2 * Copyright (c) 2017 Mellanox Technologies Inc. All rights reserved.
3 * Copyright (c) 2010 Voltaire Inc. All rights reserved.
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:
11 * Redistribution and use in source and binary forms, with or
12 * without modification, are permitted provided that the following
15 * - Redistributions of source code must retain the above
16 * copyright notice, this list of conditions and the following
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.
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
34 #define pr_fmt(fmt) "%s:%s: " fmt, KBUILD_MODNAME, __func__
36 #include <linux/export.h>
37 #include <net/netlink.h>
38 #include <net/net_namespace.h>
40 #include <rdma/rdma_netlink.h>
41 #include <linux/module.h>
42 #include "core_priv.h"
44 #include "core_priv.h"
46 static DEFINE_MUTEX(rdma_nl_mutex
);
47 static struct sock
*nls
;
49 const struct rdma_nl_cbs
*cb_table
;
50 } rdma_nl_types
[RDMA_NL_NUM_CLIENTS
];
52 int rdma_nl_chk_listeners(unsigned int group
)
54 return (netlink_has_listeners(nls
, group
)) ? 0 : -1;
56 EXPORT_SYMBOL(rdma_nl_chk_listeners
);
58 static bool is_nl_msg_valid(unsigned int type
, unsigned int op
)
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
,
68 * This BUILD_BUG_ON is intended to catch addition of new
69 * RDMA netlink protocol without updating the array above.
71 BUILD_BUG_ON(RDMA_NL_NUM_CLIENTS
!= 6);
73 if (type
>= RDMA_NL_NUM_CLIENTS
)
76 return (op
< max_num_ops
[type
]) ? true : false;
79 static bool is_nl_valid(unsigned int type
, unsigned int op
)
81 const struct rdma_nl_cbs
*cb_table
;
83 if (!is_nl_msg_valid(type
, op
))
86 cb_table
= rdma_nl_types
[type
].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
;
96 if (!cb_table
|| (!cb_table
[op
].dump
&& !cb_table
[op
].doit
))
101 void rdma_nl_register(unsigned int index
,
102 const struct rdma_nl_cbs cb_table
[])
104 mutex_lock(&rdma_nl_mutex
);
105 if (!is_nl_msg_valid(index
, 0)) {
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.
112 mutex_unlock(&rdma_nl_mutex
);
114 "The not-valid %u index was supplied to RDMA netlink\n",
119 if (rdma_nl_types
[index
].cb_table
) {
120 mutex_unlock(&rdma_nl_mutex
);
122 "The %u index is already registered in RDMA netlink\n",
127 rdma_nl_types
[index
].cb_table
= cb_table
;
128 mutex_unlock(&rdma_nl_mutex
);
130 EXPORT_SYMBOL(rdma_nl_register
);
132 void rdma_nl_unregister(unsigned int index
)
134 mutex_lock(&rdma_nl_mutex
);
135 rdma_nl_types
[index
].cb_table
= NULL
;
136 mutex_unlock(&rdma_nl_mutex
);
138 EXPORT_SYMBOL(rdma_nl_unregister
);
140 void *ibnl_put_msg(struct sk_buff
*skb
, struct nlmsghdr
**nlh
, int seq
,
141 int len
, int client
, int op
, int flags
)
143 *nlh
= nlmsg_put(skb
, 0, seq
, RDMA_NL_GET_TYPE(client
, op
), len
, flags
);
146 return nlmsg_data(*nlh
);
148 EXPORT_SYMBOL(ibnl_put_msg
);
150 int ibnl_put_attr(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
151 int len
, void *data
, int type
)
153 if (nla_put(skb
, type
, len
, data
)) {
154 nlmsg_cancel(skb
, nlh
);
159 EXPORT_SYMBOL(ibnl_put_attr
);
161 static int rdma_nl_rcv_msg(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
162 struct netlink_ext_ack
*extack
)
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
;
169 if (!is_nl_valid(index
, op
))
172 cb_table
= rdma_nl_types
[index
].cb_table
;
174 if ((cb_table
[op
].flags
& RDMA_NL_ADMIN_PERM
) &&
175 !netlink_capable(skb
, CAP_NET_ADMIN
))
179 * LS responses overload the 0x100 (NLM_F_ROOT) flag. Don't
180 * mistakenly call the .dump() function.
182 if (index
== RDMA_NL_LS
) {
183 if (cb_table
[op
].doit
)
184 return cb_table
[op
].doit(skb
, nlh
, extack
);
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
,
194 return netlink_dump_start(nls
, skb
, nlh
, &c
);
198 if (cb_table
[op
].doit
)
199 return cb_table
[op
].doit(skb
, nlh
, extack
);
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.
210 static int rdma_nl_rcv_skb(struct sk_buff
*skb
, int (*cb
)(struct sk_buff
*,
212 struct netlink_ext_ack
*))
214 struct netlink_ext_ack extack
= {};
215 struct nlmsghdr
*nlh
;
218 while (skb
->len
>= nlmsg_total_size(0)) {
221 nlh
= nlmsg_hdr(skb
);
224 if (nlh
->nlmsg_len
< NLMSG_HDRLEN
|| skb
->len
< nlh
->nlmsg_len
)
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
234 if (!(nlh
->nlmsg_flags
& NLM_F_REQUEST
) &&
235 (RDMA_NL_GET_CLIENT(nlh
->nlmsg_type
) != RDMA_NL_LS
))
238 /* Skip control messages */
239 if (nlh
->nlmsg_type
< NLMSG_MIN_TYPE
)
242 err
= cb(skb
, nlh
, &extack
);
247 if (nlh
->nlmsg_flags
& NLM_F_ACK
|| err
)
248 netlink_ack(skb
, nlh
, err
, &extack
);
251 msglen
= NLMSG_ALIGN(nlh
->nlmsg_len
);
252 if (msglen
> skb
->len
)
254 skb_pull(skb
, msglen
);
260 static void rdma_nl_rcv(struct sk_buff
*skb
)
262 mutex_lock(&rdma_nl_mutex
);
263 rdma_nl_rcv_skb(skb
, &rdma_nl_rcv_msg
);
264 mutex_unlock(&rdma_nl_mutex
);
267 int rdma_nl_unicast(struct sk_buff
*skb
, u32 pid
)
271 err
= netlink_unicast(nls
, skb
, pid
, MSG_DONTWAIT
);
272 return (err
< 0) ? err
: 0;
274 EXPORT_SYMBOL(rdma_nl_unicast
);
276 int rdma_nl_unicast_wait(struct sk_buff
*skb
, __u32 pid
)
280 err
= netlink_unicast(nls
, skb
, pid
, 0);
281 return (err
< 0) ? err
: 0;
283 EXPORT_SYMBOL(rdma_nl_unicast_wait
);
285 int rdma_nl_multicast(struct sk_buff
*skb
, unsigned int group
, gfp_t flags
)
287 return nlmsg_multicast(nls
, skb
, 0, group
, flags
);
289 EXPORT_SYMBOL(rdma_nl_multicast
);
291 int __init
rdma_nl_init(void)
293 struct netlink_kernel_cfg cfg
= {
294 .input
= rdma_nl_rcv
,
297 nls
= netlink_kernel_create(&init_net
, NETLINK_RDMA
, &cfg
);
301 nls
->sk_sndtimeo
= 10 * HZ
;
305 void rdma_nl_exit(void)
309 for (idx
= 0; idx
< RDMA_NL_NUM_CLIENTS
; idx
++)
310 rdma_nl_unregister(idx
);
312 netlink_kernel_release(nls
);
315 MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK
, NETLINK_RDMA
);