control than others I guess.
>
> Basically these patches are reducing functionality from an existing
> control. The g[sr]o_max_size values were applied to all incoming or
> outgoing traffic.

Yes, and we need to change that, otherwise we are stuck at 65536,
because legacy.

> The patches are adding a special control that only applies to a subset of ipv6 traffic.

Exactly. This is not an accident.

> Instead of taking that route I
> would rather have the max_size values allowed to exceed the legacy
> limits, and in those cases that cannot support the new sizes we
> default back to the legacy maxes.

Please send a tested patch. I think it will break drivers.

We spent months doing extensive tests, and I do not see any reason to spend more
time on something that you suggest that I feel is wrong.

> Doing that I feel like we would get
> much more consistent behavior and if somebody is wanting to use these
> values for their original intended purpose which was limiting the
> traffic they will be able to affect all traffic, not just the
> non-ipv6/tcp traffic.

Some people (not us) want to add BIG-TCP with IPv4 as well in a future
evolution.