Re: [PATCH v3] add equivalent of BIT(x) for bitfields

[Jakub Kicinski <jakub.kicinski@netronome.com>]

On Wed, 7 Dec 2016 11:00:57 +0100, Sebastian Frias wrote:
> On 07/12/16 09:42, Kalle Valo wrote:
> > Sebastian Frias <sf84@laposte.net> writes:
> >   
> >> Introduce GENVALUE(msb, lsb, value) macro to ease dealing with
> >> continuous bitfields, just as BIT(x) does for single bits.
> >>
> >> GENVALUE_ULL(msb, lsb, value) macro is also added.
> >>
> >> This is useful mostly for creating values to be packed together
> >> via OR operations, ex:
> >>
> >>    u32 val = 0x11110000;
> >>    val |= GENVALUE(19, 12, 0x5a);
> >>
> >> now 'val = 0x1115a000'
> >>
> >>
> >> Signed-off-by: Sebastian Frias <sf84@laposte.net>
> >> Link: https://marc.info/?l=linux-kernel&m=148094498711000&w=2
> >> ---
> >>
> >> Change in v2:
> >> - rename the macro to GENVALUE as proposed by Linus
> >> - longer comment attempts to show use case for the macro as
> >> proposed by Borislav
> >>
> >> Change in v3:
> >> - use BUILD_BUG_ON_ZERO() to break if some input parameters
> >> (essentially 'lsb' but also 'msb') are not constants as
> >> proposed by Linus.
> >> Indeed, 'lsb' is used twice so it cannot have side-effects;
> >> 'msb' is subjected to same constraints for consistency.  
> > 
> > (I missed there was v3 already, but I'll repeat what I said in v1.)
> > 
> > Please check FIELD_PREP() from include/linux/bitfield.h, doesn't it
> > already do the same?  
> 
> Indeed, it appears to do the same :-)
> Any reason why "include/linux/bitfield.h" is not included by default in
> bitops.h?

Hi!

The code is in a separate header because of circular dependencies
(coming from bug.h including bitops.h, IIRC).  You could possibly add an
include of bitfield.h in bitops.h if you're very careful, I haven't
tried TBH :)