1/*
2 * Copyright (c) 2016 Adrian Chadd <adrian@FreeBSD.org>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 */
26
27#include <sys/cdefs.h>
28__FBSDID("$FreeBSD$");
29
30#include <sys/param.h>
31#include <sys/types.h>
32#include <sys/socket.h>
33#include <sys/cpuset.h>
34#include <sys/sysctl.h>
35
36#include <stdio.h>
37#include <stdlib.h>
38#include <unistd.h>
39#include <strings.h>
40#include <err.h>
41#include <fcntl.h>
42#include <string.h>
43#include <errno.h>
44
45#include <netinet/in.h>
46
47#include "librss.h"
48
49int
50rss_sock_set_bindmulti(int fd, int af, int val)
51{
52	int opt;
53	socklen_t optlen;
54	int retval;
55
56	/* Set bindmulti */
57	opt = val;
58	optlen = sizeof(opt);
59	retval = setsockopt(fd,
60	    af == AF_INET ? IPPROTO_IP : IPPROTO_IPV6,
61	    af == AF_INET ? IP_BINDMULTI : IPV6_BINDMULTI,
62	    &opt,
63	    optlen);
64	if (retval < 0) {
65		warn("%s: setsockopt(IP_BINDMULTI)", __func__);
66		return (-1);
67	}
68	return (0);
69}
70
71int
72rss_sock_set_rss_bucket(int fd, int af, int rss_bucket)
73{
74	int opt;
75	socklen_t optlen;
76	int retval;
77	int f, p;
78
79	switch (af) {
80	case AF_INET:
81		p = IPPROTO_IP;
82		f = IP_RSS_LISTEN_BUCKET;
83		break;
84	case AF_INET6:
85		p = IPPROTO_IPV6;
86		f = IPV6_RSS_LISTEN_BUCKET;
87		break;
88	default:
89		return (-1);
90	}
91
92	/* Set RSS bucket */
93	opt = rss_bucket;
94	optlen = sizeof(opt);
95	retval = setsockopt(fd, p, f, &opt, optlen);
96	if (retval < 0) {
97		warn("%s: setsockopt(IP_RSS_LISTEN_BUCKET)", __func__);
98		return (-1);
99	}
100	return (0);
101}
102
103int
104rss_sock_set_recvrss(int fd, int af, int val)
105{
106	int opt, retval;
107	socklen_t optlen;
108	int f1, f2, p;
109
110	switch (af) {
111	case AF_INET:
112		p = IPPROTO_IP;
113		f1 = IP_RECVFLOWID;
114		f2 = IP_RECVRSSBUCKETID;
115		break;
116	case AF_INET6:
117		p = IPPROTO_IPV6;
118		f1 = IPV6_RECVFLOWID;
119		f2 = IPV6_RECVRSSBUCKETID;
120		break;
121	default:
122		return (-1);
123	}
124
125	/* Enable/disable flowid */
126	opt = val;
127	optlen = sizeof(opt);
128	retval = setsockopt(fd, p, f1, &opt, optlen);
129	if (retval < 0) {
130		warn("%s: setsockopt(IP_RECVFLOWID)", __func__);
131		return (-1);
132	}
133
134	/* Enable/disable RSS bucket reception */
135	opt = val;
136	optlen = sizeof(opt);
137	retval = setsockopt(fd, p, f2, &opt, optlen);
138	if (retval < 0) {
139		warn("%s: setsockopt(IP_RECVRSSBUCKETID)", __func__);
140		return (-1);
141	}
142
143	return (0);
144}
145
146static int
147rss_getsysctlint(const char *s)
148{
149	int val, retval;
150	size_t rlen;
151
152	rlen = sizeof(int);
153	retval = sysctlbyname(s, &val, &rlen, NULL, 0);
154	if (retval < 0) {
155		warn("sysctlbyname (%s)", s);
156		return (-1);
157	}
158
159	return (val);
160}
161
162static int
163rss_getbucketmap(int *bucket_map, int nbuckets)
164{
165	/* XXX I'm lazy; so static string it is */
166	char bstr[2048];
167	int retval;
168	size_t rlen;
169	char *s, *ss;
170	int r, b, c;
171
172	/* Paranoia */
173	memset(bstr, '\0', sizeof(bstr));
174
175	rlen = sizeof(bstr) - 1;
176	retval = sysctlbyname("net.inet.rss.bucket_mapping", bstr, &rlen, NULL, 0);
177	if (retval < 0) {
178		warn("sysctlbyname (net.inet.rss.bucket_mapping)");
179		return (-1);
180	}
181
182	ss = bstr;
183	while ((s = strsep(&ss, " ")) != NULL) {
184		r = sscanf(s, "%d:%d", &b, &c);
185		if (r != 2) {
186			fprintf(stderr, "%s: string (%s) not parsable\n",
187			    __func__,
188			    s);
189			return (-1);
190		}
191		if (b > nbuckets) {
192			fprintf(stderr, "%s: bucket %d > nbuckets %d\n",
193			    __func__,
194			    b,
195			    nbuckets);
196			return (-1);
197		}
198		/* XXX no maxcpu check */
199		bucket_map[b] = c;
200	}
201	return (0);
202}
203
204struct rss_config *
205rss_config_get(void)
206{
207	struct rss_config *rc = NULL;
208
209	rc = calloc(1, sizeof(*rc));
210	if (rc == NULL) {
211		warn("%s: calloc", __func__);
212		goto error;
213	}
214
215	rc->rss_ncpus = rss_getsysctlint("net.inet.rss.ncpus");
216	if (rc->rss_ncpus < 0) {
217		fprintf(stderr, "%s: couldn't fetch net.inet.rss.ncpus\n", __func__);
218		goto error;
219	}
220
221	rc->rss_nbuckets = rss_getsysctlint("net.inet.rss.buckets");
222	if (rc->rss_nbuckets < 0) {
223		fprintf(stderr, "%s: couldn't fetch net.inet.rss.nbuckets\n", __func__);
224		goto error;
225	}
226
227	rc->rss_basecpu = rss_getsysctlint("net.inet.rss.basecpu");
228	if (rc->rss_basecpu< 0) {
229		fprintf(stderr, "%s: couldn't fetch net.inet.rss.basecpu\n", __func__);
230		goto error;
231	}
232
233	rc->rss_bucket_map = calloc(rc->rss_nbuckets, sizeof(int));
234	if (rc->rss_bucket_map == NULL) {
235		warn("%s: calloc (rss buckets; %d entries)", __func__, rc->rss_nbuckets);
236		goto error;
237	}
238
239	if (rss_getbucketmap(rc->rss_bucket_map, rc->rss_nbuckets) != 0) {
240		fprintf(stderr, "%s: rss_getbucketmap failed\n", __func__);
241		goto error;
242	}
243
244	return (rc);
245
246error:
247	if (rc != NULL) {
248		free(rc->rss_bucket_map);
249		free(rc);
250	}
251	return (NULL);
252}
253
254void
255rss_config_free(struct rss_config *rc)
256{
257
258	if ((rc != NULL) && rc->rss_bucket_map)
259		free(rc->rss_bucket_map);
260	if (rc != NULL)
261		free(rc);
262}
263
264int
265rss_config_get_bucket_count(struct rss_config *rc)
266{
267
268	if (rc == NULL)
269		return (-1);
270	return (rc->rss_nbuckets);
271}
272
273int
274rss_get_bucket_cpuset(struct rss_config *rc, rss_bucket_type_t btype,
275    int bucket, cpuset_t *cs)
276{
277
278	if (bucket < 0 || bucket >= rc->rss_nbuckets) {
279		errno = EINVAL;
280		return (-1);
281	}
282
283	/*
284	 * For now all buckets are the same, but eventually we'll want
285	 * to allow administrators to set separate RSS cpusets for
286	 * {kernel,user} {tx, rx} combinations.
287	 */
288	if (btype <= RSS_BUCKET_TYPE_NONE || btype > RSS_BUCKET_TYPE_MAX) {
289		errno = ENOTSUP;
290		return (-1);
291	}
292
293	CPU_ZERO(cs);
294	CPU_SET(rc->rss_bucket_map[bucket], cs);
295
296	return (0);
297}
298
299int
300rss_set_bucket_rebalance_cb(rss_bucket_rebalance_cb_t *cb, void *cbdata)
301{
302
303	(void) cb;
304	(void) cbdata;
305
306	/*
307	 * For now there's no rebalance callback, so
308	 * just return 0 and ignore it.
309	 */
310	return (0);
311}
312