Re: [PATCHv2 2/2] mac80211: add support for mcs masks

[Simon Wunderlich <simon.wunderlich@s2003.tu-chemnitz.de>]

[-- Attachment #1: Type: text/plain, Size: 1547 bytes --]

On Tue, Jan 10, 2012 at 06:35:11PM +0100, Johannes Berg wrote:
> On Tue, 2012-01-10 at 18:28 +0100, Simon Wunderlich wrote:
> > Hello Johannes,
> 
> > > All this logic is pretty complex, maybe we could translate the bitmap to
> > > an array of unsigned longs and use test_bit() instead of open-coding it
> > > for 8-bit words?
> > 
> > Mhm, I would rather not add further copies into this fast path function, and
> > simply casting will result in bad behaviour (alignment, endianess). 
> 
> Well you don't want to copy it, you want to maintain the data that way,
> I think? Ok I think I see -- you're concerned about the masking etc.
> Dunno. Maybe at least there's a way to abstract out all the
> calculations?
> 

I have looked into bitmap.h which offers a great set of functions for
unsigned long arrays, but unfortunately I couldn't find anything equivalent
for u8. Writing our own u8 bitmap function set would be overkill here (IMHO),
and abstracting/hiding these functions into defines or changing everything to
unsigned long to be consistent (including ht_caps.mcs.rx_mask) would be ugly
as well.

I also don't consider this piece of code to be beautiful, but would suggest
to keep it for now - or maybe someone else has a better idea.

> > Generally
> > changing the mcs_mask to u32 is possible,
> 
> No, u64 might work today but maybe somebody will want to use higher
> MCSes in the future?

Yeah, u64 is possible on some architectures too - most of my routers are
32 bit however. :)

Cheers,
	Simon

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]