linux-kernel - [RFC] Time/clocksource:The clock source was misjudged as unstable

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>] [day] [month] [year] [list]

Message-Id: <20210819041008.14693-1-yanghui.def@bytedance.com>
Date:   Thu, 19 Aug 2021 12:10:08 +0800
From:   yanghui <yanghui.def@...edance.com>
To:     john.stultz@...aro.org
Cc:     tglx@...utronix.de, sboyd@...nel.org, linux-kernel@...r.kernel.org,
        songmuchun@...edance.com, duanxiongchun@...edance.com,
        yanghui <yanghui.def@...edance.com>
Subject: [RFC] Time/clocksource:The clock source was misjudged as unstable

We use clocksource_watchdog() to monitor whether the current clocksource is
stable every WATCHDOG_INTERVAL. But clocksource_watchdog() can't be schedule
in time when current CPU occurs Softlockup. This will bring following 
clocksource misjudgment problem:

if the clocksource_watchdog() schedule time longer then WATCHDOG_INTERVAL,
the value of abs(cs_nsec - wd_nsec) will be enlarged a lot, but both of
the clocksource is normal at this time. And if the value is bigger than
md, system will misjudge that current clocksource is unstable then to select
another clocksource.

So we think this is a situation that Softlockup causes the clocksource
to be misjudged. We want to know is there have some good idea to solve
this problem ? Actually we want to use hrtimer to replace normal timer.
But hrtimer is difficult to execute in turn on each CPU when the system 
is just boot up. Is there a better way to solve this problem ?

Thanks

Signed-off-by: yanghui <yanghui.def@...edance.com>
---

-- 
2.20.1