[deliverable/linux.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 struct tc_action_ops act_bpf_ops;

static int tcf_bpf(struct sk_buff *skb, const struct tc_action *act,
		   struct tcf_result *res)
{
	struct tcf_bpf *prog = to_bpf(act);
	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);
		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)
{
	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 = 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 = 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,
				      &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;
	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(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 DB	31	u32 bpf_fd;
a8cb5f55 DB	32	u16 bpf_num_ops;
f4eaed28	33	bool is_ebpf;
a8cb5f55	34	};
d23b8ad8	35
ddf97ccd	36	static int bpf_net_id;
a85a970a	37	static struct tc_action_ops act_bpf_ops;
ddf97ccd	38
a8cb5f55	39	static int tcf_bpf(struct sk_buff skb, const struct tc_action act,
d23b8ad8 JP	40	struct tcf_result *res)
d23b8ad8 JP	41	{
a85a970a	42	struct tcf_bpf *prog = to_bpf(act);
cff82457	43	struct bpf_prog *filter;
ced585c8	44	int action, filter_res;
3431205e	45	bool at_ingress = G_TC_AT(skb->tc_verd) & AT_INGRESS;
d23b8ad8	46
a166151c AS	47	if (unlikely(!skb_mac_header_was_set(skb)))
	48	return TC_ACT_UNSPEC;
	49
cff82457 AS	50	tcf_lastuse_update(&prog->tcf_tm);
cff82457 AS	51	bstats_cpu_update(this_cpu_ptr(prog->common.cpu_bstats), skb);
d23b8ad8	52
a8cb5f55	53	rcu_read_lock();
cff82457	54	filter = rcu_dereference(prog->filter);
3431205e AS	55	if (at_ingress) {
3431205e AS	56	__skb_push(skb, skb->mac_len);
db58ba45	57	bpf_compute_data_end(skb);
cff82457	58	filter_res = BPF_PROG_RUN(filter, skb);
3431205e AS	59	__skb_pull(skb, skb->mac_len);
3431205e AS	60	} else {
db58ba45	61	bpf_compute_data_end(skb);
cff82457	62	filter_res = BPF_PROG_RUN(filter, skb);
3431205e	63	}
a8cb5f55	64	rcu_read_unlock();
ced585c8 DB	65
	66	/* A BPF program may overwrite the default action opcode.
	67	* Similarly as in cls_bpf, if filter_res == -1 we use the
	68	* default action specified from tc.
	69	*
	70	* In case a different well-known TC_ACT opcode has been
	71	* returned, it will overwrite the default one.
	72	*
	73	* For everything else that is unkown, TC_ACT_UNSPEC is
	74	* returned.
	75	*/
	76	switch (filter_res) {
	77	case TC_ACT_PIPE:
	78	case TC_ACT_RECLASSIFY:
	79	case TC_ACT_OK:
27b29f63	80	case TC_ACT_REDIRECT:
ced585c8 DB	81	action = filter_res;
	82	break;
	83	case TC_ACT_SHOT:
	84	action = filter_res;
cff82457	85	qstats_drop_inc(this_cpu_ptr(prog->common.cpu_qstats));
ced585c8 DB	86	break;
ced585c8 DB	87	case TC_ACT_UNSPEC:
a8cb5f55	88	action = prog->tcf_action;
ced585c8 DB	89	break;
	90	default:
	91	action = TC_ACT_UNSPEC;
	92	break;
d23b8ad8 JP	93	}
d23b8ad8 JP	94
d23b8ad8 JP	95	return action;
	96	}
	97
a8cb5f55 DB	98	static bool tcf_bpf_is_ebpf(const struct tcf_bpf *prog)
	99	{
	100	return !prog->bpf_ops;
	101	}
	102
	103	static int tcf_bpf_dump_bpf_info(const struct tcf_bpf *prog,
	104	struct sk_buff *skb)
	105	{
	106	struct nlattr *nla;
	107
	108	if (nla_put_u16(skb, TCA_ACT_BPF_OPS_LEN, prog->bpf_num_ops))
	109	return -EMSGSIZE;
	110
	111	nla = nla_reserve(skb, TCA_ACT_BPF_OPS, prog->bpf_num_ops *
	112	sizeof(struct sock_filter));
	113	if (nla == NULL)
	114	return -EMSGSIZE;
	115
	116	memcpy(nla_data(nla), prog->bpf_ops, nla_len(nla));
	117
	118	return 0;
	119	}
	120
	121	static int tcf_bpf_dump_ebpf_info(const struct tcf_bpf *prog,
	122	struct sk_buff *skb)
	123	{
	124	if (nla_put_u32(skb, TCA_ACT_BPF_FD, prog->bpf_fd))
	125	return -EMSGSIZE;
	126
	127	if (prog->bpf_name &&
	128	nla_put_string(skb, TCA_ACT_BPF_NAME, prog->bpf_name))
	129	return -EMSGSIZE;
	130
	131	return 0;
	132	}
	133
	134	static int tcf_bpf_dump(struct sk_buff skb, struct tc_action act,
d23b8ad8 JP	135	int bind, int ref)
	136	{
	137	unsigned char *tp = skb_tail_pointer(skb);
a85a970a	138	struct tcf_bpf *prog = to_bpf(act);
d23b8ad8	139	struct tc_act_bpf opt = {
a8cb5f55 DB	140	.index = prog->tcf_index,
	141	.refcnt = prog->tcf_refcnt - ref,
	142	.bindcnt = prog->tcf_bindcnt - bind,
	143	.action = prog->tcf_action,
d23b8ad8	144	};
a8cb5f55 DB	145	struct tcf_t tm;
a8cb5f55 DB	146	int ret;
d23b8ad8 JP	147
	148	if (nla_put(skb, TCA_ACT_BPF_PARMS, sizeof(opt), &opt))
	149	goto nla_put_failure;
	150
a8cb5f55 DB	151	if (tcf_bpf_is_ebpf(prog))
	152	ret = tcf_bpf_dump_ebpf_info(prog, skb);
	153	else
	154	ret = tcf_bpf_dump_bpf_info(prog, skb);
	155	if (ret)
d23b8ad8 JP	156	goto nla_put_failure;
d23b8ad8 JP	157
48d8ee16	158	tcf_tm_dump(&tm, &prog->tcf_tm);
9854518e ND	159	if (nla_put_64bit(skb, TCA_ACT_BPF_TM, sizeof(tm), &tm,
9854518e ND	160	TCA_ACT_BPF_PAD))
d23b8ad8	161	goto nla_put_failure;
a8cb5f55	162
d23b8ad8 JP	163	return skb->len;
	164
	165	nla_put_failure:
	166	nlmsg_trim(skb, tp);
	167	return -1;
	168	}
	169
	170	static const struct nla_policy act_bpf_policy[TCA_ACT_BPF_MAX + 1] = {
	171	[TCA_ACT_BPF_PARMS] = { .len = sizeof(struct tc_act_bpf) },
a8cb5f55	172	[TCA_ACT_BPF_FD] = { .type = NLA_U32 },
0b0f43fe JHS	173	[TCA_ACT_BPF_NAME] = { .type = NLA_NUL_STRING,
0b0f43fe JHS	174	.len = ACT_BPF_NAME_LEN },
d23b8ad8 JP	175	[TCA_ACT_BPF_OPS_LEN] = { .type = NLA_U16 },
	176	[TCA_ACT_BPF_OPS] = { .type = NLA_BINARY,
	177	.len = sizeof(struct sock_filter) * BPF_MAXINSNS },
	178	};
	179
a8cb5f55	180	static int tcf_bpf_init_from_ops(struct nlattr *tb, struct tcf_bpf_cfg cfg)
d23b8ad8	181	{
d23b8ad8	182	struct sock_filter *bpf_ops;
a8cb5f55	183	struct sock_fprog_kern fprog_tmp;
d23b8ad8	184	struct bpf_prog *fp;
a8cb5f55	185	u16 bpf_size, bpf_num_ops;
d23b8ad8 JP	186	int ret;
d23b8ad8 JP	187
d23b8ad8 JP	188	bpf_num_ops = nla_get_u16(tb[TCA_ACT_BPF_OPS_LEN]);
	189	if (bpf_num_ops > BPF_MAXINSNS \|\| bpf_num_ops == 0)
	190	return -EINVAL;
	191
	192	bpf_size = bpf_num_ops * sizeof(*bpf_ops);
fd3e646c DB	193	if (bpf_size != nla_len(tb[TCA_ACT_BPF_OPS]))
	194	return -EINVAL;
	195
d23b8ad8	196	bpf_ops = kzalloc(bpf_size, GFP_KERNEL);
a8cb5f55	197	if (bpf_ops == NULL)
d23b8ad8 JP	198	return -ENOMEM;
	199
	200	memcpy(bpf_ops, nla_data(tb[TCA_ACT_BPF_OPS]), bpf_size);
	201
a8cb5f55 DB	202	fprog_tmp.len = bpf_num_ops;
a8cb5f55 DB	203	fprog_tmp.filter = bpf_ops;
d23b8ad8	204
a8cb5f55 DB	205	ret = bpf_prog_create(&fp, &fprog_tmp);
	206	if (ret < 0) {
	207	kfree(bpf_ops);
	208	return ret;
	209	}
d23b8ad8	210
a8cb5f55 DB	211	cfg->bpf_ops = bpf_ops;
	212	cfg->bpf_num_ops = bpf_num_ops;
	213	cfg->filter = fp;
f4eaed28	214	cfg->is_ebpf = false;
a8cb5f55 DB	215
	216	return 0;
	217	}
	218
	219	static int tcf_bpf_init_from_efd(struct nlattr *tb, struct tcf_bpf_cfg cfg)
	220	{
	221	struct bpf_prog *fp;
	222	char *name = NULL;
	223	u32 bpf_fd;
	224
	225	bpf_fd = nla_get_u32(tb[TCA_ACT_BPF_FD]);
	226
113214be	227	fp = bpf_prog_get_type(bpf_fd, BPF_PROG_TYPE_SCHED_ACT);
a8cb5f55 DB	228	if (IS_ERR(fp))
	229	return PTR_ERR(fp);
	230
a8cb5f55 DB	231	if (tb[TCA_ACT_BPF_NAME]) {
	232	name = kmemdup(nla_data(tb[TCA_ACT_BPF_NAME]),
	233	nla_len(tb[TCA_ACT_BPF_NAME]),
	234	GFP_KERNEL);
	235	if (!name) {
	236	bpf_prog_put(fp);
	237	return -ENOMEM;
	238	}
	239	}
	240
	241	cfg->bpf_fd = bpf_fd;
	242	cfg->bpf_name = name;
	243	cfg->filter = fp;
f4eaed28	244	cfg->is_ebpf = true;
a8cb5f55 DB	245
	246	return 0;
	247	}
	248
f4eaed28 DB	249	static void tcf_bpf_cfg_cleanup(const struct tcf_bpf_cfg *cfg)
	250	{
	251	if (cfg->is_ebpf)
	252	bpf_prog_put(cfg->filter);
	253	else
	254	bpf_prog_destroy(cfg->filter);
	255
	256	kfree(cfg->bpf_ops);
	257	kfree(cfg->bpf_name);
	258	}
	259
	260	static void tcf_bpf_prog_fill_cfg(const struct tcf_bpf *prog,
	261	struct tcf_bpf_cfg *cfg)
	262	{
	263	cfg->is_ebpf = tcf_bpf_is_ebpf(prog);
cff82457 AS	264	/* updates to prog->filter are prevented, since it's called either
	265	* with rtnl lock or during final cleanup in rcu callback
	266	*/
	267	cfg->filter = rcu_dereference_protected(prog->filter, 1);
f4eaed28 DB	268
	269	cfg->bpf_ops = prog->bpf_ops;
	270	cfg->bpf_name = prog->bpf_name;
	271	}
	272
a8cb5f55	273	static int tcf_bpf_init(struct net net, struct nlattr nla,
a85a970a	274	struct nlattr est, struct tc_action *act,
a8cb5f55 DB	275	int replace, int bind)
a8cb5f55 DB	276	{
ddf97ccd	277	struct tc_action_net *tn = net_generic(net, bpf_net_id);
a8cb5f55	278	struct nlattr *tb[TCA_ACT_BPF_MAX + 1];
f4eaed28	279	struct tcf_bpf_cfg cfg, old;
a8cb5f55 DB	280	struct tc_act_bpf *parm;
a8cb5f55 DB	281	struct tcf_bpf *prog;
a8cb5f55	282	bool is_bpf, is_ebpf;
a5c90b29	283	int ret, res = 0;
a8cb5f55 DB	284
	285	if (!nla)
	286	return -EINVAL;
	287
	288	ret = nla_parse_nested(tb, TCA_ACT_BPF_MAX, nla, act_bpf_policy);
	289	if (ret < 0)
	290	return ret;
	291
a5c90b29	292	if (!tb[TCA_ACT_BPF_PARMS])
a8cb5f55 DB	293	return -EINVAL;
	294
	295	parm = nla_data(tb[TCA_ACT_BPF_PARMS]);
	296
ddf97ccd WC	297	if (!tcf_hash_check(tn, parm->index, act, bind)) {
ddf97ccd WC	298	ret = tcf_hash_create(tn, parm->index, est, act,
a85a970a	299	&act_bpf_ops, bind, true);
a8cb5f55	300	if (ret < 0)
a5c90b29	301	return ret;
d23b8ad8	302
a5c90b29	303	res = ACT_P_CREATED;
d23b8ad8	304	} else {
a8cb5f55	305	/* Don't override defaults. */
d23b8ad8	306	if (bind)
a5c90b29	307	return 0;
a8cb5f55	308
a85a970a	309	tcf_hash_release(*act, bind);
a5c90b29 DB	310	if (!replace)
a5c90b29 DB	311	return -EEXIST;
d23b8ad8 JP	312	}
d23b8ad8 JP	313
a5c90b29 DB	314	is_bpf = tb[TCA_ACT_BPF_OPS_LEN] && tb[TCA_ACT_BPF_OPS];
	315	is_ebpf = tb[TCA_ACT_BPF_FD];
	316
	317	if ((!is_bpf && !is_ebpf) \|\| (is_bpf && is_ebpf)) {
	318	ret = -EINVAL;
	319	goto out;
	320	}
	321
	322	memset(&cfg, 0, sizeof(cfg));
	323
	324	ret = is_bpf ? tcf_bpf_init_from_ops(tb, &cfg) :
	325	tcf_bpf_init_from_efd(tb, &cfg);
	326	if (ret < 0)
	327	goto out;
	328
a85a970a	329	prog = to_bpf(*act);
cff82457	330	ASSERT_RTNL();
a8cb5f55	331
faa54be4	332	if (res != ACT_P_CREATED)
f4eaed28 DB	333	tcf_bpf_prog_fill_cfg(prog, &old);
f4eaed28 DB	334
a8cb5f55 DB	335	prog->bpf_ops = cfg.bpf_ops;
	336	prog->bpf_name = cfg.bpf_name;
	337
	338	if (cfg.bpf_num_ops)
	339	prog->bpf_num_ops = cfg.bpf_num_ops;
	340	if (cfg.bpf_fd)
	341	prog->bpf_fd = cfg.bpf_fd;
	342
	343	prog->tcf_action = parm->action;
cff82457	344	rcu_assign_pointer(prog->filter, cfg.filter);
d23b8ad8	345
cff82457	346	if (res == ACT_P_CREATED) {
a85a970a	347	tcf_hash_insert(tn, *act);
cff82457 AS	348	} else {
	349	/* make sure the program being replaced is no longer executing */
	350	synchronize_rcu();
f4eaed28	351	tcf_bpf_cfg_cleanup(&old);
cff82457	352	}
a8cb5f55	353
a5c90b29 DB	354	return res;
	355	out:
	356	if (res == ACT_P_CREATED)
a85a970a	357	tcf_hash_cleanup(*act, est);
d23b8ad8	358
d23b8ad8 JP	359	return ret;
	360	}
	361
a8cb5f55	362	static void tcf_bpf_cleanup(struct tc_action *act, int bind)
d23b8ad8	363	{
f4eaed28	364	struct tcf_bpf_cfg tmp;
ddf06c1e	365
a85a970a	366	tcf_bpf_prog_fill_cfg(to_bpf(act), &tmp);
f4eaed28	367	tcf_bpf_cfg_cleanup(&tmp);
d23b8ad8 JP	368	}
d23b8ad8 JP	369
ddf97ccd WC	370	static int tcf_bpf_walker(struct net net, struct sk_buff skb,
ddf97ccd WC	371	struct netlink_callback *cb, int type,
a85a970a	372	const struct tc_action_ops *ops)
ddf97ccd WC	373	{
	374	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	375
a85a970a	376	return tcf_generic_walker(tn, skb, cb, type, ops);
ddf97ccd WC	377	}
ddf97ccd WC	378
a85a970a	379	static int tcf_bpf_search(struct net net, struct tc_action *a, u32 index)
ddf97ccd WC	380	{
	381	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	382
	383	return tcf_hash_search(tn, a, index);
	384	}
	385
a8cb5f55 DB	386	static struct tc_action_ops act_bpf_ops __read_mostly = {
	387	.kind = "bpf",
	388	.type = TCA_ACT_BPF,
	389	.owner = THIS_MODULE,
	390	.act = tcf_bpf,
	391	.dump = tcf_bpf_dump,
	392	.cleanup = tcf_bpf_cleanup,
	393	.init = tcf_bpf_init,
ddf97ccd WC	394	.walk = tcf_bpf_walker,
ddf97ccd WC	395	.lookup = tcf_bpf_search,
a85a970a	396	.size = sizeof(struct tcf_bpf),
ddf97ccd WC	397	};
	398
	399	static __net_init int bpf_init_net(struct net *net)
	400	{
	401	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	402
	403	return tc_action_net_init(tn, &act_bpf_ops, BPF_TAB_MASK);
	404	}
	405
	406	static void __net_exit bpf_exit_net(struct net *net)
	407	{
	408	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	409
	410	tc_action_net_exit(tn);
	411	}
	412
	413	static struct pernet_operations bpf_net_ops = {
	414	.init = bpf_init_net,
	415	.exit = bpf_exit_net,
	416	.id = &bpf_net_id,
	417	.size = sizeof(struct tc_action_net),
d23b8ad8 JP	418	};
	419
	420	static int __init bpf_init_module(void)
	421	{
ddf97ccd	422	return tcf_register_action(&act_bpf_ops, &bpf_net_ops);
d23b8ad8 JP	423	}
	424
	425	static void __exit bpf_cleanup_module(void)
	426	{
ddf97ccd	427	tcf_unregister_action(&act_bpf_ops, &bpf_net_ops);
d23b8ad8 JP	428	}
	429
	430	module_init(bpf_init_module);
	431	module_exit(bpf_cleanup_module);
	432
	433	MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
	434	MODULE_DESCRIPTION("TC BPF based action");
	435	MODULE_LICENSE("GPL v2");