netdev - Re: [PATCH net] net_sched: sch_fq: don't follow the fast path if Tx is behind now

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: <CANn89iJp09omWVAk43Aw2mPsii4QsgYxOSpjDdDyYeGjVQSGRw@mail.gmail.com>
Date: Fri, 22 Nov 2024 19:09:51 +0100
From: Eric Dumazet <edumazet@...gle.com>
To: Jakub Kicinski <kuba@...nel.org>
Cc: netdev@...r.kernel.org, davem@...emloft.net, pabeni@...hat.com, 
	jhs@...atatu.com, xiyou.wangcong@...il.com, jiri@...nulli.us
Subject: Re: [PATCH net] net_sched: sch_fq: don't follow the fast path if Tx
 is behind now

On Fri, Nov 22, 2024 at 6:31 PM Jakub Kicinski <kuba@...nel.org> wrote:
>
> On Fri, 22 Nov 2024 17:44:33 +0100 Eric Dumazet wrote:
> > Interesting... I guess we could also call fq_check_throttled() to
> > refresh a better view of the qdisc state ?
> >
> > But perhaps your patch is simpler. I guess it could be reduced to
> >
> > if (q->time_next_delayed_flow <= now + q->offload_horizon)
> >       return false;
> >
> > (Note the + q->offload_horizon)
> >
> > I do not think testing q->throttled_flows is strictly needed :
> > If 0, then q->time_next_delayed_flow is set to ~0ULL.
>
> Makes sense, I'll respin using your check tomorrow.

Great. I confirm that the fix reduces the TcpExtTCPSACKReorder SNMP
counter increase we had recently.

Also "ss -temoi" was showing suspect reordering:300 values I had no
time yet to investigate.

Thanks a lot for finding this !