linux-kernel - Re: [PATCH] timers: Use only bucket expiry for base->next

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:   Tue, 14 Jul 2020 09:08:11 +0200 (CEST)
From:   Anna-Maria Behnsen <anna-maria@...utronix.de>
To:     Frederic Weisbecker <frederic@...nel.org>
cc:     linux-kernel@...r.kernel.org,
        Peter Zijlstra <peterz@...radead.org>,
        Juri Lelli <juri.lelli@...hat.com>,
        Thomas Gleixner <tglx@...utronix.de>
Subject: Re: [PATCH] timers: Use only bucket expiry for base->next_expiry
 value

Hi Frederic,

On Fri, 10 Jul 2020, Frederic Weisbecker wrote:

> Hi Anna-Maria,
> 
> Nice change, it indeed makes more sense that way.
> Just a few details below:
> 
> On Fri, Jul 10, 2020 at 05:46:22PM +0200, Anna-Maria Behnsen wrote:
> > Use the expiry value of the bucket into which the timer is queued to
> > do the new first timer check. This fixes the base->clk going backward
> > problem and also prevents unnecessary softirq invocations when the
> > timer->expiry is not equal to the bucket expiry time in case of a new
> > first timer which is queued in a secondary wheel level.
> 
> I think there shouldn't be such unecessary softirq invocations. Either they
> fire at the bucket expiry time or the timer expiry time, it doesn't make
> much difference.

It will make a difference but only with your queue (I had the changes of
your queue already in mind, when writing the commit message)... I will
remove this.

> More important below:
> 
> > -static int calc_wheel_index(unsigned long expires, unsigned long clk)
> > +static int calc_wheel_index(unsigned long expires, unsigned long clk,
> > +			    unsigned long *bucket_expiry)
> >  {
> >  	unsigned long delta = expires - clk;
> >  	unsigned int idx;
> >  
> >  	if (delta < LVL_START(1)) {
> > -		idx = calc_index(expires, 0);
> > +		idx = calc_index(expires, 0, bucket_expiry);
> >  	} else if (delta < LVL_START(2)) {
> > -		idx = calc_index(expires, 1);
> > +		idx = calc_index(expires, 1, bucket_expiry);
> >  	} else if (delta < LVL_START(3)) {
> > -		idx = calc_index(expires, 2);
> > +		idx = calc_index(expires, 2, bucket_expiry);
> >  	} else if (delta < LVL_START(4)) {
> > -		idx = calc_index(expires, 3);
> > +		idx = calc_index(expires, 3, bucket_expiry);
> >  	} else if (delta < LVL_START(5)) {
> > -		idx = calc_index(expires, 4);
> > +		idx = calc_index(expires, 4, bucket_expiry);
> >  	} else if (delta < LVL_START(6)) {
> > -		idx = calc_index(expires, 5);
> > +		idx = calc_index(expires, 5, bucket_expiry);
> >  	} else if (delta < LVL_START(7)) {
> > -		idx = calc_index(expires, 6);
> > +		idx = calc_index(expires, 6, bucket_expiry);
> >  	} else if (LVL_DEPTH > 8 && delta < LVL_START(8)) {
> > -		idx = calc_index(expires, 7);
> > +		idx = calc_index(expires, 7, bucket_expiry);
> >  	} else if ((long) delta < 0) {
> >  		idx = clk & LVL_MASK;
> 
> You also need to handle that part. That's in fact the critical one  :)
> 

damn... too many idx here...

Thanks,

	Anna-Maria