linux-kernel - Re: [PATCH] timekeeping: Use proper timekeeper for debug code

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]

Message-ID: <20170628223153.GA2589@lianli.shorne-pla.net>
Date:   Thu, 29 Jun 2017 07:31:53 +0900
From:   Stafford Horne <shorne@...il.com>
To:     John Stultz <john.stultz@...aro.org>
Cc:     LKML <linux-kernel@...r.kernel.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Stephen Boyd <sboyd@...eaurora.org>
Subject: Re: [PATCH] timekeeping: Use proper timekeeper for debug code

On Wed, Jun 28, 2017 at 10:39:21AM -0700, John Stultz wrote:
> On Wed, Jun 28, 2017 at 6:21 AM, Stafford Horne <shorne@...il.com> wrote:
> > When CONFIG_DEBUG_TIMEKEEPING is enabled the timekeeping_check_update()
> > function will update status like last_warning and underflow_seen on the
> > timekeeper.
> >
> > If there are issues found this state is used to rate limit the warnings
> > that get printed.
> >
> > This rate limiting doesn't really really work if stored in real_tk as
> > the shadow timekeeper is overwritten onto real_tk at the end of every
> > update_wall_time() call, resetting last_warning and other statuses.
> >
> > Fix rate limiting by using the shadow_timekeeper for
> > timekeeping_check_update().
> >
> > Fixes: commit 57d05a93ada7 ("time: Rework debugging variables so they aren't global")
> > Signed-off-by: Stafford Horne <shorne@...il.com>
> 
> Interesting!  Thanks for submitting this! Out of curiosity, can you
> also talk about the underlying issue that made you come across this
> issue?

I am working on OpenRISC SMP support.  There was one issue where the
per_cpu internal timers (used as clocksource) were not in sync, this
pointed it out.

There is another issue right now when switching from jiffies to the
openrisc clocksource.  Which is maybe ok because each have different
starting points.

[    0.160000] clocksource: Switched to clocksource openrisc_timer
[    0.220000] INFO: timekeeping: Cycle offset (4294173293) is larger than the 'openrisc_timer' clock's 50% safety margin (2147483647)
[    0.220000]       timekeeping: Your kernel is still fine, but is feeling a bit nervous

Let me know if you want me to add this to the commit message in a v2.

-Stafford