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 2010 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
24 * Copyright 2012 OmniTI Computer Consulting, Inc All rights reserved.
25 * Copyright 2018 Joyent, Inc.
26 */
27
28 #ifndef _SYS_AGGR_IMPL_H
29 #define _SYS_AGGR_IMPL_H
30
31 #include <sys/types.h>
32 #include <sys/cred.h>
33 #include <sys/mac_ether.h>
34 #include <sys/mac_provider.h>
35 #include <sys/mac_client.h>
36 #include <sys/mac_client_priv.h>
37 #include <sys/aggr_lacp.h>
38
39 #ifdef __cplusplus
40 extern "C" {
41 #endif
42
43 #ifdef _KERNEL
44
45 #define AGGR_MINOR_CTL 1 /* control interface minor */
46
47 /* flags for aggr_grp_modify() */
48 #define AGGR_MODIFY_POLICY 0x01
49 #define AGGR_MODIFY_MAC 0x02
50 #define AGGR_MODIFY_LACP_MODE 0x04
51 #define AGGR_MODIFY_LACP_TIMER 0x08
52
53 /*
54 * Possible value of aggr_rseudo_rx_ring_t.arr_flags. Set when the ring entry
55 * in the pseudo RX group is used.
56 */
57 #define MAC_PSEUDO_RING_INUSE 0x01
58
59 #define MAX_GROUPS_PER_PORT 128
60
61 /*
62 * VLAN filters placed on the Rx pseudo group.
63 */
64 typedef struct aggr_vlan {
65 list_node_t av_link;
66 uint16_t av_vid; /* VLAN ID */
67 uint_t av_refs; /* num aggr clients using this VID */
68 } aggr_vlan_t;
69
70 typedef struct aggr_unicst_addr_s {
71 uint8_t aua_addr[ETHERADDRL];
72 struct aggr_unicst_addr_s *aua_next;
73 } aggr_unicst_addr_t;
74
75 typedef struct aggr_pseudo_rx_ring_s {
76 mac_ring_handle_t arr_rh; /* set by aggr_fill_ring() */
77 struct aggr_port_s *arr_port;
78 struct aggr_pseudo_rx_group_s *arr_grp;
79 mac_ring_handle_t arr_hw_rh;
80 uint_t arr_flags;
81 uint64_t arr_gen;
82 } aggr_pseudo_rx_ring_t;
83
84 /*
85 * An aggr pseudo group abstracts the underlying ports' HW groups. For
86 * example, if each port has 8 groups (mac_group_t), then the aggr
87 * will create 8 pseudo groups. Each pseudo group represents a
88 * collection of HW groups: one group from each port. If you have
89 * three ports then the pseudo group stands in for three HW groups.
90 */
91 typedef struct aggr_pseudo_rx_group_s {
92 uint_t arg_index;
93 struct aggr_grp_s *arg_grp; /* filled in by aggr_fill_group() */
94 mac_group_handle_t arg_gh; /* filled in by aggr_fill_group() */
95 aggr_unicst_addr_t *arg_macaddr;
96 aggr_pseudo_rx_ring_t arg_rings[MAX_RINGS_PER_GROUP];
97 uint_t arg_ring_cnt;
98 uint_t arg_untagged; /* num clients untagged */
99 list_t arg_vlans; /* VLANs on this group */
100 } aggr_pseudo_rx_group_t;
101
102 typedef struct aggr_pseudo_tx_ring_s {
103 mac_ring_handle_t atr_rh; /* filled in by aggr_fill_ring() */
104 struct aggr_port_s *atr_port;
105 mac_ring_handle_t atr_hw_rh;
106 uint_t atr_flags;
107 } aggr_pseudo_tx_ring_t;
108
109 typedef struct aggr_pseudo_tx_group_s {
110 mac_group_handle_t atg_gh; /* filled in by aggr_fill_group() */
111 uint_t atg_ring_cnt;
112 aggr_pseudo_tx_ring_t atg_rings[MAX_RINGS_PER_GROUP];
113 } aggr_pseudo_tx_group_t;
114
115 /*
116 * A link aggregation MAC port.
117 * Note that lp_next is protected by the lg_lock of the group the
118 * port is part of.
119 */
120 typedef struct aggr_port_s {
121 struct aggr_port_s *lp_next;
122 struct aggr_grp_s *lp_grp; /* back ptr to group */
123 datalink_id_t lp_linkid;
124 uint16_t lp_portid;
125 uint8_t lp_addr[ETHERADDRL]; /* port MAC address */
126 uint32_t lp_refs; /* refcount */
127 aggr_port_state_t lp_state;
128 uint32_t lp_started : 1,
129 lp_tx_enabled : 1,
130 lp_collector_enabled : 1,
131 lp_promisc_on : 1,
132 lp_no_link_update : 1,
133 lp_tx_grp_added : 1,
134 lp_closing : 1,
135 lp_pad_bits : 25;
136 mac_handle_t lp_mh;
137
138 mac_client_handle_t lp_mch;
139
140 const mac_info_t *lp_mip;
141 mac_notify_handle_t lp_mnh;
142 uint_t lp_tx_idx; /* idx in group's tx array */
143 uint64_t lp_ifspeed;
144 link_state_t lp_link_state;
145 link_duplex_t lp_link_duplex;
146 uint64_t lp_stat[MAC_NSTAT];
147 uint64_t lp_ether_stat[ETHER_NSTAT];
148 aggr_lacp_port_t lp_lacp; /* LACP state */
149 lacp_stats_t lp_lacp_stats;
150 uint32_t lp_margin;
151
152 mac_unicast_handle_t lp_mah;
153
154 /* List of non-primary addresses that requires promiscous mode set */
155 aggr_unicst_addr_t *lp_prom_addr;
156
157 /*
158 * References to the underlying HW Rx groups of this port.
159 * Used by aggr to program HW classification for the pseudo
160 * groups.
161 */
162 mac_group_handle_t lp_hwghs[MAX_GROUPS_PER_PORT];
163
164 int lp_tx_ring_cnt;
165 /* handles of the underlying HW TX rings */
166 mac_ring_handle_t *lp_tx_rings;
167 /*
168 * Handles of the pseudo TX rings. Each of them maps to
169 * corresponding hardware TX ring in lp_tx_rings[]. A
170 * pseudo TX ring is presented to aggr primary mac
171 * client even when underlying NIC has no TX ring.
172 */
173 mac_ring_handle_t *lp_pseudo_tx_rings;
174 void *lp_tx_notify_mh;
175 } aggr_port_t;
176
177 /*
178 * A link aggregation group.
179 *
180 * The following per-group flags are defined:
181 *
182 * - lg_addr_fixed: set when the MAC address has been explicitely set
183 * when the group was created, or by a m_unicst_set() request.
184 * If this flag is not set, the MAC address of the group will be
185 * set to the first port that is added to the group.
186 *
187 * - lg_add_set: used only when lg_addr_fixed is not set. Captures whether
188 * the MAC address was initialized according to the members of the group.
189 * When set, the lg_port field points to the port from which the
190 * MAC address was initialized.
191 *
192 */
193 typedef struct aggr_grp_s {
194 datalink_id_t lg_linkid;
195 uint16_t lg_key; /* key (group port number) */
196 uint32_t lg_refs; /* refcount */
197 uint16_t lg_nports; /* number of MAC ports */
198 uint8_t lg_addr[ETHERADDRL]; /* group MAC address */
199 uint16_t
200 lg_closing : 1,
201 lg_addr_fixed : 1, /* fixed MAC address? */
202 lg_started : 1, /* group started? */
203 lg_promisc : 1, /* in promiscuous mode? */
204 lg_zcopy : 1,
205 lg_vlan : 1,
206 lg_force : 1,
207 lg_lso : 1,
208 lg_pad_bits : 8;
209 aggr_port_t *lg_ports; /* list of configured ports */
210 aggr_port_t *lg_mac_addr_port; /* using address of this port */
211 mac_handle_t lg_mh;
212 zoneid_t lg_zoneid;
213 uint_t lg_nattached_ports;
214 krwlock_t lg_tx_lock;
215 uint_t lg_ntx_ports;
216 aggr_port_t **lg_tx_ports; /* array of tx ports */
217 uint_t lg_tx_ports_size; /* size of lg_tx_ports */
218 uint32_t lg_tx_policy; /* outbound policy */
219 uint8_t lg_mac_tx_policy;
220 link_state_t lg_link_state;
221
222
223 /*
224 * The lg_stat_lock must be held when accessing these fields.
225 */
226 kmutex_t lg_stat_lock;
227 uint64_t lg_ifspeed;
228 link_duplex_t lg_link_duplex;
229 uint64_t lg_stat[MAC_NSTAT];
230 uint64_t lg_ether_stat[ETHER_NSTAT];
231
232 aggr_lacp_mode_t lg_lacp_mode; /* off, active, or passive */
233 Agg_t aggr; /* 802.3ad data */
234 uint32_t lg_hcksum_txflags;
235 uint_t lg_max_sdu;
236 uint32_t lg_margin;
237 mac_capab_lso_t lg_cap_lso;
238
239 /*
240 * The following fields are used by the LACP packets processing.
241 * Specifically, as the LACP packets processing is not performance
242 * critical, all LACP packets will be handled by a dedicated thread
243 * instead of in the mac_rx() call. This is to avoid the dead lock
244 * with mac_unicast_remove(), which holding the mac perimeter of the
245 * aggr, and wait for the mr_refcnt of the RX ring to drop to zero.
246 */
247 kmutex_t lg_lacp_lock;
248 kcondvar_t lg_lacp_cv;
249 mblk_t *lg_lacp_head;
250 mblk_t *lg_lacp_tail;
251 kthread_t *lg_lacp_rx_thread;
252 boolean_t lg_lacp_done;
253
254 uint_t lg_rx_group_count;
255 aggr_pseudo_rx_group_t lg_rx_groups[MAX_GROUPS_PER_PORT];
256
257 aggr_pseudo_tx_group_t lg_tx_group;
258
259 kmutex_t lg_tx_flowctl_lock;
260 kcondvar_t lg_tx_flowctl_cv;
261 uint_t lg_tx_blocked_cnt;
262 mac_ring_handle_t *lg_tx_blocked_rings;
263 kthread_t *lg_tx_notify_thread;
264 boolean_t lg_tx_notify_done;
265
266 /*
267 * The following fields are used by aggr to wait for all the
268 * aggr_port_notify_cb() and aggr_port_timer_thread() to finish
269 * before it calls mac_unregister() when the aggr is deleted.
270 */
271 kmutex_t lg_port_lock;
272 kcondvar_t lg_port_cv;
273 int lg_port_ref;
274 } aggr_grp_t;
275
276 #define AGGR_GRP_REFHOLD(grp) { \
277 atomic_inc_32(&(grp)->lg_refs); \
278 ASSERT((grp)->lg_refs != 0); \
279 }
280
281 #define AGGR_GRP_REFRELE(grp) { \
282 ASSERT((grp)->lg_refs != 0); \
283 membar_exit(); \
284 if (atomic_dec_32_nv(&(grp)->lg_refs) == 0) \
285 aggr_grp_free(grp); \
286 }
287
288 #define AGGR_PORT_REFHOLD(port) { \
289 atomic_inc_32(&(port)->lp_refs); \
290 ASSERT((port)->lp_refs != 0); \
291 }
292
293 #define AGGR_PORT_REFRELE(port) { \
294 ASSERT((port)->lp_refs != 0); \
295 membar_exit(); \
296 if (atomic_dec_32_nv(&(port)->lp_refs) == 0) \
297 aggr_port_free(port); \
298 }
299
300 extern dev_info_t *aggr_dip;
301 extern int aggr_ioc_init(void);
302 extern void aggr_ioc_fini(void);
303
304 typedef int (*aggr_grp_info_new_grp_fn_t)(void *, datalink_id_t, uint32_t,
305 uchar_t *, boolean_t, boolean_t, uint32_t, uint32_t, aggr_lacp_mode_t,
306 aggr_lacp_timer_t);
307 typedef int (*aggr_grp_info_new_port_fn_t)(void *, datalink_id_t, uchar_t *,
308 aggr_port_state_t, aggr_lacp_state_t *);
309
310 extern void aggr_grp_init(void);
311 extern void aggr_grp_fini(void);
312 extern int aggr_grp_create(datalink_id_t, uint32_t, uint_t, laioc_port_t *,
313 uint32_t, boolean_t, boolean_t, uchar_t *, aggr_lacp_mode_t,
314 aggr_lacp_timer_t, cred_t *);
315 extern int aggr_grp_delete(datalink_id_t, cred_t *);
316 extern void aggr_grp_free(aggr_grp_t *);
317
318 extern int aggr_grp_info(datalink_id_t, void *, aggr_grp_info_new_grp_fn_t,
319 aggr_grp_info_new_port_fn_t, cred_t *);
320 extern void aggr_grp_notify(aggr_grp_t *, uint32_t);
321 extern boolean_t aggr_grp_attach_port(aggr_grp_t *, aggr_port_t *);
322 extern boolean_t aggr_grp_detach_port(aggr_grp_t *, aggr_port_t *);
323 extern void aggr_grp_port_mac_changed(aggr_grp_t *, aggr_port_t *,
324 boolean_t *, boolean_t *);
325 extern int aggr_grp_add_ports(datalink_id_t, uint_t, boolean_t,
326 laioc_port_t *);
327 extern int aggr_grp_rem_ports(datalink_id_t, uint_t, laioc_port_t *);
328 extern boolean_t aggr_grp_update_ports_mac(aggr_grp_t *);
329 extern int aggr_grp_modify(datalink_id_t, uint8_t, uint32_t, boolean_t,
330 const uchar_t *, aggr_lacp_mode_t, aggr_lacp_timer_t);
331 extern void aggr_grp_multicst_port(aggr_port_t *, boolean_t);
332 extern uint_t aggr_grp_count(void);
333 extern void aggr_grp_update_default(aggr_grp_t *);
334
335 extern void aggr_port_init(void);
336 extern void aggr_port_fini(void);
337 extern int aggr_port_create(aggr_grp_t *, const datalink_id_t, boolean_t,
338 aggr_port_t **);
339 extern void aggr_port_delete(aggr_port_t *);
340 extern void aggr_port_free(aggr_port_t *);
341 extern int aggr_port_start(aggr_port_t *);
342 extern void aggr_port_stop(aggr_port_t *);
343 extern int aggr_port_promisc(aggr_port_t *, boolean_t);
344 extern int aggr_port_unicst(aggr_port_t *);
345 extern int aggr_port_multicst(void *, boolean_t, const uint8_t *);
346 extern uint64_t aggr_port_stat(aggr_port_t *, uint_t);
347 extern boolean_t aggr_port_notify_link(aggr_grp_t *, aggr_port_t *);
348 extern void aggr_port_init_callbacks(aggr_port_t *);
349
350 extern void aggr_recv_cb(void *, mac_resource_handle_t, mblk_t *, boolean_t);
351
352 extern void aggr_tx_ring_update(void *, uintptr_t);
353 extern void aggr_tx_notify_thread(void *);
354 extern void aggr_send_port_enable(aggr_port_t *);
355 extern void aggr_send_port_disable(aggr_port_t *);
356 extern void aggr_send_update_policy(aggr_grp_t *, uint32_t);
357
358 extern void aggr_lacp_init(void);
359 extern void aggr_lacp_fini(void);
360 extern void aggr_lacp_init_port(aggr_port_t *);
361 extern void aggr_lacp_init_grp(aggr_grp_t *);
362 extern void aggr_lacp_set_mode(aggr_grp_t *, aggr_lacp_mode_t,
363 aggr_lacp_timer_t);
364 extern void aggr_lacp_update_mode(aggr_grp_t *, aggr_lacp_mode_t);
365 extern void aggr_lacp_update_timer(aggr_grp_t *, aggr_lacp_timer_t);
366 extern void aggr_lacp_rx_enqueue(aggr_port_t *, mblk_t *);
367 extern void aggr_lacp_port_attached(aggr_port_t *);
368 extern void aggr_lacp_port_detached(aggr_port_t *);
369 extern void aggr_port_lacp_set_mode(aggr_grp_t *, aggr_port_t *);
370
371 extern void aggr_lacp_rx_thread(void *);
372 extern void aggr_recv_lacp(aggr_port_t *, mac_resource_handle_t, mblk_t *);
373
374 extern void aggr_grp_port_hold(aggr_port_t *);
375 extern void aggr_grp_port_rele(aggr_port_t *);
376 extern void aggr_grp_port_wait(aggr_grp_t *);
377
378 extern int aggr_port_addmac(aggr_port_t *, uint_t, const uint8_t *);
379 extern void aggr_port_remmac(aggr_port_t *, uint_t, const uint8_t *);
380
381 extern int aggr_port_addvlan(aggr_port_t *, uint_t, uint16_t);
382 extern int aggr_port_remvlan(aggr_port_t *, uint_t, uint16_t);
383
384 extern mblk_t *aggr_ring_tx(void *, mblk_t *);
385 extern mblk_t *aggr_find_tx_ring(void *, mblk_t *,
386 uintptr_t, mac_ring_handle_t *);
387
388 #endif /* _KERNEL */
389
390 #ifdef __cplusplus
391 }
392 #endif
393
394 #endif /* _SYS_AGGR_IMPL_H */