[linux-2.6-block.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;
	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_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_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;
	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 = BPF_PROG_RUN(prog->filter, skb);

		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, u32 classid)
{
	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;
	prog->res.classid = classid;

	return 0;
}

static int cls_bpf_prog_from_efd(struct nlattr **tb,
				 struct cls_bpf_prog *prog, u32 classid)
{
	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;
	prog->res.classid = classid;

	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)
{
	struct tcf_exts exts;
	bool is_bpf, is_ebpf;
	u32 classid;
	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) ||
	    !tb[TCA_BPF_CLASSID])
		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;

	classid = nla_get_u32(tb[TCA_BPF_CLASSID]);

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

	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(&prog->link, &oldprog->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;
	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 (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;

	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 DB	40	struct tcf_result res;
e2e9b654 DB	41	struct tcf_exts exts;
7d1d65cb	42	u32 handle;
e2e9b654 DB	43	union {
	44	u32 bpf_fd;
	45	u16 bpf_num_ops;
	46	};
	47	struct sock_filter *bpf_ops;
	48	const char *bpf_name;
1f947bf1 JF	49	struct tcf_proto *tp;
1f947bf1 JF	50	struct rcu_head rcu;
7d1d65cb DB	51	};
	52
	53	static const struct nla_policy bpf_policy[TCA_BPF_MAX + 1] = {
	54	[TCA_BPF_CLASSID] = { .type = NLA_U32 },
e2e9b654 DB	55	[TCA_BPF_FD] = { .type = NLA_U32 },
e2e9b654 DB	56	[TCA_BPF_NAME] = { .type = NLA_NUL_STRING, .len = CLS_BPF_NAME_LEN },
7d1d65cb DB	57	[TCA_BPF_OPS_LEN] = { .type = NLA_U16 },
	58	[TCA_BPF_OPS] = { .type = NLA_BINARY,
	59	.len = sizeof(struct sock_filter) * BPF_MAXINSNS },
	60	};
	61
7d1d65cb DB	62	static int cls_bpf_classify(struct sk_buff skb, const struct tcf_proto tp,
	63	struct tcf_result *res)
	64	{
80dcbd12	65	struct cls_bpf_head *head = rcu_dereference_bh(tp->root);
7d1d65cb	66	struct cls_bpf_prog *prog;
54720df1	67	int ret = -1;
7d1d65cb	68
a166151c AS	69	if (unlikely(!skb_mac_header_was_set(skb)))
	70	return -1;
	71
54720df1 DB	72	/* Needed here for accessing maps. */
54720df1 DB	73	rcu_read_lock();
1f947bf1	74	list_for_each_entry_rcu(prog, &head->plist, link) {
7ae457c1	75	int filter_res = BPF_PROG_RUN(prog->filter, skb);
7d1d65cb DB	76
	77	if (filter_res == 0)
	78	continue;
	79
	80	*res = prog->res;
	81	if (filter_res != -1)
	82	res->classid = filter_res;
	83
	84	ret = tcf_exts_exec(skb, &prog->exts, res);
	85	if (ret < 0)
	86	continue;
	87
54720df1	88	break;
7d1d65cb	89	}
54720df1	90	rcu_read_unlock();
7d1d65cb	91
54720df1	92	return ret;
7d1d65cb DB	93	}
7d1d65cb DB	94
e2e9b654 DB	95	static bool cls_bpf_is_ebpf(const struct cls_bpf_prog *prog)
	96	{
	97	return !prog->bpf_ops;
	98	}
	99
7d1d65cb DB	100	static int cls_bpf_init(struct tcf_proto *tp)
	101	{
	102	struct cls_bpf_head *head;
	103
	104	head = kzalloc(sizeof(*head), GFP_KERNEL);
	105	if (head == NULL)
	106	return -ENOBUFS;
	107
1f947bf1 JF	108	INIT_LIST_HEAD_RCU(&head->plist);
1f947bf1 JF	109	rcu_assign_pointer(tp->root, head);
7d1d65cb DB	110
	111	return 0;
	112	}
	113
	114	static void cls_bpf_delete_prog(struct tcf_proto tp, struct cls_bpf_prog prog)
	115	{
18d0264f	116	tcf_exts_destroy(&prog->exts);
7d1d65cb	117
e2e9b654 DB	118	if (cls_bpf_is_ebpf(prog))
	119	bpf_prog_put(prog->filter);
	120	else
	121	bpf_prog_destroy(prog->filter);
7d1d65cb	122
e2e9b654	123	kfree(prog->bpf_name);
7d1d65cb DB	124	kfree(prog->bpf_ops);
	125	kfree(prog);
	126	}
	127
1f947bf1 JF	128	static void __cls_bpf_delete_prog(struct rcu_head *rcu)
	129	{
	130	struct cls_bpf_prog *prog = container_of(rcu, struct cls_bpf_prog, rcu);
	131
	132	cls_bpf_delete_prog(prog->tp, prog);
	133	}
	134
7d1d65cb DB	135	static int cls_bpf_delete(struct tcf_proto *tp, unsigned long arg)
7d1d65cb DB	136	{
472f5837	137	struct cls_bpf_prog prog = (struct cls_bpf_prog ) arg;
7d1d65cb	138
472f5837 JP	139	list_del_rcu(&prog->link);
	140	tcf_unbind_filter(tp, &prog->res);
	141	call_rcu(&prog->rcu, __cls_bpf_delete_prog);
e2e9b654	142
472f5837	143	return 0;
7d1d65cb DB	144	}
7d1d65cb DB	145
1e052be6	146	static bool cls_bpf_destroy(struct tcf_proto *tp, bool force)
7d1d65cb	147	{
1f947bf1	148	struct cls_bpf_head *head = rtnl_dereference(tp->root);
7d1d65cb DB	149	struct cls_bpf_prog prog, tmp;
7d1d65cb DB	150
1e052be6 CW	151	if (!force && !list_empty(&head->plist))
	152	return false;
	153
7d1d65cb	154	list_for_each_entry_safe(prog, tmp, &head->plist, link) {
1f947bf1	155	list_del_rcu(&prog->link);
18cdb37e	156	tcf_unbind_filter(tp, &prog->res);
1f947bf1	157	call_rcu(&prog->rcu, __cls_bpf_delete_prog);
7d1d65cb DB	158	}
7d1d65cb DB	159
1f947bf1 JF	160	RCU_INIT_POINTER(tp->root, NULL);
1f947bf1 JF	161	kfree_rcu(head, rcu);
1e052be6	162	return true;
7d1d65cb DB	163	}
	164
	165	static unsigned long cls_bpf_get(struct tcf_proto *tp, u32 handle)
	166	{
1f947bf1	167	struct cls_bpf_head *head = rtnl_dereference(tp->root);
7d1d65cb DB	168	struct cls_bpf_prog *prog;
	169	unsigned long ret = 0UL;
	170
	171	if (head == NULL)
	172	return 0UL;
	173
3fe6b49e	174	list_for_each_entry(prog, &head->plist, link) {
7d1d65cb DB	175	if (prog->handle == handle) {
	176	ret = (unsigned long) prog;
	177	break;
	178	}
	179	}
	180
	181	return ret;
	182	}
	183
e2e9b654 DB	184	static int cls_bpf_prog_from_ops(struct nlattr **tb,
e2e9b654 DB	185	struct cls_bpf_prog *prog, u32 classid)
7d1d65cb	186	{
1f947bf1	187	struct sock_filter *bpf_ops;
e2e9b654	188	struct sock_fprog_kern fprog_tmp;
1f947bf1	189	struct bpf_prog *fp;
33e9fcc6	190	u16 bpf_size, bpf_num_ops;
7d1d65cb DB	191	int ret;
7d1d65cb DB	192
33e9fcc6	193	bpf_num_ops = nla_get_u16(tb[TCA_BPF_OPS_LEN]);
e2e9b654 DB	194	if (bpf_num_ops > BPF_MAXINSNS \|\| bpf_num_ops == 0)
e2e9b654 DB	195	return -EINVAL;
7d1d65cb	196
33e9fcc6	197	bpf_size = bpf_num_ops * sizeof(*bpf_ops);
e2e9b654 DB	198	if (bpf_size != nla_len(tb[TCA_BPF_OPS]))
e2e9b654 DB	199	return -EINVAL;
7913ecf6	200
7d1d65cb	201	bpf_ops = kzalloc(bpf_size, GFP_KERNEL);
e2e9b654 DB	202	if (bpf_ops == NULL)
e2e9b654 DB	203	return -ENOMEM;
7d1d65cb DB	204
	205	memcpy(bpf_ops, nla_data(tb[TCA_BPF_OPS]), bpf_size);
	206
e2e9b654 DB	207	fprog_tmp.len = bpf_num_ops;
e2e9b654 DB	208	fprog_tmp.filter = bpf_ops;
7d1d65cb	209
e2e9b654 DB	210	ret = bpf_prog_create(&fp, &fprog_tmp);
	211	if (ret < 0) {
	212	kfree(bpf_ops);
	213	return ret;
	214	}
7d1d65cb	215
7d1d65cb	216	prog->bpf_ops = bpf_ops;
e2e9b654 DB	217	prog->bpf_num_ops = bpf_num_ops;
	218	prog->bpf_name = NULL;
	219
7d1d65cb DB	220	prog->filter = fp;
7d1d65cb DB	221	prog->res.classid = classid;
7d1d65cb	222
e2e9b654 DB	223	return 0;
	224	}
	225
	226	static int cls_bpf_prog_from_efd(struct nlattr **tb,
	227	struct cls_bpf_prog *prog, u32 classid)
	228	{
	229	struct bpf_prog *fp;
	230	char *name = NULL;
	231	u32 bpf_fd;
	232
	233	bpf_fd = nla_get_u32(tb[TCA_BPF_FD]);
	234
	235	fp = bpf_prog_get(bpf_fd);
	236	if (IS_ERR(fp))
	237	return PTR_ERR(fp);
	238
	239	if (fp->type != BPF_PROG_TYPE_SCHED_CLS) {
	240	bpf_prog_put(fp);
	241	return -EINVAL;
	242	}
	243
	244	if (tb[TCA_BPF_NAME]) {
	245	name = kmemdup(nla_data(tb[TCA_BPF_NAME]),
	246	nla_len(tb[TCA_BPF_NAME]),
	247	GFP_KERNEL);
	248	if (!name) {
	249	bpf_prog_put(fp);
	250	return -ENOMEM;
	251	}
	252	}
	253
	254	prog->bpf_ops = NULL;
	255	prog->bpf_fd = bpf_fd;
	256	prog->bpf_name = name;
	257
	258	prog->filter = fp;
	259	prog->res.classid = classid;
	260
	261	return 0;
	262	}
	263
	264	static int cls_bpf_modify_existing(struct net net, struct tcf_proto tp,
	265	struct cls_bpf_prog *prog,
	266	unsigned long base, struct nlattr **tb,
	267	struct nlattr *est, bool ovr)
	268	{
	269	struct tcf_exts exts;
	270	bool is_bpf, is_ebpf;
	271	u32 classid;
	272	int ret;
	273
	274	is_bpf = tb[TCA_BPF_OPS_LEN] && tb[TCA_BPF_OPS];
	275	is_ebpf = tb[TCA_BPF_FD];
	276
	277	if ((!is_bpf && !is_ebpf) \|\| (is_bpf && is_ebpf) \|\|
	278	!tb[TCA_BPF_CLASSID])
	279	return -EINVAL;
	280
	281	tcf_exts_init(&exts, TCA_BPF_ACT, TCA_BPF_POLICE);
	282	ret = tcf_exts_validate(net, tp, tb, est, &exts, ovr);
	283	if (ret < 0)
	284	return ret;
	285
	286	classid = nla_get_u32(tb[TCA_BPF_CLASSID]);
287
288	ret = is_bpf ? cls_bpf_prog_from_ops(tb, prog, classid) :
289	cls_bpf_prog_from_efd(tb, prog, classid);
290	if (ret < 0) {
291	tcf_exts_destroy(&exts);
292	return ret;
293	}
294
7d1d65cb DB	295	tcf_bind_filter(tp, &prog->res, base);
	296	tcf_exts_change(tp, &prog->exts, &exts);
	297
7d1d65cb	298	return 0;
7d1d65cb DB	299	}
	300
	301	static u32 cls_bpf_grab_new_handle(struct tcf_proto *tp,
	302	struct cls_bpf_head *head)
	303	{
	304	unsigned int i = 0x80000000;
3f2ab135	305	u32 handle;
7d1d65cb DB	306
	307	do {
	308	if (++head->hgen == 0x7FFFFFFF)
	309	head->hgen = 1;
	310	} while (--i > 0 && cls_bpf_get(tp, head->hgen));
3f2ab135 DB	311
3f2ab135 DB	312	if (unlikely(i == 0)) {
7d1d65cb	313	pr_err("Insufficient number of handles\n");
3f2ab135 DB	314	handle = 0;
	315	} else {
	316	handle = head->hgen;
	317	}
7d1d65cb	318
3f2ab135	319	return handle;
7d1d65cb DB	320	}
	321
	322	static int cls_bpf_change(struct net net, struct sk_buff in_skb,
	323	struct tcf_proto *tp, unsigned long base,
	324	u32 handle, struct nlattr **tca,
2f7ef2f8	325	unsigned long *arg, bool ovr)
7d1d65cb	326	{
1f947bf1 JF	327	struct cls_bpf_head *head = rtnl_dereference(tp->root);
1f947bf1 JF	328	struct cls_bpf_prog oldprog = (struct cls_bpf_prog ) *arg;
7d1d65cb	329	struct nlattr *tb[TCA_BPF_MAX + 1];
1f947bf1	330	struct cls_bpf_prog *prog;
7d1d65cb DB	331	int ret;
	332
	333	if (tca[TCA_OPTIONS] == NULL)
	334	return -EINVAL;
	335
	336	ret = nla_parse_nested(tb, TCA_BPF_MAX, tca[TCA_OPTIONS], bpf_policy);
	337	if (ret < 0)
	338	return ret;
	339
7d1d65cb	340	prog = kzalloc(sizeof(*prog), GFP_KERNEL);
1f947bf1	341	if (!prog)
7d1d65cb DB	342	return -ENOBUFS;
7d1d65cb DB	343
5da57f42	344	tcf_exts_init(&prog->exts, TCA_BPF_ACT, TCA_BPF_POLICE);
1f947bf1 JF	345
	346	if (oldprog) {
	347	if (handle && oldprog->handle != handle) {
	348	ret = -EINVAL;
	349	goto errout;
	350	}
	351	}
	352
7d1d65cb DB	353	if (handle == 0)
	354	prog->handle = cls_bpf_grab_new_handle(tp, head);
	355	else
	356	prog->handle = handle;
	357	if (prog->handle == 0) {
	358	ret = -EINVAL;
	359	goto errout;
	360	}
	361
2f7ef2f8	362	ret = cls_bpf_modify_existing(net, tp, prog, base, tb, tca[TCA_RATE], ovr);
7d1d65cb DB	363	if (ret < 0)
	364	goto errout;
	365
1f947bf1 JF	366	if (oldprog) {
1f947bf1 JF	367	list_replace_rcu(&prog->link, &oldprog->link);
18cdb37e	368	tcf_unbind_filter(tp, &oldprog->res);
1f947bf1 JF	369	call_rcu(&oldprog->rcu, __cls_bpf_delete_prog);
	370	} else {
	371	list_add_rcu(&prog->link, &head->plist);
	372	}
7d1d65cb DB	373
7d1d65cb DB	374	*arg = (unsigned long) prog;
7d1d65cb DB	375	return 0;
7d1d65cb DB	376	errout:
1f947bf1	377	kfree(prog);
7d1d65cb DB	378
	379	return ret;
	380	}
	381
e2e9b654 DB	382	static int cls_bpf_dump_bpf_info(const struct cls_bpf_prog *prog,
	383	struct sk_buff *skb)
	384	{
	385	struct nlattr *nla;
	386
	387	if (nla_put_u16(skb, TCA_BPF_OPS_LEN, prog->bpf_num_ops))
	388	return -EMSGSIZE;
	389
	390	nla = nla_reserve(skb, TCA_BPF_OPS, prog->bpf_num_ops *
	391	sizeof(struct sock_filter));
	392	if (nla == NULL)
	393	return -EMSGSIZE;
	394
	395	memcpy(nla_data(nla), prog->bpf_ops, nla_len(nla));
	396
	397	return 0;
	398	}
	399
	400	static int cls_bpf_dump_ebpf_info(const struct cls_bpf_prog *prog,
	401	struct sk_buff *skb)
	402	{
	403	if (nla_put_u32(skb, TCA_BPF_FD, prog->bpf_fd))
	404	return -EMSGSIZE;
	405
	406	if (prog->bpf_name &&
	407	nla_put_string(skb, TCA_BPF_NAME, prog->bpf_name))
	408	return -EMSGSIZE;
	409
	410	return 0;
	411	}
	412
832d1d5b	413	static int cls_bpf_dump(struct net net, struct tcf_proto tp, unsigned long fh,
7d1d65cb DB	414	struct sk_buff skb, struct tcmsg tm)
	415	{
	416	struct cls_bpf_prog prog = (struct cls_bpf_prog ) fh;
e2e9b654 DB	417	struct nlattr *nest;
e2e9b654 DB	418	int ret;
7d1d65cb DB	419
	420	if (prog == NULL)
	421	return skb->len;
	422
	423	tm->tcm_handle = prog->handle;
	424
	425	nest = nla_nest_start(skb, TCA_OPTIONS);
	426	if (nest == NULL)
	427	goto nla_put_failure;
	428
	429	if (nla_put_u32(skb, TCA_BPF_CLASSID, prog->res.classid))
	430	goto nla_put_failure;
7d1d65cb	431
e2e9b654 DB	432	if (cls_bpf_is_ebpf(prog))
	433	ret = cls_bpf_dump_ebpf_info(prog, skb);
	434	else
	435	ret = cls_bpf_dump_bpf_info(prog, skb);
	436	if (ret)
7d1d65cb DB	437	goto nla_put_failure;
7d1d65cb DB	438
5da57f42	439	if (tcf_exts_dump(skb, &prog->exts) < 0)
7d1d65cb DB	440	goto nla_put_failure;
	441
	442	nla_nest_end(skb, nest);
	443
5da57f42	444	if (tcf_exts_dump_stats(skb, &prog->exts) < 0)
7d1d65cb DB	445	goto nla_put_failure;
	446
	447	return skb->len;
	448
	449	nla_put_failure:
	450	nla_nest_cancel(skb, nest);
	451	return -1;
	452	}
	453
	454	static void cls_bpf_walk(struct tcf_proto tp, struct tcf_walker arg)
	455	{
1f947bf1	456	struct cls_bpf_head *head = rtnl_dereference(tp->root);
7d1d65cb DB	457	struct cls_bpf_prog *prog;
7d1d65cb DB	458
3fe6b49e	459	list_for_each_entry(prog, &head->plist, link) {
7d1d65cb DB	460	if (arg->count < arg->skip)
	461	goto skip;
	462	if (arg->fn(tp, (unsigned long) prog, arg) < 0) {
	463	arg->stop = 1;
	464	break;
	465	}
	466	skip:
	467	arg->count++;
	468	}
	469	}
	470
	471	static struct tcf_proto_ops cls_bpf_ops __read_mostly = {
	472	.kind = "bpf",
	473	.owner = THIS_MODULE,
	474	.classify = cls_bpf_classify,
	475	.init = cls_bpf_init,
	476	.destroy = cls_bpf_destroy,
	477	.get = cls_bpf_get,
7d1d65cb DB	478	.change = cls_bpf_change,
	479	.delete = cls_bpf_delete,
	480	.walk = cls_bpf_walk,
	481	.dump = cls_bpf_dump,
	482	};
	483
	484	static int __init cls_bpf_init_mod(void)
	485	{
	486	return register_tcf_proto_ops(&cls_bpf_ops);
	487	}
	488
	489	static void __exit cls_bpf_exit_mod(void)
	490	{
	491	unregister_tcf_proto_ops(&cls_bpf_ops);
	492	}
	493
	494	module_init(cls_bpf_init_mod);
	495	module_exit(cls_bpf_exit_mod);