[linux-2.6-block.git] / net / sched / act_police.c

// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * net/sched/act_police.c	Input police filter
 *
 * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
 * 		J Hadi Salim (action changes)
 */

#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 <linux/rtnetlink.h>
#include <linux/init.h>
#include <linux/slab.h>
#include <net/act_api.h>
#include <net/netlink.h>
#include <net/pkt_cls.h>
#include <net/tc_act/tc_police.h>

/* Each policer is serialized by its individual spinlock */

static unsigned int police_net_id;
static struct tc_action_ops act_police_ops;

static int tcf_police_walker(struct net *net, struct sk_buff *skb,
				 struct netlink_callback *cb, int type,
				 const struct tc_action_ops *ops,
				 struct netlink_ext_ack *extack)
{
	struct tc_action_net *tn = net_generic(net, police_net_id);

	return tcf_generic_walker(tn, skb, cb, type, ops, extack);
}

static const struct nla_policy police_policy[TCA_POLICE_MAX + 1] = {
	[TCA_POLICE_RATE]	= { .len = TC_RTAB_SIZE },
	[TCA_POLICE_PEAKRATE]	= { .len = TC_RTAB_SIZE },
	[TCA_POLICE_AVRATE]	= { .type = NLA_U32 },
	[TCA_POLICE_RESULT]	= { .type = NLA_U32 },
	[TCA_POLICE_RATE64]     = { .type = NLA_U64 },
	[TCA_POLICE_PEAKRATE64] = { .type = NLA_U64 },
	[TCA_POLICE_PKTRATE64]  = { .type = NLA_U64, .min = 1 },
	[TCA_POLICE_PKTBURST64] = { .type = NLA_U64, .min = 1 },
};

static int tcf_police_init(struct net *net, struct nlattr *nla,
			       struct nlattr *est, struct tc_action **a,
			       int ovr, int bind, bool rtnl_held,
			       struct tcf_proto *tp, u32 flags,
			       struct netlink_ext_ack *extack)
{
	int ret = 0, tcfp_result = TC_ACT_OK, err, size;
	struct nlattr *tb[TCA_POLICE_MAX + 1];
	struct tcf_chain *goto_ch = NULL;
	struct tc_police *parm;
	struct tcf_police *police;
	struct qdisc_rate_table *R_tab = NULL, *P_tab = NULL;
	struct tc_action_net *tn = net_generic(net, police_net_id);
	struct tcf_police_params *new;
	bool exists = false;
	u32 index;
	u64 rate64, prate64;
	u64 pps, ppsburst;

	if (nla == NULL)
		return -EINVAL;

	err = nla_parse_nested_deprecated(tb, TCA_POLICE_MAX, nla,
					  police_policy, NULL);
	if (err < 0)
		return err;

	if (tb[TCA_POLICE_TBF] == NULL)
		return -EINVAL;
	size = nla_len(tb[TCA_POLICE_TBF]);
	if (size != sizeof(*parm) && size != sizeof(struct tc_police_compat))
		return -EINVAL;

	parm = nla_data(tb[TCA_POLICE_TBF]);
	index = parm->index;
	err = tcf_idr_check_alloc(tn, &index, a, bind);
	if (err < 0)
		return err;
	exists = err;
	if (exists && bind)
		return 0;

	if (!exists) {
		ret = tcf_idr_create(tn, index, NULL, a,
				     &act_police_ops, bind, true, 0);
		if (ret) {
			tcf_idr_cleanup(tn, index);
			return ret;
		}
		ret = ACT_P_CREATED;
		spin_lock_init(&(to_police(*a)->tcfp_lock));
	} else if (!ovr) {
		tcf_idr_release(*a, bind);
		return -EEXIST;
	}
	err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
	if (err < 0)
		goto release_idr;

	police = to_police(*a);
	if (parm->rate.rate) {
		err = -ENOMEM;
		R_tab = qdisc_get_rtab(&parm->rate, tb[TCA_POLICE_RATE], NULL);
		if (R_tab == NULL)
			goto failure;

		if (parm->peakrate.rate) {
			P_tab = qdisc_get_rtab(&parm->peakrate,
					       tb[TCA_POLICE_PEAKRATE], NULL);
			if (P_tab == NULL)
				goto failure;
		}
	}

	if (est) {
		err = gen_replace_estimator(&police->tcf_bstats,
					    police->common.cpu_bstats,
					    &police->tcf_rate_est,
					    &police->tcf_lock,
					    NULL, est);
		if (err)
			goto failure;
	} else if (tb[TCA_POLICE_AVRATE] &&
		   (ret == ACT_P_CREATED ||
		    !gen_estimator_active(&police->tcf_rate_est))) {
		err = -EINVAL;
		goto failure;
	}

	if (tb[TCA_POLICE_RESULT]) {
		tcfp_result = nla_get_u32(tb[TCA_POLICE_RESULT]);
		if (TC_ACT_EXT_CMP(tcfp_result, TC_ACT_GOTO_CHAIN)) {
			NL_SET_ERR_MSG(extack,
				       "goto chain not allowed on fallback");
			err = -EINVAL;
			goto failure;
		}
	}

	if ((tb[TCA_POLICE_PKTRATE64] && !tb[TCA_POLICE_PKTBURST64]) ||
	    (!tb[TCA_POLICE_PKTRATE64] && tb[TCA_POLICE_PKTBURST64])) {
		NL_SET_ERR_MSG(extack,
			       "Both or neither packet-per-second burst and rate must be provided");
		err = -EINVAL;
		goto failure;
	}

	if (tb[TCA_POLICE_PKTRATE64] && R_tab) {
		NL_SET_ERR_MSG(extack,
			       "packet-per-second and byte-per-second rate limits not allowed in same action");
		err = -EINVAL;
		goto failure;
	}

	new = kzalloc(sizeof(*new), GFP_KERNEL);
	if (unlikely(!new)) {
		err = -ENOMEM;
		goto failure;
	}

	/* No failure allowed after this point */
	new->tcfp_result = tcfp_result;
	new->tcfp_mtu = parm->mtu;
	if (!new->tcfp_mtu) {
		new->tcfp_mtu = ~0;
		if (R_tab)
			new->tcfp_mtu = 255 << R_tab->rate.cell_log;
	}
	if (R_tab) {
		new->rate_present = true;
		rate64 = tb[TCA_POLICE_RATE64] ?
			 nla_get_u64(tb[TCA_POLICE_RATE64]) : 0;
		psched_ratecfg_precompute(&new->rate, &R_tab->rate, rate64);
		qdisc_put_rtab(R_tab);
	} else {
		new->rate_present = false;
	}
	if (P_tab) {
		new->peak_present = true;
		prate64 = tb[TCA_POLICE_PEAKRATE64] ?
			  nla_get_u64(tb[TCA_POLICE_PEAKRATE64]) : 0;
		psched_ratecfg_precompute(&new->peak, &P_tab->rate, prate64);
		qdisc_put_rtab(P_tab);
	} else {
		new->peak_present = false;
	}

	new->tcfp_burst = PSCHED_TICKS2NS(parm->burst);
	if (new->peak_present)
		new->tcfp_mtu_ptoks = (s64)psched_l2t_ns(&new->peak,
							 new->tcfp_mtu);

	if (tb[TCA_POLICE_AVRATE])
		new->tcfp_ewma_rate = nla_get_u32(tb[TCA_POLICE_AVRATE]);

	if (tb[TCA_POLICE_PKTRATE64]) {
		pps = nla_get_u64(tb[TCA_POLICE_PKTRATE64]);
		ppsburst = nla_get_u64(tb[TCA_POLICE_PKTBURST64]);
		new->pps_present = true;
		new->tcfp_pkt_burst = PSCHED_TICKS2NS(ppsburst);
		psched_ppscfg_precompute(&new->ppsrate, pps);
	}

	spin_lock_bh(&police->tcf_lock);
	spin_lock_bh(&police->tcfp_lock);
	police->tcfp_t_c = ktime_get_ns();
	police->tcfp_toks = new->tcfp_burst;
	if (new->peak_present)
		police->tcfp_ptoks = new->tcfp_mtu_ptoks;
	spin_unlock_bh(&police->tcfp_lock);
	goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
	new = rcu_replace_pointer(police->params,
				  new,
				  lockdep_is_held(&police->tcf_lock));
	spin_unlock_bh(&police->tcf_lock);

	if (goto_ch)
		tcf_chain_put_by_act(goto_ch);
	if (new)
		kfree_rcu(new, rcu);

	return ret;

failure:
	qdisc_put_rtab(P_tab);
	qdisc_put_rtab(R_tab);
	if (goto_ch)
		tcf_chain_put_by_act(goto_ch);
release_idr:
	tcf_idr_release(*a, bind);
	return err;
}

static int tcf_police_act(struct sk_buff *skb, const struct tc_action *a,
			  struct tcf_result *res)
{
	struct tcf_police *police = to_police(a);
	s64 now, toks, ppstoks = 0, ptoks = 0;
	struct tcf_police_params *p;
	int ret;

	tcf_lastuse_update(&police->tcf_tm);
	bstats_cpu_update(this_cpu_ptr(police->common.cpu_bstats), skb);

	ret = READ_ONCE(police->tcf_action);
	p = rcu_dereference_bh(police->params);

	if (p->tcfp_ewma_rate) {
		struct gnet_stats_rate_est64 sample;

		if (!gen_estimator_read(&police->tcf_rate_est, &sample) ||
		    sample.bps >= p->tcfp_ewma_rate)
			goto inc_overlimits;
	}

	if (qdisc_pkt_len(skb) <= p->tcfp_mtu) {
		if (!p->rate_present && !p->pps_present) {
			ret = p->tcfp_result;
			goto end;
		}

		now = ktime_get_ns();
		spin_lock_bh(&police->tcfp_lock);
		toks = min_t(s64, now - police->tcfp_t_c, p->tcfp_burst);
		if (p->peak_present) {
			ptoks = toks + police->tcfp_ptoks;
			if (ptoks > p->tcfp_mtu_ptoks)
				ptoks = p->tcfp_mtu_ptoks;
			ptoks -= (s64)psched_l2t_ns(&p->peak,
						    qdisc_pkt_len(skb));
		}
		if (p->rate_present) {
			toks += police->tcfp_toks;
			if (toks > p->tcfp_burst)
				toks = p->tcfp_burst;
			toks -= (s64)psched_l2t_ns(&p->rate, qdisc_pkt_len(skb));
		} else if (p->pps_present) {
			ppstoks = min_t(s64, now - police->tcfp_t_c, p->tcfp_pkt_burst);
			ppstoks += police->tcfp_pkttoks;
			if (ppstoks > p->tcfp_pkt_burst)
				ppstoks = p->tcfp_pkt_burst;
			ppstoks -= (s64)psched_pkt2t_ns(&p->ppsrate, 1);
		}
		if ((toks | ptoks | ppstoks) >= 0) {
			police->tcfp_t_c = now;
			police->tcfp_toks = toks;
			police->tcfp_ptoks = ptoks;
			police->tcfp_pkttoks = ppstoks;
			spin_unlock_bh(&police->tcfp_lock);
			ret = p->tcfp_result;
			goto inc_drops;
		}
		spin_unlock_bh(&police->tcfp_lock);
	}

inc_overlimits:
	qstats_overlimit_inc(this_cpu_ptr(police->common.cpu_qstats));
inc_drops:
	if (ret == TC_ACT_SHOT)
		qstats_drop_inc(this_cpu_ptr(police->common.cpu_qstats));
end:
	return ret;
}

static void tcf_police_cleanup(struct tc_action *a)
{
	struct tcf_police *police = to_police(a);
	struct tcf_police_params *p;

	p = rcu_dereference_protected(police->params, 1);
	if (p)
		kfree_rcu(p, rcu);
}

static void tcf_police_stats_update(struct tc_action *a,
				    u64 bytes, u64 packets, u64 drops,
				    u64 lastuse, bool hw)
{
	struct tcf_police *police = to_police(a);
	struct tcf_t *tm = &police->tcf_tm;

	tcf_action_update_stats(a, bytes, packets, drops, hw);
	tm->lastuse = max_t(u64, tm->lastuse, lastuse);
}

static int tcf_police_dump(struct sk_buff *skb, struct tc_action *a,
			       int bind, int ref)
{
	unsigned char *b = skb_tail_pointer(skb);
	struct tcf_police *police = to_police(a);
	struct tcf_police_params *p;
	struct tc_police opt = {
		.index = police->tcf_index,
		.refcnt = refcount_read(&police->tcf_refcnt) - ref,
		.bindcnt = atomic_read(&police->tcf_bindcnt) - bind,
	};
	struct tcf_t t;

	spin_lock_bh(&police->tcf_lock);
	opt.action = police->tcf_action;
	p = rcu_dereference_protected(police->params,
				      lockdep_is_held(&police->tcf_lock));
	opt.mtu = p->tcfp_mtu;
	opt.burst = PSCHED_NS2TICKS(p->tcfp_burst);
	if (p->rate_present) {
		psched_ratecfg_getrate(&opt.rate, &p->rate);
		if ((police->params->rate.rate_bytes_ps >= (1ULL << 32)) &&
		    nla_put_u64_64bit(skb, TCA_POLICE_RATE64,
				      police->params->rate.rate_bytes_ps,
				      TCA_POLICE_PAD))
			goto nla_put_failure;
	}
	if (p->peak_present) {
		psched_ratecfg_getrate(&opt.peakrate, &p->peak);
		if ((police->params->peak.rate_bytes_ps >= (1ULL << 32)) &&
		    nla_put_u64_64bit(skb, TCA_POLICE_PEAKRATE64,
				      police->params->peak.rate_bytes_ps,
				      TCA_POLICE_PAD))
			goto nla_put_failure;
	}
	if (p->pps_present) {
		if (nla_put_u64_64bit(skb, TCA_POLICE_PKTRATE64,
				      police->params->ppsrate.rate_pkts_ps,
				      TCA_POLICE_PAD))
			goto nla_put_failure;
		if (nla_put_u64_64bit(skb, TCA_POLICE_PKTBURST64,
				      PSCHED_NS2TICKS(p->tcfp_pkt_burst),
				      TCA_POLICE_PAD))
			goto nla_put_failure;
	}
	if (nla_put(skb, TCA_POLICE_TBF, sizeof(opt), &opt))
		goto nla_put_failure;
	if (p->tcfp_result &&
	    nla_put_u32(skb, TCA_POLICE_RESULT, p->tcfp_result))
		goto nla_put_failure;
	if (p->tcfp_ewma_rate &&
	    nla_put_u32(skb, TCA_POLICE_AVRATE, p->tcfp_ewma_rate))
		goto nla_put_failure;

	tcf_tm_dump(&t, &police->tcf_tm);
	if (nla_put_64bit(skb, TCA_POLICE_TM, sizeof(t), &t, TCA_POLICE_PAD))
		goto nla_put_failure;
	spin_unlock_bh(&police->tcf_lock);

	return skb->len;

nla_put_failure:
	spin_unlock_bh(&police->tcf_lock);
	nlmsg_trim(skb, b);
	return -1;
}

static int tcf_police_search(struct net *net, struct tc_action **a, u32 index)
{
	struct tc_action_net *tn = net_generic(net, police_net_id);

	return tcf_idr_search(tn, a, index);
}

MODULE_AUTHOR("Alexey Kuznetsov");
MODULE_DESCRIPTION("Policing actions");
MODULE_LICENSE("GPL");

static struct tc_action_ops act_police_ops = {
	.kind		=	"police",
	.id		=	TCA_ID_POLICE,
	.owner		=	THIS_MODULE,
	.stats_update	=	tcf_police_stats_update,
	.act		=	tcf_police_act,
	.dump		=	tcf_police_dump,
	.init		=	tcf_police_init,
	.walk		=	tcf_police_walker,
	.lookup		=	tcf_police_search,
	.cleanup	=	tcf_police_cleanup,
	.size		=	sizeof(struct tcf_police),
};

static __net_init int police_init_net(struct net *net)
{
	struct tc_action_net *tn = net_generic(net, police_net_id);

	return tc_action_net_init(net, tn, &act_police_ops);
}

static void __net_exit police_exit_net(struct list_head *net_list)
{
	tc_action_net_exit(net_list, police_net_id);
}

static struct pernet_operations police_net_ops = {
	.init = police_init_net,
	.exit_batch = police_exit_net,
	.id   = &police_net_id,
	.size = sizeof(struct tc_action_net),
};

static int __init police_init_module(void)
{
	return tcf_register_action(&act_police_ops, &police_net_ops);
}

static void __exit police_cleanup_module(void)
{
	tcf_unregister_action(&act_police_ops, &police_net_ops);
}

module_init(police_init_module);
module_exit(police_cleanup_module);
Commit	Line	Data
2874c5fd	1	// SPDX-License-Identifier: GPL-2.0-or-later
1da177e4	2	/*
0c6965dd	3	* net/sched/act_police.c Input police filter
1da177e4	4	*
1da177e4 LT	5	* Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
	6	* J Hadi Salim (action changes)
	7	*/
	8
1da177e4 LT	9	#include <linux/module.h>
	10	#include <linux/types.h>
	11	#include <linux/kernel.h>
1da177e4	12	#include <linux/string.h>
1da177e4	13	#include <linux/errno.h>
1da177e4	14	#include <linux/skbuff.h>
1da177e4 LT	15	#include <linux/rtnetlink.h>
1da177e4 LT	16	#include <linux/init.h>
5a0e3ad6	17	#include <linux/slab.h>
1da177e4	18	#include <net/act_api.h>
dc5fc579	19	#include <net/netlink.h>
d6124d6b	20	#include <net/pkt_cls.h>
fa762da9	21	#include <net/tc_act/tc_police.h>
1e9b3d53	22
e9ce1cd3	23	/* Each policer is serialized by its individual spinlock */
1da177e4	24
c7d03a00	25	static unsigned int police_net_id;
a85a970a	26	static struct tc_action_ops act_police_ops;
ddf97ccd	27
2ac06347	28	static int tcf_police_walker(struct net net, struct sk_buff skb,
ddf97ccd	29	struct netlink_callback *cb, int type,
41780105 AA	30	const struct tc_action_ops *ops,
41780105 AA	31	struct netlink_ext_ack *extack)
1da177e4	32	{
ddf97ccd	33	struct tc_action_net *tn = net_generic(net, police_net_id);
1da177e4	34
b3620145	35	return tcf_generic_walker(tn, skb, cb, type, ops, extack);
1da177e4	36	}
1da177e4	37
53b2bf3f PM	38	static const struct nla_policy police_policy[TCA_POLICE_MAX + 1] = {
	39	[TCA_POLICE_RATE] = { .len = TC_RTAB_SIZE },
	40	[TCA_POLICE_PEAKRATE] = { .len = TC_RTAB_SIZE },
	41	[TCA_POLICE_AVRATE] = { .type = NLA_U32 },
	42	[TCA_POLICE_RESULT] = { .type = NLA_U32 },
d1967e49 DD	43	[TCA_POLICE_RATE64] = { .type = NLA_U64 },
d1967e49 DD	44	[TCA_POLICE_PEAKRATE64] = { .type = NLA_U64 },
2ffe0395 BZ	45	[TCA_POLICE_PKTRATE64] = { .type = NLA_U64, .min = 1 },
2ffe0395 BZ	46	[TCA_POLICE_PKTBURST64] = { .type = NLA_U64, .min = 1 },
53b2bf3f PM	47	};
53b2bf3f PM	48
2ac06347	49	static int tcf_police_init(struct net net, struct nlattr nla,
a85a970a	50	struct nlattr est, struct tc_action *a,
789871bb	51	int ovr, int bind, bool rtnl_held,
abbb0d33	52	struct tcf_proto *tp, u32 flags,
589dad6d	53	struct netlink_ext_ack *extack)
1da177e4	54	{
fd6d4338	55	int ret = 0, tcfp_result = TC_ACT_OK, err, size;
7ba699c6	56	struct nlattr *tb[TCA_POLICE_MAX + 1];
d6124d6b	57	struct tcf_chain *goto_ch = NULL;
1da177e4	58	struct tc_police *parm;
e9ce1cd3	59	struct tcf_police *police;
1da177e4	60	struct qdisc_rate_table R_tab = NULL, P_tab = NULL;
ddf97ccd	61	struct tc_action_net *tn = net_generic(net, police_net_id);
2d550dba	62	struct tcf_police_params *new;
0852e455	63	bool exists = false;
7be8ef2c	64	u32 index;
d1967e49	65	u64 rate64, prate64;
2ffe0395	66	u64 pps, ppsburst;
1da177e4	67
cee63723	68	if (nla == NULL)
1da177e4 LT	69	return -EINVAL;
1da177e4 LT	70
8cb08174 JB	71	err = nla_parse_nested_deprecated(tb, TCA_POLICE_MAX, nla,
8cb08174 JB	72	police_policy, NULL);
cee63723 PM	73	if (err < 0)
	74	return err;
	75
7ba699c6	76	if (tb[TCA_POLICE_TBF] == NULL)
1e9b3d53	77	return -EINVAL;
7ba699c6	78	size = nla_len(tb[TCA_POLICE_TBF]);
1e9b3d53	79	if (size != sizeof(*parm) && size != sizeof(struct tc_police_compat))
1da177e4	80	return -EINVAL;
0852e455	81
7ba699c6	82	parm = nla_data(tb[TCA_POLICE_TBF]);
7be8ef2c DL	83	index = parm->index;
7be8ef2c DL	84	err = tcf_idr_check_alloc(tn, &index, a, bind);
0190c1d4 VB	85	if (err < 0)
	86	return err;
	87	exists = err;
0852e455 WC	88	if (exists && bind)
0852e455 WC	89	return 0;
1da177e4	90
0852e455	91	if (!exists) {
7be8ef2c	92	ret = tcf_idr_create(tn, index, NULL, a,
e3822678	93	&act_police_ops, bind, true, 0);
0190c1d4	94	if (ret) {
7be8ef2c	95	tcf_idr_cleanup(tn, index);
a03e6fe5	96	return ret;
0190c1d4	97	}
a03e6fe5	98	ret = ACT_P_CREATED;
484afd1b	99	spin_lock_init(&(to_police(*a)->tcfp_lock));
4e8ddd7f	100	} else if (!ovr) {
65a206c0	101	tcf_idr_release(*a, bind);
4e8ddd7f	102	return -EEXIST;
1da177e4	103	}
d6124d6b DC	104	err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
	105	if (err < 0)
	106	goto release_idr;
1da177e4	107
a85a970a	108	police = to_police(*a);
1da177e4 LT	109	if (parm->rate.rate) {
1da177e4 LT	110	err = -ENOMEM;
e9bc3fa2	111	R_tab = qdisc_get_rtab(&parm->rate, tb[TCA_POLICE_RATE], NULL);
1da177e4 LT	112	if (R_tab == NULL)
1da177e4 LT	113	goto failure;
c1b56878	114
1da177e4 LT	115	if (parm->peakrate.rate) {
1da177e4 LT	116	P_tab = qdisc_get_rtab(&parm->peakrate,
e9bc3fa2	117	tb[TCA_POLICE_PEAKRATE], NULL);
71bcb09a	118	if (P_tab == NULL)
1da177e4	119	goto failure;
1da177e4 LT	120	}
1da177e4 LT	121	}
71bcb09a	122
71bcb09a	123	if (est) {
93be42f9 DC	124	err = gen_replace_estimator(&police->tcf_bstats,
93be42f9 DC	125	police->common.cpu_bstats,
71bcb09a	126	&police->tcf_rate_est,
edb09eb1 ED	127	&police->tcf_lock,
edb09eb1 ED	128	NULL, est);
71bcb09a	129	if (err)
74030603	130	goto failure;
a883bf56 JP	131	} else if (tb[TCA_POLICE_AVRATE] &&
a883bf56 JP	132	(ret == ACT_P_CREATED \|\|
1c0d32fd	133	!gen_estimator_active(&police->tcf_rate_est))) {
a883bf56	134	err = -EINVAL;
74030603	135	goto failure;
71bcb09a SH	136	}
71bcb09a SH	137
fd6d4338 DC	138	if (tb[TCA_POLICE_RESULT]) {
	139	tcfp_result = nla_get_u32(tb[TCA_POLICE_RESULT]);
	140	if (TC_ACT_EXT_CMP(tcfp_result, TC_ACT_GOTO_CHAIN)) {
	141	NL_SET_ERR_MSG(extack,
	142	"goto chain not allowed on fallback");
	143	err = -EINVAL;
	144	goto failure;
	145	}
	146	}
	147
2ffe0395 BZ	148	if ((tb[TCA_POLICE_PKTRATE64] && !tb[TCA_POLICE_PKTBURST64]) \|\|
	149	(!tb[TCA_POLICE_PKTRATE64] && tb[TCA_POLICE_PKTBURST64])) {
	150	NL_SET_ERR_MSG(extack,
	151	"Both or neither packet-per-second burst and rate must be provided");
	152	err = -EINVAL;
	153	goto failure;
	154	}
	155
	156	if (tb[TCA_POLICE_PKTRATE64] && R_tab) {
	157	NL_SET_ERR_MSG(extack,
	158	"packet-per-second and byte-per-second rate limits not allowed in same action");
	159	err = -EINVAL;
	160	goto failure;
	161	}
	162
2d550dba DC	163	new = kzalloc(sizeof(*new), GFP_KERNEL);
	164	if (unlikely(!new)) {
	165	err = -ENOMEM;
	166	goto failure;
	167	}
	168
71bcb09a	169	/* No failure allowed after this point */
fd6d4338	170	new->tcfp_result = tcfp_result;
2d550dba DC	171	new->tcfp_mtu = parm->mtu;
	172	if (!new->tcfp_mtu) {
	173	new->tcfp_mtu = ~0;
c6d14ff1	174	if (R_tab)
2d550dba	175	new->tcfp_mtu = 255 << R_tab->rate.cell_log;
c6d14ff1 JP	176	}
c6d14ff1 JP	177	if (R_tab) {
2d550dba	178	new->rate_present = true;
d1967e49 DD	179	rate64 = tb[TCA_POLICE_RATE64] ?
	180	nla_get_u64(tb[TCA_POLICE_RATE64]) : 0;
	181	psched_ratecfg_precompute(&new->rate, &R_tab->rate, rate64);
c6d14ff1 JP	182	qdisc_put_rtab(R_tab);
c6d14ff1 JP	183	} else {
2d550dba	184	new->rate_present = false;
1da177e4	185	}
c6d14ff1	186	if (P_tab) {
2d550dba	187	new->peak_present = true;
d1967e49 DD	188	prate64 = tb[TCA_POLICE_PEAKRATE64] ?
	189	nla_get_u64(tb[TCA_POLICE_PEAKRATE64]) : 0;
	190	psched_ratecfg_precompute(&new->peak, &P_tab->rate, prate64);
c6d14ff1 JP	191	qdisc_put_rtab(P_tab);
c6d14ff1 JP	192	} else {
2d550dba	193	new->peak_present = false;
1da177e4 LT	194	}
1da177e4 LT	195
2d550dba	196	new->tcfp_burst = PSCHED_TICKS2NS(parm->burst);
f2cbd485	197	if (new->peak_present)
2d550dba DC	198	new->tcfp_mtu_ptoks = (s64)psched_l2t_ns(&new->peak,
2d550dba DC	199	new->tcfp_mtu);
1da177e4	200
7ba699c6	201	if (tb[TCA_POLICE_AVRATE])
2d550dba	202	new->tcfp_ewma_rate = nla_get_u32(tb[TCA_POLICE_AVRATE]);
1da177e4	203
2ffe0395 BZ	204	if (tb[TCA_POLICE_PKTRATE64]) {
	205	pps = nla_get_u64(tb[TCA_POLICE_PKTRATE64]);
	206	ppsburst = nla_get_u64(tb[TCA_POLICE_PKTBURST64]);
	207	new->pps_present = true;
	208	new->tcfp_pkt_burst = PSCHED_TICKS2NS(ppsburst);
	209	psched_ppscfg_precompute(&new->ppsrate, pps);
	210	}
	211
2d550dba	212	spin_lock_bh(&police->tcf_lock);
f2cbd485 DC	213	spin_lock_bh(&police->tcfp_lock);
	214	police->tcfp_t_c = ktime_get_ns();
	215	police->tcfp_toks = new->tcfp_burst;
	216	if (new->peak_present)
	217	police->tcfp_ptoks = new->tcfp_mtu_ptoks;
	218	spin_unlock_bh(&police->tcfp_lock);
d6124d6b	219	goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
445d3749 PM	220	new = rcu_replace_pointer(police->params,
	221	new,
	222	lockdep_is_held(&police->tcf_lock));
e9ce1cd3	223	spin_unlock_bh(&police->tcf_lock);
1da177e4	224
d6124d6b DC	225	if (goto_ch)
d6124d6b DC	226	tcf_chain_put_by_act(goto_ch);
2d550dba DC	227	if (new)
2d550dba DC	228	kfree_rcu(new, rcu);
1da177e4	229
1da177e4 LT	230	return ret;
	231
	232	failure:
3b69a4c9 YY	233	qdisc_put_rtab(P_tab);
3b69a4c9 YY	234	qdisc_put_rtab(R_tab);
d6124d6b DC	235	if (goto_ch)
	236	tcf_chain_put_by_act(goto_ch);
	237	release_idr:
4e8ddd7f	238	tcf_idr_release(*a, bind);
1da177e4 LT	239	return err;
	240	}
	241
2ac06347	242	static int tcf_police_act(struct sk_buff skb, const struct tc_action a,
10297b99	243	struct tcf_result *res)
1da177e4	244	{
a85a970a	245	struct tcf_police *police = to_police(a);
2ffe0395	246	s64 now, toks, ppstoks = 0, ptoks = 0;
2d550dba	247	struct tcf_police_params *p;
93be42f9	248	int ret;
1da177e4	249
3d3ed181	250	tcf_lastuse_update(&police->tcf_tm);
93be42f9	251	bstats_cpu_update(this_cpu_ptr(police->common.cpu_bstats), skb);
1da177e4	252
2d550dba DC	253	ret = READ_ONCE(police->tcf_action);
	254	p = rcu_dereference_bh(police->params);
	255
	256	if (p->tcfp_ewma_rate) {
1c0d32fd ED	257	struct gnet_stats_rate_est64 sample;
	258
	259	if (!gen_estimator_read(&police->tcf_rate_est, &sample) \|\|
2d550dba	260	sample.bps >= p->tcfp_ewma_rate)
93be42f9	261	goto inc_overlimits;
1da177e4	262	}
1da177e4	263
2d550dba	264	if (qdisc_pkt_len(skb) <= p->tcfp_mtu) {
2ffe0395	265	if (!p->rate_present && !p->pps_present) {
2d550dba DC	266	ret = p->tcfp_result;
2d550dba DC	267	goto end;
1da177e4 LT	268	}
1da177e4 LT	269
d2de875c	270	now = ktime_get_ns();
f2cbd485 DC	271	spin_lock_bh(&police->tcfp_lock);
f2cbd485 DC	272	toks = min_t(s64, now - police->tcfp_t_c, p->tcfp_burst);
2d550dba	273	if (p->peak_present) {
f2cbd485	274	ptoks = toks + police->tcfp_ptoks;
2d550dba DC	275	if (ptoks > p->tcfp_mtu_ptoks)
	276	ptoks = p->tcfp_mtu_ptoks;
	277	ptoks -= (s64)psched_l2t_ns(&p->peak,
	278	qdisc_pkt_len(skb));
1da177e4	279	}
2ffe0395 BZ	280	if (p->rate_present) {
	281	toks += police->tcfp_toks;
	282	if (toks > p->tcfp_burst)
	283	toks = p->tcfp_burst;
	284	toks -= (s64)psched_l2t_ns(&p->rate, qdisc_pkt_len(skb));
	285	} else if (p->pps_present) {
	286	ppstoks = min_t(s64, now - police->tcfp_t_c, p->tcfp_pkt_burst);
	287	ppstoks += police->tcfp_pkttoks;
	288	if (ppstoks > p->tcfp_pkt_burst)
	289	ppstoks = p->tcfp_pkt_burst;
	290	ppstoks -= (s64)psched_pkt2t_ns(&p->ppsrate, 1);
	291	}
	292	if ((toks \| ptoks \| ppstoks) >= 0) {
f2cbd485 DC	293	police->tcfp_t_c = now;
	294	police->tcfp_toks = toks;
	295	police->tcfp_ptoks = ptoks;
2ffe0395	296	police->tcfp_pkttoks = ppstoks;
f2cbd485	297	spin_unlock_bh(&police->tcfp_lock);
2d550dba	298	ret = p->tcfp_result;
93be42f9	299	goto inc_drops;
1da177e4	300	}
f2cbd485	301	spin_unlock_bh(&police->tcfp_lock);
1da177e4	302	}
93be42f9 DC	303
	304	inc_overlimits:
	305	qstats_overlimit_inc(this_cpu_ptr(police->common.cpu_qstats));
	306	inc_drops:
	307	if (ret == TC_ACT_SHOT)
	308	qstats_drop_inc(this_cpu_ptr(police->common.cpu_qstats));
2d550dba	309	end:
93be42f9	310	return ret;
1da177e4 LT	311	}
1da177e4 LT	312
2d550dba DC	313	static void tcf_police_cleanup(struct tc_action *a)
	314	{
	315	struct tcf_police *police = to_police(a);
	316	struct tcf_police_params *p;
	317
	318	p = rcu_dereference_protected(police->params, 1);
	319	if (p)
	320	kfree_rcu(p, rcu);
	321	}
	322
12f02b6b	323	static void tcf_police_stats_update(struct tc_action *a,
4b61d3e8	324	u64 bytes, u64 packets, u64 drops,
12f02b6b PJV	325	u64 lastuse, bool hw)
	326	{
	327	struct tcf_police *police = to_police(a);
	328	struct tcf_t *tm = &police->tcf_tm;
	329
4b61d3e8	330	tcf_action_update_stats(a, bytes, packets, drops, hw);
12f02b6b PJV	331	tm->lastuse = max_t(u64, tm->lastuse, lastuse);
	332	}
	333
2ac06347	334	static int tcf_police_dump(struct sk_buff skb, struct tc_action a,
5a7a5555	335	int bind, int ref)
1da177e4	336	{
27a884dc	337	unsigned char *b = skb_tail_pointer(skb);
a85a970a	338	struct tcf_police *police = to_police(a);
2d550dba	339	struct tcf_police_params *p;
0f04cfd0 JM	340	struct tc_police opt = {
0f04cfd0 JM	341	.index = police->tcf_index,
036bb443 VB	342	.refcnt = refcount_read(&police->tcf_refcnt) - ref,
036bb443 VB	343	.bindcnt = atomic_read(&police->tcf_bindcnt) - bind,
0f04cfd0	344	};
3d3ed181	345	struct tcf_t t;
0f04cfd0	346
e329bc42 VB	347	spin_lock_bh(&police->tcf_lock);
e329bc42 VB	348	opt.action = police->tcf_action;
2d550dba DC	349	p = rcu_dereference_protected(police->params,
	350	lockdep_is_held(&police->tcf_lock));
	351	opt.mtu = p->tcfp_mtu;
	352	opt.burst = PSCHED_NS2TICKS(p->tcfp_burst);
d1967e49	353	if (p->rate_present) {
2d550dba	354	psched_ratecfg_getrate(&opt.rate, &p->rate);
d1967e49 DD	355	if ((police->params->rate.rate_bytes_ps >= (1ULL << 32)) &&
	356	nla_put_u64_64bit(skb, TCA_POLICE_RATE64,
	357	police->params->rate.rate_bytes_ps,
	358	TCA_POLICE_PAD))
	359	goto nla_put_failure;
	360	}
	361	if (p->peak_present) {
2d550dba	362	psched_ratecfg_getrate(&opt.peakrate, &p->peak);
d1967e49 DD	363	if ((police->params->peak.rate_bytes_ps >= (1ULL << 32)) &&
	364	nla_put_u64_64bit(skb, TCA_POLICE_PEAKRATE64,
	365	police->params->peak.rate_bytes_ps,
	366	TCA_POLICE_PAD))
	367	goto nla_put_failure;
	368	}
2ffe0395 BZ	369	if (p->pps_present) {
	370	if (nla_put_u64_64bit(skb, TCA_POLICE_PKTRATE64,
	371	police->params->ppsrate.rate_pkts_ps,
	372	TCA_POLICE_PAD))
	373	goto nla_put_failure;
	374	if (nla_put_u64_64bit(skb, TCA_POLICE_PKTBURST64,
	375	PSCHED_NS2TICKS(p->tcfp_pkt_burst),
	376	TCA_POLICE_PAD))
	377	goto nla_put_failure;
	378	}
1b34ec43 DM	379	if (nla_put(skb, TCA_POLICE_TBF, sizeof(opt), &opt))
1b34ec43 DM	380	goto nla_put_failure;
2d550dba DC	381	if (p->tcfp_result &&
2d550dba DC	382	nla_put_u32(skb, TCA_POLICE_RESULT, p->tcfp_result))
1b34ec43	383	goto nla_put_failure;
2d550dba DC	384	if (p->tcfp_ewma_rate &&
2d550dba DC	385	nla_put_u32(skb, TCA_POLICE_AVRATE, p->tcfp_ewma_rate))
1b34ec43	386	goto nla_put_failure;
3d3ed181	387
985fd98a	388	tcf_tm_dump(&t, &police->tcf_tm);
3d3ed181 JHS	389	if (nla_put_64bit(skb, TCA_POLICE_TM, sizeof(t), &t, TCA_POLICE_PAD))
3d3ed181 JHS	390	goto nla_put_failure;
e329bc42	391	spin_unlock_bh(&police->tcf_lock);
3d3ed181	392
1da177e4 LT	393	return skb->len;
1da177e4 LT	394
7ba699c6	395	nla_put_failure:
e329bc42	396	spin_unlock_bh(&police->tcf_lock);
dc5fc579	397	nlmsg_trim(skb, b);
1da177e4 LT	398	return -1;
	399	}
	400
f061b48c	401	static int tcf_police_search(struct net net, struct tc_action *a, u32 index)
ddf97ccd WC	402	{
	403	struct tc_action_net *tn = net_generic(net, police_net_id);
	404
65a206c0	405	return tcf_idr_search(tn, a, index);
ddf97ccd WC	406	}
ddf97ccd WC	407
1da177e4 LT	408	MODULE_AUTHOR("Alexey Kuznetsov");
	409	MODULE_DESCRIPTION("Policing actions");
	410	MODULE_LICENSE("GPL");
	411
	412	static struct tc_action_ops act_police_ops = {
	413	.kind = "police",
eddd2cf1	414	.id = TCA_ID_POLICE,
1da177e4	415	.owner = THIS_MODULE,
12f02b6b	416	.stats_update = tcf_police_stats_update,
2ac06347 JHS	417	.act = tcf_police_act,
	418	.dump = tcf_police_dump,
	419	.init = tcf_police_init,
	420	.walk = tcf_police_walker,
ddf97ccd	421	.lookup = tcf_police_search,
2d550dba	422	.cleanup = tcf_police_cleanup,
a85a970a	423	.size = sizeof(struct tcf_police),
ddf97ccd WC	424	};
	425
	426	static __net_init int police_init_net(struct net *net)
	427	{
	428	struct tc_action_net *tn = net_generic(net, police_net_id);
	429
981471bd	430	return tc_action_net_init(net, tn, &act_police_ops);
ddf97ccd WC	431	}
ddf97ccd WC	432
039af9c6	433	static void __net_exit police_exit_net(struct list_head *net_list)
ddf97ccd	434	{
039af9c6	435	tc_action_net_exit(net_list, police_net_id);
ddf97ccd WC	436	}
	437
	438	static struct pernet_operations police_net_ops = {
	439	.init = police_init_net,
039af9c6	440	.exit_batch = police_exit_net,
ddf97ccd WC	441	.id = &police_net_id,
ddf97ccd WC	442	.size = sizeof(struct tc_action_net),
1da177e4 LT	443	};
1da177e4 LT	444
5a7a5555	445	static int __init police_init_module(void)
1da177e4	446	{
ddf97ccd	447	return tcf_register_action(&act_police_ops, &police_net_ops);
1da177e4 LT	448	}
1da177e4 LT	449
5a7a5555	450	static void __exit police_cleanup_module(void)
1da177e4	451	{
ddf97ccd	452	tcf_unregister_action(&act_police_ops, &police_net_ops);
1da177e4 LT	453	}
	454
	455	module_init(police_init_module);
	456	module_exit(police_cleanup_module);