netdev - RE: [PATCH net-next 1/2] random32: add prandom_u32_lt_N and convert "misuses" of reciprocal

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:	Thu, 16 Jan 2014 09:28:33 +0000
From:	David Laight <David.Laight@...LAB.COM>
To:	'Eric Dumazet' <eric.dumazet@...il.com>,
	Daniel Borkmann <dborkman@...hat.com>
CC:	"davem@...emloft.net" <davem@...emloft.net>,
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	Jakub Zawadzki <darkjames-ws@...kjames.pl>,
	Hannes Frederic Sowa <hannes@...essinduktion.org>
Subject: RE: [PATCH net-next 1/2] random32: add prandom_u32_lt_N and convert
 "misuses" of reciprocal_divide

From: Eric Dumazet
> On Thu, 2014-01-16 at 00:23 +0100, Daniel Borkmann wrote:
> 
> > @@ -1220,7 +1219,7 @@ static unsigned int fanout_demux_hash(struct packet_fanout *f,
> >  				      struct sk_buff *skb,
> >  				      unsigned int num)
> >  {
> > -	return reciprocal_divide(skb->rxhash, num);
> > +	return (u32)(((u64) skb->rxhash * num) >> 32);
> >  }
> >
> 
> This is unfortunate.
> 
> (This reverts one of your patch : f55d112e529386 )
> 
> Please add a helper to explain what's going on here, and on many other
> spots we do this computation (as in get_rps_cpu()).
> Few people really understand this.
> 
> Or keep reciprocal_divide() as is, and introduce a new set of functions
> for people really wanting the precise divides.

Maybe rename the current reciprocal_divide() as well?

In the above code it isn't immediately obvious that the result of
a division is as useful as that of a modulus!
OTOH the high 32 bits of a 32x32 multiply are (slightly) more obviously
in the required range.

	David