[deliverable/linux.git] / net / sched / cls_bpf.c

/*
 * Berkeley Packet Filter based traffic classifier
 *
 * Might be used to classify traffic through flexible, user-defined and
 * possibly JIT-ed BPF filters for traffic control as an alternative to
 * ematches.
 *
 * (C) 2013 Daniel Borkmann <dborkman@redhat.com>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 */

#include <linux/module.h>
#include <linux/types.h>
#include <linux/skbuff.h>
#include <linux/filter.h>
#include <linux/bpf.h>

#include <net/rtnetlink.h>
#include <net/pkt_cls.h>
#include <net/sock.h>

MODULE_LICENSE("GPL");
MODULE_AUTHOR("Daniel Borkmann <dborkman@redhat.com>");
MODULE_DESCRIPTION("TC BPF based classifier");

#define CLS_BPF_NAME_LEN	256

struct cls_bpf_head {
	struct list_head plist;
	u32 hgen;
	struct rcu_head rcu;
};

struct cls_bpf_prog {
	struct bpf_prog *filter;
	struct list_head link;
	struct tcf_result res;
	bool exts_integrated;
	struct tcf_exts exts;
	u32 handle;
	union {
		u32 bpf_fd;
		u16 bpf_num_ops;
	};
	struct sock_filter *bpf_ops;
	const char *bpf_name;
	struct tcf_proto *tp;
	struct rcu_head rcu;
};

static const struct nla_policy bpf_policy[TCA_BPF_MAX + 1] = {
	[TCA_BPF_CLASSID]	= { .type = NLA_U32 },
	[TCA_BPF_FLAGS]		= { .type = NLA_U32 },
	[TCA_BPF_FD]		= { .type = NLA_U32 },
	[TCA_BPF_NAME]		= { .type = NLA_NUL_STRING, .len = CLS_BPF_NAME_LEN },
	[TCA_BPF_OPS_LEN]	= { .type = NLA_U16 },
	[TCA_BPF_OPS]		= { .type = NLA_BINARY,
				    .len = sizeof(struct sock_filter) * BPF_MAXINSNS },
};

static int cls_bpf_exec_opcode(int code)
{
	switch (code) {
	case TC_ACT_OK:
	case TC_ACT_SHOT:
	case TC_ACT_STOLEN:
	case TC_ACT_REDIRECT:
	case TC_ACT_UNSPEC:
		return code;
	default:
		return TC_ACT_UNSPEC;
	}
}

static int cls_bpf_classify(struct sk_buff *skb, const struct tcf_proto *tp,
			    struct tcf_result *res)
{
	struct cls_bpf_head *head = rcu_dereference_bh(tp->root);
	struct cls_bpf_prog *prog;
#ifdef CONFIG_NET_CLS_ACT
	bool at_ingress = G_TC_AT(skb->tc_verd) & AT_INGRESS;
#else
	bool at_ingress = false;
#endif
	int ret = -1;

	if (unlikely(!skb_mac_header_was_set(skb)))
		return -1;

	/* Needed here for accessing maps. */
	rcu_read_lock();
	list_for_each_entry_rcu(prog, &head->plist, link) {
		int filter_res;

		qdisc_skb_cb(skb)->tc_classid = prog->res.classid;

		if (at_ingress) {
			/* It is safe to push/pull even if skb_shared() */
			__skb_push(skb, skb->mac_len);
			filter_res = BPF_PROG_RUN(prog->filter, skb);
			__skb_pull(skb, skb->mac_len);
		} else {
			filter_res = BPF_PROG_RUN(prog->filter, skb);
		}

		if (prog->exts_integrated) {
			res->class = prog->res.class;
			res->classid = qdisc_skb_cb(skb)->tc_classid;

			ret = cls_bpf_exec_opcode(filter_res);
			if (ret == TC_ACT_UNSPEC)
				continue;
			break;
		}

		if (filter_res == 0)
			continue;

		*res = prog->res;
		if (filter_res != -1)
			res->classid = filter_res;

		ret = tcf_exts_exec(skb, &prog->exts, res);
		if (ret < 0)
			continue;

		break;
	}
	rcu_read_unlock();

	return ret;
}

static bool cls_bpf_is_ebpf(const struct cls_bpf_prog *prog)
{
	return !prog->bpf_ops;
}

static int cls_bpf_init(struct tcf_proto *tp)
{
	struct cls_bpf_head *head;

	head = kzalloc(sizeof(*head), GFP_KERNEL);
	if (head == NULL)
		return -ENOBUFS;

	INIT_LIST_HEAD_RCU(&head->plist);
	rcu_assign_pointer(tp->root, head);

	return 0;
}

static void cls_bpf_delete_prog(struct tcf_proto *tp, struct cls_bpf_prog *prog)
{
	tcf_exts_destroy(&prog->exts);

	if (cls_bpf_is_ebpf(prog))
		bpf_prog_put(prog->filter);
	else
		bpf_prog_destroy(prog->filter);

	kfree(prog->bpf_name);
	kfree(prog->bpf_ops);
	kfree(prog);
}

static void __cls_bpf_delete_prog(struct rcu_head *rcu)
{
	struct cls_bpf_prog *prog = container_of(rcu, struct cls_bpf_prog, rcu);

	cls_bpf_delete_prog(prog->tp, prog);
}

static int cls_bpf_delete(struct tcf_proto *tp, unsigned long arg)
{
	struct cls_bpf_prog *prog = (struct cls_bpf_prog *) arg;

	list_del_rcu(&prog->link);
	tcf_unbind_filter(tp, &prog->res);
	call_rcu(&prog->rcu, __cls_bpf_delete_prog);

	return 0;
}

static bool cls_bpf_destroy(struct tcf_proto *tp, bool force)
{
	struct cls_bpf_head *head = rtnl_dereference(tp->root);
	struct cls_bpf_prog *prog, *tmp;

	if (!force && !list_empty(&head->plist))
		return false;

	list_for_each_entry_safe(prog, tmp, &head->plist, link) {
		list_del_rcu(&prog->link);
		tcf_unbind_filter(tp, &prog->res);
		call_rcu(&prog->rcu, __cls_bpf_delete_prog);
	}

	RCU_INIT_POINTER(tp->root, NULL);
	kfree_rcu(head, rcu);
	return true;
}

static unsigned long cls_bpf_get(struct tcf_proto *tp, u32 handle)
{
	struct cls_bpf_head *head = rtnl_dereference(tp->root);
	struct cls_bpf_prog *prog;
	unsigned long ret = 0UL;

	if (head == NULL)
		return 0UL;

	list_for_each_entry(prog, &head->plist, link) {
		if (prog->handle == handle) {
			ret = (unsigned long) prog;
			break;
		}
	}

	return ret;
}

static int cls_bpf_prog_from_ops(struct nlattr **tb, struct cls_bpf_prog *prog)
{
	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_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_BPF_OPS]))
		return -EINVAL;

	bpf_ops = kzalloc(bpf_size, GFP_KERNEL);
	if (bpf_ops == NULL)
		return -ENOMEM;

	memcpy(bpf_ops, nla_data(tb[TCA_BPF_OPS]), bpf_size);

	fprog_tmp.len = bpf_num_ops;
	fprog_tmp.filter = bpf_ops;

	ret = bpf_prog_create(&fp, &fprog_tmp);
	if (ret < 0) {
		kfree(bpf_ops);
		return ret;
	}

	prog->bpf_ops = bpf_ops;
	prog->bpf_num_ops = bpf_num_ops;
	prog->bpf_name = NULL;
	prog->filter = fp;

	return 0;
}

static int cls_bpf_prog_from_efd(struct nlattr **tb, struct cls_bpf_prog *prog,
				 const struct tcf_proto *tp)
{
	struct bpf_prog *fp;
	char *name = NULL;
	u32 bpf_fd;

	bpf_fd = nla_get_u32(tb[TCA_BPF_FD]);

	fp = bpf_prog_get(bpf_fd);
	if (IS_ERR(fp))
		return PTR_ERR(fp);

	if (fp->type != BPF_PROG_TYPE_SCHED_CLS) {
		bpf_prog_put(fp);
		return -EINVAL;
	}

	if (tb[TCA_BPF_NAME]) {
		name = kmemdup(nla_data(tb[TCA_BPF_NAME]),
			       nla_len(tb[TCA_BPF_NAME]),
			       GFP_KERNEL);
		if (!name) {
			bpf_prog_put(fp);
			return -ENOMEM;
		}
	}

	prog->bpf_ops = NULL;
	prog->bpf_fd = bpf_fd;
	prog->bpf_name = name;
	prog->filter = fp;

	if (fp->dst_needed)
		netif_keep_dst(qdisc_dev(tp->q));

	return 0;
}

static int cls_bpf_modify_existing(struct net *net, struct tcf_proto *tp,
				   struct cls_bpf_prog *prog,
				   unsigned long base, struct nlattr **tb,
				   struct nlattr *est, bool ovr)
{
	bool is_bpf, is_ebpf, have_exts = false;
	struct tcf_exts exts;
	int ret;

	is_bpf = tb[TCA_BPF_OPS_LEN] && tb[TCA_BPF_OPS];
	is_ebpf = tb[TCA_BPF_FD];
	if ((!is_bpf && !is_ebpf) || (is_bpf && is_ebpf))
		return -EINVAL;

	tcf_exts_init(&exts, TCA_BPF_ACT, TCA_BPF_POLICE);
	ret = tcf_exts_validate(net, tp, tb, est, &exts, ovr);
	if (ret < 0)
		return ret;

	if (tb[TCA_BPF_FLAGS]) {
		u32 bpf_flags = nla_get_u32(tb[TCA_BPF_FLAGS]);

		if (bpf_flags & ~TCA_BPF_FLAG_ACT_DIRECT) {
			tcf_exts_destroy(&exts);
			return -EINVAL;
		}

		have_exts = bpf_flags & TCA_BPF_FLAG_ACT_DIRECT;
	}

	prog->exts_integrated = have_exts;

	ret = is_bpf ? cls_bpf_prog_from_ops(tb, prog) :
		       cls_bpf_prog_from_efd(tb, prog, tp);
	if (ret < 0) {
		tcf_exts_destroy(&exts);
		return ret;
	}

	if (tb[TCA_BPF_CLASSID]) {
		prog->res.classid = nla_get_u32(tb[TCA_BPF_CLASSID]);
		tcf_bind_filter(tp, &prog->res, base);
	}

	tcf_exts_change(tp, &prog->exts, &exts);
	return 0;
}

static u32 cls_bpf_grab_new_handle(struct tcf_proto *tp,
				   struct cls_bpf_head *head)
{
	unsigned int i = 0x80000000;
	u32 handle;

	do {
		if (++head->hgen == 0x7FFFFFFF)
			head->hgen = 1;
	} while (--i > 0 && cls_bpf_get(tp, head->hgen));

	if (unlikely(i == 0)) {
		pr_err("Insufficient number of handles\n");
		handle = 0;
	} else {
		handle = head->hgen;
	}

	return handle;
}

static int cls_bpf_change(struct net *net, struct sk_buff *in_skb,
			  struct tcf_proto *tp, unsigned long base,
			  u32 handle, struct nlattr **tca,
			  unsigned long *arg, bool ovr)
{
	struct cls_bpf_head *head = rtnl_dereference(tp->root);
	struct cls_bpf_prog *oldprog = (struct cls_bpf_prog *) *arg;
	struct nlattr *tb[TCA_BPF_MAX + 1];
	struct cls_bpf_prog *prog;
	int ret;

	if (tca[TCA_OPTIONS] == NULL)
		return -EINVAL;

	ret = nla_parse_nested(tb, TCA_BPF_MAX, tca[TCA_OPTIONS], bpf_policy);
	if (ret < 0)
		return ret;

	prog = kzalloc(sizeof(*prog), GFP_KERNEL);
	if (!prog)
		return -ENOBUFS;

	tcf_exts_init(&prog->exts, TCA_BPF_ACT, TCA_BPF_POLICE);

	if (oldprog) {
		if (handle && oldprog->handle != handle) {
			ret = -EINVAL;
			goto errout;
		}
	}

	if (handle == 0)
		prog->handle = cls_bpf_grab_new_handle(tp, head);
	else
		prog->handle = handle;
	if (prog->handle == 0) {
		ret = -EINVAL;
		goto errout;
	}

	ret = cls_bpf_modify_existing(net, tp, prog, base, tb, tca[TCA_RATE], ovr);
	if (ret < 0)
		goto errout;

	if (oldprog) {
		list_replace_rcu(&oldprog->link, &prog->link);
		tcf_unbind_filter(tp, &oldprog->res);
		call_rcu(&oldprog->rcu, __cls_bpf_delete_prog);
	} else {
		list_add_rcu(&prog->link, &head->plist);
	}

	*arg = (unsigned long) prog;
	return 0;
errout:
	kfree(prog);

	return ret;
}

static int cls_bpf_dump_bpf_info(const struct cls_bpf_prog *prog,
				 struct sk_buff *skb)
{
	struct nlattr *nla;

	if (nla_put_u16(skb, TCA_BPF_OPS_LEN, prog->bpf_num_ops))
		return -EMSGSIZE;

	nla = nla_reserve(skb, TCA_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 cls_bpf_dump_ebpf_info(const struct cls_bpf_prog *prog,
				  struct sk_buff *skb)
{
	if (nla_put_u32(skb, TCA_BPF_FD, prog->bpf_fd))
		return -EMSGSIZE;

	if (prog->bpf_name &&
	    nla_put_string(skb, TCA_BPF_NAME, prog->bpf_name))
		return -EMSGSIZE;

	return 0;
}

static int cls_bpf_dump(struct net *net, struct tcf_proto *tp, unsigned long fh,
			struct sk_buff *skb, struct tcmsg *tm)
{
	struct cls_bpf_prog *prog = (struct cls_bpf_prog *) fh;
	struct nlattr *nest;
	u32 bpf_flags = 0;
	int ret;

	if (prog == NULL)
		return skb->len;

	tm->tcm_handle = prog->handle;

	nest = nla_nest_start(skb, TCA_OPTIONS);
	if (nest == NULL)
		goto nla_put_failure;

	if (prog->res.classid &&
	    nla_put_u32(skb, TCA_BPF_CLASSID, prog->res.classid))
		goto nla_put_failure;

	if (cls_bpf_is_ebpf(prog))
		ret = cls_bpf_dump_ebpf_info(prog, skb);
	else
		ret = cls_bpf_dump_bpf_info(prog, skb);
	if (ret)
		goto nla_put_failure;

	if (tcf_exts_dump(skb, &prog->exts) < 0)
		goto nla_put_failure;

	if (prog->exts_integrated)
		bpf_flags |= TCA_BPF_FLAG_ACT_DIRECT;
	if (bpf_flags && nla_put_u32(skb, TCA_BPF_FLAGS, bpf_flags))
		goto nla_put_failure;

	nla_nest_end(skb, nest);

	if (tcf_exts_dump_stats(skb, &prog->exts) < 0)
		goto nla_put_failure;

	return skb->len;

nla_put_failure:
	nla_nest_cancel(skb, nest);
	return -1;
}

static void cls_bpf_walk(struct tcf_proto *tp, struct tcf_walker *arg)
{
	struct cls_bpf_head *head = rtnl_dereference(tp->root);
	struct cls_bpf_prog *prog;

	list_for_each_entry(prog, &head->plist, link) {
		if (arg->count < arg->skip)
			goto skip;
		if (arg->fn(tp, (unsigned long) prog, arg) < 0) {
			arg->stop = 1;
			break;
		}
skip:
		arg->count++;
	}
}

static struct tcf_proto_ops cls_bpf_ops __read_mostly = {
	.kind		=	"bpf",
	.owner		=	THIS_MODULE,
	.classify	=	cls_bpf_classify,
	.init		=	cls_bpf_init,
	.destroy	=	cls_bpf_destroy,
	.get		=	cls_bpf_get,
	.change		=	cls_bpf_change,
	.delete		=	cls_bpf_delete,
	.walk		=	cls_bpf_walk,
	.dump		=	cls_bpf_dump,
};

static int __init cls_bpf_init_mod(void)
{
	return register_tcf_proto_ops(&cls_bpf_ops);
}

static void __exit cls_bpf_exit_mod(void)
{
	unregister_tcf_proto_ops(&cls_bpf_ops);
}

module_init(cls_bpf_init_mod);
module_exit(cls_bpf_exit_mod);
Commit	Line	Data
7d1d65cb DB	1	/*
	2	* Berkeley Packet Filter based traffic classifier
	3	*
	4	* Might be used to classify traffic through flexible, user-defined and
	5	* possibly JIT-ed BPF filters for traffic control as an alternative to
	6	* ematches.
	7	*
	8	* (C) 2013 Daniel Borkmann <dborkman@redhat.com>
	9	*
	10	* This program is free software; you can redistribute it and/or modify
	11	* it under the terms of the GNU General Public License version 2 as
	12	* published by the Free Software Foundation.
	13	*/
	14
	15	#include <linux/module.h>
	16	#include <linux/types.h>
	17	#include <linux/skbuff.h>
	18	#include <linux/filter.h>
e2e9b654 DB	19	#include <linux/bpf.h>
e2e9b654 DB	20
7d1d65cb DB	21	#include <net/rtnetlink.h>
	22	#include <net/pkt_cls.h>
	23	#include <net/sock.h>
	24
	25	MODULE_LICENSE("GPL");
	26	MODULE_AUTHOR("Daniel Borkmann <dborkman@redhat.com>");
	27	MODULE_DESCRIPTION("TC BPF based classifier");
	28
e2e9b654 DB	29	#define CLS_BPF_NAME_LEN 256
e2e9b654 DB	30
7d1d65cb DB	31	struct cls_bpf_head {
	32	struct list_head plist;
	33	u32 hgen;
1f947bf1	34	struct rcu_head rcu;
7d1d65cb DB	35	};
	36
	37	struct cls_bpf_prog {
7ae457c1	38	struct bpf_prog *filter;
7d1d65cb	39	struct list_head link;
e2e9b654	40	struct tcf_result res;
045efa82	41	bool exts_integrated;
e2e9b654	42	struct tcf_exts exts;
7d1d65cb	43	u32 handle;
e2e9b654 DB	44	union {
	45	u32 bpf_fd;
	46	u16 bpf_num_ops;
	47	};
	48	struct sock_filter *bpf_ops;
	49	const char *bpf_name;
1f947bf1 JF	50	struct tcf_proto *tp;
1f947bf1 JF	51	struct rcu_head rcu;
7d1d65cb DB	52	};
	53
	54	static const struct nla_policy bpf_policy[TCA_BPF_MAX + 1] = {
	55	[TCA_BPF_CLASSID] = { .type = NLA_U32 },
045efa82	56	[TCA_BPF_FLAGS] = { .type = NLA_U32 },
e2e9b654 DB	57	[TCA_BPF_FD] = { .type = NLA_U32 },
e2e9b654 DB	58	[TCA_BPF_NAME] = { .type = NLA_NUL_STRING, .len = CLS_BPF_NAME_LEN },
7d1d65cb DB	59	[TCA_BPF_OPS_LEN] = { .type = NLA_U16 },
	60	[TCA_BPF_OPS] = { .type = NLA_BINARY,
	61	.len = sizeof(struct sock_filter) * BPF_MAXINSNS },
	62	};
	63
045efa82 DB	64	static int cls_bpf_exec_opcode(int code)
	65	{
	66	switch (code) {
	67	case TC_ACT_OK:
045efa82	68	case TC_ACT_SHOT:
045efa82	69	case TC_ACT_STOLEN:
27b29f63	70	case TC_ACT_REDIRECT:
045efa82 DB	71	case TC_ACT_UNSPEC:
	72	return code;
	73	default:
	74	return TC_ACT_UNSPEC;
	75	}
	76	}
	77
7d1d65cb DB	78	static int cls_bpf_classify(struct sk_buff skb, const struct tcf_proto tp,
	79	struct tcf_result *res)
	80	{
80dcbd12	81	struct cls_bpf_head *head = rcu_dereference_bh(tp->root);
7d1d65cb	82	struct cls_bpf_prog *prog;
3431205e AS	83	#ifdef CONFIG_NET_CLS_ACT
	84	bool at_ingress = G_TC_AT(skb->tc_verd) & AT_INGRESS;
	85	#else
	86	bool at_ingress = false;
	87	#endif
54720df1	88	int ret = -1;
7d1d65cb	89
a166151c AS	90	if (unlikely(!skb_mac_header_was_set(skb)))
	91	return -1;
	92
54720df1 DB	93	/* Needed here for accessing maps. */
54720df1 DB	94	rcu_read_lock();
1f947bf1	95	list_for_each_entry_rcu(prog, &head->plist, link) {
3431205e AS	96	int filter_res;
3431205e AS	97
045efa82 DB	98	qdisc_skb_cb(skb)->tc_classid = prog->res.classid;
045efa82 DB	99
3431205e AS	100	if (at_ingress) {
	101	/* It is safe to push/pull even if skb_shared() */
	102	__skb_push(skb, skb->mac_len);
	103	filter_res = BPF_PROG_RUN(prog->filter, skb);
	104	__skb_pull(skb, skb->mac_len);
	105	} else {
	106	filter_res = BPF_PROG_RUN(prog->filter, skb);
	107	}
7d1d65cb	108
045efa82 DB	109	if (prog->exts_integrated) {
	110	res->class = prog->res.class;
	111	res->classid = qdisc_skb_cb(skb)->tc_classid;
	112
	113	ret = cls_bpf_exec_opcode(filter_res);
	114	if (ret == TC_ACT_UNSPEC)
	115	continue;
	116	break;
	117	}
	118
7d1d65cb DB	119	if (filter_res == 0)
	120	continue;
	121
	122	*res = prog->res;
	123	if (filter_res != -1)
	124	res->classid = filter_res;
	125
	126	ret = tcf_exts_exec(skb, &prog->exts, res);
	127	if (ret < 0)
	128	continue;
	129
54720df1	130	break;
7d1d65cb	131	}
54720df1	132	rcu_read_unlock();
7d1d65cb	133
54720df1	134	return ret;
7d1d65cb DB	135	}
7d1d65cb DB	136
e2e9b654 DB	137	static bool cls_bpf_is_ebpf(const struct cls_bpf_prog *prog)
	138	{
	139	return !prog->bpf_ops;
	140	}
	141
7d1d65cb DB	142	static int cls_bpf_init(struct tcf_proto *tp)
	143	{
	144	struct cls_bpf_head *head;
	145
	146	head = kzalloc(sizeof(*head), GFP_KERNEL);
	147	if (head == NULL)
	148	return -ENOBUFS;
	149
1f947bf1 JF	150	INIT_LIST_HEAD_RCU(&head->plist);
1f947bf1 JF	151	rcu_assign_pointer(tp->root, head);
7d1d65cb DB	152
	153	return 0;
	154	}
	155
	156	static void cls_bpf_delete_prog(struct tcf_proto tp, struct cls_bpf_prog prog)
	157	{
18d0264f	158	tcf_exts_destroy(&prog->exts);
7d1d65cb	159
e2e9b654 DB	160	if (cls_bpf_is_ebpf(prog))
	161	bpf_prog_put(prog->filter);
	162	else
	163	bpf_prog_destroy(prog->filter);
7d1d65cb	164
e2e9b654	165	kfree(prog->bpf_name);
7d1d65cb DB	166	kfree(prog->bpf_ops);
	167	kfree(prog);
	168	}
	169
1f947bf1 JF	170	static void __cls_bpf_delete_prog(struct rcu_head *rcu)
	171	{
	172	struct cls_bpf_prog *prog = container_of(rcu, struct cls_bpf_prog, rcu);
	173
	174	cls_bpf_delete_prog(prog->tp, prog);
	175	}
	176
7d1d65cb DB	177	static int cls_bpf_delete(struct tcf_proto *tp, unsigned long arg)
7d1d65cb DB	178	{
472f5837	179	struct cls_bpf_prog prog = (struct cls_bpf_prog ) arg;
7d1d65cb	180
472f5837 JP	181	list_del_rcu(&prog->link);
	182	tcf_unbind_filter(tp, &prog->res);
	183	call_rcu(&prog->rcu, __cls_bpf_delete_prog);
e2e9b654	184
472f5837	185	return 0;
7d1d65cb DB	186	}
7d1d65cb DB	187
1e052be6	188	static bool cls_bpf_destroy(struct tcf_proto *tp, bool force)
7d1d65cb	189	{
1f947bf1	190	struct cls_bpf_head *head = rtnl_dereference(tp->root);
7d1d65cb DB	191	struct cls_bpf_prog prog, tmp;
7d1d65cb DB	192
1e052be6 CW	193	if (!force && !list_empty(&head->plist))
	194	return false;
	195
7d1d65cb	196	list_for_each_entry_safe(prog, tmp, &head->plist, link) {
1f947bf1	197	list_del_rcu(&prog->link);
18cdb37e	198	tcf_unbind_filter(tp, &prog->res);
1f947bf1	199	call_rcu(&prog->rcu, __cls_bpf_delete_prog);
7d1d65cb DB	200	}
7d1d65cb DB	201
1f947bf1 JF	202	RCU_INIT_POINTER(tp->root, NULL);
1f947bf1 JF	203	kfree_rcu(head, rcu);
1e052be6	204	return true;
7d1d65cb DB	205	}
	206
	207	static unsigned long cls_bpf_get(struct tcf_proto *tp, u32 handle)
	208	{
1f947bf1	209	struct cls_bpf_head *head = rtnl_dereference(tp->root);
7d1d65cb DB	210	struct cls_bpf_prog *prog;
	211	unsigned long ret = 0UL;
	212
	213	if (head == NULL)
	214	return 0UL;
	215
3fe6b49e	216	list_for_each_entry(prog, &head->plist, link) {
7d1d65cb DB	217	if (prog->handle == handle) {
	218	ret = (unsigned long) prog;
	219	break;
	220	}
	221	}
	222
	223	return ret;
	224	}
	225
045efa82	226	static int cls_bpf_prog_from_ops(struct nlattr *tb, struct cls_bpf_prog prog)
7d1d65cb	227	{
1f947bf1	228	struct sock_filter *bpf_ops;
e2e9b654	229	struct sock_fprog_kern fprog_tmp;
1f947bf1	230	struct bpf_prog *fp;
33e9fcc6	231	u16 bpf_size, bpf_num_ops;
7d1d65cb DB	232	int ret;
7d1d65cb DB	233
33e9fcc6	234	bpf_num_ops = nla_get_u16(tb[TCA_BPF_OPS_LEN]);
e2e9b654 DB	235	if (bpf_num_ops > BPF_MAXINSNS \|\| bpf_num_ops == 0)
e2e9b654 DB	236	return -EINVAL;
7d1d65cb	237
33e9fcc6	238	bpf_size = bpf_num_ops * sizeof(*bpf_ops);
e2e9b654 DB	239	if (bpf_size != nla_len(tb[TCA_BPF_OPS]))
e2e9b654 DB	240	return -EINVAL;
7913ecf6	241
7d1d65cb	242	bpf_ops = kzalloc(bpf_size, GFP_KERNEL);
e2e9b654 DB	243	if (bpf_ops == NULL)
e2e9b654 DB	244	return -ENOMEM;
7d1d65cb DB	245
	246	memcpy(bpf_ops, nla_data(tb[TCA_BPF_OPS]), bpf_size);
	247
e2e9b654 DB	248	fprog_tmp.len = bpf_num_ops;
e2e9b654 DB	249	fprog_tmp.filter = bpf_ops;
7d1d65cb	250
e2e9b654 DB	251	ret = bpf_prog_create(&fp, &fprog_tmp);
	252	if (ret < 0) {
	253	kfree(bpf_ops);
	254	return ret;
	255	}
7d1d65cb	256
7d1d65cb	257	prog->bpf_ops = bpf_ops;
e2e9b654 DB	258	prog->bpf_num_ops = bpf_num_ops;
e2e9b654 DB	259	prog->bpf_name = NULL;
7d1d65cb	260	prog->filter = fp;
7d1d65cb	261
e2e9b654 DB	262	return 0;
	263	}
	264
c46646d0 DB	265	static int cls_bpf_prog_from_efd(struct nlattr *tb, struct cls_bpf_prog prog,
c46646d0 DB	266	const struct tcf_proto *tp)
e2e9b654 DB	267	{
	268	struct bpf_prog *fp;
	269	char *name = NULL;
	270	u32 bpf_fd;
	271
	272	bpf_fd = nla_get_u32(tb[TCA_BPF_FD]);
	273
	274	fp = bpf_prog_get(bpf_fd);
	275	if (IS_ERR(fp))
	276	return PTR_ERR(fp);
	277
	278	if (fp->type != BPF_PROG_TYPE_SCHED_CLS) {
	279	bpf_prog_put(fp);
	280	return -EINVAL;
	281	}
	282
	283	if (tb[TCA_BPF_NAME]) {
	284	name = kmemdup(nla_data(tb[TCA_BPF_NAME]),
	285	nla_len(tb[TCA_BPF_NAME]),
	286	GFP_KERNEL);
	287	if (!name) {
	288	bpf_prog_put(fp);
	289	return -ENOMEM;
	290	}
	291	}
	292
	293	prog->bpf_ops = NULL;
	294	prog->bpf_fd = bpf_fd;
	295	prog->bpf_name = name;
e2e9b654	296	prog->filter = fp;
e2e9b654	297
c46646d0 DB	298	if (fp->dst_needed)
	299	netif_keep_dst(qdisc_dev(tp->q));
	300
e2e9b654 DB	301	return 0;
	302	}
	303
	304	static int cls_bpf_modify_existing(struct net net, struct tcf_proto tp,
	305	struct cls_bpf_prog *prog,
	306	unsigned long base, struct nlattr **tb,
	307	struct nlattr *est, bool ovr)
	308	{
045efa82	309	bool is_bpf, is_ebpf, have_exts = false;
e2e9b654	310	struct tcf_exts exts;
e2e9b654 DB	311	int ret;
	312
	313	is_bpf = tb[TCA_BPF_OPS_LEN] && tb[TCA_BPF_OPS];
	314	is_ebpf = tb[TCA_BPF_FD];
ef146fa4	315	if ((!is_bpf && !is_ebpf) \|\| (is_bpf && is_ebpf))
e2e9b654 DB	316	return -EINVAL;
	317
	318	tcf_exts_init(&exts, TCA_BPF_ACT, TCA_BPF_POLICE);
	319	ret = tcf_exts_validate(net, tp, tb, est, &exts, ovr);
	320	if (ret < 0)
	321	return ret;
	322
045efa82 DB	323	if (tb[TCA_BPF_FLAGS]) {
	324	u32 bpf_flags = nla_get_u32(tb[TCA_BPF_FLAGS]);
	325
	326	if (bpf_flags & ~TCA_BPF_FLAG_ACT_DIRECT) {
	327	tcf_exts_destroy(&exts);
	328	return -EINVAL;
	329	}
	330
	331	have_exts = bpf_flags & TCA_BPF_FLAG_ACT_DIRECT;
	332	}
	333
045efa82	334	prog->exts_integrated = have_exts;
e2e9b654	335
045efa82	336	ret = is_bpf ? cls_bpf_prog_from_ops(tb, prog) :
c46646d0	337	cls_bpf_prog_from_efd(tb, prog, tp);
e2e9b654 DB	338	if (ret < 0) {
	339	tcf_exts_destroy(&exts);
	340	return ret;
	341	}
	342
ef146fa4 DB	343	if (tb[TCA_BPF_CLASSID]) {
	344	prog->res.classid = nla_get_u32(tb[TCA_BPF_CLASSID]);
	345	tcf_bind_filter(tp, &prog->res, base);
	346	}
7d1d65cb	347
ef146fa4	348	tcf_exts_change(tp, &prog->exts, &exts);
7d1d65cb	349	return 0;
7d1d65cb DB	350	}
	351
	352	static u32 cls_bpf_grab_new_handle(struct tcf_proto *tp,
	353	struct cls_bpf_head *head)
	354	{
	355	unsigned int i = 0x80000000;
3f2ab135	356	u32 handle;
7d1d65cb DB	357
	358	do {
	359	if (++head->hgen == 0x7FFFFFFF)
	360	head->hgen = 1;
	361	} while (--i > 0 && cls_bpf_get(tp, head->hgen));
3f2ab135 DB	362
3f2ab135 DB	363	if (unlikely(i == 0)) {
7d1d65cb	364	pr_err("Insufficient number of handles\n");
3f2ab135 DB	365	handle = 0;
	366	} else {
	367	handle = head->hgen;
	368	}
7d1d65cb	369
3f2ab135	370	return handle;
7d1d65cb DB	371	}
	372
	373	static int cls_bpf_change(struct net net, struct sk_buff in_skb,
	374	struct tcf_proto *tp, unsigned long base,
	375	u32 handle, struct nlattr **tca,
2f7ef2f8	376	unsigned long *arg, bool ovr)
7d1d65cb	377	{
1f947bf1 JF	378	struct cls_bpf_head *head = rtnl_dereference(tp->root);
1f947bf1 JF	379	struct cls_bpf_prog oldprog = (struct cls_bpf_prog ) *arg;
7d1d65cb	380	struct nlattr *tb[TCA_BPF_MAX + 1];
1f947bf1	381	struct cls_bpf_prog *prog;
7d1d65cb DB	382	int ret;
	383
	384	if (tca[TCA_OPTIONS] == NULL)
	385	return -EINVAL;
	386
	387	ret = nla_parse_nested(tb, TCA_BPF_MAX, tca[TCA_OPTIONS], bpf_policy);
	388	if (ret < 0)
	389	return ret;
	390
7d1d65cb	391	prog = kzalloc(sizeof(*prog), GFP_KERNEL);
1f947bf1	392	if (!prog)
7d1d65cb DB	393	return -ENOBUFS;
7d1d65cb DB	394
5da57f42	395	tcf_exts_init(&prog->exts, TCA_BPF_ACT, TCA_BPF_POLICE);
1f947bf1 JF	396
	397	if (oldprog) {
	398	if (handle && oldprog->handle != handle) {
	399	ret = -EINVAL;
	400	goto errout;
	401	}
	402	}
	403
7d1d65cb DB	404	if (handle == 0)
	405	prog->handle = cls_bpf_grab_new_handle(tp, head);
	406	else
	407	prog->handle = handle;
	408	if (prog->handle == 0) {
	409	ret = -EINVAL;
	410	goto errout;
	411	}
	412
2f7ef2f8	413	ret = cls_bpf_modify_existing(net, tp, prog, base, tb, tca[TCA_RATE], ovr);
7d1d65cb DB	414	if (ret < 0)
	415	goto errout;
	416
1f947bf1	417	if (oldprog) {
f6bfc46d	418	list_replace_rcu(&oldprog->link, &prog->link);
18cdb37e	419	tcf_unbind_filter(tp, &oldprog->res);
1f947bf1 JF	420	call_rcu(&oldprog->rcu, __cls_bpf_delete_prog);
	421	} else {
	422	list_add_rcu(&prog->link, &head->plist);
	423	}
7d1d65cb DB	424
7d1d65cb DB	425	*arg = (unsigned long) prog;
7d1d65cb DB	426	return 0;
7d1d65cb DB	427	errout:
1f947bf1	428	kfree(prog);
7d1d65cb DB	429
	430	return ret;
	431	}
	432
e2e9b654 DB	433	static int cls_bpf_dump_bpf_info(const struct cls_bpf_prog *prog,
	434	struct sk_buff *skb)
	435	{
	436	struct nlattr *nla;
	437
	438	if (nla_put_u16(skb, TCA_BPF_OPS_LEN, prog->bpf_num_ops))
	439	return -EMSGSIZE;
	440
	441	nla = nla_reserve(skb, TCA_BPF_OPS, prog->bpf_num_ops *
	442	sizeof(struct sock_filter));
	443	if (nla == NULL)
	444	return -EMSGSIZE;
	445
	446	memcpy(nla_data(nla), prog->bpf_ops, nla_len(nla));
	447
	448	return 0;
	449	}
	450
	451	static int cls_bpf_dump_ebpf_info(const struct cls_bpf_prog *prog,
	452	struct sk_buff *skb)
	453	{
	454	if (nla_put_u32(skb, TCA_BPF_FD, prog->bpf_fd))
	455	return -EMSGSIZE;
	456
	457	if (prog->bpf_name &&
	458	nla_put_string(skb, TCA_BPF_NAME, prog->bpf_name))
	459	return -EMSGSIZE;
	460
	461	return 0;
	462	}
	463
832d1d5b	464	static int cls_bpf_dump(struct net net, struct tcf_proto tp, unsigned long fh,
7d1d65cb DB	465	struct sk_buff skb, struct tcmsg tm)
	466	{
	467	struct cls_bpf_prog prog = (struct cls_bpf_prog ) fh;
e2e9b654	468	struct nlattr *nest;
bf007d1c	469	u32 bpf_flags = 0;
e2e9b654	470	int ret;
7d1d65cb DB	471
	472	if (prog == NULL)
	473	return skb->len;
	474
	475	tm->tcm_handle = prog->handle;
	476
	477	nest = nla_nest_start(skb, TCA_OPTIONS);
	478	if (nest == NULL)
	479	goto nla_put_failure;
	480
ef146fa4 DB	481	if (prog->res.classid &&
ef146fa4 DB	482	nla_put_u32(skb, TCA_BPF_CLASSID, prog->res.classid))
7d1d65cb	483	goto nla_put_failure;
7d1d65cb	484
e2e9b654 DB	485	if (cls_bpf_is_ebpf(prog))
	486	ret = cls_bpf_dump_ebpf_info(prog, skb);
	487	else
	488	ret = cls_bpf_dump_bpf_info(prog, skb);
	489	if (ret)
7d1d65cb DB	490	goto nla_put_failure;
7d1d65cb DB	491
5da57f42	492	if (tcf_exts_dump(skb, &prog->exts) < 0)
7d1d65cb DB	493	goto nla_put_failure;
7d1d65cb DB	494
bf007d1c DB	495	if (prog->exts_integrated)
	496	bpf_flags \|= TCA_BPF_FLAG_ACT_DIRECT;
	497	if (bpf_flags && nla_put_u32(skb, TCA_BPF_FLAGS, bpf_flags))
	498	goto nla_put_failure;
	499
7d1d65cb DB	500	nla_nest_end(skb, nest);
7d1d65cb DB	501
5da57f42	502	if (tcf_exts_dump_stats(skb, &prog->exts) < 0)
7d1d65cb DB	503	goto nla_put_failure;
	504
	505	return skb->len;
	506
	507	nla_put_failure:
	508	nla_nest_cancel(skb, nest);
	509	return -1;
	510	}
	511
	512	static void cls_bpf_walk(struct tcf_proto tp, struct tcf_walker arg)
	513	{
1f947bf1	514	struct cls_bpf_head *head = rtnl_dereference(tp->root);
7d1d65cb DB	515	struct cls_bpf_prog *prog;
7d1d65cb DB	516
3fe6b49e	517	list_for_each_entry(prog, &head->plist, link) {
7d1d65cb DB	518	if (arg->count < arg->skip)
	519	goto skip;
	520	if (arg->fn(tp, (unsigned long) prog, arg) < 0) {
	521	arg->stop = 1;
	522	break;
	523	}
	524	skip:
	525	arg->count++;
	526	}
	527	}
	528
	529	static struct tcf_proto_ops cls_bpf_ops __read_mostly = {
	530	.kind = "bpf",
	531	.owner = THIS_MODULE,
	532	.classify = cls_bpf_classify,
	533	.init = cls_bpf_init,
	534	.destroy = cls_bpf_destroy,
	535	.get = cls_bpf_get,
7d1d65cb DB	536	.change = cls_bpf_change,
	537	.delete = cls_bpf_delete,
	538	.walk = cls_bpf_walk,
	539	.dump = cls_bpf_dump,
	540	};
	541
	542	static int __init cls_bpf_init_mod(void)
	543	{
	544	return register_tcf_proto_ops(&cls_bpf_ops);
	545	}
	546
	547	static void __exit cls_bpf_exit_mod(void)
	548	{
	549	unregister_tcf_proto_ops(&cls_bpf_ops);
	550	}
	551
	552	module_init(cls_bpf_init_mod);
	553	module_exit(cls_bpf_exit_mod);