Re: [Cake] Cake on openwrt - falling behind

[Dave Taht <dave.taht@gmail.com>]

On Mon, Jul 2, 2018 at 12:13 PM Toke Høiland-Jørgensen <toke@toke.dk> wrote:
>
> Pete Heist <pete@heistp.net> writes:
>
> >> On Jul 2, 2018, at 2:03 PM, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
> >>
> >> Pete Heist <pete@heistp.net> writes:
> >>
> >>> But, um, I find it curious that tb[TCA_PAD] has valid looking values
> >>> in it, and if I just go:
> >>>
> >>> tb[TCA_STATS2] = tb[TCA_PAD]
> >>>
> >>> right after parse_rtattr is called, I start getting tin stats printed
> >>> that look valid. There should be zeroes or invalid values at
> >>> tb[TCA_PAD], as that’s just supposed to be padding, right?
> >>
> >> Hmm, that's interesting. Sounds like you are on the right track. What
> >> are the numerical values of TCA_STATS2 and TCA_PAD in the kernel and
> >> iproute2, respectively?
> >
> > I never would’ve guessed they could be different when compiled at once
> > :) but true, there are at least five different versions of rtnetlink.h
> > under build_dir based on their md5sums, and I see one belongs to
> > iproute2-full. In all of those, it looks in the source like
> > TCA_STATS2=7 and TCA_PAD=9.
>
> Aha! I think I figured out what is going on:
>
> The gen_stats facility will add an nlattr header at the beginning of the
> qdisc stats, which is the toplevel TLV that contains all stats (and that
> we put our stats inside). It stores a reference to this header, and when
> all the per-qdisc callbacks have finished adding their stats, it goes
> back and fixes up the length of the containing header.
>
> The problem is that on architectures that need padding, the padding TLV
> is added *first*, which means that the nlattr pointer that is stored
> before the callbacks are performed points to the padding TLV and not the
> stats TLV. And so, when the header is fixed up, the result (from the
> parser's perspective) is just a very big padding TLV.
>
> The options TLV is before the stats TLV, so the bug only occurs if the
> options happen to have a length that means the stats will need padding.
> Which is why messing with the number of options "fixes" the bug.
>
> Could you try applying the patch below (to the kernel) and see if that
> resolves the issue, please?

This seems like it will introduce problems with stuff that isn't or is
legitimately broken in the first place, pointing to potentially random
data in the wrong place.

would a workaround be adding more padding to the cake stats output so
it's always even?

why does it work as written on arm?

>
> -Toke
>
>
> @@ -77,8 +77,12 @@ gnet_stats_start_copy_compat(struct sk_buff *skb, int type, int tc_stats_type,
>                 d->lock = lock;
>                 spin_lock_bh(lock);
>         }
> -       if (d->tail)
> -               return gnet_stats_copy(d, type, NULL, 0, padattr);
> +       if (d->tail) {
> +               int ret = gnet_stats_copy(d, type, NULL, 0, padattr);
> +               if (!ret)
> +                       d->tail = ((struct nlattr *)skb_tail_pointer(skb)) - 1;
> +               return ret;
> +       }
>
>         return 0;
>  }
> _______________________________________________
> Cake mailing list
> Cake@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cake



-- 

Dave Täht
CEO, TekLibre, LLC
http://www.teklibre.com
Tel: 1-669-226-2619