xref: /titanic_52/usr/src/uts/common/inet/ip_ire.h (revision 0f2c99a46e005b1add7df43157ee8516e585157a)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 /* Copyright (c) 1990 Mentat Inc. */
26 
27 #ifndef	_INET_IP_IRE_H
28 #define	_INET_IP_IRE_H
29 
30 #pragma ident	"%Z%%M%	%I%	%E% SMI"
31 
32 #ifdef	__cplusplus
33 extern "C" {
34 #endif
35 
36 #define	IPV6_LL_PREFIXLEN	10	/* Number of bits in link-local pref */
37 
38 #define	IP_CACHE_TABLE_SIZE	256
39 #define	IP_MRTUN_TABLE_SIZE	256	/* Mobile IP reverse tunnel table */
40 					/* size. Only used by mipagent */
41 #define	IP_SRCIF_TABLE_SIZE	256	/* Per interface routing table size */
42 #define	IP_MASK_TABLE_SIZE	(IP_ABITS + 1)		/* 33 ptrs */
43 
44 #define	IP6_FTABLE_HASH_SIZE	32	/* size of each hash table in ptrs */
45 #define	IP6_CACHE_TABLE_SIZE	256
46 #define	IP6_MASK_TABLE_SIZE	(IPV6_ABITS + 1)	/* 129 ptrs */
47 
48 /*
49  * We use the common modulo hash function.  In ip_ire_init(), we make
50  * sure that the cache table size is always a power of 2.  That's why
51  * we can use & instead of %.  Also note that we try hard to make sure
52  * the lower bits of an address capture most info from the whole address.
53  * The reason being that since our hash table is probably a lot smaller
54  * than 2^32 buckets so the lower bits are the most important.
55  */
56 #define	IRE_ADDR_HASH(addr, table_size) \
57 	(((addr) ^ ((addr) >> 8) ^ ((addr) >> 16) ^ ((addr) >> 24)) &	\
58 	((table_size) - 1))
59 
60 /*
61  * To make a byte-order neutral hash for IPv6, just take all the
62  * bytes in the bottom 32 bits into account.
63  */
64 #define	IRE_ADDR_HASH_V6(addr, table_size) 				\
65 	IRE_ADDR_HASH((addr).s6_addr32[3], table_size)
66 
67 /* This assumes that the ftable size is a power of 2. */
68 #define	IRE_ADDR_MASK_HASH_V6(addr, mask, table_size) 			\
69 	((((addr).s6_addr8[8] & (mask).s6_addr8[8]) ^ 			\
70 	((addr).s6_addr8[9] & (mask).s6_addr8[9]) ^			\
71 	((addr).s6_addr8[10] & (mask).s6_addr8[10]) ^ 			\
72 	((addr).s6_addr8[13] & (mask).s6_addr8[13]) ^ 			\
73 	((addr).s6_addr8[14] & (mask).s6_addr8[14]) ^ 			\
74 	((addr).s6_addr8[15] & (mask).s6_addr8[15])) & ((table_size) - 1))
75 
76 /*
77  * match parameter definitions for
78  * IRE lookup routines.
79  */
80 #define	MATCH_IRE_DSTONLY	0x0000	/* Match just the address */
81 #define	MATCH_IRE_TYPE		0x0001	/* Match IRE type */
82 #define	MATCH_IRE_SRC		0x0002	/* Match IRE source address */
83 #define	MATCH_IRE_MASK		0x0004	/* Match IRE mask */
84 #define	MATCH_IRE_WQ		0x0008	/* Match IRE Write Q */
85 #define	MATCH_IRE_GW		0x0010	/* Match IRE gateway */
86 #define	MATCH_IRE_IPIF		0x0020	/* Match IRE ipif */
87 #define	MATCH_IRE_RECURSIVE	0x0040	/* Do recursive lookup if necessary */
88 #define	MATCH_IRE_DEFAULT	0x0080	/* Return default route if no route */
89 					/* found. */
90 #define	MATCH_IRE_RJ_BHOLE	0x0100	/* During lookup if we hit an ire */
91 					/* with RTF_REJECT or RTF_BLACKHOLE, */
92 					/* return the ire. No recursive */
93 					/* lookup should be done. */
94 #define	MATCH_IRE_IHANDLE	0x0200	/* Match IRE on ihandle */
95 #define	MATCH_IRE_MARK_HIDDEN	0x0400	/* Match IRE ire_marks with */
96 					/* IRE_MARK_HIDDEN. */
97 /*
98  * MATCH_IRE_ILL is used whenever we want to specifically match an IRE
99  * whose ire_ipif->ipif_ill or (ill_t *)ire_stq->q_ptr matches a given
100  * ill. When MATCH_IRE_ILL is used to locate an IRE_CACHE, it implies
101  * that the packet will not be load balanced. This is normally used
102  * by in.mpathd to send out failure detection probes.
103  *
104  * MATCH_IRE_ILL_GROUP is used whenever we are not specific about which
105  * interface (ill) the packet should be sent out. This implies that the
106  * packets will be subjected to load balancing and it might go out on
107  * any interface in the group. When there is only interface in the group,
108  * MATCH_IRE_ILL_GROUP becomes MATCH_IRE_ILL. Most of the code uses
109  * MATCH_IRE_ILL_GROUP and MATCH_IRE_ILL is used in very few cases where
110  * we want to disable load balancing.
111  *
112  * MATCH_IRE_PARENT is used whenever we unconditionally want to get the
113  * parent IRE (sire) while recursively searching IREs for an offsubnet
114  * destination. With this flag, even if no IRE_CACHETABLE or IRE_INTERFACE
115  * is found to help resolving IRE_OFFSUBNET in lookup routines, the
116  * IRE_OFFSUBNET sire, if any, is returned to the caller.
117  */
118 #define	MATCH_IRE_ILL_GROUP	0x0800	/* Match IRE on ill or the ill_group. */
119 #define	MATCH_IRE_ILL		0x1000	/* Match IRE on the ill only */
120 
121 #define	MATCH_IRE_PARENT	0x2000	/* Match parent ire, if any, */
122 					/* even if ire is not matched. */
123 #define	MATCH_IRE_ZONEONLY	0x4000	/* Match IREs in specified zone, ie */
124 					/* don't match IRE_LOCALs from other */
125 					/* zones or shared IREs */
126 #define	MATCH_IRE_MARK_PRIVATE_ADDR	0x8000	/* Match IRE ire_marks with */
127 						/* IRE_MARK_PRIVATE_ADDR. */
128 #define	MATCH_IRE_SECATTR	0x10000	/* Match gateway security attributes */
129 #define	MATCH_IRE_COMPLETE	0x20000	/* ire_ftable_lookup() can return */
130 					/* IRE_CACHE entry only if it is  */
131 					/* ND_REACHABLE			  */
132 
133 /*
134  * Any ire to nce association is long term, and
135  * the refhold and refrele may be done by different
136  * threads. So all cases of making or breaking ire to
137  * nce association should all effectively use the NOTR variants.
138  * To understand the *effectively* part read on.
139  *
140  * ndp_lookup() and ndp_add() implicitly does NCE_REFHOLD. So wherever we
141  * make ire to nce association after calling these functions,
142  * we effectively want to end up with NCE_REFHOLD_NOTR,
143  * We call this macro to achieve this effect. This macro changes
144  * a NCE_REFHOLD to a NCE_REFHOLD_NOTR. The macro's NCE_REFRELE
145  * cancels off ndp_lookup[ndp_add]'s implicit NCE_REFHOLD, and what
146  * you are left with is a NCE_REFHOLD_NOTR
147  */
148 #define	NCE_REFHOLD_TO_REFHOLD_NOTR(nce) {	\
149 	NCE_REFHOLD_NOTR(nce);			\
150 	NCE_REFRELE(nce);			\
151 }
152 
153 /*
154  * find the next ire_t entry in the ire_next chain starting at ire
155  * that is not CONDEMNED.  ire is set to NULL if we reach the end of the list.
156  * Caller must hold the ire_bucket lock.
157  */
158 
159 #define	IRE_FIND_NEXT_ORIGIN(ire) {					\
160 	while ((ire) != NULL && ((ire)->ire_marks & IRE_MARK_CONDEMNED))\
161 		(ire) = (ire)->ire_next;				\
162 }
163 
164 
165 /* Structure for ire_cache_count() */
166 typedef struct {
167 	int	icc_total;	/* Total number of IRE_CACHE */
168 	int	icc_unused;	/* # off/no PMTU unused since last reclaim */
169 	int	icc_offlink;	/* # offlink without PMTU information */
170 	int	icc_pmtu;	/* # offlink with PMTU information */
171 	int	icc_onlink;	/* # onlink */
172 } ire_cache_count_t;
173 
174 /*
175  * Structure for ire_cache_reclaim(). Each field is a fraction i.e. 1 meaning
176  * reclaim all, N meaning reclaim 1/Nth of all entries, 0 meaning reclaim none.
177  *
178  * The comment below (and for other netstack_t references) refers
179  * to the fact that we only do netstack_hold in particular cases,
180  * such as the references from open streams (ill_t and conn_t's
181  * pointers). Internally within IP we rely on IP's ability to cleanup e.g.
182  * ire_t's when an ill goes away.
183  */
184 typedef struct {
185 	int	icr_unused;	/* Fraction for unused since last reclaim */
186 	int	icr_offlink;	/* Fraction for offlink without PMTU info */
187 	int	icr_pmtu;	/* Fraction for offlink with PMTU info */
188 	int	icr_onlink;	/* Fraction for onlink */
189 	ip_stack_t *icr_ipst;	/* Does not have a netstack_hold */
190 } ire_cache_reclaim_t;
191 
192 /*
193  * We use atomics so that we get an accurate accounting on the ires.
194  * Otherwise we can't determine leaks correctly.
195  */
196 #define	BUMP_IRE_STATS(ire_stats, x) atomic_add_64(&(ire_stats).x, 1)
197 
198 #ifdef _KERNEL
199 struct ts_label_s;
200 
201 extern	ipaddr_t	ip_plen_to_mask(uint_t);
202 extern	in6_addr_t	*ip_plen_to_mask_v6(uint_t, in6_addr_t *);
203 
204 extern	int	ip_ire_advise(queue_t *, mblk_t *, cred_t *);
205 extern	int	ip_ire_delete(queue_t *, mblk_t *, cred_t *);
206 extern	boolean_t ip_ire_clookup_and_delete(ipaddr_t, ipif_t *, ip_stack_t *);
207 extern	void	ip_ire_clookup_and_delete_v6(const in6_addr_t *,
208     ip_stack_t *);
209 
210 extern	int	ip_ire_report(queue_t *, mblk_t *, caddr_t, cred_t *);
211 extern	int	ip_ire_report_mrtun(queue_t *, mblk_t *, caddr_t, cred_t *);
212 extern	int	ip_ire_report_srcif(queue_t *, mblk_t *, caddr_t, cred_t *);
213 extern	int	ip_ire_report_v6(queue_t *, mblk_t *, caddr_t, cred_t *);
214 extern	void	ire_report_ftable(ire_t *, char *);
215 
216 extern	void	ip_ire_req(queue_t *, mblk_t *);
217 
218 extern	int	ip_mask_to_plen(ipaddr_t);
219 extern	int	ip_mask_to_plen_v6(const in6_addr_t *);
220 
221 extern	ire_t	*ipif_to_ire(const ipif_t *);
222 extern	ire_t	*ipif_to_ire_v6(const ipif_t *);
223 
224 extern	int	ire_add(ire_t **, queue_t *, mblk_t *, ipsq_func_t, boolean_t);
225 extern	int	ire_add_mrtun(ire_t **, queue_t *, mblk_t *, ipsq_func_t);
226 extern	void	ire_add_then_send(queue_t *, ire_t *, mblk_t *);
227 extern	int	ire_add_v6(ire_t **, queue_t *, mblk_t *, ipsq_func_t);
228 extern	int	ire_atomic_start(irb_t *irb_ptr, ire_t *ire, queue_t *q,
229     mblk_t *mp, ipsq_func_t func);
230 extern	void	ire_atomic_end(irb_t *irb_ptr, ire_t *ire);
231 
232 extern	void	ire_cache_count(ire_t *, char *);
233 extern	ire_t	*ire_cache_lookup(ipaddr_t, zoneid_t,
234     const struct ts_label_s *, ip_stack_t *);
235 extern	ire_t	*ire_cache_lookup_v6(const in6_addr_t *, zoneid_t,
236     const struct ts_label_s *, ip_stack_t *);
237 extern	void	ire_cache_reclaim(ire_t *, char *);
238 
239 extern	void	ire_check_bcast_present(ipif_t *, ipaddr_t, int, boolean_t *,
240     boolean_t *);
241 
242 extern	ire_t	*ire_create_mp(uchar_t *, uchar_t *, uchar_t *, uchar_t *,
243     uchar_t *, uint_t, mblk_t *, queue_t *, queue_t *, ushort_t, mblk_t *,
244     ipif_t *, ill_t *, ipaddr_t, uint32_t, uint32_t, uint32_t, const iulp_t *,
245     tsol_gc_t *, tsol_gcgrp_t *, ip_stack_t *);
246 extern	ire_t	*ire_create(uchar_t *, uchar_t *, uchar_t *, uchar_t *,
247     uchar_t *, uint_t *, mblk_t *, queue_t *, queue_t *, ushort_t, mblk_t *,
248     ipif_t *, ill_t *, ipaddr_t, uint32_t, uint32_t, uint32_t, const iulp_t *,
249     tsol_gc_t *, tsol_gcgrp_t *, ip_stack_t *);
250 
251 extern	ire_t	**ire_check_and_create_bcast(ipif_t *, ipaddr_t,
252     ire_t **, int);
253 extern	ire_t	**ire_create_bcast(ipif_t *, ipaddr_t, ire_t **);
254 extern	ire_t	*ire_init(ire_t *, uchar_t *, uchar_t *, uchar_t *,
255     uchar_t *, uchar_t *, uint_t *, mblk_t *, queue_t *, queue_t *, ushort_t,
256     mblk_t *, ipif_t *, ill_t *, ipaddr_t, uint32_t, uint32_t, uint32_t,
257     const iulp_t *, tsol_gc_t *, tsol_gcgrp_t *, ip_stack_t *);
258 
259 extern	boolean_t ire_init_common(ire_t *, uint_t *, mblk_t *, queue_t *,
260     queue_t *, ushort_t, mblk_t *, ipif_t *, ill_t *, uint32_t,
261     uint32_t, uint32_t, uchar_t, const iulp_t *, tsol_gc_t *, tsol_gcgrp_t *,
262     ip_stack_t *);
263 
264 extern	ire_t	*ire_create_v6(const in6_addr_t *, const in6_addr_t *,
265     const in6_addr_t *, const in6_addr_t *, uint_t *, mblk_t *, queue_t *,
266     queue_t *, ushort_t, mblk_t *, ipif_t *,
267     const in6_addr_t *, uint32_t, uint32_t, uint_t, const iulp_t *,
268     tsol_gc_t *, tsol_gcgrp_t *, ip_stack_t *);
269 
270 extern	ire_t	*ire_create_mp_v6(const in6_addr_t *, const in6_addr_t *,
271     const in6_addr_t *, const in6_addr_t *, mblk_t *, queue_t *,
272     queue_t *, ushort_t, mblk_t *, ipif_t *,
273     const in6_addr_t *, uint32_t, uint32_t, uint_t, const iulp_t *,
274     tsol_gc_t *, tsol_gcgrp_t *, ip_stack_t *);
275 
276 extern	ire_t	*ire_init_v6(ire_t *, const in6_addr_t *, const in6_addr_t *,
277     const in6_addr_t *, const in6_addr_t *, uint_t *, mblk_t *, queue_t *,
278     queue_t *, ushort_t, mblk_t *, ipif_t *,
279     const in6_addr_t *, uint32_t, uint32_t, uint_t, const iulp_t *,
280     tsol_gc_t *, tsol_gcgrp_t *, ip_stack_t *);
281 
282 extern	void	ire_clookup_delete_cache_gw(ipaddr_t, zoneid_t,
283     ip_stack_t *);
284 extern	void	ire_clookup_delete_cache_gw_v6(const in6_addr_t *, zoneid_t,
285     ip_stack_t *);
286 
287 extern	ire_t	*ire_ctable_lookup(ipaddr_t, ipaddr_t, int, const ipif_t *,
288     zoneid_t, const struct ts_label_s *, int, ip_stack_t *);
289 
290 extern	ire_t	*ire_ctable_lookup_v6(const in6_addr_t *, const in6_addr_t *,
291     int, const ipif_t *, zoneid_t, const struct ts_label_s *, int,
292     ip_stack_t *);
293 
294 extern	void	ire_delete(ire_t *);
295 extern	void	ire_delete_cache_gw(ire_t *, char *);
296 extern	void	ire_delete_cache_gw_v6(ire_t *, char *);
297 extern	void	ire_delete_cache_v6(ire_t *, char *);
298 extern	void	ire_delete_srcif(ire_t *);
299 extern	void	ire_delete_v6(ire_t *);
300 
301 extern	void	ire_expire(ire_t *, char *);
302 
303 extern	void	ire_flush_cache_v4(ire_t *, int);
304 extern	void	ire_flush_cache_v6(ire_t *, int);
305 
306 extern	ire_t	*ire_ftable_lookup_v6(const in6_addr_t *, const in6_addr_t *,
307     const in6_addr_t *, int, const ipif_t *, ire_t **, zoneid_t,
308     uint32_t, const struct ts_label_s *, int, ip_stack_t *);
309 
310 extern	ire_t	*ire_ihandle_lookup_onlink(ire_t *);
311 extern	ire_t	*ire_ihandle_lookup_offlink(ire_t *, ire_t *);
312 extern	ire_t	*ire_ihandle_lookup_offlink_v6(ire_t *, ire_t *);
313 
314 extern	boolean_t	ire_local_same_ill_group(ire_t *, ire_t *);
315 extern	boolean_t	ire_local_ok_across_zones(ire_t *, zoneid_t, void *,
316     const struct ts_label_s *, ip_stack_t *);
317 
318 extern	ire_t 	*ire_lookup_local(zoneid_t, ip_stack_t *);
319 extern	ire_t 	*ire_lookup_local_v6(zoneid_t, ip_stack_t *);
320 
321 extern  ire_t	*ire_lookup_multi(ipaddr_t, zoneid_t, ip_stack_t *);
322 extern  ire_t	*ire_lookup_multi_v6(const in6_addr_t *, zoneid_t,
323     ip_stack_t *);
324 
325 extern ire_t	*ire_mrtun_lookup(ipaddr_t, ill_t *);
326 
327 extern	void	ire_refrele(ire_t *);
328 extern	void	ire_refrele_notr(ire_t *);
329 extern	ire_t	*ire_route_lookup(ipaddr_t, ipaddr_t, ipaddr_t, int,
330     const ipif_t *, ire_t **, zoneid_t, const struct ts_label_s *, int,
331     ip_stack_t *);
332 
333 extern	ire_t	*ire_route_lookup_v6(const in6_addr_t *, const in6_addr_t *,
334     const in6_addr_t *, int, const ipif_t *, ire_t **, zoneid_t,
335     const struct ts_label_s *, int, ip_stack_t *);
336 
337 extern	ire_t	*ire_srcif_table_lookup(ipaddr_t, int, ipif_t *, ill_t *, int);
338 extern ill_t	*ire_to_ill(const ire_t *);
339 
340 extern	void	ire_walk(pfv_t, void *, ip_stack_t *);
341 extern	void	ire_walk_ill(uint_t, uint_t, pfv_t, void *, ill_t *);
342 extern	void	ire_walk_ill_mrtun(uint_t, uint_t, pfv_t, void *, ill_t *,
343     ip_stack_t *);
344 extern	void	ire_walk_ill_v4(uint_t, uint_t, pfv_t, void *, ill_t *);
345 extern	void	ire_walk_ill_v6(uint_t, uint_t, pfv_t, void *, ill_t *);
346 extern	void	ire_walk_v4(pfv_t, void *, zoneid_t, ip_stack_t *);
347 extern  void	ire_walk_ill_tables(uint_t match_flags, uint_t ire_type,
348     pfv_t func, void *arg, size_t ftbl_sz, size_t htbl_sz,
349     irb_t **ipftbl, size_t ctbl_sz, irb_t *ipctbl, ill_t *ill,
350     zoneid_t zoneid, ip_stack_t *);
351 extern	void	ire_walk_srcif_table_v4(pfv_t, void *, ip_stack_t *);
352 extern	void	ire_walk_v6(pfv_t, void *, zoneid_t, ip_stack_t *);
353 
354 extern boolean_t	ire_multirt_lookup(ire_t **, ire_t **, uint32_t,
355     const struct ts_label_s *, ip_stack_t *);
356 extern boolean_t	ire_multirt_need_resolve(ipaddr_t,
357     const struct ts_label_s *, ip_stack_t *);
358 extern boolean_t	ire_multirt_lookup_v6(ire_t **, ire_t **, uint32_t,
359     const struct ts_label_s *, ip_stack_t *);
360 extern boolean_t	ire_multirt_need_resolve_v6(const in6_addr_t *,
361     const struct ts_label_s *, ip_stack_t *);
362 
363 extern ire_t	*ipif_lookup_multi_ire(ipif_t *, ipaddr_t);
364 extern ire_t	*ipif_lookup_multi_ire_v6(ipif_t *, const in6_addr_t *);
365 
366 extern ire_t	*ire_get_next_bcast_ire(ire_t *, ire_t *);
367 extern ire_t	*ire_get_next_default_ire(ire_t *, ire_t *);
368 
369 extern  void	ire_arpresolve(ire_t *,  ill_t *);
370 extern  void	ire_freemblk(ire_t *);
371 extern boolean_t	ire_match_args(ire_t *, ipaddr_t, ipaddr_t, ipaddr_t,
372     int, const ipif_t *, zoneid_t, uint32_t, const struct ts_label_s *, int);
373 extern  int	ire_nce_init(ire_t *, mblk_t *, mblk_t *);
374 extern  boolean_t	ire_walk_ill_match(uint_t, uint_t, ire_t *, ill_t *,
375     zoneid_t, ip_stack_t *);
376 
377 #endif /* _KERNEL */
378 
379 #ifdef	__cplusplus
380 }
381 #endif
382 
383 #endif	/* _INET_IP_IRE_H */
384