Florian Westphal <fw@xxxxxxxxx> wrote:
> Brian Davidson <davidson.brian@xxxxxxxxx> wrote:
> > Using 'ip6 dscp' in a map lookup does not give expected results. It
> > seems to always match the zero value (cs0). It appears in the first
> > rule that the byteorder is not being converted between the two bitwise
> > operations, which is what happens when using ip6 dscp directly in the
> > second rule.
> >
> > # nft -f - <<EOF
> > add table ip6 t
> > add chain ip6 t c
> > add map ip6 t mapv6 { typeof ip6 dscp : meta mark; }
> > EOF
> >
> > # nft -d netlink add rule ip6 t c meta mark set ip6 dscp map @mapv6
> > ip6 t c
> > [ payload load 2b @ network header + 0 => reg 1 ]
> > [ bitwise reg 1 = ( reg 1 & 0x0000c00f ) ^ 0x00000000 ]
> > [ bitwise reg 1 = ( reg 1 >> 0x00000006 ) ]
> > [ lookup reg 1 set mapv6 dreg 1 ]
> > [ meta set mark with reg 1 ]
> >
> > # nft -d netlink add rule ip6 t c meta mark set ip6 dscp
> > ip6 t c
> > [ payload load 2b @ network header + 0 => reg 1 ]
> > [ bitwise reg 1 = ( reg 1 & 0x0000c00f ) ^ 0x00000000 ]
> > [ byteorder reg 1 = ntoh(reg 1, 2, 2) ]
> > [ bitwise reg 1 = ( reg 1 >> 0x00000006 ) ]
> > [ meta set mark with reg 1 ]
>
> Uhm. Pablo, any idea why the byte-swap-or-not logic depends
> on something *other* than if the mask length is > 8 bit or not?
>
> diff --git a/src/evaluate.c b/src/evaluate.c
> --- a/src/evaluate.c
> +++ b/src/evaluate.c
> @@ -545,7 +545,7 @@ static void expr_evaluate_bits(struct eval_ctx *ctx, struct expr **exprp)
> and->len = masklen;
>
> if (shift) {
> - if (ctx->stmt_len > 0 && div_round_up(masklen, BITS_PER_BYTE) > 1) {
> + if (masklen > BITS_PER_BYTE) {
I think this is right but binop xfer
won't remove the inserted byteorder conversion
in case the shift is to be removed by adjusting
a constant right hand side value.
I will have a look.
