[mirror_ubuntu-artful-kernel.git] / net / sched / act_bpf.c

/*
 * Copyright (c) 2015 Jiri Pirko <jiri@resnulli.us>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 */

#include <linux/module.h>
#include <linux/init.h>
#include <linux/kernel.h>
#include <linux/skbuff.h>
#include <linux/rtnetlink.h>
#include <linux/filter.h>
#include <linux/bpf.h>

#include <net/netlink.h>
#include <net/pkt_sched.h>

#include <linux/tc_act/tc_bpf.h>
#include <net/tc_act/tc_bpf.h>

#define BPF_TAB_MASK		15
#define ACT_BPF_NAME_LEN	256

struct tcf_bpf_cfg {
	struct bpf_prog *filter;
	struct sock_filter *bpf_ops;
	const char *bpf_name;
	u16 bpf_num_ops;
	bool is_ebpf;
};

static unsigned int bpf_net_id;
static struct tc_action_ops act_bpf_ops;

static int tcf_bpf(struct sk_buff *skb, const struct tc_action *act,
		   struct tcf_result *res)
{
	bool at_ingress = skb_at_tc_ingress(skb);
	struct tcf_bpf *prog = to_bpf(act);
	struct bpf_prog *filter;
	int action, filter_res;

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

	rcu_read_lock();
	filter = rcu_dereference(prog->filter);
	if (at_ingress) {
		__skb_push(skb, skb->mac_len);
		bpf_compute_data_end(skb);
		filter_res = BPF_PROG_RUN(filter, skb);
		__skb_pull(skb, skb->mac_len);
	} else {
		bpf_compute_data_end(skb);
		filter_res = BPF_PROG_RUN(filter, skb);
	}
	rcu_read_unlock();

	/* A BPF program may overwrite the default action opcode.
	 * Similarly as in cls_bpf, if filter_res == -1 we use the
	 * default action specified from tc.
	 *
	 * In case a different well-known TC_ACT opcode has been
	 * returned, it will overwrite the default one.
	 *
	 * For everything else that is unkown, TC_ACT_UNSPEC is
	 * returned.
	 */
	switch (filter_res) {
	case TC_ACT_PIPE:
	case TC_ACT_RECLASSIFY:
	case TC_ACT_OK:
	case TC_ACT_REDIRECT:
		action = filter_res;
		break;
	case TC_ACT_SHOT:
		action = filter_res;
		qstats_drop_inc(this_cpu_ptr(prog->common.cpu_qstats));
		break;
	case TC_ACT_UNSPEC:
		action = prog->tcf_action;
		break;
	default:
		action = TC_ACT_UNSPEC;
		break;
	}

	return action;
}

static bool tcf_bpf_is_ebpf(const struct tcf_bpf *prog)
{
	return !prog->bpf_ops;
}

static int tcf_bpf_dump_bpf_info(const struct tcf_bpf *prog,
				 struct sk_buff *skb)
{
	struct nlattr *nla;

	if (nla_put_u16(skb, TCA_ACT_BPF_OPS_LEN, prog->bpf_num_ops))
		return -EMSGSIZE;

	nla = nla_reserve(skb, TCA_ACT_BPF_OPS, prog->bpf_num_ops *
			  sizeof(struct sock_filter));
	if (nla == NULL)
		return -EMSGSIZE;

	memcpy(nla_data(nla), prog->bpf_ops, nla_len(nla));

	return 0;
}

static int tcf_bpf_dump_ebpf_info(const struct tcf_bpf *prog,
				  struct sk_buff *skb)
{
	struct nlattr *nla;

	if (prog->bpf_name &&
	    nla_put_string(skb, TCA_ACT_BPF_NAME, prog->bpf_name))
		return -EMSGSIZE;

	nla = nla_reserve(skb, TCA_ACT_BPF_DIGEST,
			  sizeof(prog->filter->digest));
	if (nla == NULL)
		return -EMSGSIZE;

	memcpy(nla_data(nla), prog->filter->digest, nla_len(nla));

	return 0;
}

static int tcf_bpf_dump(struct sk_buff *skb, struct tc_action *act,
			int bind, int ref)
{
	unsigned char *tp = skb_tail_pointer(skb);
	struct tcf_bpf *prog = to_bpf(act);
	struct tc_act_bpf opt = {
		.index   = prog->tcf_index,
		.refcnt  = prog->tcf_refcnt - ref,
		.bindcnt = prog->tcf_bindcnt - bind,
		.action  = prog->tcf_action,
	};
	struct tcf_t tm;
	int ret;

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

	if (tcf_bpf_is_ebpf(prog))
		ret = tcf_bpf_dump_ebpf_info(prog, skb);
	else
		ret = tcf_bpf_dump_bpf_info(prog, skb);
	if (ret)
		goto nla_put_failure;

	tcf_tm_dump(&tm, &prog->tcf_tm);
	if (nla_put_64bit(skb, TCA_ACT_BPF_TM, sizeof(tm), &tm,
			  TCA_ACT_BPF_PAD))
		goto nla_put_failure;

	return skb->len;

nla_put_failure:
	nlmsg_trim(skb, tp);
	return -1;
}

static const struct nla_policy act_bpf_policy[TCA_ACT_BPF_MAX + 1] = {
	[TCA_ACT_BPF_PARMS]	= { .len = sizeof(struct tc_act_bpf) },
	[TCA_ACT_BPF_FD]	= { .type = NLA_U32 },
	[TCA_ACT_BPF_NAME]	= { .type = NLA_NUL_STRING,
				    .len = ACT_BPF_NAME_LEN },
	[TCA_ACT_BPF_OPS_LEN]	= { .type = NLA_U16 },
	[TCA_ACT_BPF_OPS]	= { .type = NLA_BINARY,
				    .len = sizeof(struct sock_filter) * BPF_MAXINSNS },
};

static int tcf_bpf_init_from_ops(struct nlattr **tb, struct tcf_bpf_cfg *cfg)
{
	struct sock_filter *bpf_ops;
	struct sock_fprog_kern fprog_tmp;
	struct bpf_prog *fp;
	u16 bpf_size, bpf_num_ops;
	int ret;

	bpf_num_ops = nla_get_u16(tb[TCA_ACT_BPF_OPS_LEN]);
	if (bpf_num_ops	> BPF_MAXINSNS || bpf_num_ops == 0)
		return -EINVAL;

	bpf_size = bpf_num_ops * sizeof(*bpf_ops);
	if (bpf_size != nla_len(tb[TCA_ACT_BPF_OPS]))
		return -EINVAL;

	bpf_ops = kzalloc(bpf_size, GFP_KERNEL);
	if (bpf_ops == NULL)
		return -ENOMEM;

	memcpy(bpf_ops, nla_data(tb[TCA_ACT_BPF_OPS]), bpf_size);

	fprog_tmp.len = bpf_num_ops;
	fprog_tmp.filter = bpf_ops;

	ret = bpf_prog_create(&fp, &fprog_tmp);
	if (ret < 0) {
		kfree(bpf_ops);
		return ret;
	}

	cfg->bpf_ops = bpf_ops;
	cfg->bpf_num_ops = bpf_num_ops;
	cfg->filter = fp;
	cfg->is_ebpf = false;

	return 0;
}

static int tcf_bpf_init_from_efd(struct nlattr **tb, struct tcf_bpf_cfg *cfg)
{
	struct bpf_prog *fp;
	char *name = NULL;
	u32 bpf_fd;

	bpf_fd = nla_get_u32(tb[TCA_ACT_BPF_FD]);

	fp = bpf_prog_get_type(bpf_fd, BPF_PROG_TYPE_SCHED_ACT);
	if (IS_ERR(fp))
		return PTR_ERR(fp);

	if (tb[TCA_ACT_BPF_NAME]) {
		name = nla_memdup(tb[TCA_ACT_BPF_NAME], GFP_KERNEL);
		if (!name) {
			bpf_prog_put(fp);
			return -ENOMEM;
		}
	}

	cfg->bpf_name = name;
	cfg->filter = fp;
	cfg->is_ebpf = true;

	return 0;
}

static void tcf_bpf_cfg_cleanup(const struct tcf_bpf_cfg *cfg)
{
	if (cfg->is_ebpf)
		bpf_prog_put(cfg->filter);
	else
		bpf_prog_destroy(cfg->filter);

	kfree(cfg->bpf_ops);
	kfree(cfg->bpf_name);
}

static void tcf_bpf_prog_fill_cfg(const struct tcf_bpf *prog,
				  struct tcf_bpf_cfg *cfg)
{
	cfg->is_ebpf = tcf_bpf_is_ebpf(prog);
	/* updates to prog->filter are prevented, since it's called either
	 * with rtnl lock or during final cleanup in rcu callback
	 */
	cfg->filter = rcu_dereference_protected(prog->filter, 1);

	cfg->bpf_ops = prog->bpf_ops;
	cfg->bpf_name = prog->bpf_name;
}

static int tcf_bpf_init(struct net *net, struct nlattr *nla,
			struct nlattr *est, struct tc_action **act,
			int replace, int bind)
{
	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	struct nlattr *tb[TCA_ACT_BPF_MAX + 1];
	struct tcf_bpf_cfg cfg, old;
	struct tc_act_bpf *parm;
	struct tcf_bpf *prog;
	bool is_bpf, is_ebpf;
	int ret, res = 0;

	if (!nla)
		return -EINVAL;

	ret = nla_parse_nested(tb, TCA_ACT_BPF_MAX, nla, act_bpf_policy);
	if (ret < 0)
		return ret;

	if (!tb[TCA_ACT_BPF_PARMS])
		return -EINVAL;

	parm = nla_data(tb[TCA_ACT_BPF_PARMS]);

	if (!tcf_hash_check(tn, parm->index, act, bind)) {
		ret = tcf_hash_create(tn, parm->index, est, act,
				      &act_bpf_ops, bind, true);
		if (ret < 0)
			return ret;

		res = ACT_P_CREATED;
	} else {
		/* Don't override defaults. */
		if (bind)
			return 0;

		tcf_hash_release(*act, bind);
		if (!replace)
			return -EEXIST;
	}

	is_bpf = tb[TCA_ACT_BPF_OPS_LEN] && tb[TCA_ACT_BPF_OPS];
	is_ebpf = tb[TCA_ACT_BPF_FD];

	if ((!is_bpf && !is_ebpf) || (is_bpf && is_ebpf)) {
		ret = -EINVAL;
		goto out;
	}

	memset(&cfg, 0, sizeof(cfg));

	ret = is_bpf ? tcf_bpf_init_from_ops(tb, &cfg) :
		       tcf_bpf_init_from_efd(tb, &cfg);
	if (ret < 0)
		goto out;

	prog = to_bpf(*act);
	ASSERT_RTNL();

	if (res != ACT_P_CREATED)
		tcf_bpf_prog_fill_cfg(prog, &old);

	prog->bpf_ops = cfg.bpf_ops;
	prog->bpf_name = cfg.bpf_name;

	if (cfg.bpf_num_ops)
		prog->bpf_num_ops = cfg.bpf_num_ops;

	prog->tcf_action = parm->action;
	rcu_assign_pointer(prog->filter, cfg.filter);

	if (res == ACT_P_CREATED) {
		tcf_hash_insert(tn, *act);
	} else {
		/* make sure the program being replaced is no longer executing */
		synchronize_rcu();
		tcf_bpf_cfg_cleanup(&old);
	}

	return res;
out:
	if (res == ACT_P_CREATED)
		tcf_hash_cleanup(*act, est);

	return ret;
}

static void tcf_bpf_cleanup(struct tc_action *act, int bind)
{
	struct tcf_bpf_cfg tmp;

	tcf_bpf_prog_fill_cfg(to_bpf(act), &tmp);
	tcf_bpf_cfg_cleanup(&tmp);
}

static int tcf_bpf_walker(struct net *net, struct sk_buff *skb,
			  struct netlink_callback *cb, int type,
			  const struct tc_action_ops *ops)
{
	struct tc_action_net *tn = net_generic(net, bpf_net_id);

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

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

	return tcf_hash_search(tn, a, index);
}

static struct tc_action_ops act_bpf_ops __read_mostly = {
	.kind		=	"bpf",
	.type		=	TCA_ACT_BPF,
	.owner		=	THIS_MODULE,
	.act		=	tcf_bpf,
	.dump		=	tcf_bpf_dump,
	.cleanup	=	tcf_bpf_cleanup,
	.init		=	tcf_bpf_init,
	.walk		=	tcf_bpf_walker,
	.lookup		=	tcf_bpf_search,
	.size		=	sizeof(struct tcf_bpf),
};

static __net_init int bpf_init_net(struct net *net)
{
	struct tc_action_net *tn = net_generic(net, bpf_net_id);

	return tc_action_net_init(tn, &act_bpf_ops, BPF_TAB_MASK);
}

static void __net_exit bpf_exit_net(struct net *net)
{
	struct tc_action_net *tn = net_generic(net, bpf_net_id);

	tc_action_net_exit(tn);
}

static struct pernet_operations bpf_net_ops = {
	.init = bpf_init_net,
	.exit = bpf_exit_net,
	.id   = &bpf_net_id,
	.size = sizeof(struct tc_action_net),
};

static int __init bpf_init_module(void)
{
	return tcf_register_action(&act_bpf_ops, &bpf_net_ops);
}

static void __exit bpf_cleanup_module(void)
{
	tcf_unregister_action(&act_bpf_ops, &bpf_net_ops);
}

module_init(bpf_init_module);
module_exit(bpf_cleanup_module);

MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
MODULE_DESCRIPTION("TC BPF based action");
MODULE_LICENSE("GPL v2");
Commit	Line	Data
d23b8ad8 JP	1	/*
	2	* Copyright (c) 2015 Jiri Pirko <jiri@resnulli.us>
	3	*
	4	* This program is free software; you can redistribute it and/or modify
	5	* it under the terms of the GNU General Public License as published by
	6	* the Free Software Foundation; either version 2 of the License, or
	7	* (at your option) any later version.
	8	*/
	9
	10	#include <linux/module.h>
	11	#include <linux/init.h>
	12	#include <linux/kernel.h>
	13	#include <linux/skbuff.h>
	14	#include <linux/rtnetlink.h>
	15	#include <linux/filter.h>
a8cb5f55 DB	16	#include <linux/bpf.h>
a8cb5f55 DB	17
d23b8ad8 JP	18	#include <net/netlink.h>
	19	#include <net/pkt_sched.h>
	20
	21	#include <linux/tc_act/tc_bpf.h>
	22	#include <net/tc_act/tc_bpf.h>
	23
a8cb5f55 DB	24	#define BPF_TAB_MASK 15
	25	#define ACT_BPF_NAME_LEN 256
	26
	27	struct tcf_bpf_cfg {
	28	struct bpf_prog *filter;
	29	struct sock_filter *bpf_ops;
f4eaed28	30	const char *bpf_name;
a8cb5f55	31	u16 bpf_num_ops;
f4eaed28	32	bool is_ebpf;
a8cb5f55	33	};
d23b8ad8	34
c7d03a00	35	static unsigned int bpf_net_id;
a85a970a	36	static struct tc_action_ops act_bpf_ops;
ddf97ccd	37
a8cb5f55	38	static int tcf_bpf(struct sk_buff skb, const struct tc_action act,
d23b8ad8 JP	39	struct tcf_result *res)
d23b8ad8 JP	40	{
f53d8c7b	41	bool at_ingress = skb_at_tc_ingress(skb);
a85a970a	42	struct tcf_bpf *prog = to_bpf(act);
cff82457	43	struct bpf_prog *filter;
ced585c8	44	int action, filter_res;
d23b8ad8	45
cff82457 AS	46	tcf_lastuse_update(&prog->tcf_tm);
cff82457 AS	47	bstats_cpu_update(this_cpu_ptr(prog->common.cpu_bstats), skb);
d23b8ad8	48
a8cb5f55	49	rcu_read_lock();
cff82457	50	filter = rcu_dereference(prog->filter);
3431205e AS	51	if (at_ingress) {
3431205e AS	52	__skb_push(skb, skb->mac_len);
db58ba45	53	bpf_compute_data_end(skb);
cff82457	54	filter_res = BPF_PROG_RUN(filter, skb);
3431205e AS	55	__skb_pull(skb, skb->mac_len);
3431205e AS	56	} else {
db58ba45	57	bpf_compute_data_end(skb);
cff82457	58	filter_res = BPF_PROG_RUN(filter, skb);
3431205e	59	}
a8cb5f55	60	rcu_read_unlock();
ced585c8 DB	61
	62	/* A BPF program may overwrite the default action opcode.
	63	* Similarly as in cls_bpf, if filter_res == -1 we use the
	64	* default action specified from tc.
	65	*
	66	* In case a different well-known TC_ACT opcode has been
	67	* returned, it will overwrite the default one.
	68	*
	69	* For everything else that is unkown, TC_ACT_UNSPEC is
	70	* returned.
	71	*/
	72	switch (filter_res) {
	73	case TC_ACT_PIPE:
	74	case TC_ACT_RECLASSIFY:
	75	case TC_ACT_OK:
27b29f63	76	case TC_ACT_REDIRECT:
ced585c8 DB	77	action = filter_res;
	78	break;
	79	case TC_ACT_SHOT:
	80	action = filter_res;
cff82457	81	qstats_drop_inc(this_cpu_ptr(prog->common.cpu_qstats));
ced585c8 DB	82	break;
ced585c8 DB	83	case TC_ACT_UNSPEC:
a8cb5f55	84	action = prog->tcf_action;
ced585c8 DB	85	break;
	86	default:
	87	action = TC_ACT_UNSPEC;
	88	break;
d23b8ad8 JP	89	}
d23b8ad8 JP	90
d23b8ad8 JP	91	return action;
	92	}
	93
a8cb5f55 DB	94	static bool tcf_bpf_is_ebpf(const struct tcf_bpf *prog)
	95	{
	96	return !prog->bpf_ops;
	97	}
	98
	99	static int tcf_bpf_dump_bpf_info(const struct tcf_bpf *prog,
	100	struct sk_buff *skb)
	101	{
	102	struct nlattr *nla;
	103
	104	if (nla_put_u16(skb, TCA_ACT_BPF_OPS_LEN, prog->bpf_num_ops))
	105	return -EMSGSIZE;
	106
	107	nla = nla_reserve(skb, TCA_ACT_BPF_OPS, prog->bpf_num_ops *
	108	sizeof(struct sock_filter));
	109	if (nla == NULL)
	110	return -EMSGSIZE;
	111
	112	memcpy(nla_data(nla), prog->bpf_ops, nla_len(nla));
	113
	114	return 0;
	115	}
	116
	117	static int tcf_bpf_dump_ebpf_info(const struct tcf_bpf *prog,
	118	struct sk_buff *skb)
	119	{
7bd509e3 DB	120	struct nlattr *nla;
7bd509e3 DB	121
a8cb5f55 DB	122	if (prog->bpf_name &&
	123	nla_put_string(skb, TCA_ACT_BPF_NAME, prog->bpf_name))
	124	return -EMSGSIZE;
	125
7bd509e3 DB	126	nla = nla_reserve(skb, TCA_ACT_BPF_DIGEST,
	127	sizeof(prog->filter->digest));
	128	if (nla == NULL)
	129	return -EMSGSIZE;
	130
	131	memcpy(nla_data(nla), prog->filter->digest, nla_len(nla));
	132
a8cb5f55 DB	133	return 0;
	134	}
	135
	136	static int tcf_bpf_dump(struct sk_buff skb, struct tc_action act,
d23b8ad8 JP	137	int bind, int ref)
	138	{
	139	unsigned char *tp = skb_tail_pointer(skb);
a85a970a	140	struct tcf_bpf *prog = to_bpf(act);
d23b8ad8	141	struct tc_act_bpf opt = {
a8cb5f55 DB	142	.index = prog->tcf_index,
	143	.refcnt = prog->tcf_refcnt - ref,
	144	.bindcnt = prog->tcf_bindcnt - bind,
	145	.action = prog->tcf_action,
d23b8ad8	146	};
a8cb5f55 DB	147	struct tcf_t tm;
a8cb5f55 DB	148	int ret;
d23b8ad8 JP	149
	150	if (nla_put(skb, TCA_ACT_BPF_PARMS, sizeof(opt), &opt))
	151	goto nla_put_failure;
	152
a8cb5f55 DB	153	if (tcf_bpf_is_ebpf(prog))
	154	ret = tcf_bpf_dump_ebpf_info(prog, skb);
	155	else
	156	ret = tcf_bpf_dump_bpf_info(prog, skb);
	157	if (ret)
d23b8ad8 JP	158	goto nla_put_failure;
d23b8ad8 JP	159
48d8ee16	160	tcf_tm_dump(&tm, &prog->tcf_tm);
9854518e ND	161	if (nla_put_64bit(skb, TCA_ACT_BPF_TM, sizeof(tm), &tm,
9854518e ND	162	TCA_ACT_BPF_PAD))
d23b8ad8	163	goto nla_put_failure;
a8cb5f55	164
d23b8ad8 JP	165	return skb->len;
	166
	167	nla_put_failure:
	168	nlmsg_trim(skb, tp);
	169	return -1;
	170	}
	171
	172	static const struct nla_policy act_bpf_policy[TCA_ACT_BPF_MAX + 1] = {
	173	[TCA_ACT_BPF_PARMS] = { .len = sizeof(struct tc_act_bpf) },
a8cb5f55	174	[TCA_ACT_BPF_FD] = { .type = NLA_U32 },
0b0f43fe JHS	175	[TCA_ACT_BPF_NAME] = { .type = NLA_NUL_STRING,
0b0f43fe JHS	176	.len = ACT_BPF_NAME_LEN },
d23b8ad8 JP	177	[TCA_ACT_BPF_OPS_LEN] = { .type = NLA_U16 },
	178	[TCA_ACT_BPF_OPS] = { .type = NLA_BINARY,
	179	.len = sizeof(struct sock_filter) * BPF_MAXINSNS },
	180	};
	181
a8cb5f55	182	static int tcf_bpf_init_from_ops(struct nlattr *tb, struct tcf_bpf_cfg cfg)
d23b8ad8	183	{
d23b8ad8	184	struct sock_filter *bpf_ops;
a8cb5f55	185	struct sock_fprog_kern fprog_tmp;
d23b8ad8	186	struct bpf_prog *fp;
a8cb5f55	187	u16 bpf_size, bpf_num_ops;
d23b8ad8 JP	188	int ret;
d23b8ad8 JP	189
d23b8ad8 JP	190	bpf_num_ops = nla_get_u16(tb[TCA_ACT_BPF_OPS_LEN]);
	191	if (bpf_num_ops > BPF_MAXINSNS \|\| bpf_num_ops == 0)
	192	return -EINVAL;
	193
	194	bpf_size = bpf_num_ops * sizeof(*bpf_ops);
fd3e646c DB	195	if (bpf_size != nla_len(tb[TCA_ACT_BPF_OPS]))
	196	return -EINVAL;
	197
d23b8ad8	198	bpf_ops = kzalloc(bpf_size, GFP_KERNEL);
a8cb5f55	199	if (bpf_ops == NULL)
d23b8ad8 JP	200	return -ENOMEM;
	201
	202	memcpy(bpf_ops, nla_data(tb[TCA_ACT_BPF_OPS]), bpf_size);
	203
a8cb5f55 DB	204	fprog_tmp.len = bpf_num_ops;
a8cb5f55 DB	205	fprog_tmp.filter = bpf_ops;
d23b8ad8	206
a8cb5f55 DB	207	ret = bpf_prog_create(&fp, &fprog_tmp);
	208	if (ret < 0) {
	209	kfree(bpf_ops);
	210	return ret;
	211	}
d23b8ad8	212
a8cb5f55 DB	213	cfg->bpf_ops = bpf_ops;
	214	cfg->bpf_num_ops = bpf_num_ops;
	215	cfg->filter = fp;
f4eaed28	216	cfg->is_ebpf = false;
a8cb5f55 DB	217
	218	return 0;
	219	}
	220
	221	static int tcf_bpf_init_from_efd(struct nlattr *tb, struct tcf_bpf_cfg cfg)
	222	{
	223	struct bpf_prog *fp;
	224	char *name = NULL;
	225	u32 bpf_fd;
	226
	227	bpf_fd = nla_get_u32(tb[TCA_ACT_BPF_FD]);
	228
113214be	229	fp = bpf_prog_get_type(bpf_fd, BPF_PROG_TYPE_SCHED_ACT);
a8cb5f55 DB	230	if (IS_ERR(fp))
	231	return PTR_ERR(fp);
	232
a8cb5f55	233	if (tb[TCA_ACT_BPF_NAME]) {
b15ca182	234	name = nla_memdup(tb[TCA_ACT_BPF_NAME], GFP_KERNEL);
a8cb5f55 DB	235	if (!name) {
	236	bpf_prog_put(fp);
	237	return -ENOMEM;
	238	}
	239	}
	240
a8cb5f55 DB	241	cfg->bpf_name = name;
a8cb5f55 DB	242	cfg->filter = fp;
f4eaed28	243	cfg->is_ebpf = true;
a8cb5f55 DB	244
	245	return 0;
	246	}
	247
f4eaed28 DB	248	static void tcf_bpf_cfg_cleanup(const struct tcf_bpf_cfg *cfg)
	249	{
	250	if (cfg->is_ebpf)
	251	bpf_prog_put(cfg->filter);
	252	else
	253	bpf_prog_destroy(cfg->filter);
	254
	255	kfree(cfg->bpf_ops);
	256	kfree(cfg->bpf_name);
	257	}
	258
	259	static void tcf_bpf_prog_fill_cfg(const struct tcf_bpf *prog,
	260	struct tcf_bpf_cfg *cfg)
	261	{
	262	cfg->is_ebpf = tcf_bpf_is_ebpf(prog);
cff82457 AS	263	/* updates to prog->filter are prevented, since it's called either
	264	* with rtnl lock or during final cleanup in rcu callback
	265	*/
	266	cfg->filter = rcu_dereference_protected(prog->filter, 1);
f4eaed28 DB	267
	268	cfg->bpf_ops = prog->bpf_ops;
	269	cfg->bpf_name = prog->bpf_name;
	270	}
	271
a8cb5f55	272	static int tcf_bpf_init(struct net net, struct nlattr nla,
a85a970a	273	struct nlattr est, struct tc_action *act,
a8cb5f55 DB	274	int replace, int bind)
a8cb5f55 DB	275	{
ddf97ccd	276	struct tc_action_net *tn = net_generic(net, bpf_net_id);
a8cb5f55	277	struct nlattr *tb[TCA_ACT_BPF_MAX + 1];
f4eaed28	278	struct tcf_bpf_cfg cfg, old;
a8cb5f55 DB	279	struct tc_act_bpf *parm;
a8cb5f55 DB	280	struct tcf_bpf *prog;
a8cb5f55	281	bool is_bpf, is_ebpf;
a5c90b29	282	int ret, res = 0;
a8cb5f55 DB	283
	284	if (!nla)
	285	return -EINVAL;
	286
	287	ret = nla_parse_nested(tb, TCA_ACT_BPF_MAX, nla, act_bpf_policy);
	288	if (ret < 0)
	289	return ret;
	290
a5c90b29	291	if (!tb[TCA_ACT_BPF_PARMS])
a8cb5f55 DB	292	return -EINVAL;
	293
	294	parm = nla_data(tb[TCA_ACT_BPF_PARMS]);
	295
ddf97ccd WC	296	if (!tcf_hash_check(tn, parm->index, act, bind)) {
ddf97ccd WC	297	ret = tcf_hash_create(tn, parm->index, est, act,
a85a970a	298	&act_bpf_ops, bind, true);
a8cb5f55	299	if (ret < 0)
a5c90b29	300	return ret;
d23b8ad8	301
a5c90b29	302	res = ACT_P_CREATED;
d23b8ad8	303	} else {
a8cb5f55	304	/* Don't override defaults. */
d23b8ad8	305	if (bind)
a5c90b29	306	return 0;
a8cb5f55	307
a85a970a	308	tcf_hash_release(*act, bind);
a5c90b29 DB	309	if (!replace)
a5c90b29 DB	310	return -EEXIST;
d23b8ad8 JP	311	}
d23b8ad8 JP	312
a5c90b29 DB	313	is_bpf = tb[TCA_ACT_BPF_OPS_LEN] && tb[TCA_ACT_BPF_OPS];
	314	is_ebpf = tb[TCA_ACT_BPF_FD];
	315
	316	if ((!is_bpf && !is_ebpf) \|\| (is_bpf && is_ebpf)) {
	317	ret = -EINVAL;
	318	goto out;
	319	}
	320
	321	memset(&cfg, 0, sizeof(cfg));
	322
	323	ret = is_bpf ? tcf_bpf_init_from_ops(tb, &cfg) :
	324	tcf_bpf_init_from_efd(tb, &cfg);
	325	if (ret < 0)
	326	goto out;
	327
a85a970a	328	prog = to_bpf(*act);
cff82457	329	ASSERT_RTNL();
a8cb5f55	330
faa54be4	331	if (res != ACT_P_CREATED)
f4eaed28 DB	332	tcf_bpf_prog_fill_cfg(prog, &old);
f4eaed28 DB	333
a8cb5f55 DB	334	prog->bpf_ops = cfg.bpf_ops;
	335	prog->bpf_name = cfg.bpf_name;
	336
	337	if (cfg.bpf_num_ops)
	338	prog->bpf_num_ops = cfg.bpf_num_ops;
a8cb5f55 DB	339
a8cb5f55 DB	340	prog->tcf_action = parm->action;
cff82457	341	rcu_assign_pointer(prog->filter, cfg.filter);
d23b8ad8	342
cff82457	343	if (res == ACT_P_CREATED) {
a85a970a	344	tcf_hash_insert(tn, *act);
cff82457 AS	345	} else {
	346	/* make sure the program being replaced is no longer executing */
	347	synchronize_rcu();
f4eaed28	348	tcf_bpf_cfg_cleanup(&old);
cff82457	349	}
a8cb5f55	350
a5c90b29 DB	351	return res;
	352	out:
	353	if (res == ACT_P_CREATED)
a85a970a	354	tcf_hash_cleanup(*act, est);
d23b8ad8	355
d23b8ad8 JP	356	return ret;
	357	}
	358
a8cb5f55	359	static void tcf_bpf_cleanup(struct tc_action *act, int bind)
d23b8ad8	360	{
f4eaed28	361	struct tcf_bpf_cfg tmp;
ddf06c1e	362
a85a970a	363	tcf_bpf_prog_fill_cfg(to_bpf(act), &tmp);
f4eaed28	364	tcf_bpf_cfg_cleanup(&tmp);
d23b8ad8 JP	365	}
d23b8ad8 JP	366
ddf97ccd WC	367	static int tcf_bpf_walker(struct net net, struct sk_buff skb,
ddf97ccd WC	368	struct netlink_callback *cb, int type,
a85a970a	369	const struct tc_action_ops *ops)
ddf97ccd WC	370	{
	371	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	372
a85a970a	373	return tcf_generic_walker(tn, skb, cb, type, ops);
ddf97ccd WC	374	}
ddf97ccd WC	375
a85a970a	376	static int tcf_bpf_search(struct net net, struct tc_action *a, u32 index)
ddf97ccd WC	377	{
	378	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	379
	380	return tcf_hash_search(tn, a, index);
	381	}
	382
a8cb5f55 DB	383	static struct tc_action_ops act_bpf_ops __read_mostly = {
	384	.kind = "bpf",
	385	.type = TCA_ACT_BPF,
	386	.owner = THIS_MODULE,
	387	.act = tcf_bpf,
	388	.dump = tcf_bpf_dump,
	389	.cleanup = tcf_bpf_cleanup,
	390	.init = tcf_bpf_init,
ddf97ccd WC	391	.walk = tcf_bpf_walker,
ddf97ccd WC	392	.lookup = tcf_bpf_search,
a85a970a	393	.size = sizeof(struct tcf_bpf),
ddf97ccd WC	394	};
	395
	396	static __net_init int bpf_init_net(struct net *net)
	397	{
	398	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	399
	400	return tc_action_net_init(tn, &act_bpf_ops, BPF_TAB_MASK);
	401	}
	402
	403	static void __net_exit bpf_exit_net(struct net *net)
	404	{
	405	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	406
	407	tc_action_net_exit(tn);
	408	}
	409
	410	static struct pernet_operations bpf_net_ops = {
	411	.init = bpf_init_net,
	412	.exit = bpf_exit_net,
	413	.id = &bpf_net_id,
	414	.size = sizeof(struct tc_action_net),
d23b8ad8 JP	415	};
	416
	417	static int __init bpf_init_module(void)
	418	{
ddf97ccd	419	return tcf_register_action(&act_bpf_ops, &bpf_net_ops);
d23b8ad8 JP	420	}
	421
	422	static void __exit bpf_cleanup_module(void)
	423	{
ddf97ccd	424	tcf_unregister_action(&act_bpf_ops, &bpf_net_ops);
d23b8ad8 JP	425	}
	426
	427	module_init(bpf_init_module);
	428	module_exit(bpf_cleanup_module);
	429
	430	MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
	431	MODULE_DESCRIPTION("TC BPF based action");
	432	MODULE_LICENSE("GPL v2");