net: sch: prio: Add offload ability to PRIO qdisc

Add the ability to offload PRIO qdisc by using ndo_setup_tc.
There are three commands for PRIO offloading:
* TC_PRIO_REPLACE: handles set and tune
* TC_PRIO_DESTROY: handles qdisc destroy
* TC_PRIO_STATS: updates the qdiscs counters (given as reference)

Like RED qdisc, the indication of whether PRIO is being offloaded is being
set and updated as part of the dump function. It is so because the driver
could decide to offload or not based on the qdisc parent, which could
change without notifying the qdisc.
Signed-off-by: NNogah Frankel <nogahf@mellanox.com>
Reviewed-by: NYuval Mintz <yuvalm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

include/linux/netdevice.h

@@ -780,6 +780,7 @@ enum tc_setup_type {
 	TC_SETUP_BLOCK,
 	TC_SETUP_QDISC_CBS,
 	TC_SETUP_QDISC_RED,
+	TC_SETUP_QDISC_PRIO,
 };
 
 /* These structures hold the attributes of bpf state that are being passed

include/net/pkt_cls.h

@@ -761,4 +761,29 @@ struct tc_red_qopt_offload {
 	};
 };
 
+enum tc_prio_command {
+	TC_PRIO_REPLACE,
+	TC_PRIO_DESTROY,
+	TC_PRIO_STATS,
+};
+
+struct tc_prio_qopt_offload_params {
+	int bands;
+	u8 priomap[TC_PRIO_MAX + 1];
+	/* In case that a prio qdisc is offloaded and now is changed to a
+	 * non-offloadedable config, it needs to update the backlog & qlen
+	 * values to negate the HW backlog & qlen values (and only them).
+	 */
+	struct gnet_stats_queue *qstats;
+};
+
+struct tc_prio_qopt_offload {
+	enum tc_prio_command command;
+	u32 handle;
+	u32 parent;
+	union {
+		struct tc_prio_qopt_offload_params replace_params;
+		struct tc_qopt_offload_stats stats;
+	};
+};
 #endif

net/sched/sch_prio.c

@@ -142,6 +142,31 @@ prio_reset(struct Qdisc *sch)
 	sch->q.qlen = 0;
 }
 
+static int prio_offload(struct Qdisc *sch, bool enable)
+{
+	struct prio_sched_data *q = qdisc_priv(sch);
+	struct net_device *dev = qdisc_dev(sch);
+	struct tc_prio_qopt_offload opt = {
+		.handle = sch->handle,
+		.parent = sch->parent,
+	};
+
+	if (!tc_can_offload(dev) || !dev->netdev_ops->ndo_setup_tc)
+		return -EOPNOTSUPP;
+
+	if (enable) {
+		opt.command = TC_PRIO_REPLACE;
+		opt.replace_params.bands = q->bands;
+		memcpy(&opt.replace_params.priomap, q->prio2band,
+		       TC_PRIO_MAX + 1);
+		opt.replace_params.qstats = &sch->qstats;
+	} else {
+		opt.command = TC_PRIO_DESTROY;
+	}
+
+	return dev->netdev_ops->ndo_setup_tc(dev, TC_SETUP_QDISC_PRIO, &opt);
+}
+
 static void
 prio_destroy(struct Qdisc *sch)
 {
@@ -149,6 +174,7 @@ prio_destroy(struct Qdisc *sch)
 	struct prio_sched_data *q = qdisc_priv(sch);
 
 	tcf_block_put(q->block);
+	prio_offload(sch, false);
 	for (prio = 0; prio < q->bands; prio++)
 		qdisc_destroy(q->queues[prio]);
 }
@@ -204,6 +230,7 @@ static int prio_tune(struct Qdisc *sch, struct nlattr *opt,
 	}
 
 	sch_tree_unlock(sch);
+	prio_offload(sch, true);
 	return 0;
 }
 
@@ -223,15 +250,47 @@ static int prio_init(struct Qdisc *sch, struct nlattr *opt,
 	return prio_tune(sch, opt, extack);
 }
 
+static int prio_dump_offload(struct Qdisc *sch)
+{
+	struct net_device *dev = qdisc_dev(sch);
+	struct tc_prio_qopt_offload hw_stats = {
+		.handle = sch->handle,
+		.parent = sch->parent,
+		.command = TC_PRIO_STATS,
+		.stats.bstats = &sch->bstats,
+		.stats.qstats = &sch->qstats,
+	};
+	int err;
+
+	sch->flags &= ~TCQ_F_OFFLOADED;
+	if (!tc_can_offload(dev) || !dev->netdev_ops->ndo_setup_tc)
+		return 0;
+
+	err = dev->netdev_ops->ndo_setup_tc(dev, TC_SETUP_QDISC_PRIO,
+					    &hw_stats);
+	if (err == -EOPNOTSUPP)
+		return 0;
+
+	if (!err)
+		sch->flags |= TCQ_F_OFFLOADED;
+
+	return err;
+}
+
 static int prio_dump(struct Qdisc *sch, struct sk_buff *skb)
 {
 	struct prio_sched_data *q = qdisc_priv(sch);
 	unsigned char *b = skb_tail_pointer(skb);
 	struct tc_prio_qopt opt;
+	int err;
 
 	opt.bands = q->bands;
 	memcpy(&opt.priomap, q->prio2band, TC_PRIO_MAX + 1);
 
+	err = prio_dump_offload(sch);
+	if (err)
+		goto nla_put_failure;
+
 	if (nla_put(skb, TCA_OPTIONS, sizeof(opt), &opt))
 		goto nla_put_failure;