Re: [PATCH] net/sched: fq: account internal fast-path enqueues against plimit

[Eric Dumazet]

On Mon, Jun 8, 2026 at 5:39 PM Samuel Moelius
<sam.moelius@xxxxxxxxxxxxxxx> wrote:
>
> sch_fq has an internal fast path that can enqueue packets without
> applying the same packet-limit accounting as the normal path. The queue
> can grow past the configured plimit through that bypass.
>
> The packet limit must apply to all enqueue paths.
>
> Charge internal fast-path enqueues to the same plimit checks as ordinary
> enqueues.
>
> Assisted-by: Codex:gpt-5.5-cyber-preview
> Signed-off-by: Samuel Moelius <sam.moelius@xxxxxxxxxxxxxxx>
> ---
>  net/sched/sch_fq.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/net/sched/sch_fq.c b/net/sched/sch_fq.c
> index f2edcf872981..608ff3ecfb16 100644
> --- a/net/sched/sch_fq.c
> +++ b/net/sched/sch_fq.c
> @@ -548,7 +548,7 @@ static int fq_enqueue(struct sk_buff *skb, struct Qdisc *sch,
>         u8 band;
>
>         band = fq_prio2band(q->prio2band, skb->priority & TC_PRIO_MAX);
> -       if (unlikely(q->band_pkt_count[band] >= sch->limit)) {
> +       if (unlikely(sch->q.qlen >= sch->limit)) {
>                 q->stat_band_drops[band]++;
>                 return qdisc_drop_reason(skb, sch, to_free, QDISC_DROP_BAND_LIMIT);
>         }
> --
> 2.43.0
>

No, this is very much intentional, please look at commit
29f834aa326e659ed354c406056e94ea3d29706a
("net_sched: sch_fq: add 3 bands and WRR scheduling") changelog.

Your patch would totally break FQ band isolation.

Your patches are lacking to describe their targets (net or net-next).
I see no Fixes tag.

Please take a look at Documentation/process/maintainer-netdev.rst