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

// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * Copyright (c) 2015 Jiri Pirko <jiri@resnulli.us>
 */

#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 <net/pkt_cls.h>

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

#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_act(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_pointers(skb);
		filter_res = BPF_PROG_RUN(filter, skb);
		__skb_pull(skb, skb->mac_len);
	} else {
		bpf_compute_data_pointers(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;

	if (nla_put_u32(skb, TCA_ACT_BPF_ID, prog->filter->aux->id))
		return -EMSGSIZE;

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

	memcpy(nla_data(nla), prog->filter->tag, 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  = refcount_read(&prog->tcf_refcnt) - ref,
		.bindcnt = atomic_read(&prog->tcf_bindcnt) - bind,
	};
	struct tcf_t tm;
	int ret;

	spin_lock_bh(&prog->tcf_lock);
	opt.action = prog->tcf_action;
	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;

	spin_unlock_bh(&prog->tcf_lock);
	return skb->len;

nla_put_failure:
	spin_unlock_bh(&prog->tcf_lock);
	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 = kmemdup(nla_data(tb[TCA_ACT_BPF_OPS]), bpf_size, GFP_KERNEL);
	if (bpf_ops == NULL)
		return -ENOMEM;

	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)
{
	struct bpf_prog *filter = cfg->filter;

	if (filter) {
		if (cfg->is_ebpf)
			bpf_prog_put(filter);
		else
			bpf_prog_destroy(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 tcf 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, bool rtnl_held,
			struct tcf_proto *tp, u32 flags,
			struct netlink_ext_ack *extack)
{
	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	struct nlattr *tb[TCA_ACT_BPF_MAX + 1];
	struct tcf_chain *goto_ch = NULL;
	struct tcf_bpf_cfg cfg, old;
	struct tc_act_bpf *parm;
	struct tcf_bpf *prog;
	bool is_bpf, is_ebpf;
	int ret, res = 0;
	u32 index;

	if (!nla)
		return -EINVAL;

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

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

	parm = nla_data(tb[TCA_ACT_BPF_PARMS]);
	index = parm->index;
	ret = tcf_idr_check_alloc(tn, &index, act, bind);
	if (!ret) {
		ret = tcf_idr_create(tn, index, est, act,
				     &act_bpf_ops, bind, true, 0);
		if (ret < 0) {
			tcf_idr_cleanup(tn, index);
			return ret;
		}

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

		if (!replace) {
			tcf_idr_release(*act, bind);
			return -EEXIST;
		}
	} else {
		return ret;
	}

	ret = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
	if (ret < 0)
		goto release_idr;

	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 put_chain;
	}

	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 put_chain;

	prog = to_bpf(*act);

	spin_lock_bh(&prog->tcf_lock);
	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;

	goto_ch = tcf_action_set_ctrlact(*act, parm->action, goto_ch);
	rcu_assign_pointer(prog->filter, cfg.filter);
	spin_unlock_bh(&prog->tcf_lock);

	if (goto_ch)
		tcf_chain_put_by_act(goto_ch);

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

	return res;

put_chain:
	if (goto_ch)
		tcf_chain_put_by_act(goto_ch);

release_idr:
	tcf_idr_release(*act, bind);
	return ret;
}

static void tcf_bpf_cleanup(struct tc_action *act)
{
	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 netlink_ext_ack *extack)
{
	struct tc_action_net *tn = net_generic(net, bpf_net_id);

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

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_idr_search(tn, a, index);
}

static struct tc_action_ops act_bpf_ops __read_mostly = {
	.kind		=	"bpf",
	.id		=	TCA_ID_BPF,
	.owner		=	THIS_MODULE,
	.act		=	tcf_bpf_act,
	.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(net, tn, &act_bpf_ops);
}

static void __net_exit bpf_exit_net(struct list_head *net_list)
{
	tc_action_net_exit(net_list, bpf_net_id);
}

static struct pernet_operations bpf_net_ops = {
	.init = bpf_init_net,
	.exit_batch = 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
2874c5fd	1	// SPDX-License-Identifier: GPL-2.0-or-later
d23b8ad8 JP	2	/*
d23b8ad8 JP	3	* Copyright (c) 2015 Jiri Pirko <jiri@resnulli.us>
d23b8ad8 JP	4	*/
	5
	6	#include <linux/module.h>
	7	#include <linux/init.h>
	8	#include <linux/kernel.h>
	9	#include <linux/skbuff.h>
	10	#include <linux/rtnetlink.h>
	11	#include <linux/filter.h>
a8cb5f55 DB	12	#include <linux/bpf.h>
a8cb5f55 DB	13
d23b8ad8 JP	14	#include <net/netlink.h>
d23b8ad8 JP	15	#include <net/pkt_sched.h>
4e181004	16	#include <net/pkt_cls.h>
d23b8ad8 JP	17
	18	#include <linux/tc_act/tc_bpf.h>
	19	#include <net/tc_act/tc_bpf.h>
	20
a8cb5f55 DB	21	#define ACT_BPF_NAME_LEN 256
	22
	23	struct tcf_bpf_cfg {
	24	struct bpf_prog *filter;
	25	struct sock_filter *bpf_ops;
f4eaed28	26	const char *bpf_name;
a8cb5f55	27	u16 bpf_num_ops;
f4eaed28	28	bool is_ebpf;
a8cb5f55	29	};
d23b8ad8	30
c7d03a00	31	static unsigned int bpf_net_id;
a85a970a	32	static struct tc_action_ops act_bpf_ops;
ddf97ccd	33
2fbec27f JHS	34	static int tcf_bpf_act(struct sk_buff skb, const struct tc_action act,
2fbec27f JHS	35	struct tcf_result *res)
d23b8ad8	36	{
f53d8c7b	37	bool at_ingress = skb_at_tc_ingress(skb);
a85a970a	38	struct tcf_bpf *prog = to_bpf(act);
cff82457	39	struct bpf_prog *filter;
ced585c8	40	int action, filter_res;
d23b8ad8	41
cff82457 AS	42	tcf_lastuse_update(&prog->tcf_tm);
cff82457 AS	43	bstats_cpu_update(this_cpu_ptr(prog->common.cpu_bstats), skb);
d23b8ad8	44
a8cb5f55	45	rcu_read_lock();
cff82457	46	filter = rcu_dereference(prog->filter);
3431205e AS	47	if (at_ingress) {
3431205e AS	48	__skb_push(skb, skb->mac_len);
6aaae2b6	49	bpf_compute_data_pointers(skb);
cff82457	50	filter_res = BPF_PROG_RUN(filter, skb);
3431205e AS	51	__skb_pull(skb, skb->mac_len);
3431205e AS	52	} else {
6aaae2b6	53	bpf_compute_data_pointers(skb);
cff82457	54	filter_res = BPF_PROG_RUN(filter, skb);
3431205e	55	}
a8cb5f55	56	rcu_read_unlock();
ced585c8 DB	57
	58	/* A BPF program may overwrite the default action opcode.
	59	* Similarly as in cls_bpf, if filter_res == -1 we use the
	60	* default action specified from tc.
	61	*
	62	* In case a different well-known TC_ACT opcode has been
	63	* returned, it will overwrite the default one.
	64	*
	65	* For everything else that is unkown, TC_ACT_UNSPEC is
	66	* returned.
	67	*/
	68	switch (filter_res) {
	69	case TC_ACT_PIPE:
	70	case TC_ACT_RECLASSIFY:
	71	case TC_ACT_OK:
27b29f63	72	case TC_ACT_REDIRECT:
ced585c8 DB	73	action = filter_res;
	74	break;
	75	case TC_ACT_SHOT:
	76	action = filter_res;
cff82457	77	qstats_drop_inc(this_cpu_ptr(prog->common.cpu_qstats));
ced585c8 DB	78	break;
ced585c8 DB	79	case TC_ACT_UNSPEC:
a8cb5f55	80	action = prog->tcf_action;
ced585c8 DB	81	break;
	82	default:
	83	action = TC_ACT_UNSPEC;
	84	break;
d23b8ad8 JP	85	}
d23b8ad8 JP	86
d23b8ad8 JP	87	return action;
	88	}
	89
a8cb5f55 DB	90	static bool tcf_bpf_is_ebpf(const struct tcf_bpf *prog)
	91	{
	92	return !prog->bpf_ops;
	93	}
	94
	95	static int tcf_bpf_dump_bpf_info(const struct tcf_bpf *prog,
	96	struct sk_buff *skb)
	97	{
	98	struct nlattr *nla;
	99
	100	if (nla_put_u16(skb, TCA_ACT_BPF_OPS_LEN, prog->bpf_num_ops))
	101	return -EMSGSIZE;
	102
	103	nla = nla_reserve(skb, TCA_ACT_BPF_OPS, prog->bpf_num_ops *
	104	sizeof(struct sock_filter));
	105	if (nla == NULL)
	106	return -EMSGSIZE;
	107
	108	memcpy(nla_data(nla), prog->bpf_ops, nla_len(nla));
	109
	110	return 0;
	111	}
	112
	113	static int tcf_bpf_dump_ebpf_info(const struct tcf_bpf *prog,
	114	struct sk_buff *skb)
	115	{
7bd509e3 DB	116	struct nlattr *nla;
7bd509e3 DB	117
a8cb5f55 DB	118	if (prog->bpf_name &&
	119	nla_put_string(skb, TCA_ACT_BPF_NAME, prog->bpf_name))
	120	return -EMSGSIZE;
	121
e8628307 DB	122	if (nla_put_u32(skb, TCA_ACT_BPF_ID, prog->filter->aux->id))
	123	return -EMSGSIZE;
	124
f1f7714e	125	nla = nla_reserve(skb, TCA_ACT_BPF_TAG, sizeof(prog->filter->tag));
7bd509e3 DB	126	if (nla == NULL)
	127	return -EMSGSIZE;
	128
f1f7714e	129	memcpy(nla_data(nla), prog->filter->tag, nla_len(nla));
7bd509e3	130
a8cb5f55 DB	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	140	.index = prog->tcf_index,
036bb443 VB	141	.refcnt = refcount_read(&prog->tcf_refcnt) - ref,
036bb443 VB	142	.bindcnt = atomic_read(&prog->tcf_bindcnt) - bind,
d23b8ad8	143	};
a8cb5f55 DB	144	struct tcf_t tm;
a8cb5f55 DB	145	int ret;
d23b8ad8	146
653cd284	147	spin_lock_bh(&prog->tcf_lock);
2142236b	148	opt.action = prog->tcf_action;
d23b8ad8 JP	149	if (nla_put(skb, TCA_ACT_BPF_PARMS, sizeof(opt), &opt))
	150	goto nla_put_failure;
	151
a8cb5f55 DB	152	if (tcf_bpf_is_ebpf(prog))
	153	ret = tcf_bpf_dump_ebpf_info(prog, skb);
	154	else
	155	ret = tcf_bpf_dump_bpf_info(prog, skb);
	156	if (ret)
d23b8ad8 JP	157	goto nla_put_failure;
d23b8ad8 JP	158
48d8ee16	159	tcf_tm_dump(&tm, &prog->tcf_tm);
9854518e ND	160	if (nla_put_64bit(skb, TCA_ACT_BPF_TM, sizeof(tm), &tm,
9854518e ND	161	TCA_ACT_BPF_PAD))
d23b8ad8	162	goto nla_put_failure;
a8cb5f55	163
653cd284	164	spin_unlock_bh(&prog->tcf_lock);
d23b8ad8 JP	165	return skb->len;
	166
	167	nla_put_failure:
653cd284	168	spin_unlock_bh(&prog->tcf_lock);
d23b8ad8 JP	169	nlmsg_trim(skb, tp);
	170	return -1;
	171	}
	172
	173	static const struct nla_policy act_bpf_policy[TCA_ACT_BPF_MAX + 1] = {
	174	[TCA_ACT_BPF_PARMS] = { .len = sizeof(struct tc_act_bpf) },
a8cb5f55	175	[TCA_ACT_BPF_FD] = { .type = NLA_U32 },
0b0f43fe JHS	176	[TCA_ACT_BPF_NAME] = { .type = NLA_NUL_STRING,
0b0f43fe JHS	177	.len = ACT_BPF_NAME_LEN },
d23b8ad8 JP	178	[TCA_ACT_BPF_OPS_LEN] = { .type = NLA_U16 },
	179	[TCA_ACT_BPF_OPS] = { .type = NLA_BINARY,
	180	.len = sizeof(struct sock_filter) * BPF_MAXINSNS },
	181	};
	182
a8cb5f55	183	static int tcf_bpf_init_from_ops(struct nlattr *tb, struct tcf_bpf_cfg cfg)
d23b8ad8	184	{
d23b8ad8	185	struct sock_filter *bpf_ops;
a8cb5f55	186	struct sock_fprog_kern fprog_tmp;
d23b8ad8	187	struct bpf_prog *fp;
a8cb5f55	188	u16 bpf_size, bpf_num_ops;
d23b8ad8 JP	189	int ret;
d23b8ad8 JP	190
d23b8ad8 JP	191	bpf_num_ops = nla_get_u16(tb[TCA_ACT_BPF_OPS_LEN]);
	192	if (bpf_num_ops > BPF_MAXINSNS \|\| bpf_num_ops == 0)
	193	return -EINVAL;
	194
	195	bpf_size = bpf_num_ops * sizeof(*bpf_ops);
fd3e646c DB	196	if (bpf_size != nla_len(tb[TCA_ACT_BPF_OPS]))
	197	return -EINVAL;
	198
3f6bcc51	199	bpf_ops = kmemdup(nla_data(tb[TCA_ACT_BPF_OPS]), bpf_size, GFP_KERNEL);
a8cb5f55	200	if (bpf_ops == NULL)
d23b8ad8 JP	201	return -ENOMEM;
d23b8ad8 JP	202
a8cb5f55 DB	203	fprog_tmp.len = bpf_num_ops;
a8cb5f55 DB	204	fprog_tmp.filter = bpf_ops;
d23b8ad8	205
a8cb5f55 DB	206	ret = bpf_prog_create(&fp, &fprog_tmp);
	207	if (ret < 0) {
	208	kfree(bpf_ops);
	209	return ret;
	210	}
d23b8ad8	211
a8cb5f55 DB	212	cfg->bpf_ops = bpf_ops;
	213	cfg->bpf_num_ops = bpf_num_ops;
	214	cfg->filter = fp;
f4eaed28	215	cfg->is_ebpf = false;
a8cb5f55 DB	216
	217	return 0;
	218	}
	219
	220	static int tcf_bpf_init_from_efd(struct nlattr *tb, struct tcf_bpf_cfg cfg)
	221	{
	222	struct bpf_prog *fp;
	223	char *name = NULL;
	224	u32 bpf_fd;
	225
	226	bpf_fd = nla_get_u32(tb[TCA_ACT_BPF_FD]);
	227
113214be	228	fp = bpf_prog_get_type(bpf_fd, BPF_PROG_TYPE_SCHED_ACT);
a8cb5f55 DB	229	if (IS_ERR(fp))
	230	return PTR_ERR(fp);
	231
a8cb5f55	232	if (tb[TCA_ACT_BPF_NAME]) {
b15ca182	233	name = nla_memdup(tb[TCA_ACT_BPF_NAME], GFP_KERNEL);
a8cb5f55 DB	234	if (!name) {
	235	bpf_prog_put(fp);
	236	return -ENOMEM;
	237	}
	238	}
	239
a8cb5f55 DB	240	cfg->bpf_name = name;
a8cb5f55 DB	241	cfg->filter = fp;
f4eaed28	242	cfg->is_ebpf = true;
a8cb5f55 DB	243
	244	return 0;
	245	}
	246
f4eaed28 DB	247	static void tcf_bpf_cfg_cleanup(const struct tcf_bpf_cfg *cfg)
f4eaed28 DB	248	{
3239534a DC	249	struct bpf_prog *filter = cfg->filter;
	250
	251	if (filter) {
	252	if (cfg->is_ebpf)
	253	bpf_prog_put(filter);
	254	else
	255	bpf_prog_destroy(filter);
	256	}
f4eaed28 DB	257
	258	kfree(cfg->bpf_ops);
	259	kfree(cfg->bpf_name);
	260	}
	261
	262	static void tcf_bpf_prog_fill_cfg(const struct tcf_bpf *prog,
	263	struct tcf_bpf_cfg *cfg)
	264	{
	265	cfg->is_ebpf = tcf_bpf_is_ebpf(prog);
cff82457	266	/* updates to prog->filter are prevented, since it's called either
2142236b	267	* with tcf lock or during final cleanup in rcu callback
cff82457 AS	268	*/
cff82457 AS	269	cfg->filter = rcu_dereference_protected(prog->filter, 1);
f4eaed28 DB	270
	271	cfg->bpf_ops = prog->bpf_ops;
	272	cfg->bpf_name = prog->bpf_name;
	273	}
	274
a8cb5f55	275	static int tcf_bpf_init(struct net net, struct nlattr nla,
a85a970a	276	struct nlattr est, struct tc_action *act,
789871bb	277	int replace, int bind, bool rtnl_held,
abbb0d33 VB	278	struct tcf_proto *tp, u32 flags,
abbb0d33 VB	279	struct netlink_ext_ack *extack)
a8cb5f55	280	{
ddf97ccd	281	struct tc_action_net *tn = net_generic(net, bpf_net_id);
a8cb5f55	282	struct nlattr *tb[TCA_ACT_BPF_MAX + 1];
4e181004	283	struct tcf_chain *goto_ch = NULL;
f4eaed28	284	struct tcf_bpf_cfg cfg, old;
a8cb5f55 DB	285	struct tc_act_bpf *parm;
a8cb5f55 DB	286	struct tcf_bpf *prog;
a8cb5f55	287	bool is_bpf, is_ebpf;
a5c90b29	288	int ret, res = 0;
7be8ef2c	289	u32 index;
a8cb5f55 DB	290
	291	if (!nla)
	292	return -EINVAL;
	293
8cb08174 JB	294	ret = nla_parse_nested_deprecated(tb, TCA_ACT_BPF_MAX, nla,
8cb08174 JB	295	act_bpf_policy, NULL);
a8cb5f55 DB	296	if (ret < 0)
	297	return ret;
	298
a5c90b29	299	if (!tb[TCA_ACT_BPF_PARMS])
a8cb5f55 DB	300	return -EINVAL;
	301
	302	parm = nla_data(tb[TCA_ACT_BPF_PARMS]);
7be8ef2c DL	303	index = parm->index;
7be8ef2c DL	304	ret = tcf_idr_check_alloc(tn, &index, act, bind);
0190c1d4	305	if (!ret) {
7be8ef2c	306	ret = tcf_idr_create(tn, index, est, act,
e3822678	307	&act_bpf_ops, bind, true, 0);
0190c1d4	308	if (ret < 0) {
7be8ef2c	309	tcf_idr_cleanup(tn, index);
a5c90b29	310	return ret;
0190c1d4	311	}
d23b8ad8	312
a5c90b29	313	res = ACT_P_CREATED;
0190c1d4	314	} else if (ret > 0) {
a8cb5f55	315	/* Don't override defaults. */
d23b8ad8	316	if (bind)
a5c90b29	317	return 0;
a8cb5f55	318
4e8ddd7f VB	319	if (!replace) {
4e8ddd7f VB	320	tcf_idr_release(*act, bind);
a5c90b29	321	return -EEXIST;
4e8ddd7f	322	}
0190c1d4 VB	323	} else {
0190c1d4 VB	324	return ret;
d23b8ad8 JP	325	}
d23b8ad8 JP	326
4e181004 DC	327	ret = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
	328	if (ret < 0)
	329	goto release_idr;
	330
a5c90b29 DB	331	is_bpf = tb[TCA_ACT_BPF_OPS_LEN] && tb[TCA_ACT_BPF_OPS];
	332	is_ebpf = tb[TCA_ACT_BPF_FD];
	333
	334	if ((!is_bpf && !is_ebpf) \|\| (is_bpf && is_ebpf)) {
	335	ret = -EINVAL;
4e181004	336	goto put_chain;
a5c90b29 DB	337	}
	338
	339	memset(&cfg, 0, sizeof(cfg));
	340
	341	ret = is_bpf ? tcf_bpf_init_from_ops(tb, &cfg) :
	342	tcf_bpf_init_from_efd(tb, &cfg);
	343	if (ret < 0)
4e181004	344	goto put_chain;
a5c90b29	345
a85a970a	346	prog = to_bpf(*act);
a8cb5f55	347
653cd284	348	spin_lock_bh(&prog->tcf_lock);
faa54be4	349	if (res != ACT_P_CREATED)
f4eaed28 DB	350	tcf_bpf_prog_fill_cfg(prog, &old);
f4eaed28 DB	351
a8cb5f55 DB	352	prog->bpf_ops = cfg.bpf_ops;
	353	prog->bpf_name = cfg.bpf_name;
	354
	355	if (cfg.bpf_num_ops)
	356	prog->bpf_num_ops = cfg.bpf_num_ops;
a8cb5f55	357
4e181004	358	goto_ch = tcf_action_set_ctrlact(*act, parm->action, goto_ch);
cff82457	359	rcu_assign_pointer(prog->filter, cfg.filter);
653cd284	360	spin_unlock_bh(&prog->tcf_lock);
d23b8ad8	361
4e181004 DC	362	if (goto_ch)
	363	tcf_chain_put_by_act(goto_ch);
	364
cff82457	365	if (res == ACT_P_CREATED) {
65a206c0	366	tcf_idr_insert(tn, *act);
cff82457 AS	367	} else {
	368	/* make sure the program being replaced is no longer executing */
	369	synchronize_rcu();
f4eaed28	370	tcf_bpf_cfg_cleanup(&old);
cff82457	371	}
a8cb5f55	372
a5c90b29	373	return res;
d23b8ad8	374
4e181004 DC	375	put_chain:
	376	if (goto_ch)
	377	tcf_chain_put_by_act(goto_ch);
	378
	379	release_idr:
	380	tcf_idr_release(*act, bind);
d23b8ad8 JP	381	return ret;
	382	}
	383
9a63b255	384	static void tcf_bpf_cleanup(struct tc_action *act)
d23b8ad8	385	{
f4eaed28	386	struct tcf_bpf_cfg tmp;
ddf06c1e	387
a85a970a	388	tcf_bpf_prog_fill_cfg(to_bpf(act), &tmp);
f4eaed28	389	tcf_bpf_cfg_cleanup(&tmp);
d23b8ad8 JP	390	}
d23b8ad8 JP	391
ddf97ccd WC	392	static int tcf_bpf_walker(struct net net, struct sk_buff skb,
ddf97ccd WC	393	struct netlink_callback *cb, int type,
41780105 AA	394	const struct tc_action_ops *ops,
41780105 AA	395	struct netlink_ext_ack *extack)
ddf97ccd WC	396	{
	397	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	398
b3620145	399	return tcf_generic_walker(tn, skb, cb, type, ops, extack);
ddf97ccd WC	400	}
ddf97ccd WC	401
f061b48c	402	static int tcf_bpf_search(struct net net, struct tc_action *a, u32 index)
ddf97ccd WC	403	{
	404	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	405
65a206c0	406	return tcf_idr_search(tn, a, index);
ddf97ccd WC	407	}
ddf97ccd WC	408
a8cb5f55 DB	409	static struct tc_action_ops act_bpf_ops __read_mostly = {
a8cb5f55 DB	410	.kind = "bpf",
eddd2cf1	411	.id = TCA_ID_BPF,
a8cb5f55	412	.owner = THIS_MODULE,
2fbec27f	413	.act = tcf_bpf_act,
a8cb5f55 DB	414	.dump = tcf_bpf_dump,
	415	.cleanup = tcf_bpf_cleanup,
	416	.init = tcf_bpf_init,
ddf97ccd WC	417	.walk = tcf_bpf_walker,
ddf97ccd WC	418	.lookup = tcf_bpf_search,
a85a970a	419	.size = sizeof(struct tcf_bpf),
ddf97ccd WC	420	};
	421
	422	static __net_init int bpf_init_net(struct net *net)
	423	{
	424	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	425
981471bd	426	return tc_action_net_init(net, tn, &act_bpf_ops);
ddf97ccd WC	427	}
ddf97ccd WC	428
039af9c6	429	static void __net_exit bpf_exit_net(struct list_head *net_list)
ddf97ccd	430	{
039af9c6	431	tc_action_net_exit(net_list, bpf_net_id);
ddf97ccd WC	432	}
	433
	434	static struct pernet_operations bpf_net_ops = {
	435	.init = bpf_init_net,
039af9c6	436	.exit_batch = bpf_exit_net,
ddf97ccd WC	437	.id = &bpf_net_id,
ddf97ccd WC	438	.size = sizeof(struct tc_action_net),
d23b8ad8 JP	439	};
	440
	441	static int __init bpf_init_module(void)
	442	{
ddf97ccd	443	return tcf_register_action(&act_bpf_ops, &bpf_net_ops);
d23b8ad8 JP	444	}
	445
	446	static void __exit bpf_cleanup_module(void)
	447	{
ddf97ccd	448	tcf_unregister_action(&act_bpf_ops, &bpf_net_ops);
d23b8ad8 JP	449	}
	450
	451	module_init(bpf_init_module);
	452	module_exit(bpf_cleanup_module);
	453
	454	MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
	455	MODULE_DESCRIPTION("TC BPF based action");
	456	MODULE_LICENSE("GPL v2");