Loading...
1/*
2 * Copyright (c) 2017 Mellanox Technologies Inc. All rights reserved.
3 * Copyright (c) 2010 Voltaire Inc. All rights reserved.
4 *
5 * This software is available to you under a choice of one of two
6 * licenses. You may choose to be licensed under the terms of the GNU
7 * General Public License (GPL) Version 2, available from the file
8 * COPYING in the main directory of this source tree, or the
9 * OpenIB.org BSD license below:
10 *
11 * Redistribution and use in source and binary forms, with or
12 * without modification, are permitted provided that the following
13 * conditions are met:
14 *
15 * - Redistributions of source code must retain the above
16 * copyright notice, this list of conditions and the following
17 * disclaimer.
18 *
19 * - Redistributions in binary form must reproduce the above
20 * copyright notice, this list of conditions and the following
21 * disclaimer in the documentation and/or other materials
22 * provided with the distribution.
23 *
24 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
25 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
26 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
27 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
28 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
29 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
30 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
31 * SOFTWARE.
32 */
33
34#define pr_fmt(fmt) "%s:%s: " fmt, KBUILD_MODNAME, __func__
35
36#include <linux/export.h>
37#include <net/netlink.h>
38#include <net/net_namespace.h>
39#include <net/netns/generic.h>
40#include <net/sock.h>
41#include <rdma/rdma_netlink.h>
42#include <linux/module.h>
43#include "core_priv.h"
44
45static struct {
46 const struct rdma_nl_cbs *cb_table;
47 /* Synchronizes between ongoing netlink commands and netlink client
48 * unregistration.
49 */
50 struct rw_semaphore sem;
51} rdma_nl_types[RDMA_NL_NUM_CLIENTS];
52
53bool rdma_nl_chk_listeners(unsigned int group)
54{
55 struct rdma_dev_net *rnet = rdma_net_to_dev_net(&init_net);
56
57 return netlink_has_listeners(rnet->nl_sock, group);
58}
59EXPORT_SYMBOL(rdma_nl_chk_listeners);
60
61static bool is_nl_msg_valid(unsigned int type, unsigned int op)
62{
63 static const unsigned int max_num_ops[RDMA_NL_NUM_CLIENTS] = {
64 [RDMA_NL_IWCM] = RDMA_NL_IWPM_NUM_OPS,
65 [RDMA_NL_LS] = RDMA_NL_LS_NUM_OPS,
66 [RDMA_NL_NLDEV] = RDMA_NLDEV_NUM_OPS,
67 };
68
69 /*
70 * This BUILD_BUG_ON is intended to catch addition of new
71 * RDMA netlink protocol without updating the array above.
72 */
73 BUILD_BUG_ON(RDMA_NL_NUM_CLIENTS != 6);
74
75 if (type >= RDMA_NL_NUM_CLIENTS)
76 return false;
77
78 return op < max_num_ops[type];
79}
80
81static const struct rdma_nl_cbs *
82get_cb_table(const struct sk_buff *skb, unsigned int type, unsigned int op)
83{
84 const struct rdma_nl_cbs *cb_table;
85
86 /*
87 * Currently only NLDEV client is supporting netlink commands in
88 * non init_net net namespace.
89 */
90 if (sock_net(skb->sk) != &init_net && type != RDMA_NL_NLDEV)
91 return NULL;
92
93 cb_table = READ_ONCE(rdma_nl_types[type].cb_table);
94 if (!cb_table) {
95 /*
96 * Didn't get valid reference of the table, attempt module
97 * load once.
98 */
99 up_read(&rdma_nl_types[type].sem);
100
101 request_module("rdma-netlink-subsys-%u", type);
102
103 down_read(&rdma_nl_types[type].sem);
104 cb_table = READ_ONCE(rdma_nl_types[type].cb_table);
105 }
106 if (!cb_table || (!cb_table[op].dump && !cb_table[op].doit))
107 return NULL;
108 return cb_table;
109}
110
111void rdma_nl_register(unsigned int index,
112 const struct rdma_nl_cbs cb_table[])
113{
114 if (WARN_ON(!is_nl_msg_valid(index, 0)) ||
115 WARN_ON(READ_ONCE(rdma_nl_types[index].cb_table)))
116 return;
117
118 /* Pairs with the READ_ONCE in is_nl_valid() */
119 smp_store_release(&rdma_nl_types[index].cb_table, cb_table);
120}
121EXPORT_SYMBOL(rdma_nl_register);
122
123void rdma_nl_unregister(unsigned int index)
124{
125 down_write(&rdma_nl_types[index].sem);
126 rdma_nl_types[index].cb_table = NULL;
127 up_write(&rdma_nl_types[index].sem);
128}
129EXPORT_SYMBOL(rdma_nl_unregister);
130
131void *ibnl_put_msg(struct sk_buff *skb, struct nlmsghdr **nlh, int seq,
132 int len, int client, int op, int flags)
133{
134 *nlh = nlmsg_put(skb, 0, seq, RDMA_NL_GET_TYPE(client, op), len, flags);
135 if (!*nlh)
136 return NULL;
137 return nlmsg_data(*nlh);
138}
139EXPORT_SYMBOL(ibnl_put_msg);
140
141int ibnl_put_attr(struct sk_buff *skb, struct nlmsghdr *nlh,
142 int len, void *data, int type)
143{
144 if (nla_put(skb, type, len, data)) {
145 nlmsg_cancel(skb, nlh);
146 return -EMSGSIZE;
147 }
148 return 0;
149}
150EXPORT_SYMBOL(ibnl_put_attr);
151
152static int rdma_nl_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
153 struct netlink_ext_ack *extack)
154{
155 int type = nlh->nlmsg_type;
156 unsigned int index = RDMA_NL_GET_CLIENT(type);
157 unsigned int op = RDMA_NL_GET_OP(type);
158 const struct rdma_nl_cbs *cb_table;
159 int err = -EINVAL;
160
161 if (!is_nl_msg_valid(index, op))
162 return -EINVAL;
163
164 down_read(&rdma_nl_types[index].sem);
165 cb_table = get_cb_table(skb, index, op);
166 if (!cb_table)
167 goto done;
168
169 if ((cb_table[op].flags & RDMA_NL_ADMIN_PERM) &&
170 !netlink_capable(skb, CAP_NET_ADMIN)) {
171 err = -EPERM;
172 goto done;
173 }
174
175 /*
176 * LS responses overload the 0x100 (NLM_F_ROOT) flag. Don't
177 * mistakenly call the .dump() function.
178 */
179 if (index == RDMA_NL_LS) {
180 if (cb_table[op].doit)
181 err = cb_table[op].doit(skb, nlh, extack);
182 goto done;
183 }
184 /* FIXME: Convert IWCM to properly handle doit callbacks */
185 if ((nlh->nlmsg_flags & NLM_F_DUMP) || index == RDMA_NL_IWCM) {
186 struct netlink_dump_control c = {
187 .dump = cb_table[op].dump,
188 };
189 if (c.dump)
190 err = netlink_dump_start(skb->sk, skb, nlh, &c);
191 goto done;
192 }
193
194 if (cb_table[op].doit)
195 err = cb_table[op].doit(skb, nlh, extack);
196done:
197 up_read(&rdma_nl_types[index].sem);
198 return err;
199}
200
201/*
202 * This function is similar to netlink_rcv_skb with one exception:
203 * It calls to the callback for the netlink messages without NLM_F_REQUEST
204 * flag. These messages are intended for RDMA_NL_LS consumer, so it is allowed
205 * for that consumer only.
206 */
207static int rdma_nl_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
208 struct nlmsghdr *,
209 struct netlink_ext_ack *))
210{
211 struct netlink_ext_ack extack = {};
212 struct nlmsghdr *nlh;
213 int err;
214
215 while (skb->len >= nlmsg_total_size(0)) {
216 int msglen;
217
218 nlh = nlmsg_hdr(skb);
219 err = 0;
220
221 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
222 return 0;
223
224 /*
225 * Generally speaking, the only requests are handled
226 * by the kernel, but RDMA_NL_LS is different, because it
227 * runs backward netlink scheme. Kernel initiates messages
228 * and waits for reply with data to keep pathrecord cache
229 * in sync.
230 */
231 if (!(nlh->nlmsg_flags & NLM_F_REQUEST) &&
232 (RDMA_NL_GET_CLIENT(nlh->nlmsg_type) != RDMA_NL_LS))
233 goto ack;
234
235 /* Skip control messages */
236 if (nlh->nlmsg_type < NLMSG_MIN_TYPE)
237 goto ack;
238
239 err = cb(skb, nlh, &extack);
240 if (err == -EINTR)
241 goto skip;
242
243ack:
244 if (nlh->nlmsg_flags & NLM_F_ACK || err)
245 netlink_ack(skb, nlh, err, &extack);
246
247skip:
248 msglen = NLMSG_ALIGN(nlh->nlmsg_len);
249 if (msglen > skb->len)
250 msglen = skb->len;
251 skb_pull(skb, msglen);
252 }
253
254 return 0;
255}
256
257static void rdma_nl_rcv(struct sk_buff *skb)
258{
259 rdma_nl_rcv_skb(skb, &rdma_nl_rcv_msg);
260}
261
262int rdma_nl_unicast(struct net *net, struct sk_buff *skb, u32 pid)
263{
264 struct rdma_dev_net *rnet = rdma_net_to_dev_net(net);
265 int err;
266
267 err = netlink_unicast(rnet->nl_sock, skb, pid, MSG_DONTWAIT);
268 return (err < 0) ? err : 0;
269}
270EXPORT_SYMBOL(rdma_nl_unicast);
271
272int rdma_nl_unicast_wait(struct net *net, struct sk_buff *skb, __u32 pid)
273{
274 struct rdma_dev_net *rnet = rdma_net_to_dev_net(net);
275 int err;
276
277 err = netlink_unicast(rnet->nl_sock, skb, pid, 0);
278 return (err < 0) ? err : 0;
279}
280EXPORT_SYMBOL(rdma_nl_unicast_wait);
281
282int rdma_nl_multicast(struct net *net, struct sk_buff *skb,
283 unsigned int group, gfp_t flags)
284{
285 struct rdma_dev_net *rnet = rdma_net_to_dev_net(net);
286
287 return nlmsg_multicast(rnet->nl_sock, skb, 0, group, flags);
288}
289EXPORT_SYMBOL(rdma_nl_multicast);
290
291void rdma_nl_init(void)
292{
293 int idx;
294
295 for (idx = 0; idx < RDMA_NL_NUM_CLIENTS; idx++)
296 init_rwsem(&rdma_nl_types[idx].sem);
297}
298
299void rdma_nl_exit(void)
300{
301 int idx;
302
303 for (idx = 0; idx < RDMA_NL_NUM_CLIENTS; idx++)
304 WARN(rdma_nl_types[idx].cb_table,
305 "Netlink client %d wasn't released prior to unloading %s\n",
306 idx, KBUILD_MODNAME);
307}
308
309int rdma_nl_net_init(struct rdma_dev_net *rnet)
310{
311 struct net *net = read_pnet(&rnet->net);
312 struct netlink_kernel_cfg cfg = {
313 .input = rdma_nl_rcv,
314 };
315 struct sock *nls;
316
317 nls = netlink_kernel_create(net, NETLINK_RDMA, &cfg);
318 if (!nls)
319 return -ENOMEM;
320
321 nls->sk_sndtimeo = 10 * HZ;
322 rnet->nl_sock = nls;
323 return 0;
324}
325
326void rdma_nl_net_exit(struct rdma_dev_net *rnet)
327{
328 netlink_kernel_release(rnet->nl_sock);
329}
330
331MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_RDMA);
1/*
2 * Copyright (c) 2017 Mellanox Technologies Inc. All rights reserved.
3 * Copyright (c) 2010 Voltaire Inc. All rights reserved.
4 *
5 * This software is available to you under a choice of one of two
6 * licenses. You may choose to be licensed under the terms of the GNU
7 * General Public License (GPL) Version 2, available from the file
8 * COPYING in the main directory of this source tree, or the
9 * OpenIB.org BSD license below:
10 *
11 * Redistribution and use in source and binary forms, with or
12 * without modification, are permitted provided that the following
13 * conditions are met:
14 *
15 * - Redistributions of source code must retain the above
16 * copyright notice, this list of conditions and the following
17 * disclaimer.
18 *
19 * - Redistributions in binary form must reproduce the above
20 * copyright notice, this list of conditions and the following
21 * disclaimer in the documentation and/or other materials
22 * provided with the distribution.
23 *
24 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
25 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
26 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
27 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
28 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
29 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
30 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
31 * SOFTWARE.
32 */
33
34#define pr_fmt(fmt) "%s:%s: " fmt, KBUILD_MODNAME, __func__
35
36#include <linux/export.h>
37#include <net/netlink.h>
38#include <net/net_namespace.h>
39#include <net/sock.h>
40#include <rdma/rdma_netlink.h>
41#include <linux/module.h>
42#include "core_priv.h"
43
44static DEFINE_MUTEX(rdma_nl_mutex);
45static struct sock *nls;
46static struct {
47 const struct rdma_nl_cbs *cb_table;
48} rdma_nl_types[RDMA_NL_NUM_CLIENTS];
49
50int rdma_nl_chk_listeners(unsigned int group)
51{
52 return (netlink_has_listeners(nls, group)) ? 0 : -1;
53}
54EXPORT_SYMBOL(rdma_nl_chk_listeners);
55
56static bool is_nl_msg_valid(unsigned int type, unsigned int op)
57{
58 static const unsigned int max_num_ops[RDMA_NL_NUM_CLIENTS] = {
59 [RDMA_NL_RDMA_CM] = RDMA_NL_RDMA_CM_NUM_OPS,
60 [RDMA_NL_IWCM] = RDMA_NL_IWPM_NUM_OPS,
61 [RDMA_NL_LS] = RDMA_NL_LS_NUM_OPS,
62 [RDMA_NL_NLDEV] = RDMA_NLDEV_NUM_OPS,
63 };
64
65 /*
66 * This BUILD_BUG_ON is intended to catch addition of new
67 * RDMA netlink protocol without updating the array above.
68 */
69 BUILD_BUG_ON(RDMA_NL_NUM_CLIENTS != 6);
70
71 if (type >= RDMA_NL_NUM_CLIENTS)
72 return false;
73
74 return (op < max_num_ops[type]) ? true : false;
75}
76
77static bool is_nl_valid(unsigned int type, unsigned int op)
78{
79 const struct rdma_nl_cbs *cb_table;
80
81 if (!is_nl_msg_valid(type, op))
82 return false;
83
84 if (!rdma_nl_types[type].cb_table) {
85 mutex_unlock(&rdma_nl_mutex);
86 request_module("rdma-netlink-subsys-%d", type);
87 mutex_lock(&rdma_nl_mutex);
88 }
89
90 cb_table = rdma_nl_types[type].cb_table;
91
92 if (!cb_table || (!cb_table[op].dump && !cb_table[op].doit))
93 return false;
94 return true;
95}
96
97void rdma_nl_register(unsigned int index,
98 const struct rdma_nl_cbs cb_table[])
99{
100 mutex_lock(&rdma_nl_mutex);
101 if (!is_nl_msg_valid(index, 0)) {
102 /*
103 * All clients are not interesting in success/failure of
104 * this call. They want to see the print to error log and
105 * continue their initialization. Print warning for them,
106 * because it is programmer's error to be here.
107 */
108 mutex_unlock(&rdma_nl_mutex);
109 WARN(true,
110 "The not-valid %u index was supplied to RDMA netlink\n",
111 index);
112 return;
113 }
114
115 if (rdma_nl_types[index].cb_table) {
116 mutex_unlock(&rdma_nl_mutex);
117 WARN(true,
118 "The %u index is already registered in RDMA netlink\n",
119 index);
120 return;
121 }
122
123 rdma_nl_types[index].cb_table = cb_table;
124 mutex_unlock(&rdma_nl_mutex);
125}
126EXPORT_SYMBOL(rdma_nl_register);
127
128void rdma_nl_unregister(unsigned int index)
129{
130 mutex_lock(&rdma_nl_mutex);
131 rdma_nl_types[index].cb_table = NULL;
132 mutex_unlock(&rdma_nl_mutex);
133}
134EXPORT_SYMBOL(rdma_nl_unregister);
135
136void *ibnl_put_msg(struct sk_buff *skb, struct nlmsghdr **nlh, int seq,
137 int len, int client, int op, int flags)
138{
139 *nlh = nlmsg_put(skb, 0, seq, RDMA_NL_GET_TYPE(client, op), len, flags);
140 if (!*nlh)
141 return NULL;
142 return nlmsg_data(*nlh);
143}
144EXPORT_SYMBOL(ibnl_put_msg);
145
146int ibnl_put_attr(struct sk_buff *skb, struct nlmsghdr *nlh,
147 int len, void *data, int type)
148{
149 if (nla_put(skb, type, len, data)) {
150 nlmsg_cancel(skb, nlh);
151 return -EMSGSIZE;
152 }
153 return 0;
154}
155EXPORT_SYMBOL(ibnl_put_attr);
156
157static int rdma_nl_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
158 struct netlink_ext_ack *extack)
159{
160 int type = nlh->nlmsg_type;
161 unsigned int index = RDMA_NL_GET_CLIENT(type);
162 unsigned int op = RDMA_NL_GET_OP(type);
163 const struct rdma_nl_cbs *cb_table;
164
165 if (!is_nl_valid(index, op))
166 return -EINVAL;
167
168 cb_table = rdma_nl_types[index].cb_table;
169
170 if ((cb_table[op].flags & RDMA_NL_ADMIN_PERM) &&
171 !netlink_capable(skb, CAP_NET_ADMIN))
172 return -EPERM;
173
174 /*
175 * LS responses overload the 0x100 (NLM_F_ROOT) flag. Don't
176 * mistakenly call the .dump() function.
177 */
178 if (index == RDMA_NL_LS) {
179 if (cb_table[op].doit)
180 return cb_table[op].doit(skb, nlh, extack);
181 return -EINVAL;
182 }
183 /* FIXME: Convert IWCM to properly handle doit callbacks */
184 if ((nlh->nlmsg_flags & NLM_F_DUMP) || index == RDMA_NL_RDMA_CM ||
185 index == RDMA_NL_IWCM) {
186 struct netlink_dump_control c = {
187 .dump = cb_table[op].dump,
188 };
189 if (c.dump)
190 return netlink_dump_start(nls, skb, nlh, &c);
191 return -EINVAL;
192 }
193
194 if (cb_table[op].doit)
195 return cb_table[op].doit(skb, nlh, extack);
196
197 return 0;
198}
199
200/*
201 * This function is similar to netlink_rcv_skb with one exception:
202 * It calls to the callback for the netlink messages without NLM_F_REQUEST
203 * flag. These messages are intended for RDMA_NL_LS consumer, so it is allowed
204 * for that consumer only.
205 */
206static int rdma_nl_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
207 struct nlmsghdr *,
208 struct netlink_ext_ack *))
209{
210 struct netlink_ext_ack extack = {};
211 struct nlmsghdr *nlh;
212 int err;
213
214 while (skb->len >= nlmsg_total_size(0)) {
215 int msglen;
216
217 nlh = nlmsg_hdr(skb);
218 err = 0;
219
220 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
221 return 0;
222
223 /*
224 * Generally speaking, the only requests are handled
225 * by the kernel, but RDMA_NL_LS is different, because it
226 * runs backward netlink scheme. Kernel initiates messages
227 * and waits for reply with data to keep pathrecord cache
228 * in sync.
229 */
230 if (!(nlh->nlmsg_flags & NLM_F_REQUEST) &&
231 (RDMA_NL_GET_CLIENT(nlh->nlmsg_type) != RDMA_NL_LS))
232 goto ack;
233
234 /* Skip control messages */
235 if (nlh->nlmsg_type < NLMSG_MIN_TYPE)
236 goto ack;
237
238 err = cb(skb, nlh, &extack);
239 if (err == -EINTR)
240 goto skip;
241
242ack:
243 if (nlh->nlmsg_flags & NLM_F_ACK || err)
244 netlink_ack(skb, nlh, err, &extack);
245
246skip:
247 msglen = NLMSG_ALIGN(nlh->nlmsg_len);
248 if (msglen > skb->len)
249 msglen = skb->len;
250 skb_pull(skb, msglen);
251 }
252
253 return 0;
254}
255
256static void rdma_nl_rcv(struct sk_buff *skb)
257{
258 mutex_lock(&rdma_nl_mutex);
259 rdma_nl_rcv_skb(skb, &rdma_nl_rcv_msg);
260 mutex_unlock(&rdma_nl_mutex);
261}
262
263int rdma_nl_unicast(struct sk_buff *skb, u32 pid)
264{
265 int err;
266
267 err = netlink_unicast(nls, skb, pid, MSG_DONTWAIT);
268 return (err < 0) ? err : 0;
269}
270EXPORT_SYMBOL(rdma_nl_unicast);
271
272int rdma_nl_unicast_wait(struct sk_buff *skb, __u32 pid)
273{
274 int err;
275
276 err = netlink_unicast(nls, skb, pid, 0);
277 return (err < 0) ? err : 0;
278}
279EXPORT_SYMBOL(rdma_nl_unicast_wait);
280
281int rdma_nl_multicast(struct sk_buff *skb, unsigned int group, gfp_t flags)
282{
283 return nlmsg_multicast(nls, skb, 0, group, flags);
284}
285EXPORT_SYMBOL(rdma_nl_multicast);
286
287int __init rdma_nl_init(void)
288{
289 struct netlink_kernel_cfg cfg = {
290 .input = rdma_nl_rcv,
291 };
292
293 nls = netlink_kernel_create(&init_net, NETLINK_RDMA, &cfg);
294 if (!nls)
295 return -ENOMEM;
296
297 nls->sk_sndtimeo = 10 * HZ;
298 return 0;
299}
300
301void rdma_nl_exit(void)
302{
303 int idx;
304
305 for (idx = 0; idx < RDMA_NL_NUM_CLIENTS; idx++)
306 rdma_nl_unregister(idx);
307
308 netlink_kernel_release(nls);
309}
310
311MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_RDMA);