Re: [PATCH nf 1/3] netfilter: nft_set_pipapo: store index in scratch maps

[Florian Westphal <fw@strlen.de>]

Stefano Brivio <sbrivio@redhat.com> wrote:
> > In new code, k+4 is the perfect "already-aligned" address where we would
> > 'no-op' the address on a ARCH_KMALLOC_MINALIGN == 4 system.
> 
> Isn't the already aligned-address k - 4, that is, k + 28? With k + 4,
> we would have &scratch->map[0] at k + 8. But anyway:

Yes, k-4 (k+28).

[..]

> > Maybe thats what you were saying.  I could try to add/expand the
> > comments here for the alignment calculations.
> 
> ...yes, the rest is exactly what I meant. I'm not really satisfied of
> the paragraph below but maybe something on the lines of:
> 
> 	/* Align &scratch->map (not the struct itself): the extra
> 	 * %NFT_PIPAPO_ALIGN_HEADROOM bytes passed to kzalloc_node() above
> 	 * guarantee we can waste up to those bytes in order to align the map
> 	 * field regardless of its offset within the struct.
> 	 */

Thanks, thats good, I'll use that.

I think I'll also add a BUILD_BUG_ON to assert map
offset <= NFT_PIPAPO_ALIGN_HEADROOM, just in case.