1/*
2 * Copyright (c) 2015-2023 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28
29/*
30 * Copyright (C) 2011-2014 Matteo Landi, Luigi Rizzo. All rights reserved.
31 * Copyright (C) 2013-2014 Universita` di Pisa. All rights reserved.
32 *
33 * Redistribution and use in source and binary forms, with or without
34 * modification, are permitted provided that the following conditions
35 * are met:
36 * 1. Redistributions of source code must retain the above copyright
37 * notice, this list of conditions and the following disclaimer.
38 * 2. Redistributions in binary form must reproduce the above copyright
39 * notice, this list of conditions and the following disclaimer in the
40 * documentation and/or other materials provided with the distribution.
41 *
42 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
43 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
44 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
45 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
46 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
47 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
48 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
49 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
50 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
51 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
52 * SUCH DAMAGE.
53 */
54
55#ifndef _SKYWALK_NEXUS_FLOWSWITCH_H_
56#define _SKYWALK_NEXUS_FLOWSWITCH_H_
57
58#include <skywalk/os_skywalk_private.h>
59#include <net/ethernet.h>
60#include <net/if_vlan_var.h>
61#include <netinet/ip6.h>
62
63#include <skywalk/nexus/flowswitch/flow/flow_var.h>
64
65#if CONFIG_NEXUS_FLOWSWITCH
66/* Shared declarations for the flow switch. */
67#define NX_FSW_NAME "fsw" /* prefix for flow switch port name */
68
69#define NX_FSW_MAXRINGS NX_MAX_NUM_RING_PAIR
70#define NX_FSW_TXRINGSIZE 256 /* default TX ring size */
71#define NX_FSW_RXRINGSIZE 1024 /* default RX ring size */
72#define NX_FSW_AFRINGSIZE 256 /* default Alloc/Free ring size */
73
74#define NX_FSW_CHUNK 64 /* port chunk */
75#define NX_FSW_CHUNK_FREE 0xffffffffffffffff /* entire chunk is free */
76
77#define NX_FSW_VP_MIN NX_FSW_CHUNK
78#define NX_FSW_VP_MAX 4096 /* up to 4k ports */
79#define NX_FSW_VP_NOPORT (NX_FSW_VP_MAX+1)
80#define NX_FSW_VP_BROADCAST NX_FSW_VP_MAX
81
82#define NX_FSW_MINSLOTS 2 /* XXX unclear how many */
83#define NX_FSW_MAXSLOTS NX_MAX_NUM_SLOT_PER_RING /* max # of slots */
84
85#define NX_FSW_TXBATCH 64 /* default TX batch size */
86#define NX_FSW_RXBATCH 32 /* default RX batch size */
87
88#define NX_FSW_BUFSIZE (2 * 1024) /* default buffer size */
89
90#define NX_FSW_MINBUFSIZE 512 /* min buffer size */
91#define NX_FSW_MAXBUFSIZE (32 * 1024) /* max buffer size */
92#define NX_FSW_MAXBUFFERS (4 * 1024) /* max number of buffers */
93/* max number of buffers for memory constrained device */
94#define NX_FSW_MAXBUFFERS_MEM_CONSTRAINED (2 * 1024)
95/* default user buffer segment size for non memory-constrained device */
96#define NX_FSW_BUF_SEG_SIZE (32 * 1024)
97
98/* {min, def, max} values for large buffer size */
99#define NX_FSW_MIN_LARGE_BUFSIZE 0
100#define NX_FSW_DEF_LARGE_BUFSIZE (16 * 1024)
101#define NX_FSW_MAX_LARGE_BUFSIZE (32 * 1024)
102
103/*
104 * TODO: adi@apple.com -- minimum buflets for now; we will need to
105 * have a way to adjust this based on the underlying interface's
106 * parameters, e.g. jumbo MTU, large segment offload, etc.
107 */
108#define NX_FSW_UMD_SIZE _USER_PACKET_SIZE(BUFLETS_MIN)
109#define NX_FSW_KMD_SIZE _KERN_PACKET_SIZE(BUFLETS_MIN)
110
111#define NX_FSW_EVENT_RING_NUM 1 /* number of event rings */
112#define NX_FSW_EVENT_RING_SIZE 32 /* default event ring size */
113
114
115/*
116 * macro to check if TCP RX aggregation is enabled in flowswitch.
117 */
118#define NX_FSW_TCP_RX_AGG_ENABLED() (sk_fsw_rx_agg_tcp != 0)
119
120struct nx_flowswitch;
121
122/*
123 * Virtual port nexus adapter
124 */
125struct nexus_vp_adapter {
126 /*
127 * This is an overlay structure on nexus_adapter;
128 * make sure it contains 'up' as the first member.
129 */
130 struct nexus_adapter vpna_up;
131
132 /*
133 * Flow switch support:
134 *
135 * If the adapter is associated with a nexus port, vpna_fsw points
136 * to the flow switch this NA is attached to; vpna_nx_port is the
137 * port number used in the flow switch. Otherwise, vpna_fsw would
138 * be NULL and vpna_nx_port would be NEXUS_PORT_ANY.
139 */
140 struct nx_flowswitch *vpna_fsw;
141 nexus_port_t vpna_nx_port;
142 uint16_t vpna_gencnt;
143 boolean_t vpna_retry;
144 boolean_t vpna_pid_bound;
145 boolean_t vpna_defunct;
146 pid_t vpna_pid;
147};
148
149#define VPNA(_na) ((struct nexus_vp_adapter *)(_na))
150
151#define NEXUS_PROVIDER_FLOW_SWITCH "com.apple.nexus.flowswitch"
152
153/* fsw_state_flags */
154#define FSW_STATEF_QUIESCED 0x0001
155#define FSW_STATEF_NETAGENT_ADDED 0x0002
156#define FSW_STATEF_NETAGENT_ENABLED 0x0004
157
158#define FSW_QUIESCED(_fsw) \
159 (((_fsw)->fsw_state_flags & FSW_STATEF_QUIESCED) != 0)
160
161#define FSW_NETAGENT_ADDED(_fsw) \
162 (((_fsw)->fsw_state_flags & FSW_STATEF_NETAGENT_ADDED) != 0)
163
164#define FSW_NETAGENT_ENABLED(_fsw) \
165 (((_fsw)->fsw_state_flags & FSW_STATEF_NETAGENT_ENABLED) != 0)
166
167#if (DEVELOPMENT || DEBUG)
168#define FSW_RPS_MAX_NTHREADS 64
169struct fsw_rps_thread {
170 lck_mtx_t frt_lock;
171 struct nx_flowswitch *frt_fsw;
172 struct thread *frt_thread;
173 struct pktq frt_pktq;
174
175 uint32_t frt_idx;
176 uint32_t frt_flags;
177 uint32_t frt_requests;
178};
179
180#define FRT_RUNNING 0x00000001
181#define FRT_TERMINATEBLOCK 0x10000000
182#define FRT_TERMINATING 0x20000000
183#define FRT_TERMINATED 0x40000000
184#endif /* !DEVELOPMENT && !DEBUG */
185
186typedef enum {
187 FSW_TSO_MODE_NONE,
188 FSW_TSO_MODE_HW,
189 FSW_TSO_MODE_SW
190} fsw_tso_mode_t;
191
192/*
193 * nx_flowswitch is a descriptor for a flow switch instance.
194 * Interfaces for a flow switch are all in fsw_ports[].
195 * The array has fixed size, an empty entry does not terminate
196 * the search, but lookups only occur on attach/detach so we
197 * don't mind if they are slow.
198 *
199 * The flow switch is non blocking on the transmit ports: excess
200 * packets are dropped if there is no room on the output port.
201 *
202 * fsw_lock protects accesses to the fsw_ports array.
203 * This is a rw lock (or equivalent).
204 */
205struct nx_flowswitch {
206 decl_lck_rw_data(, fsw_lock);
207 uint32_t fsw_tx_rings;
208 uint32_t fsw_rx_rings;
209
210 struct kern_nexus *fsw_nx;
211
212 /* packet type enqueued by the class queues */
213 classq_pkt_type_t fsw_classq_enq_ptype;
214 boolean_t fsw_classq_enabled;
215
216 /* packet copy routines */
217 pkt_copy_from_pkt_t *fsw_pkt_copy_from_pkt;
218 pkt_copy_from_mbuf_t *fsw_pkt_copy_from_mbuf;
219 pkt_copy_to_mbuf_t *fsw_pkt_copy_to_mbuf;
220
221 uint8_t fsw_frame_headroom;
222 uint32_t fsw_src_lla_gencnt;
223 uint32_t fsw_pending_nonviable;
224 uint32_t fsw_low_power_gencnt;
225
226 /* The following are protected by fsw_lock. */
227 struct flow_mgr *fsw_flow_mgr;
228 netagent_session_t fsw_agent_session;
229 uuid_t fsw_agent_uuid;
230 struct ifnet *fsw_ifp; /* host interface */
231 struct nexus_adapter *fsw_nifna; /* netif adapter */
232 uint32_t fsw_state_flags; /* FSW_STATEF_* */
233
234
235 union {
236 uint64_t _buf[1];
237 uint8_t _eth_src[ETHER_ADDR_LEN];
238 } __fsw_slladdr __attribute((aligned(sizeof(uint64_t))));
239
240#define fsw_slla __fsw_slladdr._buf
241#define fsw_ether_shost __fsw_slladdr._eth_src
242
243 int (*fsw_resolve)(struct nx_flowswitch *, struct flow_route *,
244 struct __kern_packet *);
245 void (*fsw_frame)(struct nx_flowswitch *, struct flow_route *,
246 struct __kern_packet *);
247 sa_family_t (*fsw_demux)(struct nx_flowswitch *,
248 struct __kern_packet *);
249
250 struct fsw_stats fsw_stats;
251
252 /*
253 * The host interface attachment to the flowswitch (fsw_ifp), as well
254 * as the netagent registration, are guarded by the flowswitch's RW
255 * lock. During fsw_flow_bind() time, we need to make sure they are
256 * valid before proceeding forward, but holding that RW lock across
257 * the routine is not possible since the thread may block if there
258 * are other threads performing fsw_flow_{bind,unbind} on the same
259 * flow owner bucket. To prevent fsw_dtor() from happening while
260 * fsw_flow_bind() is in progress, we need to have it wait until all
261 * pending flow binds are done. To do this we add a busy counter
262 * incremented at flow bind time, and use the lock for synchronization.
263 */
264 decl_lck_mtx_data(, fsw_detach_barrier_lock);
265 uint32_t fsw_detach_flags; /* see fsw_DETACHF_* */
266 uint32_t fsw_detach_barriers;
267 uint32_t fsw_detach_waiters;
268
269 uint32_t fsw_ifp_dlt;
270
271 /*
272 * largest allocated packet size.
273 * used by: mbuf batch allocation logic during netif copy.
274 */
275 uint32_t fsw_rx_largest_size;
276
277 void (*fsw_ctor)(struct nx_flowswitch *, struct flow_route *);
278
279 /* store stats from na that is going to be deactivated */
280 struct __nx_stats_fsw *fsw_closed_na_stats;
281
282 /* ip fragments manager */
283 struct fsw_ip_frag_mgr *fsw_ipfm;
284
285 struct skoid fsw_skoid;
286
287 /* input network emulator */
288 struct netem *fsw_input_netem;
289
290 struct kern_channel *fsw_dev_ch;
291 struct kern_channel *fsw_host_ch;
292
293 /*
294 * The reaper thread gets scheduled on-demand, whenever there
295 * is any lingering flow entry needing to be freed or becoming
296 * nonviable. Access is protected by fsw_reap_lock.
297 */
298 decl_lck_mtx_data(, fsw_reap_lock);
299 uint32_t fsw_reap_flags; /* see fsw_REAPF_* */
300 uint32_t fsw_reap_requests;
301 struct thread *fsw_reap_thread;
302 char fsw_reap_name[MAXTHREADNAMESIZE];
303
304 uint64_t fsw_reap_last;
305 uint64_t fsw_drain_channel_chk_last;
306 uint64_t fsw_drain_netif_chk_last;
307
308 decl_lck_mtx_data(, fsw_linger_lock);
309 struct flow_entry_linger_head fsw_linger_head;
310 uint32_t fsw_linger_cnt;
311
312 fsw_tso_mode_t fsw_tso_mode;
313 uint32_t fsw_tso_sw_mtu;
314
315#if (DEVELOPMENT || DEBUG)
316 uint32_t fsw_rps_nthreads;
317 struct fsw_rps_thread *fsw_rps_threads;
318#endif /* !DEVELOPMENT && !DEBUG */
319};
320
321#define NX_FSW_PRIVATE(_nx) ((struct nx_flowswitch *)(_nx)->nx_arg)
322
323#define FSW_RWINIT(_fsw) \
324 lck_rw_init(&(_fsw)->fsw_lock, &nexus_lock_group, &nexus_lock_attr)
325#define FSW_WLOCK(_fsw) \
326 lck_rw_lock_exclusive(&(_fsw)->fsw_lock)
327#define FSW_WUNLOCK(_fsw) \
328 lck_rw_unlock_exclusive(&(_fsw)->fsw_lock)
329#define FSW_WLOCKTORLOCK(_fsw) \
330 lck_rw_lock_exclusive_to_shared(&(_fsw)->fsw_lock)
331#define FSW_RLOCK(_fsw) \
332 lck_rw_lock_shared(&(_fsw)->fsw_lock)
333#define FSW_RLOCKTOWLOCK(_fsw) \
334 lck_rw_lock_shared_to_exclusive(&(_fsw)->fsw_lock)
335#define FSW_RTRYLOCK(_fsw) \
336 lck_rw_try_lock(&(_fsw)->fsw_lock, LCK_RW_TYPE_SHARED)
337#define FSW_RUNLOCK(_fsw) \
338 lck_rw_unlock_shared(&(_fsw)->fsw_lock)
339#define FSW_UNLOCK(_fsw) \
340 lck_rw_done(&(_fsw)->fsw_lock)
341#define FSW_RWDESTROY(_fsw) \
342 lck_rw_destroy(&(_fsw)->fsw_lock, &nexus_lock_group)
343#define FSW_WLOCK_ASSERT_HELD(_fsw) \
344 LCK_RW_ASSERT(&(_fsw)->fsw_lock, LCK_RW_ASSERT_EXCLUSIVE)
345#define FSW_RLOCK_ASSERT_HELD(_fsw) \
346 LCK_RW_ASSERT(&(_fsw)->fsw_lock, LCK_RW_ASSERT_SHARED)
347#define FSW_LOCK_ASSERT_HELD(_fsw) \
348 LCK_RW_ASSERT(&(_fsw)->fsw_lock, LCK_RW_ASSERT_HELD)
349
350extern struct nxdom nx_flowswitch_dom_s;
351extern struct kern_nexus_domain_provider nx_fsw_prov_s;
352
353SYSCTL_DECL(_kern_skywalk_flowswitch);
354
355/* functions used by external modules to interface with flow switch */
356__BEGIN_DECLS
357extern int nx_fsw_na_find(struct kern_nexus *, struct kern_channel *,
358 struct chreq *, struct nxbind *, struct proc *, struct nexus_adapter **,
359 boolean_t);
360extern boolean_t nx_fsw_dom_port_is_reserved(struct kern_nexus *nx,
361 nexus_port_t nx_port);
362extern int nx_fsw_netagent_add(struct kern_nexus *nx);
363extern int nx_fsw_netagent_remove(struct kern_nexus *nx);
364extern void nx_fsw_netagent_update(struct kern_nexus *nx);
365extern void fsw_devna_rx(struct nexus_adapter *, struct __kern_packet *,
366 struct nexus_pkt_stats *);
367extern struct nx_flowswitch *fsw_ifp_to_fsw(struct ifnet *);
368extern void fsw_get_tso_capabilities(struct ifnet *, uint32_t *, uint32_t *);
369__END_DECLS
370#endif /* CONFIG_NEXUS_FLOWSWITCH */
371#endif /* _SKYWALK_NEXUS_FLOWSWITCH_H_ */
372