Re: Objection to draft-ietf-6man-rfc4291bis-07.txt

[神明達哉 <[e-mail]>]

At Mon, 27 Mar 2017 16:00:58 -0500,
Alexandre Petrescu <[e-mail]> wrote:

> > On the other hand, BSDs' implementation of the IN6_IS_ADDR_LINKLOCAL()
> > macro only checks the first 10 bits:
> >
> > #define IN6_IS_ADDR_LINKLOCAL(a)        \
> >         (((a)->s6_addr[0] == 0xfe) && (((a)->s6_addr[1] & 0xc0) == 0x80))
>
> Tatuya, masking it that way it means that fe81::/10 is also recognized
> as an LL address prefix, right?  I think it is good, but just checking.

fe81::/10 is just an "abnormal" form of fe80::/10, but I guess you
actually meant fe81::/16.  Assuming so, your understanding is correct:
an address that matches fe81::/16 is recognized as unicast link-local
wherever the IN6_IS_ADDR_LINKLOCAL macro is used for the test.  For
example, BSD kernels consider this kind of address to be unicast
link-local in the validation of received RA:

    if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
        nd6log((LOG_ERR,
            "nd6_ra_input: src %s is not link-local\n",
            ip6_sprintf(ip6bufs, &saddr6)));
        goto bad;
    }
(this code only checks the first 10 bits).

--
JINMEI, Tatuya