4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2015, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
37 #define DEBUG_SUBSYSTEM S_LNET
38 #include <linux/completion.h>
40 #include "../../include/linux/lnet/lib-lnet.h"
42 static int accept_port = 988;
43 static int accept_backlog = 127;
44 static int accept_timeout = 5;
48 struct socket *pta_sock;
49 struct completion pta_signal;
50 } lnet_acceptor_state = {
55 lnet_acceptor_port(void)
59 EXPORT_SYMBOL(lnet_acceptor_port);
62 lnet_accept_magic(__u32 magic, __u32 constant)
64 return (magic == constant ||
65 magic == __swab32(constant));
68 static char *accept = "secure";
70 module_param(accept, charp, 0444);
71 MODULE_PARM_DESC(accept, "Accept connections (secure|all|none)");
72 module_param(accept_port, int, 0444);
73 MODULE_PARM_DESC(accept_port, "Acceptor's port (same on all nodes)");
74 module_param(accept_backlog, int, 0444);
75 MODULE_PARM_DESC(accept_backlog, "Acceptor's listen backlog");
76 module_param(accept_timeout, int, 0644);
77 MODULE_PARM_DESC(accept_timeout, "Acceptor's timeout (seconds)");
79 static char *accept_type;
82 lnet_acceptor_get_tunables(void)
85 * Userland acceptor uses 'accept_type' instead of 'accept', due to
86 * conflict with 'accept(2)', but kernel acceptor still uses 'accept'
87 * for compatibility. Hence the trick.
94 lnet_acceptor_timeout(void)
96 return accept_timeout;
98 EXPORT_SYMBOL(lnet_acceptor_timeout);
101 lnet_connect_console_error(int rc, lnet_nid_t peer_nid,
102 __u32 peer_ip, int peer_port)
105 /* "normal" errors */
107 CNETERR("Connection to %s at host %pI4h on port %d was refused: check that Lustre is running on that node.\n",
108 libcfs_nid2str(peer_nid),
109 &peer_ip, peer_port);
113 CNETERR("Connection to %s at host %pI4h was unreachable: the network or that node may be down, or Lustre may be misconfigured.\n",
114 libcfs_nid2str(peer_nid), &peer_ip);
117 CNETERR("Connection to %s at host %pI4h on port %d took too long: that node may be hung or experiencing high load.\n",
118 libcfs_nid2str(peer_nid),
119 &peer_ip, peer_port);
122 LCONSOLE_ERROR_MSG(0x11b, "Connection to %s at host %pI4h on port %d was reset: is it running a compatible version of Lustre and is %s one of its NIDs?\n",
123 libcfs_nid2str(peer_nid),
125 libcfs_nid2str(peer_nid));
128 LCONSOLE_ERROR_MSG(0x11c, "Protocol error connecting to %s at host %pI4h on port %d: is it running a compatible version of Lustre?\n",
129 libcfs_nid2str(peer_nid),
130 &peer_ip, peer_port);
133 LCONSOLE_ERROR_MSG(0x11d, "No privileged ports available to connect to %s at host %pI4h on port %d\n",
134 libcfs_nid2str(peer_nid),
135 &peer_ip, peer_port);
138 LCONSOLE_ERROR_MSG(0x11e, "Unexpected error %d connecting to %s at host %pI4h on port %d\n",
139 rc, libcfs_nid2str(peer_nid),
140 &peer_ip, peer_port);
144 EXPORT_SYMBOL(lnet_connect_console_error);
147 lnet_connect(struct socket **sockp, lnet_nid_t peer_nid,
148 __u32 local_ip, __u32 peer_ip, int peer_port)
150 lnet_acceptor_connreq_t cr;
156 CLASSERT(sizeof(cr) <= 16); /* not too big to be on the stack */
158 for (port = LNET_ACCEPTOR_MAX_RESERVED_PORT;
159 port >= LNET_ACCEPTOR_MIN_RESERVED_PORT;
161 /* Iterate through reserved ports. */
163 rc = lnet_sock_connect(&sock, &fatal, local_ip, port, peer_ip,
171 CLASSERT(LNET_PROTO_ACCEPTOR_VERSION == 1);
173 cr.acr_magic = LNET_PROTO_ACCEPTOR_MAGIC;
174 cr.acr_version = LNET_PROTO_ACCEPTOR_VERSION;
175 cr.acr_nid = peer_nid;
177 if (the_lnet.ln_testprotocompat) {
178 /* single-shot proto check */
179 lnet_net_lock(LNET_LOCK_EX);
180 if (the_lnet.ln_testprotocompat & 4) {
182 the_lnet.ln_testprotocompat &= ~4;
184 if (the_lnet.ln_testprotocompat & 8) {
185 cr.acr_magic = LNET_PROTO_MAGIC;
186 the_lnet.ln_testprotocompat &= ~8;
188 lnet_net_unlock(LNET_LOCK_EX);
191 rc = lnet_sock_write(sock, &cr, sizeof(cr), accept_timeout);
205 lnet_connect_console_error(rc, peer_nid, peer_ip, peer_port);
208 EXPORT_SYMBOL(lnet_connect);
211 lnet_accept(struct socket *sock, __u32 magic)
213 lnet_acceptor_connreq_t cr;
221 LASSERT(sizeof(cr) <= 16); /* not too big for the stack */
223 rc = lnet_sock_getaddr(sock, 1, &peer_ip, &peer_port);
224 LASSERT(!rc); /* we succeeded before */
226 if (!lnet_accept_magic(magic, LNET_PROTO_ACCEPTOR_MAGIC)) {
227 if (lnet_accept_magic(magic, LNET_PROTO_MAGIC)) {
229 * future version compatibility!
230 * When LNET unifies protocols over all LNDs, the first
231 * thing sent will be a version query. I send back
232 * LNET_PROTO_ACCEPTOR_MAGIC to tell her I'm "old"
234 memset(&cr, 0, sizeof(cr));
235 cr.acr_magic = LNET_PROTO_ACCEPTOR_MAGIC;
236 cr.acr_version = LNET_PROTO_ACCEPTOR_VERSION;
237 rc = lnet_sock_write(sock, &cr, sizeof(cr),
241 CERROR("Error sending magic+version in response to LNET magic from %pI4h: %d\n",
246 if (magic == le32_to_cpu(LNET_PROTO_TCP_MAGIC))
247 str = "'old' socknal/tcpnal";
249 str = "unrecognised";
251 LCONSOLE_ERROR_MSG(0x11f, "Refusing connection from %pI4h magic %08x: %s acceptor protocol\n",
252 &peer_ip, magic, str);
256 flip = (magic != LNET_PROTO_ACCEPTOR_MAGIC);
258 rc = lnet_sock_read(sock, &cr.acr_version, sizeof(cr.acr_version),
261 CERROR("Error %d reading connection request version from %pI4h\n",
267 __swab32s(&cr.acr_version);
269 if (cr.acr_version != LNET_PROTO_ACCEPTOR_VERSION) {
271 * future version compatibility!
272 * An acceptor-specific protocol rev will first send a version
273 * query. I send back my current version to tell her I'm
276 int peer_version = cr.acr_version;
278 memset(&cr, 0, sizeof(cr));
279 cr.acr_magic = LNET_PROTO_ACCEPTOR_MAGIC;
280 cr.acr_version = LNET_PROTO_ACCEPTOR_VERSION;
282 rc = lnet_sock_write(sock, &cr, sizeof(cr), accept_timeout);
284 CERROR("Error sending magic+version in response to version %d from %pI4h: %d\n",
285 peer_version, &peer_ip, rc);
289 rc = lnet_sock_read(sock, &cr.acr_nid,
291 offsetof(lnet_acceptor_connreq_t, acr_nid),
294 CERROR("Error %d reading connection request from %pI4h\n",
300 __swab64s(&cr.acr_nid);
302 ni = lnet_net2ni(LNET_NIDNET(cr.acr_nid));
303 if (!ni || /* no matching net */
304 ni->ni_nid != cr.acr_nid) { /* right NET, wrong NID! */
307 LCONSOLE_ERROR_MSG(0x120, "Refusing connection from %pI4h for %s: No matching NI\n",
308 &peer_ip, libcfs_nid2str(cr.acr_nid));
312 if (!ni->ni_lnd->lnd_accept) {
313 /* This catches a request for the loopback LND */
315 LCONSOLE_ERROR_MSG(0x121, "Refusing connection from %pI4h for %s: NI doesn not accept IP connections\n",
316 &peer_ip, libcfs_nid2str(cr.acr_nid));
320 CDEBUG(D_NET, "Accept %s from %pI4h\n",
321 libcfs_nid2str(cr.acr_nid), &peer_ip);
323 rc = ni->ni_lnd->lnd_accept(ni, sock);
330 lnet_acceptor(void *arg)
332 struct socket *newsock;
337 int secure = (int)((long_ptr_t)arg);
339 LASSERT(!lnet_acceptor_state.pta_sock);
343 rc = lnet_sock_listen(&lnet_acceptor_state.pta_sock, 0, accept_port,
346 if (rc == -EADDRINUSE)
347 LCONSOLE_ERROR_MSG(0x122, "Can't start acceptor on port %d: port already in use\n",
350 LCONSOLE_ERROR_MSG(0x123, "Can't start acceptor on port %d: unexpected error %d\n",
353 lnet_acceptor_state.pta_sock = NULL;
355 LCONSOLE(0, "Accept %s, port %d\n", accept_type, accept_port);
358 /* set init status and unblock parent */
359 lnet_acceptor_state.pta_shutdown = rc;
360 complete(&lnet_acceptor_state.pta_signal);
365 while (!lnet_acceptor_state.pta_shutdown) {
366 rc = lnet_sock_accept(&newsock, lnet_acceptor_state.pta_sock);
369 CWARN("Accept error %d: pausing...\n", rc);
370 set_current_state(TASK_UNINTERRUPTIBLE);
371 schedule_timeout(cfs_time_seconds(1));
376 /* maybe the LNet acceptor thread has been waken */
377 if (lnet_acceptor_state.pta_shutdown) {
378 sock_release(newsock);
382 rc = lnet_sock_getaddr(newsock, 1, &peer_ip, &peer_port);
384 CERROR("Can't determine new connection's address\n");
388 if (secure && peer_port > LNET_ACCEPTOR_MAX_RESERVED_PORT) {
389 CERROR("Refusing connection from %pI4h: insecure port %d\n",
390 &peer_ip, peer_port);
394 rc = lnet_sock_read(newsock, &magic, sizeof(magic),
397 CERROR("Error %d reading connection request from %pI4h\n",
402 rc = lnet_accept(newsock, magic);
409 sock_release(newsock);
412 sock_release(lnet_acceptor_state.pta_sock);
413 lnet_acceptor_state.pta_sock = NULL;
415 CDEBUG(D_NET, "Acceptor stopping\n");
417 /* unblock lnet_acceptor_stop() */
418 complete(&lnet_acceptor_state.pta_signal);
423 accept2secure(const char *acc, long *sec)
425 if (!strcmp(acc, "secure")) {
428 } else if (!strcmp(acc, "all")) {
431 } else if (!strcmp(acc, "none")) {
435 LCONSOLE_ERROR_MSG(0x124, "Can't parse 'accept=\"%s\"'\n",
441 lnet_acceptor_start(void)
443 struct task_struct *task;
448 /* if acceptor is already running return immediately */
449 if (!lnet_acceptor_state.pta_shutdown)
452 LASSERT(!lnet_acceptor_state.pta_sock);
454 rc = lnet_acceptor_get_tunables();
458 init_completion(&lnet_acceptor_state.pta_signal);
459 rc = accept2secure(accept_type, &secure);
463 if (!lnet_count_acceptor_nis()) /* not required */
466 task = kthread_run(lnet_acceptor, (void *)(ulong_ptr_t)secure,
467 "acceptor_%03ld", secure);
470 CERROR("Can't start acceptor thread: %ld\n", rc2);
475 /* wait for acceptor to startup */
476 wait_for_completion(&lnet_acceptor_state.pta_signal);
478 if (!lnet_acceptor_state.pta_shutdown) {
480 LASSERT(lnet_acceptor_state.pta_sock);
484 LASSERT(!lnet_acceptor_state.pta_sock);
490 lnet_acceptor_stop(void)
494 if (lnet_acceptor_state.pta_shutdown) /* not running */
497 lnet_acceptor_state.pta_shutdown = 1;
499 sk = lnet_acceptor_state.pta_sock->sk;
501 /* awake any sleepers using safe method */
502 sk->sk_state_change(sk);
504 /* block until acceptor signals exit */
505 wait_for_completion(&lnet_acceptor_state.pta_signal);