[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;

	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;

		return ret;
	}

	return -1;
}

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

	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);
}

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