xref: /illumos-gate/usr/src/uts/common/inet/ipnet.h (revision b7ea883b)
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 /*
23  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 
27 #ifndef _INET_IPNET_H
28 #define	_INET_IPNET_H
29 
30 #ifdef __cplusplus
31 extern "C" {
32 #endif
33 
34 #include <sys/types.h>
35 #include <sys/netstack.h>
36 #include <sys/list.h>
37 #include <netinet/in.h>
38 #include <net/if.h>
39 #include <net/bpf.h>
40 #include <net/bpfdesc.h>
41 #include <sys/avl.h>
42 #include <sys/neti.h>
43 #include <sys/hook_event.h>
44 #include <sys/zone.h>
45 #include <sys/kstat.h>
46 
47 typedef struct ipnet_kstats_s	{
48 	kstat_named_t	ik_duplicationFail;
49 	kstat_named_t	ik_dispatchOk;
50 	kstat_named_t	ik_dispatchFail;
51 	kstat_named_t	ik_dispatchHeaderDrop;
52 	kstat_named_t	ik_dispatchDupDrop;
53 	kstat_named_t	ik_dispatchPutDrop;
54 	kstat_named_t	ik_dispatchDeliver;
55 	kstat_named_t	ik_acceptOk;
56 	kstat_named_t	ik_acceptFail;
57 } ipnet_kstats_t;
58 
59 #define	IPSK_BUMP(_x, _y)	(_x)->ips_stats._y.value.ui64++
60 
61 /*
62  * Structure used to hold information for both IPv4 and IPv6 addresses.
63  *
64  * When ifa_shared is non-NULL, it points to a "fake" ipnetif_t structure
65  * that represents the network interface for each zone that shares its
66  * network stack. This is used by BPF to build a list of interface names
67  * present in each zone. Multiple ipnetif_addr_t's may point to a single
68  * ipnetif_t using ifa_shared. The typical case is the global zone has
69  * a bge0 that other zones use as bge0:1, bge0:2, etc. In ipnet, the
70  * ipnetif_addr_t's that store the IP address for bge0:1, etc, would
71  * point to an ipnetif_t stored in the if_avl_by_shared tree that has
72  * the name "bge0".
73  */
74 typedef struct ipnetif_addr {
75 	union {
76 		ipaddr_t	ifau_ip4addr;
77 		in6_addr_t	ifau_ip6addr;
78 	} ifa_addr;
79 	ipaddr_t	ifa_brdaddr;
80 	zoneid_t	ifa_zone;
81 	uint64_t	ifa_id;
82 	list_node_t	ifa_link;
83 	struct ipnetif	*ifa_shared;
84 } ipnetif_addr_t;
85 #define	ifa_ip4addr	ifa_addr.ifau_ip4addr
86 #define	ifa_ip6addr	ifa_addr.ifau_ip6addr
87 
88 /*
89  * Structure describes the ipnet module representation of an ip interface.
90  * The structure holds both IPv4 and IPv6 addresses, the address lists are
91  * protected by a mutex. The ipnetif structures are held per stack instance
92  * within avl trees indexed on name and ip index.
93  *
94  * if_avl_by_shared is used by zones that share their instance of IP with
95  * other zones. It is used to store ipnetif_t structures. An example of this
96  * is the global zone sharing its instance of IP with other local zones.
97  * In this case, if_avl_by_shared is a tree of names that are in active use
98  * by zones using a shared instance of IP.
99  * The value in if_sharecnt represents the number of ipnetif_addr_t's that
100  * point to it.
101  */
102 typedef struct ipnetif {
103 	char		if_name[LIFNAMSIZ];
104 	uint_t		if_flags;
105 	uint_t		if_index;
106 	kmutex_t	if_addr_lock;	/* protects both addr lists */
107 	list_t		if_ip4addr_list;
108 	list_t		if_ip6addr_list;
109 	avl_node_t	if_avl_by_index;
110 	avl_node_t	if_avl_by_name;
111 	dev_t		if_dev;
112 	uint_t		if_multicnt;	/* protected by ips_event_lock */
113 	kmutex_t	if_reflock;	/* protects if_refcnt */
114 	int		if_refcnt;	/* if_reflock */
115 	zoneid_t	if_zoneid;
116 	avl_node_t	if_avl_by_shared;	/* protected by ips_avl_lock */
117 	struct ipnet_stack *if_stackp;
118 	int		if_sharecnt;	/* protected by if_reflock */
119 } ipnetif_t;
120 
121 /* if_flags */
122 #define	IPNETIF_IPV4PLUMBED	0x01
123 #define	IPNETIF_IPV6PLUMBED	0x02
124 #define	IPNETIF_IPV4ALLMULTI	0x04
125 #define	IPNETIF_IPV6ALLMULTI	0x08
126 #define	IPNETIF_LOOPBACK	0x10
127 
128 /*
129  * Structure used by the accept callback function.  This is simply an address
130  * pointer into a packet (either IPv4 or IPv6), along with an address family
131  * that denotes which pointer is valid.
132  */
133 typedef struct ipnet_addrp {
134 	sa_family_t	iap_family;
135 	union {
136 		ipaddr_t	*iapu_addr4;
137 		in6_addr_t	*iapu_addr6;
138 	} iap_addrp;
139 } ipnet_addrp_t;
140 #define	iap_addr4	iap_addrp.iapu_addr4
141 #define	iap_addr6	iap_addrp.iapu_addr6
142 
143 struct ipnet;
144 struct ipobs_hook_data;
145 typedef boolean_t ipnet_acceptfn_t(struct ipnet *, struct hook_pkt_observe_s *,
146     ipnet_addrp_t *, ipnet_addrp_t *);
147 
148 /*
149  * Per instance data for all open streams. Instance data is held on a
150  * per netstack list see struct ipnet_stack below.
151  */
152 typedef struct ipnet {
153 	queue_t		*ipnet_rq;	/* read queue pointer */
154 	minor_t		ipnet_minor;	/* minor number for this instance */
155 	ipnetif_t	*ipnet_if;	/* ipnetif for this open instance */
156 	zoneid_t	ipnet_zoneid;	/* zoneid the device was opened in */
157 	uint_t		ipnet_flags;	/* see below */
158 	t_scalar_t	ipnet_family;	/* protocol family of this instance */
159 	t_uscalar_t	ipnet_dlstate;	/* dlpi state */
160 	list_node_t	ipnet_next;	/* list next member */
161 	netstack_t	*ipnet_ns;	/* netstack of zone we were opened in */
162 	ipnet_acceptfn_t *ipnet_acceptfn; /* accept callback function pointer */
163 	hook_t		*ipnet_hook;	/* hook token to unregister */
164 	void		*ipnet_data;	/* value to pass back to bpf_itap */
165 } ipnet_t;
166 
167 /* ipnet_flags */
168 #define	IPNET_PROMISC_PHYS	0x01
169 #define	IPNET_PROMISC_MULTI	0x02
170 #define	IPNET_PROMISC_SAP	0x04
171 #define	IPNET_INFO		0x08
172 #define	IPNET_LOMODE		0x10
173 
174 /*
175  * Per-netstack data holding:
176  * - net_handle_t references for IPv4 and IPv6 for this netstack.
177  * - avl trees by name and index for ip interfaces associated with this
178  *   netstack. The trees are protected by ips_avl_lock.
179  * - ips_str_list is a list of open client streams.  ips_walkers_lock in
180  *   conjunction with ips_walkers_cv and ips_walkers_cnt synchronize access to
181  *   the list.  The count is incremented in ipnet_dispatch() at the start of a
182  *   walk and decremented when the walk is finished. If the walkers count is 0
183  *   then we cv_broadcast() waiting any threads waiting on the walkers count.
184  * - ips_event_lock synchronizes ipnet_if_init() and incoming NIC info events.
185  *   We cannot be processing any NIC info events while initializing interfaces
186  *   in ipnet_if_init().
187  *
188  * Note on lock ordering: If a thread needs to both hold the ips_event_lock
189  * and any other lock such as ips_walkers_lock, ips_avl_lock, or if_addr_lock,
190  * the ips_event_lock must be held first.  This lock ordering is mandated by
191  * ipnet_nicevent_cb() which must always grab ips_event_lock before continuing
192  * with processing NIC events.
193  */
194 typedef struct ipnet_stack {
195 	net_handle_t	ips_ndv4;
196 	net_handle_t	ips_ndv6;
197 	netstack_t	*ips_netstack;
198 	hook_t		*ips_nicevents;
199 	kmutex_t	ips_event_lock;
200 	kmutex_t	ips_avl_lock;
201 	avl_tree_t	ips_avl_by_index;
202 	avl_tree_t	ips_avl_by_name;
203 	kmutex_t	ips_walkers_lock;
204 	kcondvar_t	ips_walkers_cv;
205 	uint_t		ips_walkers_cnt;
206 	list_t		ips_str_list;
207 	kstat_t		*ips_kstatp;
208 	ipnet_kstats_t	ips_stats;
209 	avl_tree_t	ips_avl_by_shared;
210 	hook_t		*ips_hook;
211 } ipnet_stack_t;
212 
213 /*
214  * Template for dl_info_ack_t initialization.  We don't have an address, so we
215  * set the address length to just the SAP length (16 bits).  We don't really
216  * have a maximum SDU, but setting it to UINT_MAX proved problematic with
217  * applications that performed arithmetic on dl_max_sdu and wrapped around, so
218  * we sleaze out and use INT_MAX.
219  */
220 #define	IPNET_INFO_ACK_INIT {						\
221 	DL_INFO_ACK,			/* dl_primitive */		\
222 	INT_MAX,			/* dl_max_sdu */		\
223 	0,				/* dl_min_sdu */		\
224 	sizeof (uint16_t),		/* dl_addr_length */ 		\
225 	DL_IPNET,			/* dl_mac_type */		\
226 	0,				/* dl_reserved */		\
227 	0,				/* dl_current_state */		\
228 	sizeof (uint16_t),		/* dl_sap_length */ 		\
229 	DL_CLDLS,			/* dl_service_mode */		\
230 	0,				/* dl_qos_length */		\
231 	0,				/* dl_qos_offset */		\
232 	0,				/* dl_range_length */		\
233 	0,				/* dl_range_offset */		\
234 	DL_STYLE1,			/* dl_provider_style */		\
235 	0,				/* dl_addr_offset */		\
236 	DL_VERSION_2,			/* dl_version */		\
237 	0,				/* dl_brdcst_addr_length */	\
238 	0				/* dl_brdcst_addr_offset */	\
239 }
240 
241 typedef void ipnet_walkfunc_t(const char *, void *, dev_t);
242 
243 extern int	ipnet_client_open(ipnetif_t *, ipnetif_t **);
244 extern void	ipnet_client_close(ipnetif_t *);
245 extern void	ipnet_close_byhandle(ipnetif_t *);
246 extern ipnet_stack_t *ipnet_find_by_zoneid(zoneid_t zoneid);
247 extern int	ipnet_get_linkid_byname(const char *, datalink_id_t *,
248     zoneid_t);
249 extern dev_t	ipnet_if_getdev(char *, zoneid_t);
250 extern const char *ipnet_name(ipnetif_t *);
251 extern int	ipnet_open_byname(const char *, ipnetif_t **, zoneid_t);
252 extern int	ipnet_promisc_add(void *, uint_t, void *, uintptr_t *, int);
253 extern void	ipnet_promisc_remove(void *);
254 extern void	ipnet_rele(ipnet_stack_t *);
255 extern void	ipnet_set_itap(bpf_itap_fn_t);
256 extern void	ipnet_walk_if(ipnet_walkfunc_t *, void *, zoneid_t);
257 
258 extern bpf_provider_t	bpf_ipnet;
259 
260 #ifdef __cplusplus
261 }
262 #endif
263 
264 #endif /* _INET_IPNET_H */
265