[linux-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/sock.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>
#include <net/tc_wrapper.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 struct tc_action_ops act_bpf_ops;

TC_INDIRECT_SCOPE 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_update(this_cpu_ptr(prog->common.cpu_bstats), skb);

	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);
	}
	if (unlikely(!skb->tstamp && skb->mono_delivery_time))
		skb->mono_delivery_time = 0;
	if (skb_sk_is_prefetched(skb) && filter_res != TC_ACT_OK)
		skb_orphan(skb);

	/* 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 unknown, 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,
			struct tcf_proto *tp, u32 flags,
			struct netlink_ext_ack *extack)
{
	struct tc_action_net *tn = net_generic(net, act_bpf_ops.net_id);
	bool bind = flags & TCA_ACT_FLAGS_BIND;
	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, flags);
		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 ACT_P_BOUND;

		if (!(flags & TCA_ACT_FLAGS_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) {
		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) {
		/* 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 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,
	.size		=	sizeof(struct tcf_bpf),
};
MODULE_ALIAS_NET_ACT("bpf");

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

static struct pernet_operations bpf_net_ops = {
	.init = bpf_init_net,
	.exit_batch = bpf_exit_net,
	.id   = &act_bpf_ops.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	14	#include <net/netlink.h>
cf7fbe66	15	#include <net/sock.h>
d23b8ad8	16	#include <net/pkt_sched.h>
4e181004	17	#include <net/pkt_cls.h>
d23b8ad8 JP	18
	19	#include <linux/tc_act/tc_bpf.h>
	20	#include <net/tc_act/tc_bpf.h>
871cf386	21	#include <net/tc_wrapper.h>
d23b8ad8	22
a8cb5f55 DB	23	#define ACT_BPF_NAME_LEN 256
	24
	25	struct tcf_bpf_cfg {
	26	struct bpf_prog *filter;
	27	struct sock_filter *bpf_ops;
f4eaed28	28	const char *bpf_name;
a8cb5f55	29	u16 bpf_num_ops;
f4eaed28	30	bool is_ebpf;
a8cb5f55	31	};
d23b8ad8	32
a85a970a	33	static struct tc_action_ops act_bpf_ops;
ddf97ccd	34
871cf386 PT	35	TC_INDIRECT_SCOPE int tcf_bpf_act(struct sk_buff *skb,
	36	const struct tc_action *act,
	37	struct tcf_result *res)
d23b8ad8	38	{
f53d8c7b	39	bool at_ingress = skb_at_tc_ingress(skb);
a85a970a	40	struct tcf_bpf *prog = to_bpf(act);
cff82457	41	struct bpf_prog *filter;
ced585c8	42	int action, filter_res;
d23b8ad8	43
cff82457	44	tcf_lastuse_update(&prog->tcf_tm);
50dc9a85	45	bstats_update(this_cpu_ptr(prog->common.cpu_bstats), skb);
d23b8ad8	46
cff82457	47	filter = rcu_dereference(prog->filter);
3431205e AS	48	if (at_ingress) {
3431205e AS	49	__skb_push(skb, skb->mac_len);
6aaae2b6	50	bpf_compute_data_pointers(skb);
fb7dd8bc	51	filter_res = bpf_prog_run(filter, skb);
3431205e AS	52	__skb_pull(skb, skb->mac_len);
3431205e AS	53	} else {
6aaae2b6	54	bpf_compute_data_pointers(skb);
fb7dd8bc	55	filter_res = bpf_prog_run(filter, skb);
3431205e	56	}
7449197d MKL	57	if (unlikely(!skb->tstamp && skb->mono_delivery_time))
7449197d MKL	58	skb->mono_delivery_time = 0;
cf7fbe66 JS	59	if (skb_sk_is_prefetched(skb) && filter_res != TC_ACT_OK)
cf7fbe66 JS	60	skb_orphan(skb);
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	*
e5a4b17d	69	* For everything else that is unknown, TC_ACT_UNSPEC is
ced585c8 DB	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
e8628307 DB	126	if (nla_put_u32(skb, TCA_ACT_BPF_ID, prog->filter->aux->id))
	127	return -EMSGSIZE;
	128
f1f7714e	129	nla = nla_reserve(skb, TCA_ACT_BPF_TAG, sizeof(prog->filter->tag));
7bd509e3 DB	130	if (nla == NULL)
	131	return -EMSGSIZE;
	132
f1f7714e	133	memcpy(nla_data(nla), prog->filter->tag, nla_len(nla));
7bd509e3	134
a8cb5f55 DB	135	return 0;
	136	}
	137
	138	static int tcf_bpf_dump(struct sk_buff skb, struct tc_action act,
d23b8ad8 JP	139	int bind, int ref)
	140	{
	141	unsigned char *tp = skb_tail_pointer(skb);
a85a970a	142	struct tcf_bpf *prog = to_bpf(act);
d23b8ad8	143	struct tc_act_bpf opt = {
a8cb5f55	144	.index = prog->tcf_index,
036bb443 VB	145	.refcnt = refcount_read(&prog->tcf_refcnt) - ref,
036bb443 VB	146	.bindcnt = atomic_read(&prog->tcf_bindcnt) - bind,
d23b8ad8	147	};
a8cb5f55 DB	148	struct tcf_t tm;
a8cb5f55 DB	149	int ret;
d23b8ad8	150
653cd284	151	spin_lock_bh(&prog->tcf_lock);
2142236b	152	opt.action = prog->tcf_action;
d23b8ad8 JP	153	if (nla_put(skb, TCA_ACT_BPF_PARMS, sizeof(opt), &opt))
	154	goto nla_put_failure;
	155
a8cb5f55 DB	156	if (tcf_bpf_is_ebpf(prog))
	157	ret = tcf_bpf_dump_ebpf_info(prog, skb);
	158	else
	159	ret = tcf_bpf_dump_bpf_info(prog, skb);
	160	if (ret)
d23b8ad8 JP	161	goto nla_put_failure;
d23b8ad8 JP	162
48d8ee16	163	tcf_tm_dump(&tm, &prog->tcf_tm);
9854518e ND	164	if (nla_put_64bit(skb, TCA_ACT_BPF_TM, sizeof(tm), &tm,
9854518e ND	165	TCA_ACT_BPF_PAD))
d23b8ad8	166	goto nla_put_failure;
a8cb5f55	167
653cd284	168	spin_unlock_bh(&prog->tcf_lock);
d23b8ad8 JP	169	return skb->len;
	170
	171	nla_put_failure:
653cd284	172	spin_unlock_bh(&prog->tcf_lock);
d23b8ad8 JP	173	nlmsg_trim(skb, tp);
	174	return -1;
	175	}
	176
	177	static const struct nla_policy act_bpf_policy[TCA_ACT_BPF_MAX + 1] = {
	178	[TCA_ACT_BPF_PARMS] = { .len = sizeof(struct tc_act_bpf) },
a8cb5f55	179	[TCA_ACT_BPF_FD] = { .type = NLA_U32 },
0b0f43fe JHS	180	[TCA_ACT_BPF_NAME] = { .type = NLA_NUL_STRING,
0b0f43fe JHS	181	.len = ACT_BPF_NAME_LEN },
d23b8ad8 JP	182	[TCA_ACT_BPF_OPS_LEN] = { .type = NLA_U16 },
	183	[TCA_ACT_BPF_OPS] = { .type = NLA_BINARY,
	184	.len = sizeof(struct sock_filter) * BPF_MAXINSNS },
	185	};
	186
a8cb5f55	187	static int tcf_bpf_init_from_ops(struct nlattr *tb, struct tcf_bpf_cfg cfg)
d23b8ad8	188	{
d23b8ad8	189	struct sock_filter *bpf_ops;
a8cb5f55	190	struct sock_fprog_kern fprog_tmp;
d23b8ad8	191	struct bpf_prog *fp;
a8cb5f55	192	u16 bpf_size, bpf_num_ops;
d23b8ad8 JP	193	int ret;
d23b8ad8 JP	194
d23b8ad8 JP	195	bpf_num_ops = nla_get_u16(tb[TCA_ACT_BPF_OPS_LEN]);
	196	if (bpf_num_ops > BPF_MAXINSNS \|\| bpf_num_ops == 0)
	197	return -EINVAL;
	198
	199	bpf_size = bpf_num_ops * sizeof(*bpf_ops);
fd3e646c DB	200	if (bpf_size != nla_len(tb[TCA_ACT_BPF_OPS]))
	201	return -EINVAL;
	202
3f6bcc51	203	bpf_ops = kmemdup(nla_data(tb[TCA_ACT_BPF_OPS]), bpf_size, GFP_KERNEL);
a8cb5f55	204	if (bpf_ops == NULL)
d23b8ad8 JP	205	return -ENOMEM;
d23b8ad8 JP	206
a8cb5f55 DB	207	fprog_tmp.len = bpf_num_ops;
a8cb5f55 DB	208	fprog_tmp.filter = bpf_ops;
d23b8ad8	209
a8cb5f55 DB	210	ret = bpf_prog_create(&fp, &fprog_tmp);
	211	if (ret < 0) {
	212	kfree(bpf_ops);
	213	return ret;
	214	}
d23b8ad8	215
a8cb5f55 DB	216	cfg->bpf_ops = bpf_ops;
	217	cfg->bpf_num_ops = bpf_num_ops;
	218	cfg->filter = fp;
f4eaed28	219	cfg->is_ebpf = false;
a8cb5f55 DB	220
	221	return 0;
	222	}
	223
	224	static int tcf_bpf_init_from_efd(struct nlattr *tb, struct tcf_bpf_cfg cfg)
	225	{
	226	struct bpf_prog *fp;
	227	char *name = NULL;
	228	u32 bpf_fd;
	229
	230	bpf_fd = nla_get_u32(tb[TCA_ACT_BPF_FD]);
	231
113214be	232	fp = bpf_prog_get_type(bpf_fd, BPF_PROG_TYPE_SCHED_ACT);
a8cb5f55 DB	233	if (IS_ERR(fp))
	234	return PTR_ERR(fp);
	235
a8cb5f55	236	if (tb[TCA_ACT_BPF_NAME]) {
b15ca182	237	name = nla_memdup(tb[TCA_ACT_BPF_NAME], GFP_KERNEL);
a8cb5f55 DB	238	if (!name) {
	239	bpf_prog_put(fp);
	240	return -ENOMEM;
	241	}
	242	}
	243
a8cb5f55 DB	244	cfg->bpf_name = name;
a8cb5f55 DB	245	cfg->filter = fp;
f4eaed28	246	cfg->is_ebpf = true;
a8cb5f55 DB	247
	248	return 0;
	249	}
	250
f4eaed28 DB	251	static void tcf_bpf_cfg_cleanup(const struct tcf_bpf_cfg *cfg)
f4eaed28 DB	252	{
3239534a DC	253	struct bpf_prog *filter = cfg->filter;
	254
	255	if (filter) {
	256	if (cfg->is_ebpf)
	257	bpf_prog_put(filter);
	258	else
	259	bpf_prog_destroy(filter);
	260	}
f4eaed28 DB	261
	262	kfree(cfg->bpf_ops);
	263	kfree(cfg->bpf_name);
	264	}
	265
	266	static void tcf_bpf_prog_fill_cfg(const struct tcf_bpf *prog,
	267	struct tcf_bpf_cfg *cfg)
	268	{
	269	cfg->is_ebpf = tcf_bpf_is_ebpf(prog);
cff82457	270	/* updates to prog->filter are prevented, since it's called either
2142236b	271	* with tcf lock or during final cleanup in rcu callback
cff82457 AS	272	*/
cff82457 AS	273	cfg->filter = rcu_dereference_protected(prog->filter, 1);
f4eaed28 DB	274
	275	cfg->bpf_ops = prog->bpf_ops;
	276	cfg->bpf_name = prog->bpf_name;
	277	}
	278
a8cb5f55	279	static int tcf_bpf_init(struct net net, struct nlattr nla,
a85a970a	280	struct nlattr est, struct tc_action *act,
abbb0d33 VB	281	struct tcf_proto *tp, u32 flags,
abbb0d33 VB	282	struct netlink_ext_ack *extack)
a8cb5f55	283	{
acd0a7ab	284	struct tc_action_net *tn = net_generic(net, act_bpf_ops.net_id);
695176bf	285	bool bind = flags & TCA_ACT_FLAGS_BIND;
a8cb5f55	286	struct nlattr *tb[TCA_ACT_BPF_MAX + 1];
4e181004	287	struct tcf_chain *goto_ch = NULL;
f4eaed28	288	struct tcf_bpf_cfg cfg, old;
a8cb5f55 DB	289	struct tc_act_bpf *parm;
a8cb5f55 DB	290	struct tcf_bpf *prog;
a8cb5f55	291	bool is_bpf, is_ebpf;
a5c90b29	292	int ret, res = 0;
7be8ef2c	293	u32 index;
a8cb5f55 DB	294
	295	if (!nla)
	296	return -EINVAL;
	297
8cb08174 JB	298	ret = nla_parse_nested_deprecated(tb, TCA_ACT_BPF_MAX, nla,
8cb08174 JB	299	act_bpf_policy, NULL);
a8cb5f55 DB	300	if (ret < 0)
	301	return ret;
	302
a5c90b29	303	if (!tb[TCA_ACT_BPF_PARMS])
a8cb5f55 DB	304	return -EINVAL;
	305
	306	parm = nla_data(tb[TCA_ACT_BPF_PARMS]);
7be8ef2c DL	307	index = parm->index;
7be8ef2c DL	308	ret = tcf_idr_check_alloc(tn, &index, act, bind);
0190c1d4	309	if (!ret) {
7be8ef2c	310	ret = tcf_idr_create(tn, index, est, act,
40bd094d	311	&act_bpf_ops, bind, true, flags);
0190c1d4	312	if (ret < 0) {
7be8ef2c	313	tcf_idr_cleanup(tn, index);
a5c90b29	314	return ret;
0190c1d4	315	}
d23b8ad8	316
a5c90b29	317	res = ACT_P_CREATED;
0190c1d4	318	} else if (ret > 0) {
a8cb5f55	319	/* Don't override defaults. */
d23b8ad8	320	if (bind)
c2a67de9	321	return ACT_P_BOUND;
a8cb5f55	322
695176bf	323	if (!(flags & TCA_ACT_FLAGS_REPLACE)) {
4e8ddd7f	324	tcf_idr_release(*act, bind);
a5c90b29	325	return -EEXIST;
4e8ddd7f	326	}
0190c1d4 VB	327	} else {
0190c1d4 VB	328	return ret;
d23b8ad8 JP	329	}
d23b8ad8 JP	330
4e181004 DC	331	ret = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
	332	if (ret < 0)
	333	goto release_idr;
	334
a5c90b29 DB	335	is_bpf = tb[TCA_ACT_BPF_OPS_LEN] && tb[TCA_ACT_BPF_OPS];
	336	is_ebpf = tb[TCA_ACT_BPF_FD];
	337
8fff09ef	338	if (is_bpf == is_ebpf) {
a5c90b29	339	ret = -EINVAL;
4e181004	340	goto put_chain;
a5c90b29 DB	341	}
	342
	343	memset(&cfg, 0, sizeof(cfg));
	344
	345	ret = is_bpf ? tcf_bpf_init_from_ops(tb, &cfg) :
	346	tcf_bpf_init_from_efd(tb, &cfg);
	347	if (ret < 0)
4e181004	348	goto put_chain;
a5c90b29	349
a85a970a	350	prog = to_bpf(*act);
a8cb5f55	351
653cd284	352	spin_lock_bh(&prog->tcf_lock);
faa54be4	353	if (res != ACT_P_CREATED)
f4eaed28 DB	354	tcf_bpf_prog_fill_cfg(prog, &old);
f4eaed28 DB	355
a8cb5f55 DB	356	prog->bpf_ops = cfg.bpf_ops;
	357	prog->bpf_name = cfg.bpf_name;
	358
	359	if (cfg.bpf_num_ops)
	360	prog->bpf_num_ops = cfg.bpf_num_ops;
a8cb5f55	361
4e181004	362	goto_ch = tcf_action_set_ctrlact(*act, parm->action, goto_ch);
cff82457	363	rcu_assign_pointer(prog->filter, cfg.filter);
653cd284	364	spin_unlock_bh(&prog->tcf_lock);
d23b8ad8	365
4e181004 DC	366	if (goto_ch)
	367	tcf_chain_put_by_act(goto_ch);
	368
e49d8c22	369	if (res != ACT_P_CREATED) {
cff82457 AS	370	/* make sure the program being replaced is no longer executing */
cff82457 AS	371	synchronize_rcu();
f4eaed28	372	tcf_bpf_cfg_cleanup(&old);
cff82457	373	}
a8cb5f55	374
a5c90b29	375	return res;
d23b8ad8	376
4e181004 DC	377	put_chain:
	378	if (goto_ch)
	379	tcf_chain_put_by_act(goto_ch);
	380
	381	release_idr:
	382	tcf_idr_release(*act, bind);
d23b8ad8 JP	383	return ret;
	384	}
	385
9a63b255	386	static void tcf_bpf_cleanup(struct tc_action *act)
d23b8ad8	387	{
f4eaed28	388	struct tcf_bpf_cfg tmp;
ddf06c1e	389
a85a970a	390	tcf_bpf_prog_fill_cfg(to_bpf(act), &tmp);
f4eaed28	391	tcf_bpf_cfg_cleanup(&tmp);
d23b8ad8 JP	392	}
d23b8ad8 JP	393
a8cb5f55 DB	394	static struct tc_action_ops act_bpf_ops __read_mostly = {
a8cb5f55 DB	395	.kind = "bpf",
eddd2cf1	396	.id = TCA_ID_BPF,
a8cb5f55	397	.owner = THIS_MODULE,
2fbec27f	398	.act = tcf_bpf_act,
a8cb5f55 DB	399	.dump = tcf_bpf_dump,
	400	.cleanup = tcf_bpf_cleanup,
	401	.init = tcf_bpf_init,
a85a970a	402	.size = sizeof(struct tcf_bpf),
ddf97ccd	403	};
241a94ab	404	MODULE_ALIAS_NET_ACT("bpf");
ddf97ccd WC	405
	406	static __net_init int bpf_init_net(struct net *net)
	407	{
acd0a7ab	408	struct tc_action_net *tn = net_generic(net, act_bpf_ops.net_id);
ddf97ccd	409
981471bd	410	return tc_action_net_init(net, tn, &act_bpf_ops);
ddf97ccd WC	411	}
ddf97ccd WC	412
039af9c6	413	static void __net_exit bpf_exit_net(struct list_head *net_list)
ddf97ccd	414	{
acd0a7ab	415	tc_action_net_exit(net_list, act_bpf_ops.net_id);
ddf97ccd WC	416	}
	417
	418	static struct pernet_operations bpf_net_ops = {
	419	.init = bpf_init_net,
039af9c6	420	.exit_batch = bpf_exit_net,
acd0a7ab	421	.id = &act_bpf_ops.net_id,
ddf97ccd	422	.size = sizeof(struct tc_action_net),
d23b8ad8 JP	423	};
	424
	425	static int __init bpf_init_module(void)
	426	{
ddf97ccd	427	return tcf_register_action(&act_bpf_ops, &bpf_net_ops);
d23b8ad8 JP	428	}
	429
	430	static void __exit bpf_cleanup_module(void)
	431	{
ddf97ccd	432	tcf_unregister_action(&act_bpf_ops, &bpf_net_ops);
d23b8ad8 JP	433	}
	434
	435	module_init(bpf_init_module);
	436	module_exit(bpf_cleanup_module);
	437
	438	MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
	439	MODULE_DESCRIPTION("TC BPF based action");
	440	MODULE_LICENSE("GPL v2");