linux-kernel - Re: [PATCH 1/2] Make x86_64 udelay() round up instead of down.

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]

Date:	Fri, 10 Nov 2006 23:16:19 +0100
From:	Blaisorblade <blaisorblade@...oo.it>
To:	Pavel Machek <pavel@....cz>
Cc:	Andrew Morton <akpm@...l.org>, Andi Kleen <ak@....de>,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/2] Make x86_64 udelay() round up instead of down.

Hi Pavel!

On Wednesday 01 November 2006 17:30, Pavel Machek wrote:
> Hi!
>
> > From: Paolo 'Blaisorblade' Giarrusso <blaisorblade@...oo.it>
> >
> > Port two patches from i386 to x86_64 delay.c to make sure all rounding is
> > done upward instead of downward.
> >
> > There is no sign in commit messages that the mismatch was done on
> > purpose, and "delay() guarantees sleeping at least for the specified
> > time" is still a valid rule IMHO.
> >
> > diff --git a/arch/x86_64/lib/delay.c b/arch/x86_64/lib/delay.c
> > index 50be909..7514df0 100644
> > --- a/arch/x86_64/lib/delay.c
> > +++ b/arch/x86_64/lib/delay.c
> > @@ -40,13 +40,13 @@ EXPORT_SYMBOL(__delay);
> >
> >  inline void __const_udelay(unsigned long xloops)
> >  {
> > -	__delay((xloops * HZ *
> > cpu_data[raw_smp_processor_id()].loops_per_jiffy) >> 32);
> > +	__delay((xloops * HZ *
> > cpu_data[raw_smp_processor_id()].loops_per_jiffy) >> 32 + 1);

Btw, that's read as >> (32 + 1). My fault.
> Well, if this should be *rounding* up, you should do
>
> (xloops * HZ * cpu_data[raw_smp_processor_id()].loops_per_jiffy +
> 0xffffffff) >> 32

Indeed, I did it so following i386 code (where fixing this is awkward - doable 
with carry sums or 64-bit sums), but it's better this way.

I'd slightly prefer (for readability) (xloops * HZ * 
cpu_data[raw_smp_processor_id()].loops_per_jiffy +
(1<<32)-1) >> 32, or even

DIV_ROUND_UP(xloops * HZ * cpu_data[raw_smp_processor_id()].loops_per_jiffy, 1 
<< 32)

but kernel.h macros such as that seem to be underused (everything reimplements 
them as min() and max() before 2.6) and should be made more consistent in 
naming:

#define ALIGN(x,a) (((x)+(a)-1UL)&~((a)-1UL))
#define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
#define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))

I hope that gcc optimizes well x / (1<<32) to x >> 32, but without assurance I 
won't use DIV_ROUND_UP in timer code.

 (I also thought there was a bug, so I had a deep look and the difference 
between -1 in roundup and -1UL is correct due to sign extension ruining 
bitmasks, even if it seemed a bad mismatch).
-- 
Inform me of my mistakes, so I can keep imitating Homer Simpson's "Doh!".
Paolo Giarrusso, aka Blaisorblade
http://www.user-mode-linux.org/~blaisorblade

Chiacchiera con i tuoi amici in tempo reale! 
 http://it.yahoo.com/mail_it/foot/*http://it.messenger.yahoo.com 
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/