On Tue, Apr 03, 2018 at 01:30:33PM +0200, Florian Westphal wrote:
> Pablo Neira Ayuso <pablo@xxxxxxxxxxxxx> wrote:
> > On Mon, Apr 02, 2018 at 10:30:12PM +0200, Florian Westphal wrote:
> > > Pablo Neira Ayuso <pablo@xxxxxxxxxxxxx> wrote:
> > > > On Sun, Apr 01, 2018 at 12:47:47AM +0200, Florian Westphal wrote:
> > > > > this makes following failing test case work:
> > > > > ip6 dscp vmap { 0x04, ..
> > > > >
> > > > > problem was that the 6bit dscp value spans a byte boundary,
> > > > > so payload postprocessing increases the size to a 2 byte access
> > > > > (plus a binop mask operation to discard the unwanted bits).
> > > > >
> > > > > But because set key size was already set based on the previous
> > > > > payload size, we get a set with a keysize of 1 that is then
> > > > > populated with keys of size 2 (which causes a kernel error).
> > > > >
> > > > > So, fixup the set size so kernel is told to expect 2-byte keys
> > > > > in this case.
> > > > >
> > > > > Signed-off-by: Florian Westphal <fw@xxxxxxxxx>
> > > > > ---
> > > > > diff --git a/src/evaluate.c b/src/evaluate.c
> > > > > index d224f0f3c2c1..a1193f7d8456 100644
> > > > > --- a/src/evaluate.c
> > > > > +++ b/src/evaluate.c
> > > > > @@ -1279,6 +1279,8 @@ static int expr_evaluate_map(struct eval_ctx *ctx, struct expr **expr)
> > > > > if (expr_evaluate(ctx, &map->mappings->set->init) < 0)
> > > > > return -1;
> > > > > expr_set_context(&ctx->ectx, ctx->set->key->dtype, ctx->set->key->len);
> > > > > + if (ctx->set->key->len < map->left->len)
> > > > > + ctx->set->key->len = map->left->len;
> > > >
> > > > Will this work with concatenations?
> > >
> > > No, concat is still broken, binop_transfer() doesn't handle them.
> > >
> > > I'll look into it.
> >
> > http://patchwork.ozlabs.org/patch/894505/
> > http://patchwork.ozlabs.org/patch/894509/
> >
> > concatenations are still not working, because we hit:
> >
> > BUG: invalid binary operation 5
> > nft: netlink_linearize.c:567: netlink_gen_binop: Assertion `0' failed.
> > Aborted
> >
> > But that is exactly the same problem that Sabrina has been reporting
> > lately, ie. binop_transfer() needs to learn how to deal with
> > binary operations (as in Sabrina's case) and concatenations (this
> > particular case).
> >
> > If you don't mind, I'll push out this. Thanks.
>
> Yes, push this.
>
> WARNING: I am rewriting binop_transfer handling RIGHT NOW.
> (Well, simplifying is more like it).
>
> I think i will have concat working soon, and even if it doesn't code
> will be much simpler (after some refactoring, i can now handle SET_REF
> with recursive call back to binop_transfer().
>
> So please give me a few hours before you start working on same thing :-)
Great. I won't work on this :-)
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
