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

/*
 * Copyright (c) 2015 Jiri Pirko <jiri@resnulli.us>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 */

#include <linux/module.h>
#include <linux/init.h>
#include <linux/kernel.h>
#include <linux/skbuff.h>
#include <linux/rtnetlink.h>
#include <linux/filter.h>
#include <linux/bpf.h>

#include <net/netlink.h>
#include <net/pkt_sched.h>

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

#define 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, 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_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, NULL);
	if (ret < 0)
		return ret;

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

	parm = nla_data(tb[TCA_ACT_BPF_PARMS]);

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

	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);

	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;

	prog->tcf_action = parm->action;
	rcu_assign_pointer(prog->filter, cfg.filter);
	spin_unlock_bh(&prog->tcf_lock);

	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;
out:
	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(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
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 ACT_BPF_NAME_LEN 256
	25
	26	struct tcf_bpf_cfg {
	27	struct bpf_prog *filter;
	28	struct sock_filter *bpf_ops;
f4eaed28	29	const char *bpf_name;
a8cb5f55	30	u16 bpf_num_ops;
f4eaed28	31	bool is_ebpf;
a8cb5f55	32	};
d23b8ad8	33
c7d03a00	34	static unsigned int bpf_net_id;
a85a970a	35	static struct tc_action_ops act_bpf_ops;
ddf97ccd	36
2fbec27f JHS	37	static int tcf_bpf_act(struct sk_buff skb, const struct tc_action act,
2fbec27f JHS	38	struct tcf_result *res)
d23b8ad8	39	{
f53d8c7b	40	bool at_ingress = skb_at_tc_ingress(skb);
a85a970a	41	struct tcf_bpf *prog = to_bpf(act);
cff82457	42	struct bpf_prog *filter;
ced585c8	43	int action, filter_res;
d23b8ad8	44
cff82457 AS	45	tcf_lastuse_update(&prog->tcf_tm);
cff82457 AS	46	bstats_cpu_update(this_cpu_ptr(prog->common.cpu_bstats), skb);
d23b8ad8	47
a8cb5f55	48	rcu_read_lock();
cff82457	49	filter = rcu_dereference(prog->filter);
3431205e AS	50	if (at_ingress) {
3431205e AS	51	__skb_push(skb, skb->mac_len);
6aaae2b6	52	bpf_compute_data_pointers(skb);
cff82457	53	filter_res = BPF_PROG_RUN(filter, skb);
3431205e AS	54	__skb_pull(skb, skb->mac_len);
3431205e AS	55	} else {
6aaae2b6	56	bpf_compute_data_pointers(skb);
cff82457	57	filter_res = BPF_PROG_RUN(filter, skb);
3431205e	58	}
a8cb5f55	59	rcu_read_unlock();
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	*
	68	* For everything else that is unkown, TC_ACT_UNSPEC is
	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,
789871bb	280	int replace, int bind, bool rtnl_held,
85d0966f	281	struct tcf_proto tp, struct netlink_ext_ack extack)
a8cb5f55	282	{
ddf97ccd	283	struct tc_action_net *tn = net_generic(net, bpf_net_id);
a8cb5f55	284	struct nlattr *tb[TCA_ACT_BPF_MAX + 1];
f4eaed28	285	struct tcf_bpf_cfg cfg, old;
a8cb5f55 DB	286	struct tc_act_bpf *parm;
a8cb5f55 DB	287	struct tcf_bpf *prog;
a8cb5f55	288	bool is_bpf, is_ebpf;
a5c90b29	289	int ret, res = 0;
a8cb5f55 DB	290
	291	if (!nla)
	292	return -EINVAL;
	293
fceb6435	294	ret = nla_parse_nested(tb, TCA_ACT_BPF_MAX, nla, act_bpf_policy, NULL);
a8cb5f55 DB	295	if (ret < 0)
	296	return ret;
	297
a5c90b29	298	if (!tb[TCA_ACT_BPF_PARMS])
a8cb5f55 DB	299	return -EINVAL;
	300
	301	parm = nla_data(tb[TCA_ACT_BPF_PARMS]);
	302
0190c1d4 VB	303	ret = tcf_idr_check_alloc(tn, &parm->index, act, bind);
0190c1d4 VB	304	if (!ret) {
65a206c0 CM	305	ret = tcf_idr_create(tn, parm->index, est, act,
65a206c0 CM	306	&act_bpf_ops, bind, true);
0190c1d4 VB	307	if (ret < 0) {
0190c1d4 VB	308	tcf_idr_cleanup(tn, parm->index);
a5c90b29	309	return ret;
0190c1d4	310	}
d23b8ad8	311
a5c90b29	312	res = ACT_P_CREATED;
0190c1d4	313	} else if (ret > 0) {
a8cb5f55	314	/* Don't override defaults. */
d23b8ad8	315	if (bind)
a5c90b29	316	return 0;
a8cb5f55	317
4e8ddd7f VB	318	if (!replace) {
4e8ddd7f VB	319	tcf_idr_release(*act, bind);
a5c90b29	320	return -EEXIST;
4e8ddd7f	321	}
0190c1d4 VB	322	} else {
0190c1d4 VB	323	return ret;
d23b8ad8 JP	324	}
d23b8ad8 JP	325
a5c90b29 DB	326	is_bpf = tb[TCA_ACT_BPF_OPS_LEN] && tb[TCA_ACT_BPF_OPS];
	327	is_ebpf = tb[TCA_ACT_BPF_FD];
	328
	329	if ((!is_bpf && !is_ebpf) \|\| (is_bpf && is_ebpf)) {
	330	ret = -EINVAL;
	331	goto out;
	332	}
	333
	334	memset(&cfg, 0, sizeof(cfg));
	335
	336	ret = is_bpf ? tcf_bpf_init_from_ops(tb, &cfg) :
	337	tcf_bpf_init_from_efd(tb, &cfg);
	338	if (ret < 0)
	339	goto out;
	340
a85a970a	341	prog = to_bpf(*act);
a8cb5f55	342
653cd284	343	spin_lock_bh(&prog->tcf_lock);
faa54be4	344	if (res != ACT_P_CREATED)
f4eaed28 DB	345	tcf_bpf_prog_fill_cfg(prog, &old);
f4eaed28 DB	346
a8cb5f55 DB	347	prog->bpf_ops = cfg.bpf_ops;
	348	prog->bpf_name = cfg.bpf_name;
	349
	350	if (cfg.bpf_num_ops)
	351	prog->bpf_num_ops = cfg.bpf_num_ops;
a8cb5f55 DB	352
a8cb5f55 DB	353	prog->tcf_action = parm->action;
cff82457	354	rcu_assign_pointer(prog->filter, cfg.filter);
653cd284	355	spin_unlock_bh(&prog->tcf_lock);
d23b8ad8	356
cff82457	357	if (res == ACT_P_CREATED) {
65a206c0	358	tcf_idr_insert(tn, *act);
cff82457 AS	359	} else {
	360	/* make sure the program being replaced is no longer executing */
	361	synchronize_rcu();
f4eaed28	362	tcf_bpf_cfg_cleanup(&old);
cff82457	363	}
a8cb5f55	364
a5c90b29 DB	365	return res;
a5c90b29 DB	366	out:
4e8ddd7f	367	tcf_idr_release(*act, bind);
d23b8ad8	368
d23b8ad8 JP	369	return ret;
	370	}
	371
9a63b255	372	static void tcf_bpf_cleanup(struct tc_action *act)
d23b8ad8	373	{
f4eaed28	374	struct tcf_bpf_cfg tmp;
ddf06c1e	375
a85a970a	376	tcf_bpf_prog_fill_cfg(to_bpf(act), &tmp);
f4eaed28	377	tcf_bpf_cfg_cleanup(&tmp);
d23b8ad8 JP	378	}
d23b8ad8 JP	379
ddf97ccd WC	380	static int tcf_bpf_walker(struct net net, struct sk_buff skb,
ddf97ccd WC	381	struct netlink_callback *cb, int type,
41780105 AA	382	const struct tc_action_ops *ops,
41780105 AA	383	struct netlink_ext_ack *extack)
ddf97ccd WC	384	{
	385	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	386
b3620145	387	return tcf_generic_walker(tn, skb, cb, type, ops, extack);
ddf97ccd WC	388	}
ddf97ccd WC	389
f061b48c	390	static int tcf_bpf_search(struct net net, struct tc_action *a, u32 index)
ddf97ccd WC	391	{
	392	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	393
65a206c0	394	return tcf_idr_search(tn, a, index);
ddf97ccd WC	395	}
ddf97ccd WC	396
a8cb5f55 DB	397	static struct tc_action_ops act_bpf_ops __read_mostly = {
a8cb5f55 DB	398	.kind = "bpf",
eddd2cf1	399	.id = TCA_ID_BPF,
a8cb5f55	400	.owner = THIS_MODULE,
2fbec27f	401	.act = tcf_bpf_act,
a8cb5f55 DB	402	.dump = tcf_bpf_dump,
	403	.cleanup = tcf_bpf_cleanup,
	404	.init = tcf_bpf_init,
ddf97ccd WC	405	.walk = tcf_bpf_walker,
ddf97ccd WC	406	.lookup = tcf_bpf_search,
a85a970a	407	.size = sizeof(struct tcf_bpf),
ddf97ccd WC	408	};
	409
	410	static __net_init int bpf_init_net(struct net *net)
	411	{
	412	struct tc_action_net *tn = net_generic(net, bpf_net_id);
	413
c7e460ce	414	return tc_action_net_init(tn, &act_bpf_ops);
ddf97ccd WC	415	}
ddf97ccd WC	416
039af9c6	417	static void __net_exit bpf_exit_net(struct list_head *net_list)
ddf97ccd	418	{
039af9c6	419	tc_action_net_exit(net_list, bpf_net_id);
ddf97ccd WC	420	}
	421
	422	static struct pernet_operations bpf_net_ops = {
	423	.init = bpf_init_net,
039af9c6	424	.exit_batch = bpf_exit_net,
ddf97ccd WC	425	.id = &bpf_net_id,
ddf97ccd WC	426	.size = sizeof(struct tc_action_net),
d23b8ad8 JP	427	};
	428
	429	static int __init bpf_init_module(void)
	430	{
ddf97ccd	431	return tcf_register_action(&act_bpf_ops, &bpf_net_ops);
d23b8ad8 JP	432	}
	433
	434	static void __exit bpf_cleanup_module(void)
	435	{
ddf97ccd	436	tcf_unregister_action(&act_bpf_ops, &bpf_net_ops);
d23b8ad8 JP	437	}
	438
	439	module_init(bpf_init_module);
	440	module_exit(bpf_cleanup_module);
	441
	442	MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
	443	MODULE_DESCRIPTION("TC BPF based action");
	444	MODULE_LICENSE("GPL v2");