linux-kernel - Re: [PATCH RT 2/2] softirq: split timer softirqs out of ksoftirqd

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]

Date:	Mon, 6 Jun 2016 11:21:21 -0700
From:	Brian Silverman <brian@...oton-tech.com>
To:	Sebastian Andrzej Siewior <bigeasy@...utronix.de>
Cc:	rt-users <linux-rt-users@...r.kernel.org>,
	LKML <linux-kernel@...r.kernel.org>,
	Thomas Gleixner <tglx@...utronix.de>,
	Steven Rostedt <rostedt@...dmis.org>
Subject: Re: [PATCH RT 2/2] softirq: split timer softirqs out of ksoftirqd

I've been running with this patch on ARM (dra7 SOC, which is a
dual-core A15) with a 4.1.18-rt17 kernel for a while without any
noticeable problems. Boosting the ktimersoftd processes on both cores
to a higher priority fixed some timing problems my userspace
application was seeing too.

The userspace application in question has a repeating timerfd in an
epoll and spends most of its time in blocking epoll_wait calls. Under
system load (most from other lower-RT-priority userspace processes),
it was seeing several-ms delays between the timerfd interrupt
happening and ksoftirqd running the HRTIMER softirq to wake it up.
That was with ksoftirqd running as SCHED_OTHER, which is what it
defaulted to. The priority inversion between the userspace processes
is gone after making ktimersoftd the same priority as the
highest-priority application.

Tested-by: Brian Silverman <brian@...oton-tech.com>