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
44 #include "core_priv.h"
45
46 static DEFINE_MUTEX(rdma_nl_mutex);
47 static struct sock *nls;
48 static struct {
49 const struct rdma_nl_cbs *cb_table;
50 } rdma_nl_types[RDMA_NL_NUM_CLIENTS];
51
rdma_nl_chk_listeners(unsigned int group)52 int rdma_nl_chk_listeners(unsigned int group)
53 {
54 return (netlink_has_listeners(nls, group)) ? 0 : -1;
55 }
56 EXPORT_SYMBOL(rdma_nl_chk_listeners);
57
is_nl_msg_valid(unsigned int type,unsigned int op)58 static bool is_nl_msg_valid(unsigned int type, unsigned int op)
59 {
60 static const unsigned int max_num_ops[RDMA_NL_NUM_CLIENTS] = {
61 [RDMA_NL_RDMA_CM] = RDMA_NL_RDMA_CM_NUM_OPS,
62 [RDMA_NL_IWCM] = RDMA_NL_IWPM_NUM_OPS,
63 [RDMA_NL_LS] = RDMA_NL_LS_NUM_OPS,
64 [RDMA_NL_NLDEV] = RDMA_NLDEV_NUM_OPS,
65 };
66
67 /*
68 * This BUILD_BUG_ON is intended to catch addition of new
69 * RDMA netlink protocol without updating the array above.
70 */
71 BUILD_BUG_ON(RDMA_NL_NUM_CLIENTS != 6);
72
73 if (type >= RDMA_NL_NUM_CLIENTS)
74 return false;
75
76 return (op < max_num_ops[type]) ? true : false;
77 }
78
is_nl_valid(unsigned int type,unsigned int op)79 static bool is_nl_valid(unsigned int type, unsigned int op)
80 {
81 const struct rdma_nl_cbs *cb_table;
82
83 if (!is_nl_msg_valid(type, op))
84 return false;
85
86 cb_table = rdma_nl_types[type].cb_table;
87 #ifdef CONFIG_MODULES
88 if (!cb_table) {
89 mutex_unlock(&rdma_nl_mutex);
90 request_module("rdma-netlink-subsys-%d", type);
91 mutex_lock(&rdma_nl_mutex);
92 cb_table = rdma_nl_types[type].cb_table;
93 }
94 #endif
95
96 if (!cb_table || (!cb_table[op].dump && !cb_table[op].doit))
97 return false;
98 return true;
99 }
100
rdma_nl_register(unsigned int index,const struct rdma_nl_cbs cb_table[])101 void rdma_nl_register(unsigned int index,
102 const struct rdma_nl_cbs cb_table[])
103 {
104 mutex_lock(&rdma_nl_mutex);
105 if (!is_nl_msg_valid(index, 0)) {
106 /*
107 * All clients are not interesting in success/failure of
108 * this call. They want to see the print to error log and
109 * continue their initialization. Print warning for them,
110 * because it is programmer's error to be here.
111 */
112 mutex_unlock(&rdma_nl_mutex);
113 WARN(true,
114 "The not-valid %u index was supplied to RDMA netlink\n",
115 index);
116 return;
117 }
118
119 if (rdma_nl_types[index].cb_table) {
120 mutex_unlock(&rdma_nl_mutex);
121 WARN(true,
122 "The %u index is already registered in RDMA netlink\n",
123 index);
124 return;
125 }
126
127 rdma_nl_types[index].cb_table = cb_table;
128 mutex_unlock(&rdma_nl_mutex);
129 }
130 EXPORT_SYMBOL(rdma_nl_register);
131
rdma_nl_unregister(unsigned int index)132 void rdma_nl_unregister(unsigned int index)
133 {
134 mutex_lock(&rdma_nl_mutex);
135 rdma_nl_types[index].cb_table = NULL;
136 mutex_unlock(&rdma_nl_mutex);
137 }
138 EXPORT_SYMBOL(rdma_nl_unregister);
139
ibnl_put_msg(struct sk_buff * skb,struct nlmsghdr ** nlh,int seq,int len,int client,int op,int flags)140 void *ibnl_put_msg(struct sk_buff *skb, struct nlmsghdr **nlh, int seq,
141 int len, int client, int op, int flags)
142 {
143 *nlh = nlmsg_put(skb, 0, seq, RDMA_NL_GET_TYPE(client, op), len, flags);
144 if (!*nlh)
145 return NULL;
146 return nlmsg_data(*nlh);
147 }
148 EXPORT_SYMBOL(ibnl_put_msg);
149
ibnl_put_attr(struct sk_buff * skb,struct nlmsghdr * nlh,int len,void * data,int type)150 int ibnl_put_attr(struct sk_buff *skb, struct nlmsghdr *nlh,
151 int len, void *data, int type)
152 {
153 if (nla_put(skb, type, len, data)) {
154 nlmsg_cancel(skb, nlh);
155 return -EMSGSIZE;
156 }
157 return 0;
158 }
159 EXPORT_SYMBOL(ibnl_put_attr);
160
rdma_nl_rcv_msg(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)161 static int rdma_nl_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
162 struct netlink_ext_ack *extack)
163 {
164 int type = nlh->nlmsg_type;
165 unsigned int index = RDMA_NL_GET_CLIENT(type);
166 unsigned int op = RDMA_NL_GET_OP(type);
167 const struct rdma_nl_cbs *cb_table;
168
169 if (!is_nl_valid(index, op))
170 return -EINVAL;
171
172 cb_table = rdma_nl_types[index].cb_table;
173
174 if ((cb_table[op].flags & RDMA_NL_ADMIN_PERM) &&
175 !netlink_capable(skb, CAP_NET_ADMIN))
176 return -EPERM;
177
178 /*
179 * LS responses overload the 0x100 (NLM_F_ROOT) flag. Don't
180 * mistakenly call the .dump() function.
181 */
182 if (index == RDMA_NL_LS) {
183 if (cb_table[op].doit)
184 return cb_table[op].doit(skb, nlh, extack);
185 return -EINVAL;
186 }
187 /* FIXME: Convert IWCM to properly handle doit callbacks */
188 if ((nlh->nlmsg_flags & NLM_F_DUMP) || index == RDMA_NL_RDMA_CM ||
189 index == RDMA_NL_IWCM) {
190 struct netlink_dump_control c = {
191 .dump = cb_table[op].dump,
192 };
193 if (c.dump)
194 return netlink_dump_start(nls, skb, nlh, &c);
195 return -EINVAL;
196 }
197
198 if (cb_table[op].doit)
199 return cb_table[op].doit(skb, nlh, extack);
200
201 return 0;
202 }
203
204 /*
205 * This function is similar to netlink_rcv_skb with one exception:
206 * It calls to the callback for the netlink messages without NLM_F_REQUEST
207 * flag. These messages are intended for RDMA_NL_LS consumer, so it is allowed
208 * for that consumer only.
209 */
rdma_nl_rcv_skb(struct sk_buff * skb,int (* cb)(struct sk_buff *,struct nlmsghdr *,struct netlink_ext_ack *))210 static int rdma_nl_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
211 struct nlmsghdr *,
212 struct netlink_ext_ack *))
213 {
214 struct netlink_ext_ack extack = {};
215 struct nlmsghdr *nlh;
216 int err;
217
218 while (skb->len >= nlmsg_total_size(0)) {
219 int msglen;
220
221 nlh = nlmsg_hdr(skb);
222 err = 0;
223
224 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
225 return 0;
226
227 /*
228 * Generally speaking, the only requests are handled
229 * by the kernel, but RDMA_NL_LS is different, because it
230 * runs backward netlink scheme. Kernel initiates messages
231 * and waits for reply with data to keep pathrecord cache
232 * in sync.
233 */
234 if (!(nlh->nlmsg_flags & NLM_F_REQUEST) &&
235 (RDMA_NL_GET_CLIENT(nlh->nlmsg_type) != RDMA_NL_LS))
236 goto ack;
237
238 /* Skip control messages */
239 if (nlh->nlmsg_type < NLMSG_MIN_TYPE)
240 goto ack;
241
242 err = cb(skb, nlh, &extack);
243 if (err == -EINTR)
244 goto skip;
245
246 ack:
247 if (nlh->nlmsg_flags & NLM_F_ACK || err)
248 netlink_ack(skb, nlh, err, &extack);
249
250 skip:
251 msglen = NLMSG_ALIGN(nlh->nlmsg_len);
252 if (msglen > skb->len)
253 msglen = skb->len;
254 skb_pull(skb, msglen);
255 }
256
257 return 0;
258 }
259
rdma_nl_rcv(struct sk_buff * skb)260 static void rdma_nl_rcv(struct sk_buff *skb)
261 {
262 mutex_lock(&rdma_nl_mutex);
263 rdma_nl_rcv_skb(skb, &rdma_nl_rcv_msg);
264 mutex_unlock(&rdma_nl_mutex);
265 }
266
rdma_nl_unicast(struct sk_buff * skb,u32 pid)267 int rdma_nl_unicast(struct sk_buff *skb, u32 pid)
268 {
269 int err;
270
271 err = netlink_unicast(nls, skb, pid, MSG_DONTWAIT);
272 return (err < 0) ? err : 0;
273 }
274 EXPORT_SYMBOL(rdma_nl_unicast);
275
rdma_nl_unicast_wait(struct sk_buff * skb,__u32 pid)276 int rdma_nl_unicast_wait(struct sk_buff *skb, __u32 pid)
277 {
278 int err;
279
280 err = netlink_unicast(nls, skb, pid, 0);
281 return (err < 0) ? err : 0;
282 }
283 EXPORT_SYMBOL(rdma_nl_unicast_wait);
284
rdma_nl_multicast(struct sk_buff * skb,unsigned int group,gfp_t flags)285 int rdma_nl_multicast(struct sk_buff *skb, unsigned int group, gfp_t flags)
286 {
287 return nlmsg_multicast(nls, skb, 0, group, flags);
288 }
289 EXPORT_SYMBOL(rdma_nl_multicast);
290
rdma_nl_init(void)291 int __init rdma_nl_init(void)
292 {
293 struct netlink_kernel_cfg cfg = {
294 .input = rdma_nl_rcv,
295 };
296
297 nls = netlink_kernel_create(&init_net, NETLINK_RDMA, &cfg);
298 if (!nls)
299 return -ENOMEM;
300
301 nls->sk_sndtimeo = 10 * HZ;
302 return 0;
303 }
304
rdma_nl_exit(void)305 void rdma_nl_exit(void)
306 {
307 int idx;
308
309 for (idx = 0; idx < RDMA_NL_NUM_CLIENTS; idx++)
310 rdma_nl_unregister(idx);
311
312 netlink_kernel_release(nls);
313 }
314
315 MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_RDMA);
316