[linux-block.git] / include / net / nexthop.h

/* SPDX-License-Identifier: GPL-2.0 */
/*
 * Generic nexthop implementation
 *
 * Copyright (c) 2017-19 Cumulus Networks
 * Copyright (c) 2017-19 David Ahern <dsa@cumulusnetworks.com>
 */

#ifndef __LINUX_NEXTHOP_H
#define __LINUX_NEXTHOP_H

#include <linux/netdevice.h>
#include <linux/route.h>
#include <linux/types.h>
#include <net/ip_fib.h>
#include <net/ip6_fib.h>
#include <net/netlink.h>

#define NEXTHOP_VALID_USER_FLAGS RTNH_F_ONLINK

struct nexthop;

struct nh_config {
	u32		nh_id;

	u8		nh_family;
	u8		nh_protocol;
	u8		nh_blackhole;
	u32		nh_flags;

	int		nh_ifindex;
	struct net_device *dev;

	union {
		__be32		ipv4;
		struct in6_addr	ipv6;
	} gw;

	struct nlattr	*nh_grp;
	u16		nh_grp_type;

	struct nlattr	*nh_encap;
	u16		nh_encap_type;

	u32		nlflags;
	struct nl_info	nlinfo;
};

struct nh_info {
	struct hlist_node	dev_hash;    /* entry on netns devhash */
	struct nexthop		*nh_parent;

	u8			family;
	bool			reject_nh;

	union {
		struct fib_nh_common	fib_nhc;
		struct fib_nh		fib_nh;
		struct fib6_nh		fib6_nh;
	};
};

struct nh_grp_entry {
	struct nexthop	*nh;
	u8		weight;
	atomic_t	upper_bound;

	struct list_head nh_list;
	struct nexthop	*nh_parent;  /* nexthop of group with this entry */
};

struct nh_group {
	struct nh_group		*spare; /* spare group for removals */
	u16			num_nh;
	bool			mpath;
	bool			has_v4;
	struct nh_grp_entry	nh_entries[];
};

struct nexthop {
	struct rb_node		rb_node;    /* entry on netns rbtree */
	struct list_head	fi_list;    /* v4 entries using nh */
	struct list_head	f6i_list;   /* v6 entries using nh */
	struct list_head	grp_list;   /* nh group entries using this nh */
	struct net		*net;

	u32			id;

	u8			protocol;   /* app managing this nh */
	u8			nh_flags;
	bool			is_group;

	refcount_t		refcnt;
	struct rcu_head		rcu;

	union {
		struct nh_info	__rcu *nh_info;
		struct nh_group __rcu *nh_grp;
	};
};

/* caller is holding rcu or rtnl; no reference taken to nexthop */
struct nexthop *nexthop_find_by_id(struct net *net, u32 id);
void nexthop_free_rcu(struct rcu_head *head);

static inline bool nexthop_get(struct nexthop *nh)
{
	return refcount_inc_not_zero(&nh->refcnt);
}

static inline void nexthop_put(struct nexthop *nh)
{
	if (refcount_dec_and_test(&nh->refcnt))
		call_rcu(&nh->rcu, nexthop_free_rcu);
}

static inline bool nexthop_cmp(const struct nexthop *nh1,
			       const struct nexthop *nh2)
{
	return nh1 == nh2;
}

static inline bool nexthop_is_multipath(const struct nexthop *nh)
{
	if (nh->is_group) {
		struct nh_group *nh_grp;

		nh_grp = rcu_dereference_rtnl(nh->nh_grp);
		return nh_grp->mpath;
	}
	return false;
}

struct nexthop *nexthop_select_path(struct nexthop *nh, int hash);

static inline unsigned int nexthop_num_path(const struct nexthop *nh)
{
	unsigned int rc = 1;

	if (nh->is_group) {
		struct nh_group *nh_grp;

		nh_grp = rcu_dereference_rtnl(nh->nh_grp);
		if (nh_grp->mpath)
			rc = nh_grp->num_nh;
	}

	return rc;
}

static inline
struct nexthop *nexthop_mpath_select(const struct nh_group *nhg, int nhsel)
{
	/* for_nexthops macros in fib_semantics.c grabs a pointer to
	 * the nexthop before checking nhsel
	 */
	if (nhsel >= nhg->num_nh)
		return NULL;

	return nhg->nh_entries[nhsel].nh;
}

static inline
int nexthop_mpath_fill_node(struct sk_buff *skb, struct nexthop *nh,
			    u8 rt_family)
{
	struct nh_group *nhg = rtnl_dereference(nh->nh_grp);
	int i;

	for (i = 0; i < nhg->num_nh; i++) {
		struct nexthop *nhe = nhg->nh_entries[i].nh;
		struct nh_info *nhi = rcu_dereference_rtnl(nhe->nh_info);
		struct fib_nh_common *nhc = &nhi->fib_nhc;
		int weight = nhg->nh_entries[i].weight;

		if (fib_add_nexthop(skb, nhc, weight, rt_family) < 0)
			return -EMSGSIZE;
	}

	return 0;
}

/* called with rcu lock */
static inline bool nexthop_is_blackhole(const struct nexthop *nh)
{
	const struct nh_info *nhi;

	if (nh->is_group) {
		struct nh_group *nh_grp;

		nh_grp = rcu_dereference_rtnl(nh->nh_grp);
		if (nh_grp->num_nh > 1)
			return false;

		nh = nh_grp->nh_entries[0].nh;
	}

	nhi = rcu_dereference_rtnl(nh->nh_info);
	return nhi->reject_nh;
}

static inline void nexthop_path_fib_result(struct fib_result *res, int hash)
{
	struct nh_info *nhi;
	struct nexthop *nh;

	nh = nexthop_select_path(res->fi->nh, hash);
	nhi = rcu_dereference(nh->nh_info);
	res->nhc = &nhi->fib_nhc;
}

/* called with rcu read lock or rtnl held */
static inline
struct fib_nh_common *nexthop_fib_nhc(struct nexthop *nh, int nhsel)
{
	struct nh_info *nhi;

	BUILD_BUG_ON(offsetof(struct fib_nh, nh_common) != 0);
	BUILD_BUG_ON(offsetof(struct fib6_nh, nh_common) != 0);

	if (nh->is_group) {
		struct nh_group *nh_grp;

		nh_grp = rcu_dereference_rtnl(nh->nh_grp);
		if (nh_grp->mpath) {
			nh = nexthop_mpath_select(nh_grp, nhsel);
			if (!nh)
				return NULL;
		}
	}

	nhi = rcu_dereference_rtnl(nh->nh_info);
	return &nhi->fib_nhc;
}

static inline unsigned int fib_info_num_path(const struct fib_info *fi)
{
	if (unlikely(fi->nh))
		return nexthop_num_path(fi->nh);

	return fi->fib_nhs;
}

int fib_check_nexthop(struct nexthop *nh, u8 scope,
		      struct netlink_ext_ack *extack);

static inline struct fib_nh_common *fib_info_nhc(struct fib_info *fi, int nhsel)
{
	if (unlikely(fi->nh))
		return nexthop_fib_nhc(fi->nh, nhsel);

	return &fi->fib_nh[nhsel].nh_common;
}

/* only used when fib_nh is built into fib_info */
static inline struct fib_nh *fib_info_nh(struct fib_info *fi, int nhsel)
{
	WARN_ON(fi->nh);

	return &fi->fib_nh[nhsel];
}

/*
 * IPv6 variants
 */
int fib6_check_nexthop(struct nexthop *nh, struct fib6_config *cfg,
		       struct netlink_ext_ack *extack);

static inline struct fib6_nh *nexthop_fib6_nh(struct nexthop *nh)
{
	struct nh_info *nhi;

	if (nh->is_group) {
		struct nh_group *nh_grp;

		nh_grp = rcu_dereference_rtnl(nh->nh_grp);
		nh = nexthop_mpath_select(nh_grp, 0);
		if (!nh)
			return NULL;
	}

	nhi = rcu_dereference_rtnl(nh->nh_info);
	if (nhi->family == AF_INET6)
		return &nhi->fib6_nh;

	return NULL;
}

static inline struct net_device *fib6_info_nh_dev(struct fib6_info *f6i)
{
	struct fib6_nh *fib6_nh;

	fib6_nh = f6i->nh ? nexthop_fib6_nh(f6i->nh) : f6i->fib6_nh;
	return fib6_nh->fib_nh_dev;
}

static inline void nexthop_path_fib6_result(struct fib6_result *res, int hash)
{
	struct nexthop *nh = res->f6i->nh;
	struct nh_info *nhi;

	nh = nexthop_select_path(nh, hash);

	nhi = rcu_dereference_rtnl(nh->nh_info);
	if (nhi->reject_nh) {
		res->fib6_type = RTN_BLACKHOLE;
		res->fib6_flags |= RTF_REJECT;
		res->nh = nexthop_fib6_nh(nh);
	} else {
		res->nh = &nhi->fib6_nh;
	}
}

int nexthop_for_each_fib6_nh(struct nexthop *nh,
			     int (*cb)(struct fib6_nh *nh, void *arg),
			     void *arg);
#endif
Commit	Line	Data
ab84be7e DA	1	/* SPDX-License-Identifier: GPL-2.0 */
	2	/*
	3	* Generic nexthop implementation
	4	*
	5	* Copyright (c) 2017-19 Cumulus Networks
	6	* Copyright (c) 2017-19 David Ahern <dsa@cumulusnetworks.com>
	7	*/
	8
	9	#ifndef __LINUX_NEXTHOP_H
	10	#define __LINUX_NEXTHOP_H
	11
	12	#include <linux/netdevice.h>
f88d8ea6	13	#include <linux/route.h>
ab84be7e DA	14	#include <linux/types.h>
ab84be7e DA	15	#include <net/ip_fib.h>
53010f99	16	#include <net/ip6_fib.h>
ab84be7e DA	17	#include <net/netlink.h>
	18
	19	#define NEXTHOP_VALID_USER_FLAGS RTNH_F_ONLINK
	20
	21	struct nexthop;
	22
	23	struct nh_config {
	24	u32 nh_id;
	25
	26	u8 nh_family;
	27	u8 nh_protocol;
	28	u8 nh_blackhole;
	29	u32 nh_flags;
	30
	31	int nh_ifindex;
	32	struct net_device *dev;
	33
597cfe4f DA	34	union {
597cfe4f DA	35	__be32 ipv4;
53010f99	36	struct in6_addr ipv6;
597cfe4f DA	37	} gw;
597cfe4f DA	38
430a0491 DA	39	struct nlattr *nh_grp;
	40	u16 nh_grp_type;
	41
b513bd03 DA	42	struct nlattr *nh_encap;
	43	u16 nh_encap_type;
	44
ab84be7e DA	45	u32 nlflags;
	46	struct nl_info nlinfo;
	47	};
	48
	49	struct nh_info {
	50	struct hlist_node dev_hash; /* entry on netns devhash */
	51	struct nexthop *nh_parent;
	52
	53	u8 family;
	54	bool reject_nh;
	55
	56	union {
	57	struct fib_nh_common fib_nhc;
597cfe4f	58	struct fib_nh fib_nh;
53010f99	59	struct fib6_nh fib6_nh;
ab84be7e DA	60	};
	61	};
	62
430a0491 DA	63	struct nh_grp_entry {
	64	struct nexthop *nh;
	65	u8 weight;
	66	atomic_t upper_bound;
	67
	68	struct list_head nh_list;
	69	struct nexthop nh_parent; / nexthop of group with this entry */
	70	};
	71
	72	struct nh_group {
90f33bff	73	struct nh_group spare; / spare group for removals */
430a0491 DA	74	u16 num_nh;
	75	bool mpath;
	76	bool has_v4;
97a888c2	77	struct nh_grp_entry nh_entries[];
430a0491 DA	78	};
430a0491 DA	79
ab84be7e DA	80	struct nexthop {
ab84be7e DA	81	struct rb_node rb_node; /* entry on netns rbtree */
4c7e8084	82	struct list_head fi_list; /* v4 entries using nh */
f88d8ea6	83	struct list_head f6i_list; /* v6 entries using nh */
430a0491	84	struct list_head grp_list; /* nh group entries using this nh */
ab84be7e DA	85	struct net *net;
	86
	87	u32 id;
	88
	89	u8 protocol; /* app managing this nh */
	90	u8 nh_flags;
430a0491	91	bool is_group;
ab84be7e DA	92
	93	refcount_t refcnt;
	94	struct rcu_head rcu;
	95
	96	union {
	97	struct nh_info __rcu *nh_info;
430a0491	98	struct nh_group __rcu *nh_grp;
ab84be7e DA	99	};
	100	};
	101
	102	/* caller is holding rcu or rtnl; no reference taken to nexthop */
	103	struct nexthop nexthop_find_by_id(struct net net, u32 id);
	104	void nexthop_free_rcu(struct rcu_head *head);
	105
	106	static inline bool nexthop_get(struct nexthop *nh)
	107	{
	108	return refcount_inc_not_zero(&nh->refcnt);
	109	}
	110
	111	static inline void nexthop_put(struct nexthop *nh)
	112	{
	113	if (refcount_dec_and_test(&nh->refcnt))
	114	call_rcu(&nh->rcu, nexthop_free_rcu);
	115	}
	116
4c7e8084 DA	117	static inline bool nexthop_cmp(const struct nexthop *nh1,
	118	const struct nexthop *nh2)
	119	{
	120	return nh1 == nh2;
	121	}
	122
430a0491 DA	123	static inline bool nexthop_is_multipath(const struct nexthop *nh)
	124	{
	125	if (nh->is_group) {
	126	struct nh_group *nh_grp;
	127
	128	nh_grp = rcu_dereference_rtnl(nh->nh_grp);
	129	return nh_grp->mpath;
	130	}
	131	return false;
	132	}
	133
	134	struct nexthop nexthop_select_path(struct nexthop nh, int hash);
	135
	136	static inline unsigned int nexthop_num_path(const struct nexthop *nh)
	137	{
	138	unsigned int rc = 1;
	139
0b5e2e39	140	if (nh->is_group) {
430a0491 DA	141	struct nh_group *nh_grp;
	142
	143	nh_grp = rcu_dereference_rtnl(nh->nh_grp);
0b5e2e39 DA	144	if (nh_grp->mpath)
0b5e2e39 DA	145	rc = nh_grp->num_nh;
430a0491 DA	146	}
	147
	148	return rc;
	149	}
	150
	151	static inline
0b5e2e39	152	struct nexthop nexthop_mpath_select(const struct nh_group nhg, int nhsel)
430a0491	153	{
430a0491 DA	154	/* for_nexthops macros in fib_semantics.c grabs a pointer to
	155	* the nexthop before checking nhsel
	156	*/
5270041d	157	if (nhsel >= nhg->num_nh)
430a0491 DA	158	return NULL;
	159
	160	return nhg->nh_entries[nhsel].nh;
	161	}
	162
	163	static inline
7bdf4de1 DS	164	int nexthop_mpath_fill_node(struct sk_buff skb, struct nexthop nh,
7bdf4de1 DS	165	u8 rt_family)
430a0491 DA	166	{
	167	struct nh_group *nhg = rtnl_dereference(nh->nh_grp);
	168	int i;
	169
	170	for (i = 0; i < nhg->num_nh; i++) {
	171	struct nexthop *nhe = nhg->nh_entries[i].nh;
	172	struct nh_info *nhi = rcu_dereference_rtnl(nhe->nh_info);
	173	struct fib_nh_common *nhc = &nhi->fib_nhc;
	174	int weight = nhg->nh_entries[i].weight;
	175
7bdf4de1	176	if (fib_add_nexthop(skb, nhc, weight, rt_family) < 0)
430a0491 DA	177	return -EMSGSIZE;
	178	}
	179
	180	return 0;
	181	}
	182
ab84be7e DA	183	/* called with rcu lock */
	184	static inline bool nexthop_is_blackhole(const struct nexthop *nh)
	185	{
	186	const struct nh_info *nhi;
	187
0b5e2e39 DA	188	if (nh->is_group) {
	189	struct nh_group *nh_grp;
	190
	191	nh_grp = rcu_dereference_rtnl(nh->nh_grp);
	192	if (nh_grp->num_nh > 1)
430a0491	193	return false;
0b5e2e39 DA	194
0b5e2e39 DA	195	nh = nh_grp->nh_entries[0].nh;
430a0491 DA	196	}
	197
	198	nhi = rcu_dereference_rtnl(nh->nh_info);
ab84be7e DA	199	return nhi->reject_nh;
ab84be7e DA	200	}
5481d73f	201
4c7e8084 DA	202	static inline void nexthop_path_fib_result(struct fib_result *res, int hash)
	203	{
	204	struct nh_info *nhi;
	205	struct nexthop *nh;
	206
	207	nh = nexthop_select_path(res->fi->nh, hash);
	208	nhi = rcu_dereference(nh->nh_info);
	209	res->nhc = &nhi->fib_nhc;
	210	}
	211
	212	/* called with rcu read lock or rtnl held */
	213	static inline
	214	struct fib_nh_common nexthop_fib_nhc(struct nexthop nh, int nhsel)
	215	{
	216	struct nh_info *nhi;
	217
	218	BUILD_BUG_ON(offsetof(struct fib_nh, nh_common) != 0);
	219	BUILD_BUG_ON(offsetof(struct fib6_nh, nh_common) != 0);
	220
0b5e2e39 DA	221	if (nh->is_group) {
	222	struct nh_group *nh_grp;
	223
	224	nh_grp = rcu_dereference_rtnl(nh->nh_grp);
	225	if (nh_grp->mpath) {
	226	nh = nexthop_mpath_select(nh_grp, nhsel);
	227	if (!nh)
	228	return NULL;
	229	}
4c7e8084 DA	230	}
	231
	232	nhi = rcu_dereference_rtnl(nh->nh_info);
	233	return &nhi->fib_nhc;
	234	}
	235
5481d73f DA	236	static inline unsigned int fib_info_num_path(const struct fib_info *fi)
5481d73f DA	237	{
4c7e8084 DA	238	if (unlikely(fi->nh))
	239	return nexthop_num_path(fi->nh);
	240
5481d73f DA	241	return fi->fib_nhs;
	242	}
	243
4c7e8084 DA	244	int fib_check_nexthop(struct nexthop *nh, u8 scope,
	245	struct netlink_ext_ack *extack);
	246
5481d73f DA	247	static inline struct fib_nh_common fib_info_nhc(struct fib_info fi, int nhsel)
5481d73f DA	248	{
4c7e8084 DA	249	if (unlikely(fi->nh))
	250	return nexthop_fib_nhc(fi->nh, nhsel);
	251
5481d73f DA	252	return &fi->fib_nh[nhsel].nh_common;
	253	}
	254
4c7e8084	255	/* only used when fib_nh is built into fib_info */
5481d73f DA	256	static inline struct fib_nh fib_info_nh(struct fib_info fi, int nhsel)
5481d73f DA	257	{
4c7e8084 DA	258	WARN_ON(fi->nh);
4c7e8084 DA	259
5481d73f DA	260	return &fi->fib_nh[nhsel];
5481d73f DA	261	}
f88d8ea6 DA	262
	263	/*
	264	* IPv6 variants
	265	*/
	266	int fib6_check_nexthop(struct nexthop nh, struct fib6_config cfg,
	267	struct netlink_ext_ack *extack);
	268
	269	static inline struct fib6_nh nexthop_fib6_nh(struct nexthop nh)
	270	{
	271	struct nh_info *nhi;
	272
0b5e2e39 DA	273	if (nh->is_group) {
	274	struct nh_group *nh_grp;
	275
	276	nh_grp = rcu_dereference_rtnl(nh->nh_grp);
	277	nh = nexthop_mpath_select(nh_grp, 0);
f88d8ea6 DA	278	if (!nh)
	279	return NULL;
	280	}
	281
	282	nhi = rcu_dereference_rtnl(nh->nh_info);
	283	if (nhi->family == AF_INET6)
	284	return &nhi->fib6_nh;
	285
	286	return NULL;
	287	}
	288
	289	static inline struct net_device fib6_info_nh_dev(struct fib6_info f6i)
	290	{
	291	struct fib6_nh *fib6_nh;
	292
	293	fib6_nh = f6i->nh ? nexthop_fib6_nh(f6i->nh) : f6i->fib6_nh;
	294	return fib6_nh->fib_nh_dev;
	295	}
	296
	297	static inline void nexthop_path_fib6_result(struct fib6_result *res, int hash)
	298	{
	299	struct nexthop *nh = res->f6i->nh;
	300	struct nh_info *nhi;
	301
	302	nh = nexthop_select_path(nh, hash);
	303
	304	nhi = rcu_dereference_rtnl(nh->nh_info);
	305	if (nhi->reject_nh) {
	306	res->fib6_type = RTN_BLACKHOLE;
	307	res->fib6_flags \|= RTF_REJECT;
	308	res->nh = nexthop_fib6_nh(nh);
	309	} else {
	310	res->nh = &nhi->fib6_nh;
	311	}
	312	}
f88c9aa1 DA	313
	314	int nexthop_for_each_fib6_nh(struct nexthop *nh,
	315	int (cb)(struct fib6_nh nh, void *arg),
	316	void *arg);
ab84be7e	317	#endif