[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;
	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);
	bool at_ingress = skb_at_tc_ingress(skb);
	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;

		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);
			bpf_compute_data_end(skb);
			filter_res = BPF_PROG_RUN(prog->filter, skb);
			__skb_pull(skb, skb->mac_len);
		} else {
			bpf_compute_data_end(skb);
			filter_res = BPF_PROG_RUN(prog->filter, skb);
		}

		if (prog->exts_integrated) {
			res->class   = 0;
			res->classid = TC_H_MAJ(prog->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;
		if (filter_res != -1) {
			res->class   = 0;
			res->classid = filter_res;
		} else {
			*res = prog->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 && !(tp->q->flags & TCQ_F_INGRESS))
		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);
fdc5432a	82	bool at_ingress = skb_at_tc_ingress(skb);
7d1d65cb	83	struct cls_bpf_prog *prog;
54720df1	84	int ret = -1;
7d1d65cb	85
a166151c AS	86	if (unlikely(!skb_mac_header_was_set(skb)))
	87	return -1;
	88
54720df1 DB	89	/* Needed here for accessing maps. */
54720df1 DB	90	rcu_read_lock();
1f947bf1	91	list_for_each_entry_rcu(prog, &head->plist, link) {
3431205e AS	92	int filter_res;
3431205e AS	93
045efa82 DB	94	qdisc_skb_cb(skb)->tc_classid = prog->res.classid;
045efa82 DB	95
3431205e AS	96	if (at_ingress) {
	97	/* It is safe to push/pull even if skb_shared() */
	98	__skb_push(skb, skb->mac_len);
db58ba45	99	bpf_compute_data_end(skb);
3431205e AS	100	filter_res = BPF_PROG_RUN(prog->filter, skb);
	101	__skb_pull(skb, skb->mac_len);
	102	} else {
db58ba45	103	bpf_compute_data_end(skb);
3431205e AS	104	filter_res = BPF_PROG_RUN(prog->filter, skb);
3431205e AS	105	}
7d1d65cb	106
045efa82	107	if (prog->exts_integrated) {
3a461da1 DB	108	res->class = 0;
	109	res->classid = TC_H_MAJ(prog->res.classid) \|
	110	qdisc_skb_cb(skb)->tc_classid;
045efa82 DB	111
	112	ret = cls_bpf_exec_opcode(filter_res);
	113	if (ret == TC_ACT_UNSPEC)
	114	continue;
	115	break;
	116	}
	117
7d1d65cb DB	118	if (filter_res == 0)
7d1d65cb DB	119	continue;
3a461da1 DB	120	if (filter_res != -1) {
3a461da1 DB	121	res->class = 0;
7d1d65cb	122	res->classid = filter_res;
3a461da1 DB	123	} else {
	124	*res = prog->res;
	125	}
7d1d65cb DB	126
	127	ret = tcf_exts_exec(skb, &prog->exts, res);
	128	if (ret < 0)
	129	continue;
	130
54720df1	131	break;
7d1d65cb	132	}
54720df1	133	rcu_read_unlock();
7d1d65cb	134
54720df1	135	return ret;
7d1d65cb DB	136	}
7d1d65cb DB	137
e2e9b654 DB	138	static bool cls_bpf_is_ebpf(const struct cls_bpf_prog *prog)
	139	{
	140	return !prog->bpf_ops;
	141	}
	142
7d1d65cb DB	143	static int cls_bpf_init(struct tcf_proto *tp)
	144	{
	145	struct cls_bpf_head *head;
	146
	147	head = kzalloc(sizeof(*head), GFP_KERNEL);
	148	if (head == NULL)
	149	return -ENOBUFS;
	150
1f947bf1 JF	151	INIT_LIST_HEAD_RCU(&head->plist);
1f947bf1 JF	152	rcu_assign_pointer(tp->root, head);
7d1d65cb DB	153
	154	return 0;
	155	}
	156
	157	static void cls_bpf_delete_prog(struct tcf_proto tp, struct cls_bpf_prog prog)
	158	{
18d0264f	159	tcf_exts_destroy(&prog->exts);
7d1d65cb	160
e2e9b654 DB	161	if (cls_bpf_is_ebpf(prog))
	162	bpf_prog_put(prog->filter);
	163	else
	164	bpf_prog_destroy(prog->filter);
7d1d65cb	165
e2e9b654	166	kfree(prog->bpf_name);
7d1d65cb DB	167	kfree(prog->bpf_ops);
	168	kfree(prog);
	169	}
	170
1f947bf1 JF	171	static void __cls_bpf_delete_prog(struct rcu_head *rcu)
	172	{
	173	struct cls_bpf_prog *prog = container_of(rcu, struct cls_bpf_prog, rcu);
	174
	175	cls_bpf_delete_prog(prog->tp, prog);
	176	}
	177
7d1d65cb DB	178	static int cls_bpf_delete(struct tcf_proto *tp, unsigned long arg)
7d1d65cb DB	179	{
472f5837	180	struct cls_bpf_prog prog = (struct cls_bpf_prog ) arg;
7d1d65cb	181
472f5837 JP	182	list_del_rcu(&prog->link);
	183	tcf_unbind_filter(tp, &prog->res);
	184	call_rcu(&prog->rcu, __cls_bpf_delete_prog);
e2e9b654	185
472f5837	186	return 0;
7d1d65cb DB	187	}
7d1d65cb DB	188
1e052be6	189	static bool cls_bpf_destroy(struct tcf_proto *tp, bool force)
7d1d65cb	190	{
1f947bf1	191	struct cls_bpf_head *head = rtnl_dereference(tp->root);
7d1d65cb DB	192	struct cls_bpf_prog prog, tmp;
7d1d65cb DB	193
1e052be6 CW	194	if (!force && !list_empty(&head->plist))
	195	return false;
	196
7d1d65cb	197	list_for_each_entry_safe(prog, tmp, &head->plist, link) {
1f947bf1	198	list_del_rcu(&prog->link);
18cdb37e	199	tcf_unbind_filter(tp, &prog->res);
1f947bf1	200	call_rcu(&prog->rcu, __cls_bpf_delete_prog);
7d1d65cb DB	201	}
7d1d65cb DB	202
1f947bf1 JF	203	RCU_INIT_POINTER(tp->root, NULL);
1f947bf1 JF	204	kfree_rcu(head, rcu);
1e052be6	205	return true;
7d1d65cb DB	206	}
	207
	208	static unsigned long cls_bpf_get(struct tcf_proto *tp, u32 handle)
	209	{
1f947bf1	210	struct cls_bpf_head *head = rtnl_dereference(tp->root);
7d1d65cb DB	211	struct cls_bpf_prog *prog;
	212	unsigned long ret = 0UL;
	213
	214	if (head == NULL)
	215	return 0UL;
	216
3fe6b49e	217	list_for_each_entry(prog, &head->plist, link) {
7d1d65cb DB	218	if (prog->handle == handle) {
	219	ret = (unsigned long) prog;
	220	break;
	221	}
	222	}
	223
	224	return ret;
	225	}
	226
045efa82	227	static int cls_bpf_prog_from_ops(struct nlattr *tb, struct cls_bpf_prog prog)
7d1d65cb	228	{
1f947bf1	229	struct sock_filter *bpf_ops;
e2e9b654	230	struct sock_fprog_kern fprog_tmp;
1f947bf1	231	struct bpf_prog *fp;
33e9fcc6	232	u16 bpf_size, bpf_num_ops;
7d1d65cb DB	233	int ret;
7d1d65cb DB	234
33e9fcc6	235	bpf_num_ops = nla_get_u16(tb[TCA_BPF_OPS_LEN]);
e2e9b654 DB	236	if (bpf_num_ops > BPF_MAXINSNS \|\| bpf_num_ops == 0)
e2e9b654 DB	237	return -EINVAL;
7d1d65cb	238
33e9fcc6	239	bpf_size = bpf_num_ops * sizeof(*bpf_ops);
e2e9b654 DB	240	if (bpf_size != nla_len(tb[TCA_BPF_OPS]))
e2e9b654 DB	241	return -EINVAL;
7913ecf6	242
7d1d65cb	243	bpf_ops = kzalloc(bpf_size, GFP_KERNEL);
e2e9b654 DB	244	if (bpf_ops == NULL)
e2e9b654 DB	245	return -ENOMEM;
7d1d65cb DB	246
	247	memcpy(bpf_ops, nla_data(tb[TCA_BPF_OPS]), bpf_size);
	248
e2e9b654 DB	249	fprog_tmp.len = bpf_num_ops;
e2e9b654 DB	250	fprog_tmp.filter = bpf_ops;
7d1d65cb	251
e2e9b654 DB	252	ret = bpf_prog_create(&fp, &fprog_tmp);
	253	if (ret < 0) {
	254	kfree(bpf_ops);
	255	return ret;
	256	}
7d1d65cb	257
7d1d65cb	258	prog->bpf_ops = bpf_ops;
e2e9b654 DB	259	prog->bpf_num_ops = bpf_num_ops;
e2e9b654 DB	260	prog->bpf_name = NULL;
7d1d65cb	261	prog->filter = fp;
7d1d65cb	262
e2e9b654 DB	263	return 0;
	264	}
	265
c46646d0 DB	266	static int cls_bpf_prog_from_efd(struct nlattr *tb, struct cls_bpf_prog prog,
c46646d0 DB	267	const struct tcf_proto *tp)
e2e9b654 DB	268	{
	269	struct bpf_prog *fp;
	270	char *name = NULL;
	271	u32 bpf_fd;
	272
	273	bpf_fd = nla_get_u32(tb[TCA_BPF_FD]);
	274
	275	fp = bpf_prog_get(bpf_fd);
	276	if (IS_ERR(fp))
	277	return PTR_ERR(fp);
	278
	279	if (fp->type != BPF_PROG_TYPE_SCHED_CLS) {
	280	bpf_prog_put(fp);
	281	return -EINVAL;
	282	}
	283
	284	if (tb[TCA_BPF_NAME]) {
	285	name = kmemdup(nla_data(tb[TCA_BPF_NAME]),
	286	nla_len(tb[TCA_BPF_NAME]),
	287	GFP_KERNEL);
	288	if (!name) {
	289	bpf_prog_put(fp);
	290	return -ENOMEM;
	291	}
	292	}
	293
	294	prog->bpf_ops = NULL;
	295	prog->bpf_fd = bpf_fd;
	296	prog->bpf_name = name;
e2e9b654	297	prog->filter = fp;
e2e9b654	298
1f211a1b	299	if (fp->dst_needed && !(tp->q->flags & TCQ_F_INGRESS))
c46646d0 DB	300	netif_keep_dst(qdisc_dev(tp->q));
c46646d0 DB	301
e2e9b654 DB	302	return 0;
	303	}
	304
	305	static int cls_bpf_modify_existing(struct net net, struct tcf_proto tp,
	306	struct cls_bpf_prog *prog,
	307	unsigned long base, struct nlattr **tb,
	308	struct nlattr *est, bool ovr)
	309	{
045efa82	310	bool is_bpf, is_ebpf, have_exts = false;
e2e9b654	311	struct tcf_exts exts;
e2e9b654 DB	312	int ret;
	313
	314	is_bpf = tb[TCA_BPF_OPS_LEN] && tb[TCA_BPF_OPS];
	315	is_ebpf = tb[TCA_BPF_FD];
ef146fa4	316	if ((!is_bpf && !is_ebpf) \|\| (is_bpf && is_ebpf))
e2e9b654 DB	317	return -EINVAL;
	318
	319	tcf_exts_init(&exts, TCA_BPF_ACT, TCA_BPF_POLICE);
	320	ret = tcf_exts_validate(net, tp, tb, est, &exts, ovr);
	321	if (ret < 0)
	322	return ret;
	323
045efa82 DB	324	if (tb[TCA_BPF_FLAGS]) {
	325	u32 bpf_flags = nla_get_u32(tb[TCA_BPF_FLAGS]);
	326
	327	if (bpf_flags & ~TCA_BPF_FLAG_ACT_DIRECT) {
	328	tcf_exts_destroy(&exts);
	329	return -EINVAL;
	330	}
	331
	332	have_exts = bpf_flags & TCA_BPF_FLAG_ACT_DIRECT;
	333	}
	334
045efa82	335	prog->exts_integrated = have_exts;
e2e9b654	336
045efa82	337	ret = is_bpf ? cls_bpf_prog_from_ops(tb, prog) :
c46646d0	338	cls_bpf_prog_from_efd(tb, prog, tp);
e2e9b654 DB	339	if (ret < 0) {
	340	tcf_exts_destroy(&exts);
	341	return ret;
	342	}
	343
ef146fa4 DB	344	if (tb[TCA_BPF_CLASSID]) {
	345	prog->res.classid = nla_get_u32(tb[TCA_BPF_CLASSID]);
	346	tcf_bind_filter(tp, &prog->res, base);
	347	}
7d1d65cb	348
ef146fa4	349	tcf_exts_change(tp, &prog->exts, &exts);
7d1d65cb	350	return 0;
7d1d65cb DB	351	}
	352
	353	static u32 cls_bpf_grab_new_handle(struct tcf_proto *tp,
	354	struct cls_bpf_head *head)
	355	{
	356	unsigned int i = 0x80000000;
3f2ab135	357	u32 handle;
7d1d65cb DB	358
	359	do {
	360	if (++head->hgen == 0x7FFFFFFF)
	361	head->hgen = 1;
	362	} while (--i > 0 && cls_bpf_get(tp, head->hgen));
3f2ab135 DB	363
3f2ab135 DB	364	if (unlikely(i == 0)) {
7d1d65cb	365	pr_err("Insufficient number of handles\n");
3f2ab135 DB	366	handle = 0;
	367	} else {
	368	handle = head->hgen;
	369	}
7d1d65cb	370
3f2ab135	371	return handle;
7d1d65cb DB	372	}
	373
	374	static int cls_bpf_change(struct net net, struct sk_buff in_skb,
	375	struct tcf_proto *tp, unsigned long base,
	376	u32 handle, struct nlattr **tca,
2f7ef2f8	377	unsigned long *arg, bool ovr)
7d1d65cb	378	{
1f947bf1 JF	379	struct cls_bpf_head *head = rtnl_dereference(tp->root);
1f947bf1 JF	380	struct cls_bpf_prog oldprog = (struct cls_bpf_prog ) *arg;
7d1d65cb	381	struct nlattr *tb[TCA_BPF_MAX + 1];
1f947bf1	382	struct cls_bpf_prog *prog;
7d1d65cb DB	383	int ret;
	384
	385	if (tca[TCA_OPTIONS] == NULL)
	386	return -EINVAL;
	387
	388	ret = nla_parse_nested(tb, TCA_BPF_MAX, tca[TCA_OPTIONS], bpf_policy);
	389	if (ret < 0)
	390	return ret;
	391
7d1d65cb	392	prog = kzalloc(sizeof(*prog), GFP_KERNEL);
1f947bf1	393	if (!prog)
7d1d65cb DB	394	return -ENOBUFS;
7d1d65cb DB	395
5da57f42	396	tcf_exts_init(&prog->exts, TCA_BPF_ACT, TCA_BPF_POLICE);
1f947bf1 JF	397
	398	if (oldprog) {
	399	if (handle && oldprog->handle != handle) {
	400	ret = -EINVAL;
	401	goto errout;
	402	}
	403	}
	404
7d1d65cb DB	405	if (handle == 0)
	406	prog->handle = cls_bpf_grab_new_handle(tp, head);
	407	else
	408	prog->handle = handle;
	409	if (prog->handle == 0) {
	410	ret = -EINVAL;
	411	goto errout;
	412	}
	413
2f7ef2f8	414	ret = cls_bpf_modify_existing(net, tp, prog, base, tb, tca[TCA_RATE], ovr);
7d1d65cb DB	415	if (ret < 0)
	416	goto errout;
	417
1f947bf1	418	if (oldprog) {
f6bfc46d	419	list_replace_rcu(&oldprog->link, &prog->link);
18cdb37e	420	tcf_unbind_filter(tp, &oldprog->res);
1f947bf1 JF	421	call_rcu(&oldprog->rcu, __cls_bpf_delete_prog);
	422	} else {
	423	list_add_rcu(&prog->link, &head->plist);
	424	}
7d1d65cb DB	425
7d1d65cb DB	426	*arg = (unsigned long) prog;
7d1d65cb DB	427	return 0;
7d1d65cb DB	428	errout:
1f947bf1	429	kfree(prog);
7d1d65cb DB	430
	431	return ret;
	432	}
	433
e2e9b654 DB	434	static int cls_bpf_dump_bpf_info(const struct cls_bpf_prog *prog,
	435	struct sk_buff *skb)
	436	{
	437	struct nlattr *nla;
	438
	439	if (nla_put_u16(skb, TCA_BPF_OPS_LEN, prog->bpf_num_ops))
	440	return -EMSGSIZE;
	441
	442	nla = nla_reserve(skb, TCA_BPF_OPS, prog->bpf_num_ops *
	443	sizeof(struct sock_filter));
	444	if (nla == NULL)
	445	return -EMSGSIZE;
	446
	447	memcpy(nla_data(nla), prog->bpf_ops, nla_len(nla));
	448
	449	return 0;
	450	}
	451
	452	static int cls_bpf_dump_ebpf_info(const struct cls_bpf_prog *prog,
	453	struct sk_buff *skb)
	454	{
	455	if (nla_put_u32(skb, TCA_BPF_FD, prog->bpf_fd))
	456	return -EMSGSIZE;
	457
	458	if (prog->bpf_name &&
	459	nla_put_string(skb, TCA_BPF_NAME, prog->bpf_name))
	460	return -EMSGSIZE;
	461
	462	return 0;
	463	}
	464
832d1d5b	465	static int cls_bpf_dump(struct net net, struct tcf_proto tp, unsigned long fh,
7d1d65cb DB	466	struct sk_buff skb, struct tcmsg tm)
	467	{
	468	struct cls_bpf_prog prog = (struct cls_bpf_prog ) fh;
e2e9b654	469	struct nlattr *nest;
bf007d1c	470	u32 bpf_flags = 0;
e2e9b654	471	int ret;
7d1d65cb DB	472
	473	if (prog == NULL)
	474	return skb->len;
	475
	476	tm->tcm_handle = prog->handle;
	477
	478	nest = nla_nest_start(skb, TCA_OPTIONS);
	479	if (nest == NULL)
	480	goto nla_put_failure;
	481
ef146fa4 DB	482	if (prog->res.classid &&
ef146fa4 DB	483	nla_put_u32(skb, TCA_BPF_CLASSID, prog->res.classid))
7d1d65cb	484	goto nla_put_failure;
7d1d65cb	485
e2e9b654 DB	486	if (cls_bpf_is_ebpf(prog))
	487	ret = cls_bpf_dump_ebpf_info(prog, skb);
	488	else
	489	ret = cls_bpf_dump_bpf_info(prog, skb);
	490	if (ret)
7d1d65cb DB	491	goto nla_put_failure;
7d1d65cb DB	492
5da57f42	493	if (tcf_exts_dump(skb, &prog->exts) < 0)
7d1d65cb DB	494	goto nla_put_failure;
7d1d65cb DB	495
bf007d1c DB	496	if (prog->exts_integrated)
	497	bpf_flags \|= TCA_BPF_FLAG_ACT_DIRECT;
	498	if (bpf_flags && nla_put_u32(skb, TCA_BPF_FLAGS, bpf_flags))
	499	goto nla_put_failure;
	500
7d1d65cb DB	501	nla_nest_end(skb, nest);
7d1d65cb DB	502
5da57f42	503	if (tcf_exts_dump_stats(skb, &prog->exts) < 0)
7d1d65cb DB	504	goto nla_put_failure;
	505
	506	return skb->len;
	507
	508	nla_put_failure:
	509	nla_nest_cancel(skb, nest);
	510	return -1;
	511	}
	512
	513	static void cls_bpf_walk(struct tcf_proto tp, struct tcf_walker arg)
	514	{
1f947bf1	515	struct cls_bpf_head *head = rtnl_dereference(tp->root);
7d1d65cb DB	516	struct cls_bpf_prog *prog;
7d1d65cb DB	517
3fe6b49e	518	list_for_each_entry(prog, &head->plist, link) {
7d1d65cb DB	519	if (arg->count < arg->skip)
	520	goto skip;
	521	if (arg->fn(tp, (unsigned long) prog, arg) < 0) {
	522	arg->stop = 1;
	523	break;
	524	}
	525	skip:
	526	arg->count++;
	527	}
	528	}
	529
	530	static struct tcf_proto_ops cls_bpf_ops __read_mostly = {
	531	.kind = "bpf",
	532	.owner = THIS_MODULE,
	533	.classify = cls_bpf_classify,
	534	.init = cls_bpf_init,
	535	.destroy = cls_bpf_destroy,
	536	.get = cls_bpf_get,
7d1d65cb DB	537	.change = cls_bpf_change,
	538	.delete = cls_bpf_delete,
	539	.walk = cls_bpf_walk,
	540	.dump = cls_bpf_dump,
	541	};
	542
	543	static int __init cls_bpf_init_mod(void)
	544	{
	545	return register_tcf_proto_ops(&cls_bpf_ops);
	546	}
	547
	548	static void __exit cls_bpf_exit_mod(void)
	549	{
	550	unregister_tcf_proto_ops(&cls_bpf_ops);
	551	}
	552
	553	module_init(cls_bpf_init_mod);
	554	module_exit(cls_bpf_exit_mod);