Date: 9 Jun 2015 01:39:30 -0400
Message-ID: <20150609053930.25020.qmail@ns.horizon.com>
From: "George Spelvin" <linux@horizon.com>
To: linux@horizon.com, tglx@linutronix.de
Subject: Re: [patch 2/7] timer: Remove FIFO guarantee
Cc: linux-kernel@vger.kernel.org, peterz@infradead.org,
        viresh.kumar@linaro.org
In-Reply-To: <alpine.DEB.2.11.1506081930550.4133@nanos>
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1325
Lines: 35

Thomas Gleixner wrote:
> It does. Depending on when you enqueue the timer because the thing is
> calculated from the delta (expires - jiffies).

Ah, right.  If slack > 0, the slack amount is absolute and the rounding
will be consistent.

But if slack < 0, which is the default, it's a percentage of remaining
jiffies.  Since slack only delays timeouts, an earlier-scheduled
timeout could easily be delayed more.

(There are only six calls to set_timer_slack() to change the default
to something positive in the kernel.)

>>    I'm not disagreeing with the change, but it's not clear to me that
>>    it's as safe as you think.

> After thinking more about it, I'm even more sure that any code which
> relies on the FIFO "guarantee" is broken today.

Indeed, I am completely convinced.  All I might request is a reassignment
of blame in the commit message.


Thank you for your comments on my other blue-sky ideas, too.

I need to look into why we're using wheels, and what the point is.
How much of an advantage do they have over an efficient priority queue
like a pairing heap?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/