[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>

#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_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, bpf_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 0;

		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) || (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 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	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>
	21
a8cb5f55 DB	22	#define ACT_BPF_NAME_LEN 256
	23
	24	struct tcf_bpf_cfg {
	25	struct bpf_prog *filter;
	26	struct sock_filter *bpf_ops;
f4eaed28	27	const char *bpf_name;
a8cb5f55	28	u16 bpf_num_ops;
f4eaed28	29	bool is_ebpf;
a8cb5f55	30	};
d23b8ad8	31
c7d03a00	32	static unsigned int bpf_net_id;
a85a970a	33	static struct tc_action_ops act_bpf_ops;
ddf97ccd	34
2fbec27f JHS	35	static int tcf_bpf_act(struct sk_buff skb, const struct tc_action act,
2fbec27f JHS	36	struct tcf_result *res)
d23b8ad8	37	{
f53d8c7b	38	bool at_ingress = skb_at_tc_ingress(skb);
a85a970a	39	struct tcf_bpf *prog = to_bpf(act);
cff82457	40	struct bpf_prog *filter;
ced585c8	41	int action, filter_res;
d23b8ad8	42
cff82457	43	tcf_lastuse_update(&prog->tcf_tm);
50dc9a85	44	bstats_update(this_cpu_ptr(prog->common.cpu_bstats), skb);
d23b8ad8	45
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);
fb7dd8bc	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);
fb7dd8bc	54	filter_res = bpf_prog_run(filter, skb);
3431205e	55	}
7449197d MKL	56	if (unlikely(!skb->tstamp && skb->mono_delivery_time))
7449197d MKL	57	skb->mono_delivery_time = 0;
cf7fbe66 JS	58	if (skb_sk_is_prefetched(skb) && filter_res != TC_ACT_OK)
cf7fbe66 JS	59	skb_orphan(skb);
ced585c8 DB	60
	61	/* A BPF program may overwrite the default action opcode.
	62	* Similarly as in cls_bpf, if filter_res == -1 we use the
	63	* default action specified from tc.
	64	*
	65	* In case a different well-known TC_ACT opcode has been
	66	* returned, it will overwrite the default one.
	67	*
e5a4b17d	68	* For everything else that is unknown, TC_ACT_UNSPEC is
ced585c8 DB	69	* returned.
	70	*/
	71	switch (filter_res) {
	72	case TC_ACT_PIPE:
	73	case TC_ACT_RECLASSIFY:
	74	case TC_ACT_OK:
27b29f63	75	case TC_ACT_REDIRECT:
ced585c8 DB	76	action = filter_res;
	77	break;
	78	case TC_ACT_SHOT:
	79	action = filter_res;
cff82457	80	qstats_drop_inc(this_cpu_ptr(prog->common.cpu_qstats));
ced585c8 DB	81	break;
ced585c8 DB	82	case TC_ACT_UNSPEC:
a8cb5f55	83	action = prog->tcf_action;
ced585c8 DB	84	break;
	85	default:
	86	action = TC_ACT_UNSPEC;
	87	break;
d23b8ad8 JP	88	}
d23b8ad8 JP	89
d23b8ad8 JP	90	return action;
	91	}
	92
a8cb5f55 DB	93	static bool tcf_bpf_is_ebpf(const struct tcf_bpf *prog)
	94	{
	95	return !prog->bpf_ops;
	96	}
	97
	98	static int tcf_bpf_dump_bpf_info(const struct tcf_bpf *prog,
	99	struct sk_buff *skb)
	100	{
	101	struct nlattr *nla;
	102
	103	if (nla_put_u16(skb, TCA_ACT_BPF_OPS_LEN, prog->bpf_num_ops))
	104	return -EMSGSIZE;
	105
	106	nla = nla_reserve(skb, TCA_ACT_BPF_OPS, prog->bpf_num_ops *
	107	sizeof(struct sock_filter));
	108	if (nla == NULL)
	109	return -EMSGSIZE;
	110
	111	memcpy(nla_data(nla), prog->bpf_ops, nla_len(nla));
	112
	113	return 0;
	114	}
	115
	116	static int tcf_bpf_dump_ebpf_info(const struct tcf_bpf *prog,
	117	struct sk_buff *skb)
	118	{
7bd509e3 DB	119	struct nlattr *nla;
7bd509e3 DB	120
a8cb5f55 DB	121	if (prog->bpf_name &&
	122	nla_put_string(skb, TCA_ACT_BPF_NAME, prog->bpf_name))
	123	return -EMSGSIZE;
	124
e8628307 DB	125	if (nla_put_u32(skb, TCA_ACT_BPF_ID, prog->filter->aux->id))
	126	return -EMSGSIZE;
	127
f1f7714e	128	nla = nla_reserve(skb, TCA_ACT_BPF_TAG, sizeof(prog->filter->tag));
7bd509e3 DB	129	if (nla == NULL)
	130	return -EMSGSIZE;
	131
f1f7714e	132	memcpy(nla_data(nla), prog->filter->tag, nla_len(nla));
7bd509e3	133
a8cb5f55 DB	134	return 0;
	135	}
	136
	137	static int tcf_bpf_dump(struct sk_buff skb, struct tc_action act,
d23b8ad8 JP	138	int bind, int ref)
	139	{
	140	unsigned char *tp = skb_tail_pointer(skb);
a85a970a	141	struct tcf_bpf *prog = to_bpf(act);
d23b8ad8	142	struct tc_act_bpf opt = {
a8cb5f55	143	.index = prog->tcf_index,
036bb443 VB	144	.refcnt = refcount_read(&prog->tcf_refcnt) - ref,
036bb443 VB	145	.bindcnt = atomic_read(&prog->tcf_bindcnt) - bind,
d23b8ad8	146	};
a8cb5f55 DB	147	struct tcf_t tm;
a8cb5f55 DB	148	int ret;
d23b8ad8	149
653cd284	150	spin_lock_bh(&prog->tcf_lock);
2142236b	151	opt.action = prog->tcf_action;
d23b8ad8 JP	152	if (nla_put(skb, TCA_ACT_BPF_PARMS, sizeof(opt), &opt))
	153	goto nla_put_failure;
	154
a8cb5f55 DB	155	if (tcf_bpf_is_ebpf(prog))
	156	ret = tcf_bpf_dump_ebpf_info(prog, skb);
	157	else
	158	ret = tcf_bpf_dump_bpf_info(prog, skb);
	159	if (ret)
d23b8ad8 JP	160	goto nla_put_failure;
d23b8ad8 JP	161
48d8ee16	162	tcf_tm_dump(&tm, &prog->tcf_tm);
9854518e ND	163	if (nla_put_64bit(skb, TCA_ACT_BPF_TM, sizeof(tm), &tm,
9854518e ND	164	TCA_ACT_BPF_PAD))
d23b8ad8	165	goto nla_put_failure;
a8cb5f55	166
653cd284	167	spin_unlock_bh(&prog->tcf_lock);
d23b8ad8 JP	168	return skb->len;
	169
	170	nla_put_failure:
653cd284	171	spin_unlock_bh(&prog->tcf_lock);
d23b8ad8 JP	172	nlmsg_trim(skb, tp);
	173	return -1;
	174	}
	175
	176	static const struct nla_policy act_bpf_policy[TCA_ACT_BPF_MAX + 1] = {
	177	[TCA_ACT_BPF_PARMS] = { .len = sizeof(struct tc_act_bpf) },
a8cb5f55	178	[TCA_ACT_BPF_FD] = { .type = NLA_U32 },
0b0f43fe JHS	179	[TCA_ACT_BPF_NAME] = { .type = NLA_NUL_STRING,
0b0f43fe JHS	180	.len = ACT_BPF_NAME_LEN },
d23b8ad8 JP	181	[TCA_ACT_BPF_OPS_LEN] = { .type = NLA_U16 },
	182	[TCA_ACT_BPF_OPS] = { .type = NLA_BINARY,
	183	.len = sizeof(struct sock_filter) * BPF_MAXINSNS },
	184	};
	185
a8cb5f55	186	static int tcf_bpf_init_from_ops(struct nlattr *tb, struct tcf_bpf_cfg cfg)
d23b8ad8	187	{
d23b8ad8	188	struct sock_filter *bpf_ops;
a8cb5f55	189	struct sock_fprog_kern fprog_tmp;
d23b8ad8	190	struct bpf_prog *fp;
a8cb5f55	191	u16 bpf_size, bpf_num_ops;
d23b8ad8 JP	192	int ret;
d23b8ad8 JP	193
d23b8ad8 JP	194	bpf_num_ops = nla_get_u16(tb[TCA_ACT_BPF_OPS_LEN]);
	195	if (bpf_num_ops > BPF_MAXINSNS \|\| bpf_num_ops == 0)
	196	return -EINVAL;
	197
	198	bpf_size = bpf_num_ops * sizeof(*bpf_ops);
fd3e646c DB	199	if (bpf_size != nla_len(tb[TCA_ACT_BPF_OPS]))
	200	return -EINVAL;
	201
3f6bcc51	202	bpf_ops = kmemdup(nla_data(tb[TCA_ACT_BPF_OPS]), bpf_size, GFP_KERNEL);
a8cb5f55	203	if (bpf_ops == NULL)
d23b8ad8 JP	204	return -ENOMEM;
d23b8ad8 JP	205
a8cb5f55 DB	206	fprog_tmp.len = bpf_num_ops;
a8cb5f55 DB	207	fprog_tmp.filter = bpf_ops;
d23b8ad8	208
a8cb5f55 DB	209	ret = bpf_prog_create(&fp, &fprog_tmp);
	210	if (ret < 0) {
	211	kfree(bpf_ops);
	212	return ret;
	213	}
d23b8ad8	214
a8cb5f55 DB	215	cfg->bpf_ops = bpf_ops;
	216	cfg->bpf_num_ops = bpf_num_ops;
	217	cfg->filter = fp;
f4eaed28	218	cfg->is_ebpf = false;
a8cb5f55 DB	219
	220	return 0;
	221	}
	222
	223	static int tcf_bpf_init_from_efd(struct nlattr *tb, struct tcf_bpf_cfg cfg)
	224	{
	225	struct bpf_prog *fp;
	226	char *name = NULL;
	227	u32 bpf_fd;
	228
	229	bpf_fd = nla_get_u32(tb[TCA_ACT_BPF_FD]);
	230
113214be	231	fp = bpf_prog_get_type(bpf_fd, BPF_PROG_TYPE_SCHED_ACT);
a8cb5f55 DB	232	if (IS_ERR(fp))
	233	return PTR_ERR(fp);
	234
a8cb5f55	235	if (tb[TCA_ACT_BPF_NAME]) {
b15ca182	236	name = nla_memdup(tb[TCA_ACT_BPF_NAME], GFP_KERNEL);
a8cb5f55 DB	237	if (!name) {
	238	bpf_prog_put(fp);
	239	return -ENOMEM;
	240	}
	241	}
	242
a8cb5f55 DB	243	cfg->bpf_name = name;
a8cb5f55 DB	244	cfg->filter = fp;
f4eaed28	245	cfg->is_ebpf = true;
a8cb5f55 DB	246
	247	return 0;
	248	}
	249
f4eaed28 DB	250	static void tcf_bpf_cfg_cleanup(const struct tcf_bpf_cfg *cfg)
f4eaed28 DB	251	{
3239534a DC	252	struct bpf_prog *filter = cfg->filter;
	253
	254	if (filter) {
	255	if (cfg->is_ebpf)
	256	bpf_prog_put(filter);
	257	else
	258	bpf_prog_destroy(filter);
	259	}
f4eaed28 DB	260
	261	kfree(cfg->bpf_ops);
	262	kfree(cfg->bpf_name);
	263	}
	264
	265	static void tcf_bpf_prog_fill_cfg(const struct tcf_bpf *prog,
	266	struct tcf_bpf_cfg *cfg)
	267	{
	268	cfg->is_ebpf = tcf_bpf_is_ebpf(prog);
cff82457	269	/* updates to prog->filter are prevented, since it's called either
2142236b	270	* with tcf lock or during final cleanup in rcu callback
cff82457 AS	271	*/
cff82457 AS	272	cfg->filter = rcu_dereference_protected(prog->filter, 1);
f4eaed28 DB	273
	274	cfg->bpf_ops = prog->bpf_ops;
	275	cfg->bpf_name = prog->bpf_name;
	276	}
	277
a8cb5f55	278	static int tcf_bpf_init(struct net net, struct nlattr nla,
a85a970a	279	struct nlattr est, struct tc_action *act,
abbb0d33 VB	280	struct tcf_proto *tp, u32 flags,
abbb0d33 VB	281	struct netlink_ext_ack *extack)
a8cb5f55	282	{
ddf97ccd	283	struct tc_action_net *tn = net_generic(net, bpf_net_id);
695176bf	284	bool bind = flags & TCA_ACT_FLAGS_BIND;
a8cb5f55	285	struct nlattr *tb[TCA_ACT_BPF_MAX + 1];
4e181004	286	struct tcf_chain *goto_ch = NULL;
f4eaed28	287	struct tcf_bpf_cfg cfg, old;
a8cb5f55 DB	288	struct tc_act_bpf *parm;
a8cb5f55 DB	289	struct tcf_bpf *prog;
a8cb5f55	290	bool is_bpf, is_ebpf;
a5c90b29	291	int ret, res = 0;
7be8ef2c	292	u32 index;
a8cb5f55 DB	293
	294	if (!nla)
	295	return -EINVAL;
	296
8cb08174 JB	297	ret = nla_parse_nested_deprecated(tb, TCA_ACT_BPF_MAX, nla,
8cb08174 JB	298	act_bpf_policy, NULL);
a8cb5f55 DB	299	if (ret < 0)
	300	return ret;
	301
a5c90b29	302	if (!tb[TCA_ACT_BPF_PARMS])
a8cb5f55 DB	303	return -EINVAL;
	304
	305	parm = nla_data(tb[TCA_ACT_BPF_PARMS]);
7be8ef2c DL	306	index = parm->index;
7be8ef2c DL	307	ret = tcf_idr_check_alloc(tn, &index, act, bind);
0190c1d4	308	if (!ret) {
7be8ef2c	309	ret = tcf_idr_create(tn, index, est, act,
40bd094d	310	&act_bpf_ops, bind, true, flags);
0190c1d4	311	if (ret < 0) {
7be8ef2c	312	tcf_idr_cleanup(tn, index);
a5c90b29	313	return ret;
0190c1d4	314	}
d23b8ad8	315
a5c90b29	316	res = ACT_P_CREATED;
0190c1d4	317	} else if (ret > 0) {
a8cb5f55	318	/* Don't override defaults. */
d23b8ad8	319	if (bind)
a5c90b29	320	return 0;
a8cb5f55	321
695176bf	322	if (!(flags & TCA_ACT_FLAGS_REPLACE)) {
4e8ddd7f	323	tcf_idr_release(*act, bind);
a5c90b29	324	return -EEXIST;
4e8ddd7f	325	}
0190c1d4 VB	326	} else {
0190c1d4 VB	327	return ret;
d23b8ad8 JP	328	}
d23b8ad8 JP	329
4e181004 DC	330	ret = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
	331	if (ret < 0)
	332	goto release_idr;
	333
a5c90b29 DB	334	is_bpf = tb[TCA_ACT_BPF_OPS_LEN] && tb[TCA_ACT_BPF_OPS];
	335	is_ebpf = tb[TCA_ACT_BPF_FD];
	336
	337	if ((!is_bpf && !is_ebpf) \|\| (is_bpf && is_ebpf)) {
	338	ret = -EINVAL;
4e181004	339	goto put_chain;
a5c90b29 DB	340	}
	341
	342	memset(&cfg, 0, sizeof(cfg));
	343
	344	ret = is_bpf ? tcf_bpf_init_from_ops(tb, &cfg) :
	345	tcf_bpf_init_from_efd(tb, &cfg);
	346	if (ret < 0)
4e181004	347	goto put_chain;
a5c90b29	348
a85a970a	349	prog = to_bpf(*act);
a8cb5f55	350
653cd284	351	spin_lock_bh(&prog->tcf_lock);
faa54be4	352	if (res != ACT_P_CREATED)
f4eaed28 DB	353	tcf_bpf_prog_fill_cfg(prog, &old);
f4eaed28 DB	354
a8cb5f55 DB	355	prog->bpf_ops = cfg.bpf_ops;
	356	prog->bpf_name = cfg.bpf_name;
	357
	358	if (cfg.bpf_num_ops)
	359	prog->bpf_num_ops = cfg.bpf_num_ops;
a8cb5f55	360
4e181004	361	goto_ch = tcf_action_set_ctrlact(*act, parm->action, goto_ch);
cff82457	362	rcu_assign_pointer(prog->filter, cfg.filter);
653cd284	363	spin_unlock_bh(&prog->tcf_lock);
d23b8ad8	364
4e181004 DC	365	if (goto_ch)
	366	tcf_chain_put_by_act(goto_ch);
	367
e49d8c22	368	if (res != ACT_P_CREATED) {
cff82457 AS	369	/* make sure the program being replaced is no longer executing */
cff82457 AS	370	synchronize_rcu();
f4eaed28	371	tcf_bpf_cfg_cleanup(&old);
cff82457	372	}
a8cb5f55	373
a5c90b29	374	return res;
d23b8ad8	375
4e181004 DC	376	put_chain:
	377	if (goto_ch)
	378	tcf_chain_put_by_act(goto_ch);
	379
	380	release_idr:
	381	tcf_idr_release(*act, bind);
d23b8ad8 JP	382	return ret;
	383	}
	384
9a63b255	385	static void tcf_bpf_cleanup(struct tc_action *act)
d23b8ad8	386	{
f4eaed28	387	struct tcf_bpf_cfg tmp;
ddf06c1e	388
a85a970a	389	tcf_bpf_prog_fill_cfg(to_bpf(act), &tmp);
f4eaed28	390	tcf_bpf_cfg_cleanup(&tmp);
d23b8ad8 JP	391	}
d23b8ad8 JP	392
ddf97ccd WC	393	static int tcf_bpf_walker(struct net net, struct sk_buff skb,
ddf97ccd WC	394	struct netlink_callback *cb, int type,
41780105 AA	395	const struct tc_action_ops *ops,
41780105 AA	396	struct netlink_ext_ack *extack)
ddf97ccd WC	397	{
	398	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	399
b3620145	400	return tcf_generic_walker(tn, skb, cb, type, ops, extack);
ddf97ccd WC	401	}
ddf97ccd WC	402
f061b48c	403	static int tcf_bpf_search(struct net net, struct tc_action *a, u32 index)
ddf97ccd WC	404	{
	405	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	406
65a206c0	407	return tcf_idr_search(tn, a, index);
ddf97ccd WC	408	}
ddf97ccd WC	409
a8cb5f55 DB	410	static struct tc_action_ops act_bpf_ops __read_mostly = {
a8cb5f55 DB	411	.kind = "bpf",
eddd2cf1	412	.id = TCA_ID_BPF,
a8cb5f55	413	.owner = THIS_MODULE,
2fbec27f	414	.act = tcf_bpf_act,
a8cb5f55 DB	415	.dump = tcf_bpf_dump,
	416	.cleanup = tcf_bpf_cleanup,
	417	.init = tcf_bpf_init,
ddf97ccd WC	418	.walk = tcf_bpf_walker,
ddf97ccd WC	419	.lookup = tcf_bpf_search,
a85a970a	420	.size = sizeof(struct tcf_bpf),
ddf97ccd WC	421	};
	422
	423	static __net_init int bpf_init_net(struct net *net)
	424	{
	425	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	426
981471bd	427	return tc_action_net_init(net, tn, &act_bpf_ops);
ddf97ccd WC	428	}
ddf97ccd WC	429
039af9c6	430	static void __net_exit bpf_exit_net(struct list_head *net_list)
ddf97ccd	431	{
039af9c6	432	tc_action_net_exit(net_list, bpf_net_id);
ddf97ccd WC	433	}
	434
	435	static struct pernet_operations bpf_net_ops = {
	436	.init = bpf_init_net,
039af9c6	437	.exit_batch = bpf_exit_net,
ddf97ccd WC	438	.id = &bpf_net_id,
ddf97ccd WC	439	.size = sizeof(struct tc_action_net),
d23b8ad8 JP	440	};
	441
	442	static int __init bpf_init_module(void)
	443	{
ddf97ccd	444	return tcf_register_action(&act_bpf_ops, &bpf_net_ops);
d23b8ad8 JP	445	}
	446
	447	static void __exit bpf_cleanup_module(void)
	448	{
ddf97ccd	449	tcf_unregister_action(&act_bpf_ops, &bpf_net_ops);
d23b8ad8 JP	450	}
	451
	452	module_init(bpf_init_module);
	453	module_exit(bpf_cleanup_module);
	454
	455	MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
	456	MODULE_DESCRIPTION("TC BPF based action");
	457	MODULE_LICENSE("GPL v2");