netdev - Re: qdisc_watchdog_schedule_range

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]

Date:   Wed, 12 Oct 2022 16:32:56 -0700
From:   Eric Dumazet <eric.dumazet@...il.com>
To:     Thorsten Glaser <t.glaser@...ent.de>
Cc:     netdev@...r.kernel.org
Subject: Re: qdisc_watchdog_schedule_range_ns granularity


On 10/12/22 16:15, Thorsten Glaser wrote:
> On Wed, 12 Oct 2022, Eric Dumazet wrote:
>
>> CONFIG_HIGH_RES_TIMERS=y
> It does.
>
>> I don't know how you measure this latency, but net/sched/sch_fq.c has
>> instrumentation,
> On enqueue I add now+extradelay and save that as enqueue timestamp.
> On dequeue I check that now>=timestamp then process the packet,
> measuring now-timestamp as queue delay. This is surprisingly higher.

When packets are eligible, the qdisc itself can be stopped if NIC is 
full (or BQL triggers)

net/sched/sch_fq.c is not using the skb tstamp which could very well be 
in the past,

but an internal variable (q->time_next_delayed_flow)


>
> I’ll add some printks as well, to see when I’m called next after
> such a watchdog scheduling.
>
>> Under high cpu pressure, it is possible the softirq is delayed,
>> because ksoftirqd might compete with user threads.
> Is it a good idea to renice these?


Depends if you want to starve user thread(s) under flood/attack, I guess 
you can try.


>
> Thanks,
> //mirabilos