netdev - Re: Excessive IPI interrupts with single RX queue devices and rps

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: <1387503054.19078.405.camel@edumazet-glaptop2.roam.corp.google.com>
Date:	Thu, 19 Dec 2013 17:30:54 -0800
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	Florian Fainelli <f.fainelli@...il.com>
Cc:	netdev <netdev@...r.kernel.org>, therbert@...gle.com
Subject: Re: Excessive IPI interrupts with single RX queue devices and
 rps_cpus mask set

On Thu, 2013-12-19 at 17:12 -0800, Florian Fainelli wrote:
> Hello,
> 
> After configuring the rps_cpus mask to enable all CPUs for a single RX
> queue network device, I noticed that I was getting one interrupt per
> packet received, with the interrupt count being splitted between NAPI
> interrupts and IPI SMP interrupts.
> 
> This behavior was observed on 3.8.13 as well as 3.3. The host was
> receiving a Gigabit UDP stream and the receiver was not consuming the
> SKBs. Whenever there was a listener consuming the SKBs received, the
> number of IPI/sec quickly jumped to about 4 times the number without a
> rps_cpus mask.
> 
> Shall we prevent the rps_cpu mask to be set at all for single RX queue
> devices since this seems to causes excessive IPI interrupts?

Not sure what you mean.

This is the thing that highly depends on workload.

(Most workloads perform better without RPS/RFS)

This is why it is disabled by default.



--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html