linux-kernel - Re: [RFC][PATCH] new byteorder primitives

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:   Tue, 12 Dec 2017 12:04:09 -0800
From:   Jakub Kicinski <kubakici@...pl>
To:     Al Viro <viro@...IV.linux.org.uk>
Cc:     Linus Torvalds <torvalds@...ux-foundation.org>,
        netdev@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [RFC][PATCH] new byteorder primitives -
 ..._{replace,get}_bits()

On Tue, 12 Dec 2017 19:45:32 +0000, Al Viro wrote:
> On Tue, Dec 12, 2017 at 06:20:02AM +0000, Al Viro wrote:
> 
> > Umm...  What's wrong with
> > 
> > #define FIELD_FOO 0,4
> > #define FIELD_BAR 6,12
> > #define FIELD_BAZ 18,14
> > 
> > A macro can bloody well expand to any sequence of tokens - le32_get_bits(v, FIELD_BAZ)
> > will become le32_get_bits(v, 18, 14) just fine.  What's the problem with that?  
> 
> FWIW, if you want to use the mask, __builtin_ffsll() is not the only way to do
> it - you don't need the shift.  Multiplier would do just as well, and that can
> be had easier.  If mask = (2*a + 1)<<n = ((2*a)<<n) ^ (1<<n), then
> 	mask - 1 = ((2*a) << n) + ((1<<n) - 1) = ((2*n) << n) ^ ((1<<n) - 1)
> 	mask ^ (mask - 1) = (1<<n) + ((1<<n) - 1)
> and
> 	mask & (mask ^ (mask - 1)) = 1<<n.
> 
> IOW, with
> 
> static __always_inline u64 mask_to_multiplier(u64 mask)
> {
> 	return mask & (mask ^ (mask - 1));
> }
> 
> we could do
> 
> static __always_inline __le64 le64_replace_bits(__le64 old, u64 v, u64 mask)
> {
> 	__le64 m = cpu_to_le64(mask);
> 	return (old & ~m) | (cpu_to_le64(v * mask_to_multiplier(mask)) & m);
> }
> 
> static __always_inline u64 le64_get_bits(__le64 v, u64 mask)
> {
> 	return (le64_to_cpu(v) & mask) / mask_to_multiplier(mask);
> }
> 
> etc.  Compiler will turn those into shifts...  I can live with either calling
> conventions.
> 
> Comments?

Very nice!  The compilation-time check if the value can fit in a field
covered by the mask (if they're both known) did help me catch bugs
early a few times over the years, so if it could be preserved we can
maybe even drop the FIELD_* macros and just use this approach?