netdev - Re: [RFC PATCH net-next] veth: apply qdisc backpressure on full ptr

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: <55a2ea3c-30fb-497f-b373-22394df52576@kernel.org>
Date: Tue, 8 Apr 2025 09:20:32 -0600
From: David Ahern <dsahern@...nel.org>
To: Toke Høiland-Jørgensen <toke@...hat.com>,
 Jesper Dangaard Brouer <hawk@...nel.org>, netdev@...r.kernel.org,
 Jakub Kicinski <kuba@...nel.org>
Cc: bpf@...r.kernel.org, tom@...bertland.com,
 Eric Dumazet <eric.dumazet@...il.com>, "David S. Miller"
 <davem@...emloft.net>, Paolo Abeni <pabeni@...hat.com>,
 kernel-team@...udflare.com
Subject: Re: [RFC PATCH net-next] veth: apply qdisc backpressure on full
 ptr_ring to reduce TX drops

On 4/8/25 5:23 AM, Toke Høiland-Jørgensen wrote:
> David Ahern <dsahern@...nel.org> writes:
> 
>> On 4/7/25 3:15 AM, Toke Høiland-Jørgensen wrote:
>>>> +static inline bool txq_has_qdisc(struct netdev_queue *txq)
>>>> +{
>>>> +	struct Qdisc *q;
>>>> +
>>>> +	q = rcu_dereference(txq->qdisc);
>>>> +	if (q->enqueue)
>>>> +		return true;
>>>> +	else
>>>> +		return false;
>>>> +}
>>>
>>> This seems like a pretty ugly layering violation, inspecting the qdisc
>>> like this in the driver?
>>
>> vrf driver has something very similar - been there since March 2017.
> 
> Doesn't make it any less ugly, though ;)

in my eyes, it is a thing of beauty.

> 
> And AFAICT, vrf is doing more with the information; basically picking a
> whole different TX path? Can you elaborate on the reasoning for this (do
> people actually install qdiscs on VRF devices in practice)?

It is not common AFAIK, but it is possible. I wanted to avoid the
overhead for what I think is a rare configuration.