projects/linux·blob master · refs

258 lines · 7.3 KiB

1/*
2 * Copyright (c) 2006, 2017 Oracle and/or its affiliates. All rights reserved.
3 *
4 * This software is available to you under a choice of one of two
5 * licenses. You may choose to be licensed under the terms of the GNU
6 * General Public License (GPL) Version 2, available from the file
7 * COPYING in the main directory of this source tree, or the
8 * OpenIB.org BSD license below:
9 *
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
12 * conditions are met:
13 *
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
16 * disclaimer.
17 *
18 * - Redistributions in binary form must reproduce the above
19 * copyright notice, this list of conditions and the following
20 * disclaimer in the documentation and/or other materials
21 * provided with the distribution.
22 *
23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
30 * SOFTWARE.
31 *
32 */
33#include <linux/kernel.h>
34#include <linux/slab.h>
35#include <linux/in.h>
36#include <net/net_namespace.h>
37#include <net/netns/generic.h>
38#include <linux/ipv6.h>
39
40#include "rds_single_path.h"
41#include "rds.h"
42#include "loop.h"
43
44static DEFINE_SPINLOCK(loop_conns_lock);
45static LIST_HEAD(loop_conns);
46static atomic_t rds_loop_unloading = ATOMIC_INIT(0);
47
48static void rds_loop_set_unloading(void)
49{
50 atomic_set(&rds_loop_unloading, 1);
51}
52
53static bool rds_loop_is_unloading(struct rds_connection *conn)
54{
55 return atomic_read(&rds_loop_unloading) != 0;
56}
57
58/*
59 * This 'loopback' transport is a special case for flows that originate
60 * and terminate on the same machine.
61 *
62 * Connection build-up notices if the destination address is thought of
63 * as a local address by a transport. At that time it decides to use the
64 * loopback transport instead of the bound transport of the sending socket.
65 *
66 * The loopback transport's sending path just hands the sent rds_message
67 * straight to the receiving path via an embedded rds_incoming.
68 */
69
70/*
71 * Usually a message transits both the sender and receiver's conns as it
72 * flows to the receiver. In the loopback case, though, the receive path
73 * is handed the sending conn so the sense of the addresses is reversed.
74 */
75static int rds_loop_xmit(struct rds_connection *conn, struct rds_message *rm,
76 unsigned int hdr_off, unsigned int sg,
77 unsigned int off)
78{
79 struct scatterlist *sgp = &rm->data.op_sg[sg];
80 int ret = sizeof(struct rds_header) +
81 be32_to_cpu(rm->m_inc.i_hdr.h_len);
82
83 /* Do not send cong updates to loopback */
84 if (rm->m_inc.i_hdr.h_flags & RDS_FLAG_CONG_BITMAP) {
85 rds_cong_map_updated(conn->c_fcong, ~(u64) 0);
86 ret = min_t(int, ret, sgp->length - conn->c_xmit_data_off);
87 goto out;
88 }
89
90 BUG_ON(hdr_off || sg || off);
91
92 /* rds_send_queue_rm() stored the connection path in this embedded
93 * inc; use the path init so the re-initialization keeps the field
94 * valid instead of discarding it.
95 */
96 rds_inc_path_init(&rm->m_inc, &conn->c_path[0], &conn->c_laddr);
97 /* For the embedded inc. Matching put is in loop_inc_free() */
98 rds_message_addref(rm);
99
100 rds_recv_incoming(conn, &conn->c_laddr, &conn->c_faddr, &rm->m_inc,
101 GFP_KERNEL);
102
103 rds_send_drop_acked(conn, be64_to_cpu(rm->m_inc.i_hdr.h_sequence),
104 NULL);
105
106 rds_inc_put(&rm->m_inc);
107out:
108 return ret;
109}
110
111/*
112 * See rds_loop_xmit(). Since our inc is embedded in the rm, we
113 * make sure the rm lives at least until the inc is done.
114 */
115static void rds_loop_inc_free(struct rds_incoming *inc)
116{
117 struct rds_message *rm = container_of(inc, struct rds_message, m_inc);
118
119 rds_message_put(rm);
120}
121
122/* we need to at least give the thread something to succeed */
123static int rds_loop_recv_path(struct rds_conn_path *cp)
124{
125 return 0;
126}
127
128struct rds_loop_connection {
129 struct list_head loop_node;
130 struct rds_connection *conn;
131};
132
133/*
134 * Even the loopback transport needs to keep track of its connections,
135 * so it can call rds_conn_destroy() on them on exit. N.B. there are
136 * 1+ loopback addresses (127.*.*.*) so it's not a bug to have
137 * multiple loopback conns allocated, although rather useless.
138 */
139static int rds_loop_conn_alloc(struct rds_connection *conn, gfp_t gfp)
140{
141 struct rds_loop_connection *lc;
142 unsigned long flags;
143
144 lc = kzalloc_obj(struct rds_loop_connection, gfp);
145 if (!lc)
146 return -ENOMEM;
147
148 INIT_LIST_HEAD(&lc->loop_node);
149 lc->conn = conn;
150 conn->c_transport_data = lc;
151
152 spin_lock_irqsave(&loop_conns_lock, flags);
153 list_add_tail(&lc->loop_node, &loop_conns);
154 spin_unlock_irqrestore(&loop_conns_lock, flags);
155
156 return 0;
157}
158
159static void rds_loop_conn_free(void *arg)
160{
161 struct rds_loop_connection *lc = arg;
162 unsigned long flags;
163
164 rdsdebug("lc %p\n", lc);
165 spin_lock_irqsave(&loop_conns_lock, flags);
166 list_del(&lc->loop_node);
167 spin_unlock_irqrestore(&loop_conns_lock, flags);
168 kfree(lc);
169}
170
171static int rds_loop_conn_path_connect(struct rds_conn_path *cp)
172{
173 rds_connect_complete(cp->cp_conn);
174 return 0;
175}
176
177static void rds_loop_conn_path_shutdown(struct rds_conn_path *cp)
178{
179}
180
181void rds_loop_exit(void)
182{
183 struct rds_loop_connection *lc, *_lc;
184 LIST_HEAD(tmp_list);
185
186 rds_loop_set_unloading();
187 synchronize_rcu();
188 /* avoid calling conn_destroy with irqs off */
189 spin_lock_irq(&loop_conns_lock);
190 list_splice(&loop_conns, &tmp_list);
191 INIT_LIST_HEAD(&loop_conns);
192 spin_unlock_irq(&loop_conns_lock);
193
194 list_for_each_entry_safe(lc, _lc, &tmp_list, loop_node) {
195 WARN_ON(lc->conn->c_passive);
196 rds_conn_destroy(lc->conn);
197 }
198}
199
200static void rds_loop_kill_conns(struct net *net)
201{
202 struct rds_loop_connection *lc, *_lc;
203 LIST_HEAD(tmp_list);
204
205 spin_lock_irq(&loop_conns_lock);
206 list_for_each_entry_safe(lc, _lc, &loop_conns, loop_node) {
207 struct net *c_net = read_pnet(&lc->conn->c_net);
208
209 if (net != c_net)
210 continue;
211 list_move_tail(&lc->loop_node, &tmp_list);
212 }
213 spin_unlock_irq(&loop_conns_lock);
214
215 list_for_each_entry_safe(lc, _lc, &tmp_list, loop_node) {
216 WARN_ON(lc->conn->c_passive);
217 rds_conn_destroy(lc->conn);
218 }
219}
220
221static void __net_exit rds_loop_exit_net(struct net *net)
222{
223 rds_loop_kill_conns(net);
224}
225
226static struct pernet_operations rds_loop_net_ops = {
227 .exit = rds_loop_exit_net,
228};
229
230int rds_loop_net_init(void)
231{
232 return register_pernet_device(&rds_loop_net_ops);
233}
234
235void rds_loop_net_exit(void)
236{
237 unregister_pernet_device(&rds_loop_net_ops);
238}
239
240/*
241 * This is missing .xmit_* because loop doesn't go through generic
242 * rds_send_xmit() and doesn't call rds_recv_incoming(). .listen_stop and
243 * .laddr_check are missing because transport.c doesn't iterate over
244 * rds_loop_transport.
245 */
246struct rds_transport rds_loop_transport = {
247 .xmit = rds_loop_xmit,
248 .recv_path = rds_loop_recv_path,
249 .conn_alloc = rds_loop_conn_alloc,
250 .conn_free = rds_loop_conn_free,
251 .conn_path_connect = rds_loop_conn_path_connect,
252 .conn_path_shutdown = rds_loop_conn_path_shutdown,
253 .inc_copy_to_user = rds_message_inc_copy_to_user,
254 .inc_free = rds_loop_inc_free,
255 .t_name = "loopback",
256 .t_type = RDS_TRANS_LOOP,
257 .t_unloading = rds_loop_is_unloading,
258};