[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>
#include <net/tc_wrapper.h>

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

static struct tc_action_ops act_police_ops;

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,
			       struct tcf_proto *tp, u32 flags,
			       struct netlink_ext_ack *extack)
{
	int ret = 0, tcfp_result = TC_ACT_OK, err, size;
	bool bind = flags & TCA_ACT_FLAGS_BIND;
	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, act_police_ops.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, flags);
		if (ret) {
			tcf_idr_cleanup(tn, index);
			return ret;
		}
		ret = ACT_P_CREATED;
		spin_lock_init(&(to_police(*a)->tcfp_lock));
	} else if (!(flags & TCA_ACT_FLAGS_REPLACE)) {
		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,
					    false, 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 bool tcf_police_mtu_check(struct sk_buff *skb, u32 limit)
{
	u32 len;

	if (skb_is_gso(skb))
		return skb_gso_validate_mac_len(skb, limit);

	len = qdisc_pkt_len(skb);
	if (skb_at_tc_ingress(skb))
		len += skb->mac_len;

	return len <= limit;
}

TC_INDIRECT_SCOPE 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_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 (tcf_police_mtu_check(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 ((p->rate.rate_bytes_ps >= (1ULL << 32)) &&
		    nla_put_u64_64bit(skb, TCA_POLICE_RATE64,
				      p->rate.rate_bytes_ps,
				      TCA_POLICE_PAD))
			goto nla_put_failure;
	}
	if (p->peak_present) {
		psched_ratecfg_getrate(&opt.peakrate, &p->peak);
		if ((p->peak.rate_bytes_ps >= (1ULL << 32)) &&
		    nla_put_u64_64bit(skb, TCA_POLICE_PEAKRATE64,
				      p->peak.rate_bytes_ps,
				      TCA_POLICE_PAD))
			goto nla_put_failure;
	}
	if (p->pps_present) {
		if (nla_put_u64_64bit(skb, TCA_POLICE_PKTRATE64,
				      p->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_act_to_flow_act(int tc_act, u32 *extval,
				      struct netlink_ext_ack *extack)
{
	int act_id = -EOPNOTSUPP;

	if (!TC_ACT_EXT_OPCODE(tc_act)) {
		if (tc_act == TC_ACT_OK)
			act_id = FLOW_ACTION_ACCEPT;
		else if (tc_act ==  TC_ACT_SHOT)
			act_id = FLOW_ACTION_DROP;
		else if (tc_act == TC_ACT_PIPE)
			act_id = FLOW_ACTION_PIPE;
		else if (tc_act == TC_ACT_RECLASSIFY)
			NL_SET_ERR_MSG_MOD(extack, "Offload not supported when conform/exceed action is \"reclassify\"");
		else
			NL_SET_ERR_MSG_MOD(extack, "Unsupported conform/exceed action offload");
	} else if (TC_ACT_EXT_CMP(tc_act, TC_ACT_GOTO_CHAIN)) {
		act_id = FLOW_ACTION_GOTO;
		*extval = tc_act & TC_ACT_EXT_VAL_MASK;
	} else if (TC_ACT_EXT_CMP(tc_act, TC_ACT_JUMP)) {
		act_id = FLOW_ACTION_JUMP;
		*extval = tc_act & TC_ACT_EXT_VAL_MASK;
	} else if (tc_act == TC_ACT_UNSPEC) {
		act_id = FLOW_ACTION_CONTINUE;
	} else {
		NL_SET_ERR_MSG_MOD(extack, "Unsupported conform/exceed action offload");
	}

	return act_id;
}

static int tcf_police_offload_act_setup(struct tc_action *act, void *entry_data,
					u32 *index_inc, bool bind,
					struct netlink_ext_ack *extack)
{
	if (bind) {
		struct flow_action_entry *entry = entry_data;
		struct tcf_police *police = to_police(act);
		struct tcf_police_params *p;
		int act_id;

		p = rcu_dereference_protected(police->params,
					      lockdep_is_held(&police->tcf_lock));

		entry->id = FLOW_ACTION_POLICE;
		entry->police.burst = tcf_police_burst(act);
		entry->police.rate_bytes_ps =
			tcf_police_rate_bytes_ps(act);
		entry->police.peakrate_bytes_ps = tcf_police_peakrate_bytes_ps(act);
		entry->police.avrate = tcf_police_tcfp_ewma_rate(act);
		entry->police.overhead = tcf_police_rate_overhead(act);
		entry->police.burst_pkt = tcf_police_burst_pkt(act);
		entry->police.rate_pkt_ps =
			tcf_police_rate_pkt_ps(act);
		entry->police.mtu = tcf_police_tcfp_mtu(act);

		act_id = tcf_police_act_to_flow_act(police->tcf_action,
						    &entry->police.exceed.extval,
						    extack);
		if (act_id < 0)
			return act_id;

		entry->police.exceed.act_id = act_id;

		act_id = tcf_police_act_to_flow_act(p->tcfp_result,
						    &entry->police.notexceed.extval,
						    extack);
		if (act_id < 0)
			return act_id;

		entry->police.notexceed.act_id = act_id;

		*index_inc = 1;
	} else {
		struct flow_offload_action *fl_action = entry_data;

		fl_action->id = FLOW_ACTION_POLICE;
	}

	return 0;
}

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,
	.cleanup	=	tcf_police_cleanup,
	.offload_act_setup =	tcf_police_offload_act_setup,
	.size		=	sizeof(struct tcf_police),
};

static __net_init int police_init_net(struct net *net)
{
	struct tc_action_net *tn = net_generic(net, act_police_ops.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, act_police_ops.net_id);
}

static struct pernet_operations police_net_ops = {
	.init = police_init_net,
	.exit_batch = police_exit_net,
	.id   = &act_police_ops.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>
871cf386	22	#include <net/tc_wrapper.h>
1e9b3d53	23
e9ce1cd3	24	/* Each policer is serialized by its individual spinlock */
1da177e4	25
a85a970a	26	static struct tc_action_ops act_police_ops;
ddf97ccd	27
53b2bf3f PM	28	static const struct nla_policy police_policy[TCA_POLICE_MAX + 1] = {
	29	[TCA_POLICE_RATE] = { .len = TC_RTAB_SIZE },
	30	[TCA_POLICE_PEAKRATE] = { .len = TC_RTAB_SIZE },
	31	[TCA_POLICE_AVRATE] = { .type = NLA_U32 },
	32	[TCA_POLICE_RESULT] = { .type = NLA_U32 },
d1967e49 DD	33	[TCA_POLICE_RATE64] = { .type = NLA_U64 },
d1967e49 DD	34	[TCA_POLICE_PEAKRATE64] = { .type = NLA_U64 },
2ffe0395 BZ	35	[TCA_POLICE_PKTRATE64] = { .type = NLA_U64, .min = 1 },
2ffe0395 BZ	36	[TCA_POLICE_PKTBURST64] = { .type = NLA_U64, .min = 1 },
53b2bf3f PM	37	};
53b2bf3f PM	38
2ac06347	39	static int tcf_police_init(struct net net, struct nlattr nla,
a85a970a	40	struct nlattr est, struct tc_action *a,
abbb0d33	41	struct tcf_proto *tp, u32 flags,
589dad6d	42	struct netlink_ext_ack *extack)
1da177e4	43	{
fd6d4338	44	int ret = 0, tcfp_result = TC_ACT_OK, err, size;
695176bf	45	bool bind = flags & TCA_ACT_FLAGS_BIND;
7ba699c6	46	struct nlattr *tb[TCA_POLICE_MAX + 1];
d6124d6b	47	struct tcf_chain *goto_ch = NULL;
1da177e4	48	struct tc_police *parm;
e9ce1cd3	49	struct tcf_police *police;
1da177e4	50	struct qdisc_rate_table R_tab = NULL, P_tab = NULL;
acd0a7ab	51	struct tc_action_net *tn = net_generic(net, act_police_ops.net_id);
2d550dba	52	struct tcf_police_params *new;
0852e455	53	bool exists = false;
7be8ef2c	54	u32 index;
d1967e49	55	u64 rate64, prate64;
2ffe0395	56	u64 pps, ppsburst;
1da177e4	57
cee63723	58	if (nla == NULL)
1da177e4 LT	59	return -EINVAL;
1da177e4 LT	60
8cb08174 JB	61	err = nla_parse_nested_deprecated(tb, TCA_POLICE_MAX, nla,
8cb08174 JB	62	police_policy, NULL);
cee63723 PM	63	if (err < 0)
	64	return err;
	65
7ba699c6	66	if (tb[TCA_POLICE_TBF] == NULL)
1e9b3d53	67	return -EINVAL;
7ba699c6	68	size = nla_len(tb[TCA_POLICE_TBF]);
1e9b3d53	69	if (size != sizeof(*parm) && size != sizeof(struct tc_police_compat))
1da177e4	70	return -EINVAL;
0852e455	71
7ba699c6	72	parm = nla_data(tb[TCA_POLICE_TBF]);
7be8ef2c DL	73	index = parm->index;
7be8ef2c DL	74	err = tcf_idr_check_alloc(tn, &index, a, bind);
0190c1d4 VB	75	if (err < 0)
	76	return err;
	77	exists = err;
0852e455 WC	78	if (exists && bind)
0852e455 WC	79	return 0;
1da177e4	80
0852e455	81	if (!exists) {
7be8ef2c	82	ret = tcf_idr_create(tn, index, NULL, a,
40bd094d	83	&act_police_ops, bind, true, flags);
0190c1d4	84	if (ret) {
7be8ef2c	85	tcf_idr_cleanup(tn, index);
a03e6fe5	86	return ret;
0190c1d4	87	}
a03e6fe5	88	ret = ACT_P_CREATED;
484afd1b	89	spin_lock_init(&(to_police(*a)->tcfp_lock));
695176bf	90	} else if (!(flags & TCA_ACT_FLAGS_REPLACE)) {
65a206c0	91	tcf_idr_release(*a, bind);
4e8ddd7f	92	return -EEXIST;
1da177e4	93	}
d6124d6b DC	94	err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
	95	if (err < 0)
	96	goto release_idr;
1da177e4	97
a85a970a	98	police = to_police(*a);
1da177e4 LT	99	if (parm->rate.rate) {
1da177e4 LT	100	err = -ENOMEM;
e9bc3fa2	101	R_tab = qdisc_get_rtab(&parm->rate, tb[TCA_POLICE_RATE], NULL);
1da177e4 LT	102	if (R_tab == NULL)
1da177e4 LT	103	goto failure;
c1b56878	104
1da177e4 LT	105	if (parm->peakrate.rate) {
1da177e4 LT	106	P_tab = qdisc_get_rtab(&parm->peakrate,
e9bc3fa2	107	tb[TCA_POLICE_PEAKRATE], NULL);
71bcb09a	108	if (P_tab == NULL)
1da177e4	109	goto failure;
1da177e4 LT	110	}
1da177e4 LT	111	}
71bcb09a	112
71bcb09a	113	if (est) {
93be42f9 DC	114	err = gen_replace_estimator(&police->tcf_bstats,
93be42f9 DC	115	police->common.cpu_bstats,
71bcb09a	116	&police->tcf_rate_est,
edb09eb1	117	&police->tcf_lock,
29cbcd85	118	false, est);
71bcb09a	119	if (err)
74030603	120	goto failure;
a883bf56 JP	121	} else if (tb[TCA_POLICE_AVRATE] &&
a883bf56 JP	122	(ret == ACT_P_CREATED \|\|
1c0d32fd	123	!gen_estimator_active(&police->tcf_rate_est))) {
a883bf56	124	err = -EINVAL;
74030603	125	goto failure;
71bcb09a SH	126	}
71bcb09a SH	127
fd6d4338 DC	128	if (tb[TCA_POLICE_RESULT]) {
	129	tcfp_result = nla_get_u32(tb[TCA_POLICE_RESULT]);
	130	if (TC_ACT_EXT_CMP(tcfp_result, TC_ACT_GOTO_CHAIN)) {
	131	NL_SET_ERR_MSG(extack,
	132	"goto chain not allowed on fallback");
	133	err = -EINVAL;
	134	goto failure;
	135	}
	136	}
	137
2ffe0395 BZ	138	if ((tb[TCA_POLICE_PKTRATE64] && !tb[TCA_POLICE_PKTBURST64]) \|\|
	139	(!tb[TCA_POLICE_PKTRATE64] && tb[TCA_POLICE_PKTBURST64])) {
	140	NL_SET_ERR_MSG(extack,
	141	"Both or neither packet-per-second burst and rate must be provided");
	142	err = -EINVAL;
	143	goto failure;
	144	}
	145
	146	if (tb[TCA_POLICE_PKTRATE64] && R_tab) {
	147	NL_SET_ERR_MSG(extack,
	148	"packet-per-second and byte-per-second rate limits not allowed in same action");
	149	err = -EINVAL;
	150	goto failure;
	151	}
	152
2d550dba DC	153	new = kzalloc(sizeof(*new), GFP_KERNEL);
	154	if (unlikely(!new)) {
	155	err = -ENOMEM;
	156	goto failure;
	157	}
	158
71bcb09a	159	/* No failure allowed after this point */
fd6d4338	160	new->tcfp_result = tcfp_result;
2d550dba DC	161	new->tcfp_mtu = parm->mtu;
	162	if (!new->tcfp_mtu) {
	163	new->tcfp_mtu = ~0;
c6d14ff1	164	if (R_tab)
2d550dba	165	new->tcfp_mtu = 255 << R_tab->rate.cell_log;
c6d14ff1 JP	166	}
c6d14ff1 JP	167	if (R_tab) {
2d550dba	168	new->rate_present = true;
d1967e49 DD	169	rate64 = tb[TCA_POLICE_RATE64] ?
	170	nla_get_u64(tb[TCA_POLICE_RATE64]) : 0;
	171	psched_ratecfg_precompute(&new->rate, &R_tab->rate, rate64);
c6d14ff1 JP	172	qdisc_put_rtab(R_tab);
c6d14ff1 JP	173	} else {
2d550dba	174	new->rate_present = false;
1da177e4	175	}
c6d14ff1	176	if (P_tab) {
2d550dba	177	new->peak_present = true;
d1967e49 DD	178	prate64 = tb[TCA_POLICE_PEAKRATE64] ?
	179	nla_get_u64(tb[TCA_POLICE_PEAKRATE64]) : 0;
	180	psched_ratecfg_precompute(&new->peak, &P_tab->rate, prate64);
c6d14ff1 JP	181	qdisc_put_rtab(P_tab);
c6d14ff1 JP	182	} else {
2d550dba	183	new->peak_present = false;
1da177e4 LT	184	}
1da177e4 LT	185
2d550dba	186	new->tcfp_burst = PSCHED_TICKS2NS(parm->burst);
f2cbd485	187	if (new->peak_present)
2d550dba DC	188	new->tcfp_mtu_ptoks = (s64)psched_l2t_ns(&new->peak,
2d550dba DC	189	new->tcfp_mtu);
1da177e4	190
7ba699c6	191	if (tb[TCA_POLICE_AVRATE])
2d550dba	192	new->tcfp_ewma_rate = nla_get_u32(tb[TCA_POLICE_AVRATE]);
1da177e4	193
2ffe0395 BZ	194	if (tb[TCA_POLICE_PKTRATE64]) {
	195	pps = nla_get_u64(tb[TCA_POLICE_PKTRATE64]);
	196	ppsburst = nla_get_u64(tb[TCA_POLICE_PKTBURST64]);
	197	new->pps_present = true;
	198	new->tcfp_pkt_burst = PSCHED_TICKS2NS(ppsburst);
	199	psched_ppscfg_precompute(&new->ppsrate, pps);
	200	}
	201
2d550dba	202	spin_lock_bh(&police->tcf_lock);
f2cbd485 DC	203	spin_lock_bh(&police->tcfp_lock);
	204	police->tcfp_t_c = ktime_get_ns();
	205	police->tcfp_toks = new->tcfp_burst;
	206	if (new->peak_present)
	207	police->tcfp_ptoks = new->tcfp_mtu_ptoks;
	208	spin_unlock_bh(&police->tcfp_lock);
d6124d6b	209	goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
445d3749 PM	210	new = rcu_replace_pointer(police->params,
	211	new,
	212	lockdep_is_held(&police->tcf_lock));
e9ce1cd3	213	spin_unlock_bh(&police->tcf_lock);
1da177e4	214
d6124d6b DC	215	if (goto_ch)
d6124d6b DC	216	tcf_chain_put_by_act(goto_ch);
2d550dba DC	217	if (new)
2d550dba DC	218	kfree_rcu(new, rcu);
1da177e4	219
1da177e4 LT	220	return ret;
	221
	222	failure:
3b69a4c9 YY	223	qdisc_put_rtab(P_tab);
3b69a4c9 YY	224	qdisc_put_rtab(R_tab);
d6124d6b DC	225	if (goto_ch)
	226	tcf_chain_put_by_act(goto_ch);
	227	release_idr:
4e8ddd7f	228	tcf_idr_release(*a, bind);
1da177e4 LT	229	return err;
	230	}
	231
4ddc844e DC	232	static bool tcf_police_mtu_check(struct sk_buff *skb, u32 limit)
	233	{
	234	u32 len;
	235
	236	if (skb_is_gso(skb))
	237	return skb_gso_validate_mac_len(skb, limit);
	238
	239	len = qdisc_pkt_len(skb);
	240	if (skb_at_tc_ingress(skb))
	241	len += skb->mac_len;
	242
	243	return len <= limit;
	244	}
	245
871cf386 PT	246	TC_INDIRECT_SCOPE int tcf_police_act(struct sk_buff *skb,
	247	const struct tc_action *a,
	248	struct tcf_result *res)
1da177e4	249	{
a85a970a	250	struct tcf_police *police = to_police(a);
2ffe0395	251	s64 now, toks, ppstoks = 0, ptoks = 0;
2d550dba	252	struct tcf_police_params *p;
93be42f9	253	int ret;
1da177e4	254
3d3ed181	255	tcf_lastuse_update(&police->tcf_tm);
50dc9a85	256	bstats_update(this_cpu_ptr(police->common.cpu_bstats), skb);
1da177e4	257
2d550dba DC	258	ret = READ_ONCE(police->tcf_action);
	259	p = rcu_dereference_bh(police->params);
	260
	261	if (p->tcfp_ewma_rate) {
1c0d32fd ED	262	struct gnet_stats_rate_est64 sample;
	263
	264	if (!gen_estimator_read(&police->tcf_rate_est, &sample) \|\|
2d550dba	265	sample.bps >= p->tcfp_ewma_rate)
93be42f9	266	goto inc_overlimits;
1da177e4	267	}
1da177e4	268
4ddc844e	269	if (tcf_police_mtu_check(skb, p->tcfp_mtu)) {
2ffe0395	270	if (!p->rate_present && !p->pps_present) {
2d550dba DC	271	ret = p->tcfp_result;
2d550dba DC	272	goto end;
1da177e4 LT	273	}
1da177e4 LT	274
d2de875c	275	now = ktime_get_ns();
f2cbd485 DC	276	spin_lock_bh(&police->tcfp_lock);
f2cbd485 DC	277	toks = min_t(s64, now - police->tcfp_t_c, p->tcfp_burst);
2d550dba	278	if (p->peak_present) {
f2cbd485	279	ptoks = toks + police->tcfp_ptoks;
2d550dba DC	280	if (ptoks > p->tcfp_mtu_ptoks)
	281	ptoks = p->tcfp_mtu_ptoks;
	282	ptoks -= (s64)psched_l2t_ns(&p->peak,
	283	qdisc_pkt_len(skb));
1da177e4	284	}
2ffe0395 BZ	285	if (p->rate_present) {
	286	toks += police->tcfp_toks;
	287	if (toks > p->tcfp_burst)
	288	toks = p->tcfp_burst;
	289	toks -= (s64)psched_l2t_ns(&p->rate, qdisc_pkt_len(skb));
	290	} else if (p->pps_present) {
	291	ppstoks = min_t(s64, now - police->tcfp_t_c, p->tcfp_pkt_burst);
	292	ppstoks += police->tcfp_pkttoks;
	293	if (ppstoks > p->tcfp_pkt_burst)
	294	ppstoks = p->tcfp_pkt_burst;
	295	ppstoks -= (s64)psched_pkt2t_ns(&p->ppsrate, 1);
	296	}
	297	if ((toks \| ptoks \| ppstoks) >= 0) {
f2cbd485 DC	298	police->tcfp_t_c = now;
	299	police->tcfp_toks = toks;
	300	police->tcfp_ptoks = ptoks;
2ffe0395	301	police->tcfp_pkttoks = ppstoks;
f2cbd485	302	spin_unlock_bh(&police->tcfp_lock);
2d550dba	303	ret = p->tcfp_result;
93be42f9	304	goto inc_drops;
1da177e4	305	}
f2cbd485	306	spin_unlock_bh(&police->tcfp_lock);
1da177e4	307	}
93be42f9 DC	308
	309	inc_overlimits:
	310	qstats_overlimit_inc(this_cpu_ptr(police->common.cpu_qstats));
	311	inc_drops:
	312	if (ret == TC_ACT_SHOT)
	313	qstats_drop_inc(this_cpu_ptr(police->common.cpu_qstats));
2d550dba	314	end:
93be42f9	315	return ret;
1da177e4 LT	316	}
1da177e4 LT	317
2d550dba DC	318	static void tcf_police_cleanup(struct tc_action *a)
	319	{
	320	struct tcf_police *police = to_police(a);
	321	struct tcf_police_params *p;
	322
	323	p = rcu_dereference_protected(police->params, 1);
	324	if (p)
	325	kfree_rcu(p, rcu);
	326	}
	327
12f02b6b	328	static void tcf_police_stats_update(struct tc_action *a,
4b61d3e8	329	u64 bytes, u64 packets, u64 drops,
12f02b6b PJV	330	u64 lastuse, bool hw)
	331	{
	332	struct tcf_police *police = to_police(a);
	333	struct tcf_t *tm = &police->tcf_tm;
	334
4b61d3e8	335	tcf_action_update_stats(a, bytes, packets, drops, hw);
12f02b6b PJV	336	tm->lastuse = max_t(u64, tm->lastuse, lastuse);
	337	}
	338
2ac06347	339	static int tcf_police_dump(struct sk_buff skb, struct tc_action a,
5a7a5555	340	int bind, int ref)
1da177e4	341	{
27a884dc	342	unsigned char *b = skb_tail_pointer(skb);
a85a970a	343	struct tcf_police *police = to_police(a);
2d550dba	344	struct tcf_police_params *p;
0f04cfd0 JM	345	struct tc_police opt = {
0f04cfd0 JM	346	.index = police->tcf_index,
036bb443 VB	347	.refcnt = refcount_read(&police->tcf_refcnt) - ref,
036bb443 VB	348	.bindcnt = atomic_read(&police->tcf_bindcnt) - bind,
0f04cfd0	349	};
3d3ed181	350	struct tcf_t t;
0f04cfd0	351
e329bc42 VB	352	spin_lock_bh(&police->tcf_lock);
e329bc42 VB	353	opt.action = police->tcf_action;
2d550dba DC	354	p = rcu_dereference_protected(police->params,
	355	lockdep_is_held(&police->tcf_lock));
	356	opt.mtu = p->tcfp_mtu;
	357	opt.burst = PSCHED_NS2TICKS(p->tcfp_burst);
d1967e49	358	if (p->rate_present) {
2d550dba	359	psched_ratecfg_getrate(&opt.rate, &p->rate);
682881ee	360	if ((p->rate.rate_bytes_ps >= (1ULL << 32)) &&
d1967e49	361	nla_put_u64_64bit(skb, TCA_POLICE_RATE64,
682881ee	362	p->rate.rate_bytes_ps,
d1967e49 DD	363	TCA_POLICE_PAD))
	364	goto nla_put_failure;
	365	}
	366	if (p->peak_present) {
2d550dba	367	psched_ratecfg_getrate(&opt.peakrate, &p->peak);
682881ee	368	if ((p->peak.rate_bytes_ps >= (1ULL << 32)) &&
d1967e49	369	nla_put_u64_64bit(skb, TCA_POLICE_PEAKRATE64,
682881ee	370	p->peak.rate_bytes_ps,
d1967e49 DD	371	TCA_POLICE_PAD))
	372	goto nla_put_failure;
	373	}
2ffe0395 BZ	374	if (p->pps_present) {
2ffe0395 BZ	375	if (nla_put_u64_64bit(skb, TCA_POLICE_PKTRATE64,
682881ee	376	p->ppsrate.rate_pkts_ps,
2ffe0395 BZ	377	TCA_POLICE_PAD))
	378	goto nla_put_failure;
	379	if (nla_put_u64_64bit(skb, TCA_POLICE_PKTBURST64,
	380	PSCHED_NS2TICKS(p->tcfp_pkt_burst),
	381	TCA_POLICE_PAD))
	382	goto nla_put_failure;
	383	}
1b34ec43 DM	384	if (nla_put(skb, TCA_POLICE_TBF, sizeof(opt), &opt))
1b34ec43 DM	385	goto nla_put_failure;
2d550dba DC	386	if (p->tcfp_result &&
2d550dba DC	387	nla_put_u32(skb, TCA_POLICE_RESULT, p->tcfp_result))
1b34ec43	388	goto nla_put_failure;
2d550dba DC	389	if (p->tcfp_ewma_rate &&
2d550dba DC	390	nla_put_u32(skb, TCA_POLICE_AVRATE, p->tcfp_ewma_rate))
1b34ec43	391	goto nla_put_failure;
3d3ed181	392
985fd98a	393	tcf_tm_dump(&t, &police->tcf_tm);
3d3ed181 JHS	394	if (nla_put_64bit(skb, TCA_POLICE_TM, sizeof(t), &t, TCA_POLICE_PAD))
3d3ed181 JHS	395	goto nla_put_failure;
e329bc42	396	spin_unlock_bh(&police->tcf_lock);
3d3ed181	397
1da177e4 LT	398	return skb->len;
1da177e4 LT	399
7ba699c6	400	nla_put_failure:
e329bc42	401	spin_unlock_bh(&police->tcf_lock);
dc5fc579	402	nlmsg_trim(skb, b);
1da177e4 LT	403	return -1;
	404	}
	405
b50e462b IS	406	static int tcf_police_act_to_flow_act(int tc_act, u32 *extval,
b50e462b IS	407	struct netlink_ext_ack *extack)
b8cd5831 JL	408	{
	409	int act_id = -EOPNOTSUPP;
	410
	411	if (!TC_ACT_EXT_OPCODE(tc_act)) {
	412	if (tc_act == TC_ACT_OK)
	413	act_id = FLOW_ACTION_ACCEPT;
	414	else if (tc_act == TC_ACT_SHOT)
	415	act_id = FLOW_ACTION_DROP;
	416	else if (tc_act == TC_ACT_PIPE)
	417	act_id = FLOW_ACTION_PIPE;
b50e462b IS	418	else if (tc_act == TC_ACT_RECLASSIFY)
	419	NL_SET_ERR_MSG_MOD(extack, "Offload not supported when conform/exceed action is \"reclassify\"");
	420	else
	421	NL_SET_ERR_MSG_MOD(extack, "Unsupported conform/exceed action offload");
b8cd5831 JL	422	} else if (TC_ACT_EXT_CMP(tc_act, TC_ACT_GOTO_CHAIN)) {
	423	act_id = FLOW_ACTION_GOTO;
	424	*extval = tc_act & TC_ACT_EXT_VAL_MASK;
	425	} else if (TC_ACT_EXT_CMP(tc_act, TC_ACT_JUMP)) {
	426	act_id = FLOW_ACTION_JUMP;
	427	*extval = tc_act & TC_ACT_EXT_VAL_MASK;
b50e462b	428	} else if (tc_act == TC_ACT_UNSPEC) {
052f744f	429	act_id = FLOW_ACTION_CONTINUE;
b50e462b IS	430	} else {
b50e462b IS	431	NL_SET_ERR_MSG_MOD(extack, "Unsupported conform/exceed action offload");
b8cd5831 JL	432	}
	433
	434	return act_id;
	435	}
	436
c54e1d92	437	static int tcf_police_offload_act_setup(struct tc_action act, void entry_data,
c2ccf84e IS	438	u32 *index_inc, bool bind,
c2ccf84e IS	439	struct netlink_ext_ack *extack)
c54e1d92 BZ	440	{
	441	if (bind) {
	442	struct flow_action_entry *entry = entry_data;
b8cd5831 JL	443	struct tcf_police *police = to_police(act);
	444	struct tcf_police_params *p;
	445	int act_id;
	446
	447	p = rcu_dereference_protected(police->params,
	448	lockdep_is_held(&police->tcf_lock));
c54e1d92 BZ	449
	450	entry->id = FLOW_ACTION_POLICE;
	451	entry->police.burst = tcf_police_burst(act);
	452	entry->police.rate_bytes_ps =
	453	tcf_police_rate_bytes_ps(act);
b8cd5831 JL	454	entry->police.peakrate_bytes_ps = tcf_police_peakrate_bytes_ps(act);
	455	entry->police.avrate = tcf_police_tcfp_ewma_rate(act);
	456	entry->police.overhead = tcf_police_rate_overhead(act);
c54e1d92 BZ	457	entry->police.burst_pkt = tcf_police_burst_pkt(act);
	458	entry->police.rate_pkt_ps =
	459	tcf_police_rate_pkt_ps(act);
	460	entry->police.mtu = tcf_police_tcfp_mtu(act);
b8cd5831 JL	461
b8cd5831 JL	462	act_id = tcf_police_act_to_flow_act(police->tcf_action,
b50e462b IS	463	&entry->police.exceed.extval,
b50e462b IS	464	extack);
b8cd5831 JL	465	if (act_id < 0)
	466	return act_id;
	467
	468	entry->police.exceed.act_id = act_id;
	469
	470	act_id = tcf_police_act_to_flow_act(p->tcfp_result,
b50e462b IS	471	&entry->police.notexceed.extval,
b50e462b IS	472	extack);
b8cd5831 JL	473	if (act_id < 0)
	474	return act_id;
	475
	476	entry->police.notexceed.act_id = act_id;
	477
c54e1d92 BZ	478	*index_inc = 1;
c54e1d92 BZ	479	} else {
8cbfe939 BZ	480	struct flow_offload_action *fl_action = entry_data;
	481
	482	fl_action->id = FLOW_ACTION_POLICE;
c54e1d92 BZ	483	}
	484
	485	return 0;
	486	}
	487
1da177e4 LT	488	MODULE_AUTHOR("Alexey Kuznetsov");
	489	MODULE_DESCRIPTION("Policing actions");
	490	MODULE_LICENSE("GPL");
	491
	492	static struct tc_action_ops act_police_ops = {
	493	.kind = "police",
eddd2cf1	494	.id = TCA_ID_POLICE,
1da177e4	495	.owner = THIS_MODULE,
12f02b6b	496	.stats_update = tcf_police_stats_update,
2ac06347 JHS	497	.act = tcf_police_act,
	498	.dump = tcf_police_dump,
	499	.init = tcf_police_init,
2d550dba	500	.cleanup = tcf_police_cleanup,
c54e1d92	501	.offload_act_setup = tcf_police_offload_act_setup,
a85a970a	502	.size = sizeof(struct tcf_police),
ddf97ccd WC	503	};
	504
	505	static __net_init int police_init_net(struct net *net)
	506	{
acd0a7ab	507	struct tc_action_net *tn = net_generic(net, act_police_ops.net_id);
ddf97ccd	508
981471bd	509	return tc_action_net_init(net, tn, &act_police_ops);
ddf97ccd WC	510	}
ddf97ccd WC	511
039af9c6	512	static void __net_exit police_exit_net(struct list_head *net_list)
ddf97ccd	513	{
acd0a7ab	514	tc_action_net_exit(net_list, act_police_ops.net_id);
ddf97ccd WC	515	}
	516
	517	static struct pernet_operations police_net_ops = {
	518	.init = police_init_net,
039af9c6	519	.exit_batch = police_exit_net,
acd0a7ab	520	.id = &act_police_ops.net_id,
ddf97ccd	521	.size = sizeof(struct tc_action_net),
1da177e4 LT	522	};
1da177e4 LT	523
5a7a5555	524	static int __init police_init_module(void)
1da177e4	525	{
ddf97ccd	526	return tcf_register_action(&act_police_ops, &police_net_ops);
1da177e4 LT	527	}
1da177e4 LT	528
5a7a5555	529	static void __exit police_cleanup_module(void)
1da177e4	530	{
ddf97ccd	531	tcf_unregister_action(&act_police_ops, &police_net_ops);
1da177e4 LT	532	}
	533
	534	module_init(police_init_module);
	535	module_exit(police_cleanup_module);