[mirror_ubuntu-jammy-kernel.git] / net / sched / sch_cbs.c

// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * net/sched/sch_cbs.c	Credit Based Shaper
 *
 * Authors:	Vinicius Costa Gomes <vinicius.gomes@intel.com>
 */

/* Credit Based Shaper (CBS)
 * =========================
 *
 * This is a simple rate-limiting shaper aimed at TSN applications on
 * systems with known traffic workloads.
 *
 * Its algorithm is defined by the IEEE 802.1Q-2014 Specification,
 * Section 8.6.8.2, and explained in more detail in the Annex L of the
 * same specification.
 *
 * There are four tunables to be considered:
 *
 *	'idleslope': Idleslope is the rate of credits that is
 *	accumulated (in kilobits per second) when there is at least
 *	one packet waiting for transmission. Packets are transmitted
 *	when the current value of credits is equal or greater than
 *	zero. When there is no packet to be transmitted the amount of
 *	credits is set to zero. This is the main tunable of the CBS
 *	algorithm.
 *
 *	'sendslope':
 *	Sendslope is the rate of credits that is depleted (it should be a
 *	negative number of kilobits per second) when a transmission is
 *	ocurring. It can be calculated as follows, (IEEE 802.1Q-2014 Section
 *	8.6.8.2 item g):
 *
 *	sendslope = idleslope - port_transmit_rate
 *
 *	'hicredit': Hicredit defines the maximum amount of credits (in
 *	bytes) that can be accumulated. Hicredit depends on the
 *	characteristics of interfering traffic,
 *	'max_interference_size' is the maximum size of any burst of
 *	traffic that can delay the transmission of a frame that is
 *	available for transmission for this traffic class, (IEEE
 *	802.1Q-2014 Annex L, Equation L-3):
 *
 *	hicredit = max_interference_size * (idleslope / port_transmit_rate)
 *
 *	'locredit': Locredit is the minimum amount of credits that can
 *	be reached. It is a function of the traffic flowing through
 *	this qdisc (IEEE 802.1Q-2014 Annex L, Equation L-2):
 *
 *	locredit = max_frame_size * (sendslope / port_transmit_rate)
 */

#include <linux/ethtool.h>
#include <linux/module.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/string.h>
#include <linux/errno.h>
#include <linux/skbuff.h>
#include <net/netevent.h>
#include <net/netlink.h>
#include <net/sch_generic.h>
#include <net/pkt_sched.h>

static LIST_HEAD(cbs_list);
static DEFINE_SPINLOCK(cbs_list_lock);

#define BYTES_PER_KBIT (1000LL / 8)

struct cbs_sched_data {
	bool offload;
	int queue;
	atomic64_t port_rate; /* in bytes/s */
	s64 last; /* timestamp in ns */
	s64 credits; /* in bytes */
	s32 locredit; /* in bytes */
	s32 hicredit; /* in bytes */
	s64 sendslope; /* in bytes/s */
	s64 idleslope; /* in bytes/s */
	struct qdisc_watchdog watchdog;
	int (*enqueue)(struct sk_buff *skb, struct Qdisc *sch,
		       struct sk_buff **to_free);
	struct sk_buff *(*dequeue)(struct Qdisc *sch);
	struct Qdisc *qdisc;
	struct list_head cbs_list;
};

static int cbs_child_enqueue(struct sk_buff *skb, struct Qdisc *sch,
			     struct Qdisc *child,
			     struct sk_buff **to_free)
{
	unsigned int len = qdisc_pkt_len(skb);
	int err;

	err = child->ops->enqueue(skb, child, to_free);
	if (err != NET_XMIT_SUCCESS)
		return err;

	sch->qstats.backlog += len;
	sch->q.qlen++;

	return NET_XMIT_SUCCESS;
}

static int cbs_enqueue_offload(struct sk_buff *skb, struct Qdisc *sch,
			       struct sk_buff **to_free)
{
	struct cbs_sched_data *q = qdisc_priv(sch);
	struct Qdisc *qdisc = q->qdisc;

	return cbs_child_enqueue(skb, sch, qdisc, to_free);
}

static int cbs_enqueue_soft(struct sk_buff *skb, struct Qdisc *sch,
			    struct sk_buff **to_free)
{
	struct cbs_sched_data *q = qdisc_priv(sch);
	struct Qdisc *qdisc = q->qdisc;

	if (sch->q.qlen == 0 && q->credits > 0) {
		/* We need to stop accumulating credits when there's
		 * no enqueued packets and q->credits is positive.
		 */
		q->credits = 0;
		q->last = ktime_get_ns();
	}

	return cbs_child_enqueue(skb, sch, qdisc, to_free);
}

static int cbs_enqueue(struct sk_buff *skb, struct Qdisc *sch,
		       struct sk_buff **to_free)
{
	struct cbs_sched_data *q = qdisc_priv(sch);

	return q->enqueue(skb, sch, to_free);
}

/* timediff is in ns, slope is in bytes/s */
static s64 timediff_to_credits(s64 timediff, s64 slope)
{
	return div64_s64(timediff * slope, NSEC_PER_SEC);
}

static s64 delay_from_credits(s64 credits, s64 slope)
{
	if (unlikely(slope == 0))
		return S64_MAX;

	return div64_s64(-credits * NSEC_PER_SEC, slope);
}

static s64 credits_from_len(unsigned int len, s64 slope, s64 port_rate)
{
	if (unlikely(port_rate == 0))
		return S64_MAX;

	return div64_s64(len * slope, port_rate);
}

static struct sk_buff *cbs_child_dequeue(struct Qdisc *sch, struct Qdisc *child)
{
	struct sk_buff *skb;

	skb = child->ops->dequeue(child);
	if (!skb)
		return NULL;

	qdisc_qstats_backlog_dec(sch, skb);
	qdisc_bstats_update(sch, skb);
	sch->q.qlen--;

	return skb;
}

static struct sk_buff *cbs_dequeue_soft(struct Qdisc *sch)
{
	struct cbs_sched_data *q = qdisc_priv(sch);
	struct Qdisc *qdisc = q->qdisc;
	s64 now = ktime_get_ns();
	struct sk_buff *skb;
	s64 credits;
	int len;

	/* The previous packet is still being sent */
	if (now < q->last) {
		qdisc_watchdog_schedule_ns(&q->watchdog, q->last);
		return NULL;
	}
	if (q->credits < 0) {
		credits = timediff_to_credits(now - q->last, q->idleslope);

		credits = q->credits + credits;
		q->credits = min_t(s64, credits, q->hicredit);

		if (q->credits < 0) {
			s64 delay;

			delay = delay_from_credits(q->credits, q->idleslope);
			qdisc_watchdog_schedule_ns(&q->watchdog, now + delay);

			q->last = now;

			return NULL;
		}
	}
	skb = cbs_child_dequeue(sch, qdisc);
	if (!skb)
		return NULL;

	len = qdisc_pkt_len(skb);

	/* As sendslope is a negative number, this will decrease the
	 * amount of q->credits.
	 */
	credits = credits_from_len(len, q->sendslope,
				   atomic64_read(&q->port_rate));
	credits += q->credits;

	q->credits = max_t(s64, credits, q->locredit);
	/* Estimate of the transmission of the last byte of the packet in ns */
	if (unlikely(atomic64_read(&q->port_rate) == 0))
		q->last = now;
	else
		q->last = now + div64_s64(len * NSEC_PER_SEC,
					  atomic64_read(&q->port_rate));

	return skb;
}

static struct sk_buff *cbs_dequeue_offload(struct Qdisc *sch)
{
	struct cbs_sched_data *q = qdisc_priv(sch);
	struct Qdisc *qdisc = q->qdisc;

	return cbs_child_dequeue(sch, qdisc);
}

static struct sk_buff *cbs_dequeue(struct Qdisc *sch)
{
	struct cbs_sched_data *q = qdisc_priv(sch);

	return q->dequeue(sch);
}

static const struct nla_policy cbs_policy[TCA_CBS_MAX + 1] = {
	[TCA_CBS_PARMS]	= { .len = sizeof(struct tc_cbs_qopt) },
};

static void cbs_disable_offload(struct net_device *dev,
				struct cbs_sched_data *q)
{
	struct tc_cbs_qopt_offload cbs = { };
	const struct net_device_ops *ops;
	int err;

	if (!q->offload)
		return;

	q->enqueue = cbs_enqueue_soft;
	q->dequeue = cbs_dequeue_soft;

	ops = dev->netdev_ops;
	if (!ops->ndo_setup_tc)
		return;

	cbs.queue = q->queue;
	cbs.enable = 0;

	err = ops->ndo_setup_tc(dev, TC_SETUP_QDISC_CBS, &cbs);
	if (err < 0)
		pr_warn("Couldn't disable CBS offload for queue %d\n",
			cbs.queue);
}

static int cbs_enable_offload(struct net_device *dev, struct cbs_sched_data *q,
			      const struct tc_cbs_qopt *opt,
			      struct netlink_ext_ack *extack)
{
	const struct net_device_ops *ops = dev->netdev_ops;
	struct tc_cbs_qopt_offload cbs = { };
	int err;

	if (!ops->ndo_setup_tc) {
		NL_SET_ERR_MSG(extack, "Specified device does not support cbs offload");
		return -EOPNOTSUPP;
	}

	cbs.queue = q->queue;

	cbs.enable = 1;
	cbs.hicredit = opt->hicredit;
	cbs.locredit = opt->locredit;
	cbs.idleslope = opt->idleslope;
	cbs.sendslope = opt->sendslope;

	err = ops->ndo_setup_tc(dev, TC_SETUP_QDISC_CBS, &cbs);
	if (err < 0) {
		NL_SET_ERR_MSG(extack, "Specified device failed to setup cbs hardware offload");
		return err;
	}

	q->enqueue = cbs_enqueue_offload;
	q->dequeue = cbs_dequeue_offload;

	return 0;
}

static void cbs_set_port_rate(struct net_device *dev, struct cbs_sched_data *q)
{
	struct ethtool_link_ksettings ecmd;
	int speed = SPEED_10;
	int port_rate;
	int err;

	err = __ethtool_get_link_ksettings(dev, &ecmd);
	if (err < 0)
		goto skip;

	if (ecmd.base.speed && ecmd.base.speed != SPEED_UNKNOWN)
		speed = ecmd.base.speed;

skip:
	port_rate = speed * 1000 * BYTES_PER_KBIT;

	atomic64_set(&q->port_rate, port_rate);
	netdev_dbg(dev, "cbs: set %s's port_rate to: %lld, linkspeed: %d\n",
		   dev->name, (long long)atomic64_read(&q->port_rate),
		   ecmd.base.speed);
}

static int cbs_dev_notifier(struct notifier_block *nb, unsigned long event,
			    void *ptr)
{
	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
	struct cbs_sched_data *q;
	struct net_device *qdev;
	bool found = false;

	ASSERT_RTNL();

	if (event != NETDEV_UP && event != NETDEV_CHANGE)
		return NOTIFY_DONE;

	spin_lock(&cbs_list_lock);
	list_for_each_entry(q, &cbs_list, cbs_list) {
		qdev = qdisc_dev(q->qdisc);
		if (qdev == dev) {
			found = true;
			break;
		}
	}
	spin_unlock(&cbs_list_lock);

	if (found)
		cbs_set_port_rate(dev, q);

	return NOTIFY_DONE;
}

static int cbs_change(struct Qdisc *sch, struct nlattr *opt,
		      struct netlink_ext_ack *extack)
{
	struct cbs_sched_data *q = qdisc_priv(sch);
	struct net_device *dev = qdisc_dev(sch);
	struct nlattr *tb[TCA_CBS_MAX + 1];
	struct tc_cbs_qopt *qopt;
	int err;

	err = nla_parse_nested_deprecated(tb, TCA_CBS_MAX, opt, cbs_policy,
					  extack);
	if (err < 0)
		return err;

	if (!tb[TCA_CBS_PARMS]) {
		NL_SET_ERR_MSG(extack, "Missing CBS parameter which are mandatory");
		return -EINVAL;
	}

	qopt = nla_data(tb[TCA_CBS_PARMS]);

	if (!qopt->offload) {
		cbs_set_port_rate(dev, q);
		cbs_disable_offload(dev, q);
	} else {
		err = cbs_enable_offload(dev, q, qopt, extack);
		if (err < 0)
			return err;
	}

	/* Everything went OK, save the parameters used. */
	q->hicredit = qopt->hicredit;
	q->locredit = qopt->locredit;
	q->idleslope = qopt->idleslope * BYTES_PER_KBIT;
	q->sendslope = qopt->sendslope * BYTES_PER_KBIT;
	q->offload = qopt->offload;

	return 0;
}

static int cbs_init(struct Qdisc *sch, struct nlattr *opt,
		    struct netlink_ext_ack *extack)
{
	struct cbs_sched_data *q = qdisc_priv(sch);
	struct net_device *dev = qdisc_dev(sch);

	if (!opt) {
		NL_SET_ERR_MSG(extack, "Missing CBS qdisc options  which are mandatory");
		return -EINVAL;
	}

	q->qdisc = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
				     sch->handle, extack);
	if (!q->qdisc)
		return -ENOMEM;

	spin_lock(&cbs_list_lock);
	list_add(&q->cbs_list, &cbs_list);
	spin_unlock(&cbs_list_lock);

	qdisc_hash_add(q->qdisc, false);

	q->queue = sch->dev_queue - netdev_get_tx_queue(dev, 0);

	q->enqueue = cbs_enqueue_soft;
	q->dequeue = cbs_dequeue_soft;

	qdisc_watchdog_init(&q->watchdog, sch);

	return cbs_change(sch, opt, extack);
}

static void cbs_destroy(struct Qdisc *sch)
{
	struct cbs_sched_data *q = qdisc_priv(sch);
	struct net_device *dev = qdisc_dev(sch);

	/* Nothing to do if we couldn't create the underlying qdisc */
	if (!q->qdisc)
		return;

	qdisc_watchdog_cancel(&q->watchdog);
	cbs_disable_offload(dev, q);

	spin_lock(&cbs_list_lock);
	list_del(&q->cbs_list);
	spin_unlock(&cbs_list_lock);

	qdisc_put(q->qdisc);
}

static int cbs_dump(struct Qdisc *sch, struct sk_buff *skb)
{
	struct cbs_sched_data *q = qdisc_priv(sch);
	struct tc_cbs_qopt opt = { };
	struct nlattr *nest;

	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
	if (!nest)
		goto nla_put_failure;

	opt.hicredit = q->hicredit;
	opt.locredit = q->locredit;
	opt.sendslope = div64_s64(q->sendslope, BYTES_PER_KBIT);
	opt.idleslope = div64_s64(q->idleslope, BYTES_PER_KBIT);
	opt.offload = q->offload;

	if (nla_put(skb, TCA_CBS_PARMS, sizeof(opt), &opt))
		goto nla_put_failure;

	return nla_nest_end(skb, nest);

nla_put_failure:
	nla_nest_cancel(skb, nest);
	return -1;
}

static int cbs_dump_class(struct Qdisc *sch, unsigned long cl,
			  struct sk_buff *skb, struct tcmsg *tcm)
{
	struct cbs_sched_data *q = qdisc_priv(sch);

	if (cl != 1 || !q->qdisc)	/* only one class */
		return -ENOENT;

	tcm->tcm_handle |= TC_H_MIN(1);
	tcm->tcm_info = q->qdisc->handle;

	return 0;
}

static int cbs_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new,
		     struct Qdisc **old, struct netlink_ext_ack *extack)
{
	struct cbs_sched_data *q = qdisc_priv(sch);

	if (!new) {
		new = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
					sch->handle, NULL);
		if (!new)
			new = &noop_qdisc;
	}

	*old = qdisc_replace(sch, new, &q->qdisc);
	return 0;
}

static struct Qdisc *cbs_leaf(struct Qdisc *sch, unsigned long arg)
{
	struct cbs_sched_data *q = qdisc_priv(sch);

	return q->qdisc;
}

static unsigned long cbs_find(struct Qdisc *sch, u32 classid)
{
	return 1;
}

static void cbs_walk(struct Qdisc *sch, struct qdisc_walker *walker)
{
	if (!walker->stop) {
		if (walker->count >= walker->skip) {
			if (walker->fn(sch, 1, walker) < 0) {
				walker->stop = 1;
				return;
			}
		}
		walker->count++;
	}
}

static const struct Qdisc_class_ops cbs_class_ops = {
	.graft		=	cbs_graft,
	.leaf		=	cbs_leaf,
	.find		=	cbs_find,
	.walk		=	cbs_walk,
	.dump		=	cbs_dump_class,
};

static struct Qdisc_ops cbs_qdisc_ops __read_mostly = {
	.id		=	"cbs",
	.cl_ops		=	&cbs_class_ops,
	.priv_size	=	sizeof(struct cbs_sched_data),
	.enqueue	=	cbs_enqueue,
	.dequeue	=	cbs_dequeue,
	.peek		=	qdisc_peek_dequeued,
	.init		=	cbs_init,
	.reset		=	qdisc_reset_queue,
	.destroy	=	cbs_destroy,
	.change		=	cbs_change,
	.dump		=	cbs_dump,
	.owner		=	THIS_MODULE,
};

static struct notifier_block cbs_device_notifier = {
	.notifier_call = cbs_dev_notifier,
};

static int __init cbs_module_init(void)
{
	int err;

	err = register_netdevice_notifier(&cbs_device_notifier);
	if (err)
		return err;

	err = register_qdisc(&cbs_qdisc_ops);
	if (err)
		unregister_netdevice_notifier(&cbs_device_notifier);

	return err;
}

static void __exit cbs_module_exit(void)
{
	unregister_qdisc(&cbs_qdisc_ops);
	unregister_netdevice_notifier(&cbs_device_notifier);
}
module_init(cbs_module_init)
module_exit(cbs_module_exit)
MODULE_LICENSE("GPL");
Commit	Line	Data
2874c5fd	1	// SPDX-License-Identifier: GPL-2.0-or-later
585d763a VCG	2	/*
	3	* net/sched/sch_cbs.c Credit Based Shaper
	4	*
585d763a	5	* Authors: Vinicius Costa Gomes <vinicius.gomes@intel.com>
585d763a VCG	6	*/
	7
	8	/* Credit Based Shaper (CBS)
	9	* =========================
	10	*
	11	* This is a simple rate-limiting shaper aimed at TSN applications on
	12	* systems with known traffic workloads.
	13	*
	14	* Its algorithm is defined by the IEEE 802.1Q-2014 Specification,
	15	* Section 8.6.8.2, and explained in more detail in the Annex L of the
	16	* same specification.
	17	*
	18	* There are four tunables to be considered:
	19	*
	20	* 'idleslope': Idleslope is the rate of credits that is
	21	* accumulated (in kilobits per second) when there is at least
	22	* one packet waiting for transmission. Packets are transmitted
	23	* when the current value of credits is equal or greater than
	24	* zero. When there is no packet to be transmitted the amount of
	25	* credits is set to zero. This is the main tunable of the CBS
	26	* algorithm.
	27	*
	28	* 'sendslope':
	29	* Sendslope is the rate of credits that is depleted (it should be a
	30	* negative number of kilobits per second) when a transmission is
	31	* ocurring. It can be calculated as follows, (IEEE 802.1Q-2014 Section
	32	* 8.6.8.2 item g):
	33	*
	34	* sendslope = idleslope - port_transmit_rate
	35	*
	36	* 'hicredit': Hicredit defines the maximum amount of credits (in
	37	* bytes) that can be accumulated. Hicredit depends on the
	38	* characteristics of interfering traffic,
	39	* 'max_interference_size' is the maximum size of any burst of
	40	* traffic that can delay the transmission of a frame that is
	41	* available for transmission for this traffic class, (IEEE
	42	* 802.1Q-2014 Annex L, Equation L-3):
	43	*
	44	* hicredit = max_interference_size * (idleslope / port_transmit_rate)
	45	*
	46	* 'locredit': Locredit is the minimum amount of credits that can
	47	* be reached. It is a function of the traffic flowing through
	48	* this qdisc (IEEE 802.1Q-2014 Annex L, Equation L-2):
	49	*
	50	* locredit = max_frame_size * (sendslope / port_transmit_rate)
	51	*/
	52
cc69837f	53	#include <linux/ethtool.h>
585d763a VCG	54	#include <linux/module.h>
	55	#include <linux/types.h>
	56	#include <linux/kernel.h>
	57	#include <linux/string.h>
	58	#include <linux/errno.h>
	59	#include <linux/skbuff.h>
e0a7683d	60	#include <net/netevent.h>
585d763a VCG	61	#include <net/netlink.h>
	62	#include <net/sch_generic.h>
	63	#include <net/pkt_sched.h>
	64
e0a7683d LD	65	static LIST_HEAD(cbs_list);
	66	static DEFINE_SPINLOCK(cbs_list_lock);
	67
585d763a VCG	68	#define BYTES_PER_KBIT (1000LL / 8)
	69
	70	struct cbs_sched_data {
3d0bd028 VCG	71	bool offload;
3d0bd028 VCG	72	int queue;
e0a7683d	73	atomic64_t port_rate; /* in bytes/s */
585d763a VCG	74	s64 last; /* timestamp in ns */
	75	s64 credits; /* in bytes */
	76	s32 locredit; /* in bytes */
	77	s32 hicredit; /* in bytes */
	78	s64 sendslope; /* in bytes/s */
	79	s64 idleslope; /* in bytes/s */
	80	struct qdisc_watchdog watchdog;
990e35ec VCG	81	int (enqueue)(struct sk_buff skb, struct Qdisc *sch,
990e35ec VCG	82	struct sk_buff **to_free);
585d763a	83	struct sk_buff (dequeue)(struct Qdisc *sch);
990e35ec	84	struct Qdisc *qdisc;
e0a7683d	85	struct list_head cbs_list;
585d763a VCG	86	};
585d763a VCG	87
990e35ec VCG	88	static int cbs_child_enqueue(struct sk_buff skb, struct Qdisc sch,
	89	struct Qdisc *child,
	90	struct sk_buff **to_free)
3d0bd028	91	{
f6bab199	92	unsigned int len = qdisc_pkt_len(skb);
990e35ec VCG	93	int err;
	94
	95	err = child->ops->enqueue(skb, child, to_free);
	96	if (err != NET_XMIT_SUCCESS)
	97	return err;
	98
f6bab199	99	sch->qstats.backlog += len;
990e35ec VCG	100	sch->q.qlen++;
	101
	102	return NET_XMIT_SUCCESS;
3d0bd028 VCG	103	}
3d0bd028 VCG	104
990e35ec VCG	105	static int cbs_enqueue_offload(struct sk_buff skb, struct Qdisc sch,
990e35ec VCG	106	struct sk_buff **to_free)
585d763a VCG	107	{
585d763a VCG	108	struct cbs_sched_data *q = qdisc_priv(sch);
990e35ec VCG	109	struct Qdisc *qdisc = q->qdisc;
	110
	111	return cbs_child_enqueue(skb, sch, qdisc, to_free);
	112	}
	113
	114	static int cbs_enqueue_soft(struct sk_buff skb, struct Qdisc sch,
	115	struct sk_buff **to_free)
	116	{
	117	struct cbs_sched_data *q = qdisc_priv(sch);
	118	struct Qdisc *qdisc = q->qdisc;
585d763a VCG	119
	120	if (sch->q.qlen == 0 && q->credits > 0) {
	121	/* We need to stop accumulating credits when there's
	122	* no enqueued packets and q->credits is positive.
	123	*/
	124	q->credits = 0;
	125	q->last = ktime_get_ns();
	126	}
	127
990e35ec	128	return cbs_child_enqueue(skb, sch, qdisc, to_free);
585d763a VCG	129	}
	130
	131	static int cbs_enqueue(struct sk_buff skb, struct Qdisc sch,
	132	struct sk_buff **to_free)
	133	{
	134	struct cbs_sched_data *q = qdisc_priv(sch);
	135
990e35ec	136	return q->enqueue(skb, sch, to_free);
585d763a VCG	137	}
	138
	139	/* timediff is in ns, slope is in bytes/s */
	140	static s64 timediff_to_credits(s64 timediff, s64 slope)
	141	{
	142	return div64_s64(timediff * slope, NSEC_PER_SEC);
	143	}
	144
	145	static s64 delay_from_credits(s64 credits, s64 slope)
	146	{
	147	if (unlikely(slope == 0))
	148	return S64_MAX;
	149
	150	return div64_s64(-credits * NSEC_PER_SEC, slope);
	151	}
	152
	153	static s64 credits_from_len(unsigned int len, s64 slope, s64 port_rate)
	154	{
	155	if (unlikely(port_rate == 0))
	156	return S64_MAX;
	157
	158	return div64_s64(len * slope, port_rate);
	159	}
	160
990e35ec VCG	161	static struct sk_buff cbs_child_dequeue(struct Qdisc sch, struct Qdisc *child)
	162	{
	163	struct sk_buff *skb;
	164
	165	skb = child->ops->dequeue(child);
	166	if (!skb)
	167	return NULL;
	168
	169	qdisc_qstats_backlog_dec(sch, skb);
	170	qdisc_bstats_update(sch, skb);
	171	sch->q.qlen--;
	172
	173	return skb;
	174	}
	175
585d763a VCG	176	static struct sk_buff cbs_dequeue_soft(struct Qdisc sch)
	177	{
	178	struct cbs_sched_data *q = qdisc_priv(sch);
990e35ec	179	struct Qdisc *qdisc = q->qdisc;
585d763a VCG	180	s64 now = ktime_get_ns();
	181	struct sk_buff *skb;
	182	s64 credits;
	183	int len;
	184
961d0e5b ZY	185	/* The previous packet is still being sent */
	186	if (now < q->last) {
	187	qdisc_watchdog_schedule_ns(&q->watchdog, q->last);
	188	return NULL;
	189	}
585d763a VCG	190	if (q->credits < 0) {
	191	credits = timediff_to_credits(now - q->last, q->idleslope);
	192
	193	credits = q->credits + credits;
	194	q->credits = min_t(s64, credits, q->hicredit);
	195
	196	if (q->credits < 0) {
	197	s64 delay;
	198
	199	delay = delay_from_credits(q->credits, q->idleslope);
	200	qdisc_watchdog_schedule_ns(&q->watchdog, now + delay);
	201
	202	q->last = now;
	203
	204	return NULL;
	205	}
	206	}
990e35ec	207	skb = cbs_child_dequeue(sch, qdisc);
585d763a VCG	208	if (!skb)
	209	return NULL;
	210
	211	len = qdisc_pkt_len(skb);
	212
	213	/* As sendslope is a negative number, this will decrease the
	214	* amount of q->credits.
	215	*/
e0a7683d LD	216	credits = credits_from_len(len, q->sendslope,
e0a7683d LD	217	atomic64_read(&q->port_rate));
585d763a VCG	218	credits += q->credits;
	219
	220	q->credits = max_t(s64, credits, q->locredit);
961d0e5b ZY	221	/* Estimate of the transmission of the last byte of the packet in ns */
	222	if (unlikely(atomic64_read(&q->port_rate) == 0))
	223	q->last = now;
	224	else
	225	q->last = now + div64_s64(len * NSEC_PER_SEC,
	226	atomic64_read(&q->port_rate));
585d763a VCG	227
	228	return skb;
	229	}
	230
3d0bd028 VCG	231	static struct sk_buff cbs_dequeue_offload(struct Qdisc sch)
3d0bd028 VCG	232	{
990e35ec VCG	233	struct cbs_sched_data *q = qdisc_priv(sch);
	234	struct Qdisc *qdisc = q->qdisc;
	235
	236	return cbs_child_dequeue(sch, qdisc);
3d0bd028 VCG	237	}
3d0bd028 VCG	238
585d763a VCG	239	static struct sk_buff cbs_dequeue(struct Qdisc sch)
	240	{
	241	struct cbs_sched_data *q = qdisc_priv(sch);
	242
	243	return q->dequeue(sch);
	244	}
	245
	246	static const struct nla_policy cbs_policy[TCA_CBS_MAX + 1] = {
	247	[TCA_CBS_PARMS] = { .len = sizeof(struct tc_cbs_qopt) },
	248	};
	249
3d0bd028 VCG	250	static void cbs_disable_offload(struct net_device *dev,
	251	struct cbs_sched_data *q)
	252	{
	253	struct tc_cbs_qopt_offload cbs = { };
	254	const struct net_device_ops *ops;
	255	int err;
	256
	257	if (!q->offload)
	258	return;
	259
	260	q->enqueue = cbs_enqueue_soft;
	261	q->dequeue = cbs_dequeue_soft;
	262
	263	ops = dev->netdev_ops;
	264	if (!ops->ndo_setup_tc)
	265	return;
	266
	267	cbs.queue = q->queue;
	268	cbs.enable = 0;
	269
8521db4c	270	err = ops->ndo_setup_tc(dev, TC_SETUP_QDISC_CBS, &cbs);
3d0bd028 VCG	271	if (err < 0)
	272	pr_warn("Couldn't disable CBS offload for queue %d\n",
	273	cbs.queue);
	274	}
	275
	276	static int cbs_enable_offload(struct net_device dev, struct cbs_sched_data q,
710fb396 AA	277	const struct tc_cbs_qopt *opt,
710fb396 AA	278	struct netlink_ext_ack *extack)
3d0bd028 VCG	279	{
	280	const struct net_device_ops *ops = dev->netdev_ops;
	281	struct tc_cbs_qopt_offload cbs = { };
	282	int err;
	283
710fb396 AA	284	if (!ops->ndo_setup_tc) {
710fb396 AA	285	NL_SET_ERR_MSG(extack, "Specified device does not support cbs offload");
3d0bd028	286	return -EOPNOTSUPP;
710fb396	287	}
3d0bd028 VCG	288
	289	cbs.queue = q->queue;
	290
	291	cbs.enable = 1;
	292	cbs.hicredit = opt->hicredit;
	293	cbs.locredit = opt->locredit;
	294	cbs.idleslope = opt->idleslope;
	295	cbs.sendslope = opt->sendslope;
	296
8521db4c	297	err = ops->ndo_setup_tc(dev, TC_SETUP_QDISC_CBS, &cbs);
710fb396 AA	298	if (err < 0) {
710fb396 AA	299	NL_SET_ERR_MSG(extack, "Specified device failed to setup cbs hardware offload");
3d0bd028	300	return err;
710fb396	301	}
3d0bd028 VCG	302
	303	q->enqueue = cbs_enqueue_offload;
	304	q->dequeue = cbs_dequeue_offload;
	305
	306	return 0;
	307	}
	308
e0a7683d LD	309	static void cbs_set_port_rate(struct net_device dev, struct cbs_sched_data q)
	310	{
	311	struct ethtool_link_ksettings ecmd;
1c6c09a0	312	int speed = SPEED_10;
9367fa08	313	int port_rate;
1c6c09a0 VO	314	int err;
	315
	316	err = __ethtool_get_link_ksettings(dev, &ecmd);
	317	if (err < 0)
	318	goto skip;
	319
83c8c3cf	320	if (ecmd.base.speed && ecmd.base.speed != SPEED_UNKNOWN)
1c6c09a0	321	speed = ecmd.base.speed;
e0a7683d	322
1c6c09a0 VO	323	skip:
1c6c09a0 VO	324	port_rate = speed * 1000 * BYTES_PER_KBIT;
e0a7683d LD	325
	326	atomic64_set(&q->port_rate, port_rate);
	327	netdev_dbg(dev, "cbs: set %s's port_rate to: %lld, linkspeed: %d\n",
	328	dev->name, (long long)atomic64_read(&q->port_rate),
	329	ecmd.base.speed);
	330	}
	331
	332	static int cbs_dev_notifier(struct notifier_block *nb, unsigned long event,
	333	void *ptr)
	334	{
	335	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
	336	struct cbs_sched_data *q;
	337	struct net_device *qdev;
	338	bool found = false;
	339
	340	ASSERT_RTNL();
	341
	342	if (event != NETDEV_UP && event != NETDEV_CHANGE)
	343	return NOTIFY_DONE;
	344
	345	spin_lock(&cbs_list_lock);
	346	list_for_each_entry(q, &cbs_list, cbs_list) {
	347	qdev = qdisc_dev(q->qdisc);
	348	if (qdev == dev) {
	349	found = true;
	350	break;
	351	}
	352	}
	353	spin_unlock(&cbs_list_lock);
	354
	355	if (found)
	356	cbs_set_port_rate(dev, q);
	357
	358	return NOTIFY_DONE;
	359	}
	360
2030721c AA	361	static int cbs_change(struct Qdisc sch, struct nlattr opt,
2030721c AA	362	struct netlink_ext_ack *extack)
585d763a VCG	363	{
	364	struct cbs_sched_data *q = qdisc_priv(sch);
	365	struct net_device *dev = qdisc_dev(sch);
	366	struct nlattr *tb[TCA_CBS_MAX + 1];
585d763a	367	struct tc_cbs_qopt *qopt;
585d763a VCG	368	int err;
585d763a VCG	369
8cb08174 JB	370	err = nla_parse_nested_deprecated(tb, TCA_CBS_MAX, opt, cbs_policy,
8cb08174 JB	371	extack);
585d763a VCG	372	if (err < 0)
	373	return err;
	374
710fb396 AA	375	if (!tb[TCA_CBS_PARMS]) {
710fb396 AA	376	NL_SET_ERR_MSG(extack, "Missing CBS parameter which are mandatory");
585d763a	377	return -EINVAL;
710fb396	378	}
585d763a VCG	379
	380	qopt = nla_data(tb[TCA_CBS_PARMS]);
	381
3d0bd028	382	if (!qopt->offload) {
e0a7683d	383	cbs_set_port_rate(dev, q);
3d0bd028 VCG	384	cbs_disable_offload(dev, q);
3d0bd028 VCG	385	} else {
710fb396	386	err = cbs_enable_offload(dev, q, qopt, extack);
3d0bd028 VCG	387	if (err < 0)
	388	return err;
	389	}
585d763a	390
3d0bd028	391	/* Everything went OK, save the parameters used. */
585d763a VCG	392	q->hicredit = qopt->hicredit;
	393	q->locredit = qopt->locredit;
	394	q->idleslope = qopt->idleslope * BYTES_PER_KBIT;
	395	q->sendslope = qopt->sendslope * BYTES_PER_KBIT;
3d0bd028	396	q->offload = qopt->offload;
585d763a VCG	397
	398	return 0;
	399	}
	400
e63d7dfd AA	401	static int cbs_init(struct Qdisc sch, struct nlattr opt,
e63d7dfd AA	402	struct netlink_ext_ack *extack)
585d763a VCG	403	{
585d763a VCG	404	struct cbs_sched_data *q = qdisc_priv(sch);
3d0bd028	405	struct net_device *dev = qdisc_dev(sch);
585d763a	406
710fb396 AA	407	if (!opt) {
710fb396 AA	408	NL_SET_ERR_MSG(extack, "Missing CBS qdisc options which are mandatory");
585d763a	409	return -EINVAL;
710fb396	410	}
585d763a	411
990e35ec VCG	412	q->qdisc = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
	413	sch->handle, extack);
	414	if (!q->qdisc)
	415	return -ENOMEM;
	416
3e8b9bfa VCG	417	spin_lock(&cbs_list_lock);
	418	list_add(&q->cbs_list, &cbs_list);
	419	spin_unlock(&cbs_list_lock);
	420
990e35ec VCG	421	qdisc_hash_add(q->qdisc, false);
990e35ec VCG	422
3d0bd028 VCG	423	q->queue = sch->dev_queue - netdev_get_tx_queue(dev, 0);
	424
	425	q->enqueue = cbs_enqueue_soft;
	426	q->dequeue = cbs_dequeue_soft;
	427
585d763a VCG	428	qdisc_watchdog_init(&q->watchdog, sch);
585d763a VCG	429
3e8b9bfa	430	return cbs_change(sch, opt, extack);
585d763a VCG	431	}
	432
	433	static void cbs_destroy(struct Qdisc *sch)
	434	{
	435	struct cbs_sched_data *q = qdisc_priv(sch);
3d0bd028	436	struct net_device *dev = qdisc_dev(sch);
585d763a	437
3e8b9bfa VCG	438	/* Nothing to do if we couldn't create the underlying qdisc */
	439	if (!q->qdisc)
	440	return;
3d0bd028	441
e0a7683d	442	qdisc_watchdog_cancel(&q->watchdog);
3d0bd028	443	cbs_disable_offload(dev, q);
990e35ec	444
3e8b9bfa VCG	445	spin_lock(&cbs_list_lock);
	446	list_del(&q->cbs_list);
	447	spin_unlock(&cbs_list_lock);
	448
	449	qdisc_put(q->qdisc);
585d763a VCG	450	}
	451
	452	static int cbs_dump(struct Qdisc sch, struct sk_buff skb)
	453	{
	454	struct cbs_sched_data *q = qdisc_priv(sch);
	455	struct tc_cbs_qopt opt = { };
	456	struct nlattr *nest;
	457
ae0be8de	458	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
585d763a VCG	459	if (!nest)
	460	goto nla_put_failure;
	461
	462	opt.hicredit = q->hicredit;
	463	opt.locredit = q->locredit;
	464	opt.sendslope = div64_s64(q->sendslope, BYTES_PER_KBIT);
	465	opt.idleslope = div64_s64(q->idleslope, BYTES_PER_KBIT);
3d0bd028	466	opt.offload = q->offload;
585d763a VCG	467
	468	if (nla_put(skb, TCA_CBS_PARMS, sizeof(opt), &opt))
	469	goto nla_put_failure;
	470
	471	return nla_nest_end(skb, nest);
	472
	473	nla_put_failure:
	474	nla_nest_cancel(skb, nest);
	475	return -1;
	476	}
	477
990e35ec VCG	478	static int cbs_dump_class(struct Qdisc *sch, unsigned long cl,
	479	struct sk_buff skb, struct tcmsg tcm)
	480	{
	481	struct cbs_sched_data *q = qdisc_priv(sch);
	482
	483	if (cl != 1 \|\| !q->qdisc) /* only one class */
	484	return -ENOENT;
	485
	486	tcm->tcm_handle \|= TC_H_MIN(1);
	487	tcm->tcm_info = q->qdisc->handle;
	488
	489	return 0;
	490	}
	491
	492	static int cbs_graft(struct Qdisc sch, unsigned long arg, struct Qdisc new,
	493	struct Qdisc *old, struct netlink_ext_ack extack)
	494	{
	495	struct cbs_sched_data *q = qdisc_priv(sch);
	496
	497	if (!new) {
	498	new = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
	499	sch->handle, NULL);
	500	if (!new)
	501	new = &noop_qdisc;
	502	}
	503
	504	*old = qdisc_replace(sch, new, &q->qdisc);
	505	return 0;
	506	}
	507
	508	static struct Qdisc cbs_leaf(struct Qdisc sch, unsigned long arg)
	509	{
	510	struct cbs_sched_data *q = qdisc_priv(sch);
	511
	512	return q->qdisc;
	513	}
	514
	515	static unsigned long cbs_find(struct Qdisc *sch, u32 classid)
	516	{
	517	return 1;
	518	}
	519
	520	static void cbs_walk(struct Qdisc sch, struct qdisc_walker walker)
	521	{
	522	if (!walker->stop) {
	523	if (walker->count >= walker->skip) {
	524	if (walker->fn(sch, 1, walker) < 0) {
	525	walker->stop = 1;
	526	return;
	527	}
	528	}
	529	walker->count++;
	530	}
	531	}
	532
	533	static const struct Qdisc_class_ops cbs_class_ops = {
	534	.graft = cbs_graft,
	535	.leaf = cbs_leaf,
	536	.find = cbs_find,
	537	.walk = cbs_walk,
	538	.dump = cbs_dump_class,
	539	};
	540
585d763a VCG	541	static struct Qdisc_ops cbs_qdisc_ops __read_mostly = {
585d763a VCG	542	.id = "cbs",
990e35ec	543	.cl_ops = &cbs_class_ops,
585d763a VCG	544	.priv_size = sizeof(struct cbs_sched_data),
	545	.enqueue = cbs_enqueue,
	546	.dequeue = cbs_dequeue,
	547	.peek = qdisc_peek_dequeued,
	548	.init = cbs_init,
	549	.reset = qdisc_reset_queue,
	550	.destroy = cbs_destroy,
	551	.change = cbs_change,
	552	.dump = cbs_dump,
	553	.owner = THIS_MODULE,
	554	};
	555
e0a7683d LD	556	static struct notifier_block cbs_device_notifier = {
	557	.notifier_call = cbs_dev_notifier,
	558	};
	559
585d763a VCG	560	static int __init cbs_module_init(void)
585d763a VCG	561	{
45d5cb13	562	int err;
e0a7683d	563
45d5cb13	564	err = register_netdevice_notifier(&cbs_device_notifier);
e0a7683d LD	565	if (err)
	566	return err;
	567
45d5cb13 Y	568	err = register_qdisc(&cbs_qdisc_ops);
	569	if (err)
	570	unregister_netdevice_notifier(&cbs_device_notifier);
	571
	572	return err;
585d763a VCG	573	}
	574
	575	static void __exit cbs_module_exit(void)
	576	{
	577	unregister_qdisc(&cbs_qdisc_ops);
e0a7683d	578	unregister_netdevice_notifier(&cbs_device_notifier);
585d763a VCG	579	}
	580	module_init(cbs_module_init)
	581	module_exit(cbs_module_exit)
	582	MODULE_LICENSE("GPL");