[linux-2.6-block.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;
	u32 bpf_fd;
	u16 bpf_num_ops;
	bool is_ebpf;
};

static int bpf_net_id;

static int tcf_bpf(struct sk_buff *skb, const struct tc_action *act,
		   struct tcf_result *res)
{
	struct tcf_bpf *prog = act->priv;
	struct bpf_prog *filter;
	int action, filter_res;
	bool at_ingress = G_TC_AT(skb->tc_verd) & AT_INGRESS;

	if (unlikely(!skb_mac_header_was_set(skb)))
		return TC_ACT_UNSPEC;

	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);
		filter_res = BPF_PROG_RUN(filter, skb);
		__skb_pull(skb, skb->mac_len);
	} else {
		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)
{
	if (nla_put_u32(skb, TCA_ACT_BPF_FD, prog->bpf_fd))
		return -EMSGSIZE;

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

	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 = act->priv;
	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;

	tm.install = jiffies_to_clock_t(jiffies - prog->tcf_tm.install);
	tm.lastuse = jiffies_to_clock_t(jiffies - prog->tcf_tm.lastuse);
	tm.expires = jiffies_to_clock_t(prog->tcf_tm.expires);

	if (nla_put(skb, TCA_ACT_BPF_TM, sizeof(tm), &tm))
		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(bpf_fd);
	if (IS_ERR(fp))
		return PTR_ERR(fp);

	if (fp->type != BPF_PROG_TYPE_SCHED_ACT) {
		bpf_prog_put(fp);
		return -EINVAL;
	}

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

	cfg->bpf_fd = bpf_fd;
	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,
				      sizeof(*prog), 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;
	if (cfg.bpf_fd)
		prog->bpf_fd = cfg.bpf_fd;

	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(act->priv, &tmp);
	tcf_bpf_cfg_cleanup(&tmp);
}

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

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

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,
};

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 DB	31	u32 bpf_fd;
a8cb5f55 DB	32	u16 bpf_num_ops;
f4eaed28	33	bool is_ebpf;
a8cb5f55	34	};
d23b8ad8	35
ddf97ccd WC	36	static int bpf_net_id;
ddf97ccd WC	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	{
a8cb5f55	41	struct tcf_bpf *prog = act->priv;
cff82457	42	struct bpf_prog *filter;
ced585c8	43	int action, filter_res;
3431205e	44	bool at_ingress = G_TC_AT(skb->tc_verd) & AT_INGRESS;
d23b8ad8	45
a166151c AS	46	if (unlikely(!skb_mac_header_was_set(skb)))
	47	return TC_ACT_UNSPEC;
	48
cff82457 AS	49	tcf_lastuse_update(&prog->tcf_tm);
cff82457 AS	50	bstats_cpu_update(this_cpu_ptr(prog->common.cpu_bstats), skb);
d23b8ad8	51
a8cb5f55	52	rcu_read_lock();
cff82457	53	filter = rcu_dereference(prog->filter);
3431205e AS	54	if (at_ingress) {
3431205e AS	55	__skb_push(skb, skb->mac_len);
cff82457	56	filter_res = BPF_PROG_RUN(filter, skb);
3431205e AS	57	__skb_pull(skb, skb->mac_len);
3431205e AS	58	} else {
cff82457	59	filter_res = BPF_PROG_RUN(filter, skb);
3431205e	60	}
a8cb5f55	61	rcu_read_unlock();
ced585c8 DB	62
	63	/* A BPF program may overwrite the default action opcode.
	64	* Similarly as in cls_bpf, if filter_res == -1 we use the
	65	* default action specified from tc.
	66	*
	67	* In case a different well-known TC_ACT opcode has been
	68	* returned, it will overwrite the default one.
	69	*
	70	* For everything else that is unkown, TC_ACT_UNSPEC is
	71	* returned.
	72	*/
	73	switch (filter_res) {
	74	case TC_ACT_PIPE:
	75	case TC_ACT_RECLASSIFY:
	76	case TC_ACT_OK:
27b29f63	77	case TC_ACT_REDIRECT:
ced585c8 DB	78	action = filter_res;
	79	break;
	80	case TC_ACT_SHOT:
	81	action = filter_res;
cff82457	82	qstats_drop_inc(this_cpu_ptr(prog->common.cpu_qstats));
ced585c8 DB	83	break;
ced585c8 DB	84	case TC_ACT_UNSPEC:
a8cb5f55	85	action = prog->tcf_action;
ced585c8 DB	86	break;
	87	default:
	88	action = TC_ACT_UNSPEC;
	89	break;
d23b8ad8 JP	90	}
d23b8ad8 JP	91
d23b8ad8 JP	92	return action;
	93	}
	94
a8cb5f55 DB	95	static bool tcf_bpf_is_ebpf(const struct tcf_bpf *prog)
	96	{
	97	return !prog->bpf_ops;
	98	}
	99
	100	static int tcf_bpf_dump_bpf_info(const struct tcf_bpf *prog,
	101	struct sk_buff *skb)
	102	{
	103	struct nlattr *nla;
	104
	105	if (nla_put_u16(skb, TCA_ACT_BPF_OPS_LEN, prog->bpf_num_ops))
	106	return -EMSGSIZE;
	107
	108	nla = nla_reserve(skb, TCA_ACT_BPF_OPS, prog->bpf_num_ops *
	109	sizeof(struct sock_filter));
	110	if (nla == NULL)
	111	return -EMSGSIZE;
	112
	113	memcpy(nla_data(nla), prog->bpf_ops, nla_len(nla));
	114
	115	return 0;
	116	}
	117
	118	static int tcf_bpf_dump_ebpf_info(const struct tcf_bpf *prog,
	119	struct sk_buff *skb)
	120	{
	121	if (nla_put_u32(skb, TCA_ACT_BPF_FD, prog->bpf_fd))
	122	return -EMSGSIZE;
	123
	124	if (prog->bpf_name &&
	125	nla_put_string(skb, TCA_ACT_BPF_NAME, prog->bpf_name))
	126	return -EMSGSIZE;
	127
	128	return 0;
	129	}
	130
	131	static int tcf_bpf_dump(struct sk_buff skb, struct tc_action act,
d23b8ad8 JP	132	int bind, int ref)
	133	{
	134	unsigned char *tp = skb_tail_pointer(skb);
a8cb5f55	135	struct tcf_bpf *prog = act->priv;
d23b8ad8	136	struct tc_act_bpf opt = {
a8cb5f55 DB	137	.index = prog->tcf_index,
	138	.refcnt = prog->tcf_refcnt - ref,
	139	.bindcnt = prog->tcf_bindcnt - bind,
	140	.action = prog->tcf_action,
d23b8ad8	141	};
a8cb5f55 DB	142	struct tcf_t tm;
a8cb5f55 DB	143	int ret;
d23b8ad8 JP	144
	145	if (nla_put(skb, TCA_ACT_BPF_PARMS, sizeof(opt), &opt))
	146	goto nla_put_failure;
	147
a8cb5f55 DB	148	if (tcf_bpf_is_ebpf(prog))
	149	ret = tcf_bpf_dump_ebpf_info(prog, skb);
	150	else
	151	ret = tcf_bpf_dump_bpf_info(prog, skb);
	152	if (ret)
d23b8ad8 JP	153	goto nla_put_failure;
d23b8ad8 JP	154
a8cb5f55 DB	155	tm.install = jiffies_to_clock_t(jiffies - prog->tcf_tm.install);
	156	tm.lastuse = jiffies_to_clock_t(jiffies - prog->tcf_tm.lastuse);
	157	tm.expires = jiffies_to_clock_t(prog->tcf_tm.expires);
d23b8ad8	158
a8cb5f55	159	if (nla_put(skb, TCA_ACT_BPF_TM, sizeof(tm), &tm))
d23b8ad8	160	goto nla_put_failure;
a8cb5f55	161
d23b8ad8 JP	162	return skb->len;
	163
	164	nla_put_failure:
	165	nlmsg_trim(skb, tp);
	166	return -1;
	167	}
	168
	169	static const struct nla_policy act_bpf_policy[TCA_ACT_BPF_MAX + 1] = {
	170	[TCA_ACT_BPF_PARMS] = { .len = sizeof(struct tc_act_bpf) },
a8cb5f55 DB	171	[TCA_ACT_BPF_FD] = { .type = NLA_U32 },
a8cb5f55 DB	172	[TCA_ACT_BPF_NAME] = { .type = NLA_NUL_STRING, .len = ACT_BPF_NAME_LEN },
d23b8ad8 JP	173	[TCA_ACT_BPF_OPS_LEN] = { .type = NLA_U16 },
	174	[TCA_ACT_BPF_OPS] = { .type = NLA_BINARY,
	175	.len = sizeof(struct sock_filter) * BPF_MAXINSNS },
	176	};
	177
a8cb5f55	178	static int tcf_bpf_init_from_ops(struct nlattr *tb, struct tcf_bpf_cfg cfg)
d23b8ad8	179	{
d23b8ad8	180	struct sock_filter *bpf_ops;
a8cb5f55	181	struct sock_fprog_kern fprog_tmp;
d23b8ad8	182	struct bpf_prog *fp;
a8cb5f55	183	u16 bpf_size, bpf_num_ops;
d23b8ad8 JP	184	int ret;
d23b8ad8 JP	185
d23b8ad8 JP	186	bpf_num_ops = nla_get_u16(tb[TCA_ACT_BPF_OPS_LEN]);
	187	if (bpf_num_ops > BPF_MAXINSNS \|\| bpf_num_ops == 0)
	188	return -EINVAL;
	189
	190	bpf_size = bpf_num_ops * sizeof(*bpf_ops);
fd3e646c DB	191	if (bpf_size != nla_len(tb[TCA_ACT_BPF_OPS]))
	192	return -EINVAL;
	193
d23b8ad8	194	bpf_ops = kzalloc(bpf_size, GFP_KERNEL);
a8cb5f55	195	if (bpf_ops == NULL)
d23b8ad8 JP	196	return -ENOMEM;
	197
	198	memcpy(bpf_ops, nla_data(tb[TCA_ACT_BPF_OPS]), bpf_size);
	199
a8cb5f55 DB	200	fprog_tmp.len = bpf_num_ops;
a8cb5f55 DB	201	fprog_tmp.filter = bpf_ops;
d23b8ad8	202
a8cb5f55 DB	203	ret = bpf_prog_create(&fp, &fprog_tmp);
	204	if (ret < 0) {
	205	kfree(bpf_ops);
	206	return ret;
	207	}
d23b8ad8	208
a8cb5f55 DB	209	cfg->bpf_ops = bpf_ops;
	210	cfg->bpf_num_ops = bpf_num_ops;
	211	cfg->filter = fp;
f4eaed28	212	cfg->is_ebpf = false;
a8cb5f55 DB	213
	214	return 0;
	215	}
	216
	217	static int tcf_bpf_init_from_efd(struct nlattr *tb, struct tcf_bpf_cfg cfg)
	218	{
	219	struct bpf_prog *fp;
	220	char *name = NULL;
	221	u32 bpf_fd;
	222
	223	bpf_fd = nla_get_u32(tb[TCA_ACT_BPF_FD]);
	224
	225	fp = bpf_prog_get(bpf_fd);
	226	if (IS_ERR(fp))
	227	return PTR_ERR(fp);
	228
	229	if (fp->type != BPF_PROG_TYPE_SCHED_ACT) {
	230	bpf_prog_put(fp);
	231	return -EINVAL;
	232	}
	233
	234	if (tb[TCA_ACT_BPF_NAME]) {
	235	name = kmemdup(nla_data(tb[TCA_ACT_BPF_NAME]),
	236	nla_len(tb[TCA_ACT_BPF_NAME]),
	237	GFP_KERNEL);
	238	if (!name) {
	239	bpf_prog_put(fp);
	240	return -ENOMEM;
	241	}
	242	}
	243
	244	cfg->bpf_fd = bpf_fd;
	245	cfg->bpf_name = name;
	246	cfg->filter = fp;
f4eaed28	247	cfg->is_ebpf = true;
a8cb5f55 DB	248
	249	return 0;
	250	}
	251
f4eaed28 DB	252	static void tcf_bpf_cfg_cleanup(const struct tcf_bpf_cfg *cfg)
	253	{
	254	if (cfg->is_ebpf)
	255	bpf_prog_put(cfg->filter);
	256	else
	257	bpf_prog_destroy(cfg->filter);
	258
	259	kfree(cfg->bpf_ops);
	260	kfree(cfg->bpf_name);
	261	}
	262
	263	static void tcf_bpf_prog_fill_cfg(const struct tcf_bpf *prog,
	264	struct tcf_bpf_cfg *cfg)
	265	{
	266	cfg->is_ebpf = tcf_bpf_is_ebpf(prog);
cff82457 AS	267	/* updates to prog->filter are prevented, since it's called either
	268	* with rtnl lock or during final cleanup in rcu callback
	269	*/
	270	cfg->filter = rcu_dereference_protected(prog->filter, 1);
f4eaed28 DB	271
	272	cfg->bpf_ops = prog->bpf_ops;
	273	cfg->bpf_name = prog->bpf_name;
	274	}
	275
a8cb5f55 DB	276	static int tcf_bpf_init(struct net net, struct nlattr nla,
	277	struct nlattr est, struct tc_action act,
	278	int replace, int bind)
	279	{
ddf97ccd	280	struct tc_action_net *tn = net_generic(net, bpf_net_id);
a8cb5f55	281	struct nlattr *tb[TCA_ACT_BPF_MAX + 1];
f4eaed28	282	struct tcf_bpf_cfg cfg, old;
a8cb5f55 DB	283	struct tc_act_bpf *parm;
a8cb5f55 DB	284	struct tcf_bpf *prog;
a8cb5f55	285	bool is_bpf, is_ebpf;
a5c90b29	286	int ret, res = 0;
a8cb5f55 DB	287
	288	if (!nla)
	289	return -EINVAL;
	290
	291	ret = nla_parse_nested(tb, TCA_ACT_BPF_MAX, nla, act_bpf_policy);
	292	if (ret < 0)
	293	return ret;
	294
a5c90b29	295	if (!tb[TCA_ACT_BPF_PARMS])
a8cb5f55 DB	296	return -EINVAL;
	297
	298	parm = nla_data(tb[TCA_ACT_BPF_PARMS]);
	299
ddf97ccd WC	300	if (!tcf_hash_check(tn, parm->index, act, bind)) {
ddf97ccd WC	301	ret = tcf_hash_create(tn, parm->index, est, act,
cff82457	302	sizeof(*prog), bind, true);
a8cb5f55	303	if (ret < 0)
a5c90b29	304	return ret;
d23b8ad8	305
a5c90b29	306	res = ACT_P_CREATED;
d23b8ad8	307	} else {
a8cb5f55	308	/* Don't override defaults. */
d23b8ad8	309	if (bind)
a5c90b29	310	return 0;
a8cb5f55 DB	311
a8cb5f55 DB	312	tcf_hash_release(act, bind);
a5c90b29 DB	313	if (!replace)
a5c90b29 DB	314	return -EEXIST;
d23b8ad8 JP	315	}
d23b8ad8 JP	316
a5c90b29 DB	317	is_bpf = tb[TCA_ACT_BPF_OPS_LEN] && tb[TCA_ACT_BPF_OPS];
	318	is_ebpf = tb[TCA_ACT_BPF_FD];
	319
	320	if ((!is_bpf && !is_ebpf) \|\| (is_bpf && is_ebpf)) {
	321	ret = -EINVAL;
	322	goto out;
	323	}
	324
	325	memset(&cfg, 0, sizeof(cfg));
	326
	327	ret = is_bpf ? tcf_bpf_init_from_ops(tb, &cfg) :
	328	tcf_bpf_init_from_efd(tb, &cfg);
	329	if (ret < 0)
	330	goto out;
	331
a8cb5f55	332	prog = to_bpf(act);
cff82457	333	ASSERT_RTNL();
a8cb5f55	334
faa54be4	335	if (res != ACT_P_CREATED)
f4eaed28 DB	336	tcf_bpf_prog_fill_cfg(prog, &old);
f4eaed28 DB	337
a8cb5f55 DB	338	prog->bpf_ops = cfg.bpf_ops;
	339	prog->bpf_name = cfg.bpf_name;
	340
	341	if (cfg.bpf_num_ops)
	342	prog->bpf_num_ops = cfg.bpf_num_ops;
	343	if (cfg.bpf_fd)
	344	prog->bpf_fd = cfg.bpf_fd;
	345
	346	prog->tcf_action = parm->action;
cff82457	347	rcu_assign_pointer(prog->filter, cfg.filter);
d23b8ad8	348
cff82457	349	if (res == ACT_P_CREATED) {
ddf97ccd	350	tcf_hash_insert(tn, act);
cff82457 AS	351	} else {
	352	/* make sure the program being replaced is no longer executing */
	353	synchronize_rcu();
f4eaed28	354	tcf_bpf_cfg_cleanup(&old);
cff82457	355	}
a8cb5f55	356
a5c90b29 DB	357	return res;
	358	out:
	359	if (res == ACT_P_CREATED)
	360	tcf_hash_cleanup(act, est);
d23b8ad8	361
d23b8ad8 JP	362	return ret;
	363	}
	364
a8cb5f55	365	static void tcf_bpf_cleanup(struct tc_action *act, int bind)
d23b8ad8	366	{
f4eaed28	367	struct tcf_bpf_cfg tmp;
ddf06c1e	368
f4eaed28 DB	369	tcf_bpf_prog_fill_cfg(act->priv, &tmp);
f4eaed28 DB	370	tcf_bpf_cfg_cleanup(&tmp);
d23b8ad8 JP	371	}
d23b8ad8 JP	372
ddf97ccd WC	373	static int tcf_bpf_walker(struct net net, struct sk_buff skb,
	374	struct netlink_callback *cb, int type,
	375	struct tc_action *a)
	376	{
	377	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	378
	379	return tcf_generic_walker(tn, skb, cb, type, a);
	380	}
	381
	382	static int tcf_bpf_search(struct net net, struct tc_action a, u32 index)
	383	{
	384	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	385
	386	return tcf_hash_search(tn, a, index);
	387	}
	388
a8cb5f55 DB	389	static struct tc_action_ops act_bpf_ops __read_mostly = {
	390	.kind = "bpf",
	391	.type = TCA_ACT_BPF,
	392	.owner = THIS_MODULE,
	393	.act = tcf_bpf,
	394	.dump = tcf_bpf_dump,
	395	.cleanup = tcf_bpf_cleanup,
	396	.init = tcf_bpf_init,
ddf97ccd WC	397	.walk = tcf_bpf_walker,
	398	.lookup = tcf_bpf_search,
	399	};
	400
	401	static __net_init int bpf_init_net(struct net *net)
	402	{
	403	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	404
	405	return tc_action_net_init(tn, &act_bpf_ops, BPF_TAB_MASK);
	406	}
	407
	408	static void __net_exit bpf_exit_net(struct net *net)
	409	{
	410	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	411
	412	tc_action_net_exit(tn);
	413	}
	414
	415	static struct pernet_operations bpf_net_ops = {
	416	.init = bpf_init_net,
	417	.exit = bpf_exit_net,
	418	.id = &bpf_net_id,
	419	.size = sizeof(struct tc_action_net),
d23b8ad8 JP	420	};
	421
	422	static int __init bpf_init_module(void)
	423	{
ddf97ccd	424	return tcf_register_action(&act_bpf_ops, &bpf_net_ops);
d23b8ad8 JP	425	}
	426
	427	static void __exit bpf_cleanup_module(void)
	428	{
ddf97ccd	429	tcf_unregister_action(&act_bpf_ops, &bpf_net_ops);
d23b8ad8 JP	430	}
	431
	432	module_init(bpf_init_module);
	433	module_exit(bpf_cleanup_module);
	434
	435	MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
	436	MODULE_DESCRIPTION("TC BPF based action");
	437	MODULE_LICENSE("GPL v2");