netdev - Re: [PATCH net] lib: packing: fix shift wrapping in bit

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: <Y5CMPGrSuP+0ptdP@kadam>
Date:   Wed, 7 Dec 2022 15:51:08 +0300
From:   Dan Carpenter <error27@...il.com>
To:     Vladimir Oltean <olteanv@...il.com>
Cc:     "David S. Miller" <davem@...emloft.net>, netdev@...r.kernel.org,
        kernel-janitors@...r.kernel.org
Subject: Re: [PATCH net] lib: packing: fix shift wrapping in bit_reverse()

On Wed, Dec 07, 2022 at 02:22:54PM +0200, Vladimir Oltean wrote:
> On Wed, Dec 07, 2022 at 03:21:04PM +0300, Dan Carpenter wrote:
> > On Wed, Dec 07, 2022 at 02:19:36PM +0200, Vladimir Oltean wrote:
> > > On Wed, Dec 07, 2022 at 02:23:28PM +0300, Dan Carpenter wrote:
> > > > The bit_reverse() function is clearly supposed to be able to handle
> > > > 64 bit values, but the types for "(1 << i)" and "bit << (width - i - 1)"
> > > > are not enough to handle more than 32 bits.
> > > > 
> > > > Fixes: 554aae35007e ("lib: Add support for generic packing operations")
> > > > Signed-off-by: Dan Carpenter <error27@...il.com>
> > > > ---
> > > >  lib/packing.c | 5 ++---
> > > >  1 file changed, 2 insertions(+), 3 deletions(-)
> > > > 
> > > > diff --git a/lib/packing.c b/lib/packing.c
> > > > index 9a72f4bbf0e2..9d7418052f5a 100644
> > > > --- a/lib/packing.c
> > > > +++ b/lib/packing.c
> > > > @@ -32,12 +32,11 @@ static int get_reverse_lsw32_offset(int offset, size_t len)
> > > >  static u64 bit_reverse(u64 val, unsigned int width)
> > > >  {
> > > >  	u64 new_val = 0;
> > > > -	unsigned int bit;
> > > >  	unsigned int i;
> > > >  
> > > >  	for (i = 0; i < width; i++) {
> > > > -		bit = (val & (1 << i)) != 0;
> > > > -		new_val |= (bit << (width - i - 1));
> > > > +		if (val & BIT_ULL(1))
> > > 
> > > hmm, why 1 and not i?
> > 
> > Because I'm a moron.  Let me resend.
> 
> Wait a second, I deliberately wrote the code without conditionals.
> Let me look at the code disassembly before and after the patch and see
> what they look like.

My crappy benchmark says that the if statement is faster.  22 vs 26
seconds.

regards,
dan carpenter

#include <stdio.h>
#include <limits.h>
#include <stdbool.h>
#include <string.h>

#define BIT(n) (1 << (n))
#define BIT_ULL(n) (1ULL << (n))

#define u64 unsigned long long
#define u32 unsigned int
#define u16 unsigned short
#define u8  unsigned char

static u64 bit_reverse1(u64 val, unsigned int width)
{
	u64 new_val = 0;
	unsigned int i;

	for (i = 0; i < width; i++) {
		if (val & BIT_ULL(i))
			new_val |= BIT_ULL(width - i - 1);
	}
	return new_val;
}

static u64 bit_reverse2(u64 val, unsigned int width)
{
	u64 new_val = 0;
	u64 bit;
	unsigned int i;

	for (i = 0; i < width; i++) {
		bit = (val & BIT_ULL(i)) != 0;
		new_val |= (bit << (width - i - 1));
	}
	return new_val;
}

int main(void)
{
	unsigned long long val;

	for (val = ULLONG_MAX - INT_MAX; val; val++)
		bit_reverse1(val, 2);

	return 0;
}