1 /*
2 * GPL HEADER START
3 *
4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
5 *
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.
9 *
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).
15 *
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
19 *
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
22 * have any questions.
23 *
24 * GPL HEADER END
25 */
26 /*
27 * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
29 *
30 * Copyright (c) 2012, Intel Corporation.
31 */
32 /*
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
35 */
36 #define DEBUG_SUBSYSTEM S_LNET
37
38 #include "../../../include/linux/libcfs/libcfs.h"
39
40 #include <linux/if.h>
41 #include <linux/in.h>
42 #include <linux/file.h>
43 /* For sys_open & sys_close */
44 #include <linux/syscalls.h>
45
46 static int
libcfs_sock_ioctl(int cmd,unsigned long arg)47 libcfs_sock_ioctl(int cmd, unsigned long arg)
48 {
49 mm_segment_t oldmm = get_fs();
50 struct socket *sock;
51 int rc;
52 struct file *sock_filp;
53
54 rc = sock_create (PF_INET, SOCK_STREAM, 0, &sock);
55 if (rc != 0) {
56 CERROR ("Can't create socket: %d\n", rc);
57 return rc;
58 }
59
60 sock_filp = sock_alloc_file(sock, 0, NULL);
61 if (IS_ERR(sock_filp)) {
62 sock_release(sock);
63 rc = PTR_ERR(sock_filp);
64 goto out;
65 }
66
67 set_fs(KERNEL_DS);
68 if (sock_filp->f_op->unlocked_ioctl)
69 rc = sock_filp->f_op->unlocked_ioctl(sock_filp, cmd, arg);
70 set_fs(oldmm);
71
72 fput(sock_filp);
73 out:
74 return rc;
75 }
76
77 int
libcfs_ipif_query(char * name,int * up,__u32 * ip,__u32 * mask)78 libcfs_ipif_query (char *name, int *up, __u32 *ip, __u32 *mask)
79 {
80 struct ifreq ifr;
81 int nob;
82 int rc;
83 __u32 val;
84
85 nob = strnlen(name, IFNAMSIZ);
86 if (nob == IFNAMSIZ) {
87 CERROR("Interface name %s too long\n", name);
88 return -EINVAL;
89 }
90
91 CLASSERT (sizeof(ifr.ifr_name) >= IFNAMSIZ);
92
93 strcpy(ifr.ifr_name, name);
94 rc = libcfs_sock_ioctl(SIOCGIFFLAGS, (unsigned long)&ifr);
95
96 if (rc != 0) {
97 CERROR("Can't get flags for interface %s\n", name);
98 return rc;
99 }
100
101 if ((ifr.ifr_flags & IFF_UP) == 0) {
102 CDEBUG(D_NET, "Interface %s down\n", name);
103 *up = 0;
104 *ip = *mask = 0;
105 return 0;
106 }
107
108 *up = 1;
109
110 strcpy(ifr.ifr_name, name);
111 ifr.ifr_addr.sa_family = AF_INET;
112 rc = libcfs_sock_ioctl(SIOCGIFADDR, (unsigned long)&ifr);
113
114 if (rc != 0) {
115 CERROR("Can't get IP address for interface %s\n", name);
116 return rc;
117 }
118
119 val = ((struct sockaddr_in *)&ifr.ifr_addr)->sin_addr.s_addr;
120 *ip = ntohl(val);
121
122 strcpy(ifr.ifr_name, name);
123 ifr.ifr_addr.sa_family = AF_INET;
124 rc = libcfs_sock_ioctl(SIOCGIFNETMASK, (unsigned long)&ifr);
125
126 if (rc != 0) {
127 CERROR("Can't get netmask for interface %s\n", name);
128 return rc;
129 }
130
131 val = ((struct sockaddr_in *)&ifr.ifr_netmask)->sin_addr.s_addr;
132 *mask = ntohl(val);
133
134 return 0;
135 }
136
137 EXPORT_SYMBOL(libcfs_ipif_query);
138
139 int
libcfs_ipif_enumerate(char *** namesp)140 libcfs_ipif_enumerate (char ***namesp)
141 {
142 /* Allocate and fill in 'names', returning # interfaces/error */
143 char **names;
144 int toobig;
145 int nalloc;
146 int nfound;
147 struct ifreq *ifr;
148 struct ifconf ifc;
149 int rc;
150 int nob;
151 int i;
152
153
154 nalloc = 16; /* first guess at max interfaces */
155 toobig = 0;
156 for (;;) {
157 if (nalloc * sizeof(*ifr) > PAGE_CACHE_SIZE) {
158 toobig = 1;
159 nalloc = PAGE_CACHE_SIZE/sizeof(*ifr);
160 CWARN("Too many interfaces: only enumerating first %d\n",
161 nalloc);
162 }
163
164 LIBCFS_ALLOC(ifr, nalloc * sizeof(*ifr));
165 if (ifr == NULL) {
166 CERROR ("ENOMEM enumerating up to %d interfaces\n", nalloc);
167 rc = -ENOMEM;
168 goto out0;
169 }
170
171 ifc.ifc_buf = (char *)ifr;
172 ifc.ifc_len = nalloc * sizeof(*ifr);
173
174 rc = libcfs_sock_ioctl(SIOCGIFCONF, (unsigned long)&ifc);
175
176 if (rc < 0) {
177 CERROR ("Error %d enumerating interfaces\n", rc);
178 goto out1;
179 }
180
181 LASSERT (rc == 0);
182
183 nfound = ifc.ifc_len/sizeof(*ifr);
184 LASSERT (nfound <= nalloc);
185
186 if (nfound < nalloc || toobig)
187 break;
188
189 LIBCFS_FREE(ifr, nalloc * sizeof(*ifr));
190 nalloc *= 2;
191 }
192
193 if (nfound == 0)
194 goto out1;
195
196 LIBCFS_ALLOC(names, nfound * sizeof(*names));
197 if (names == NULL) {
198 rc = -ENOMEM;
199 goto out1;
200 }
201
202 for (i = 0; i < nfound; i++) {
203
204 nob = strnlen (ifr[i].ifr_name, IFNAMSIZ);
205 if (nob == IFNAMSIZ) {
206 /* no space for terminating NULL */
207 CERROR("interface name %.*s too long (%d max)\n",
208 nob, ifr[i].ifr_name, IFNAMSIZ);
209 rc = -ENAMETOOLONG;
210 goto out2;
211 }
212
213 LIBCFS_ALLOC(names[i], IFNAMSIZ);
214 if (names[i] == NULL) {
215 rc = -ENOMEM;
216 goto out2;
217 }
218
219 memcpy(names[i], ifr[i].ifr_name, nob);
220 names[i][nob] = 0;
221 }
222
223 *namesp = names;
224 rc = nfound;
225
226 out2:
227 if (rc < 0)
228 libcfs_ipif_free_enumeration(names, nfound);
229 out1:
230 LIBCFS_FREE(ifr, nalloc * sizeof(*ifr));
231 out0:
232 return rc;
233 }
234
235 EXPORT_SYMBOL(libcfs_ipif_enumerate);
236
237 void
libcfs_ipif_free_enumeration(char ** names,int n)238 libcfs_ipif_free_enumeration (char **names, int n)
239 {
240 int i;
241
242 LASSERT (n > 0);
243
244 for (i = 0; i < n && names[i] != NULL; i++)
245 LIBCFS_FREE(names[i], IFNAMSIZ);
246
247 LIBCFS_FREE(names, n * sizeof(*names));
248 }
249
250 EXPORT_SYMBOL(libcfs_ipif_free_enumeration);
251
252 int
libcfs_sock_write(struct socket * sock,void * buffer,int nob,int timeout)253 libcfs_sock_write (struct socket *sock, void *buffer, int nob, int timeout)
254 {
255 int rc;
256 long ticks = timeout * HZ;
257 unsigned long then;
258 struct timeval tv;
259
260 LASSERT (nob > 0);
261 /* Caller may pass a zero timeout if she thinks the socket buffer is
262 * empty enough to take the whole message immediately */
263
264 for (;;) {
265 struct kvec iov = {
266 .iov_base = buffer,
267 .iov_len = nob
268 };
269 struct msghdr msg = {
270 .msg_flags = (timeout == 0) ? MSG_DONTWAIT : 0
271 };
272
273 if (timeout != 0) {
274 /* Set send timeout to remaining time */
275 tv = (struct timeval) {
276 .tv_sec = ticks / HZ,
277 .tv_usec = ((ticks % HZ) * 1000000) / HZ
278 };
279 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO,
280 (char *)&tv, sizeof(tv));
281 if (rc != 0) {
282 CERROR("Can't set socket send timeout %ld.%06d: %d\n",
283 (long)tv.tv_sec, (int)tv.tv_usec, rc);
284 return rc;
285 }
286 }
287
288 then = jiffies;
289 rc = kernel_sendmsg(sock, &msg, &iov, 1, nob);
290 ticks -= jiffies - then;
291
292 if (rc == nob)
293 return 0;
294
295 if (rc < 0)
296 return rc;
297
298 if (rc == 0) {
299 CERROR ("Unexpected zero rc\n");
300 return -ECONNABORTED;
301 }
302
303 if (ticks <= 0)
304 return -EAGAIN;
305
306 buffer = ((char *)buffer) + rc;
307 nob -= rc;
308 }
309
310 return 0;
311 }
312 EXPORT_SYMBOL(libcfs_sock_write);
313
314 int
libcfs_sock_read(struct socket * sock,void * buffer,int nob,int timeout)315 libcfs_sock_read (struct socket *sock, void *buffer, int nob, int timeout)
316 {
317 int rc;
318 long ticks = timeout * HZ;
319 unsigned long then;
320 struct timeval tv;
321
322 LASSERT (nob > 0);
323 LASSERT (ticks > 0);
324
325 for (;;) {
326 struct kvec iov = {
327 .iov_base = buffer,
328 .iov_len = nob
329 };
330 struct msghdr msg = {
331 .msg_flags = 0
332 };
333
334 /* Set receive timeout to remaining time */
335 tv = (struct timeval) {
336 .tv_sec = ticks / HZ,
337 .tv_usec = ((ticks % HZ) * 1000000) / HZ
338 };
339 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVTIMEO,
340 (char *)&tv, sizeof(tv));
341 if (rc != 0) {
342 CERROR("Can't set socket recv timeout %ld.%06d: %d\n",
343 (long)tv.tv_sec, (int)tv.tv_usec, rc);
344 return rc;
345 }
346
347 then = jiffies;
348 rc = kernel_recvmsg(sock, &msg, &iov, 1, nob, 0);
349 ticks -= jiffies - then;
350
351 if (rc < 0)
352 return rc;
353
354 if (rc == 0)
355 return -ECONNRESET;
356
357 buffer = ((char *)buffer) + rc;
358 nob -= rc;
359
360 if (nob == 0)
361 return 0;
362
363 if (ticks <= 0)
364 return -ETIMEDOUT;
365 }
366 }
367
368 EXPORT_SYMBOL(libcfs_sock_read);
369
370 static int
libcfs_sock_create(struct socket ** sockp,int * fatal,__u32 local_ip,int local_port)371 libcfs_sock_create (struct socket **sockp, int *fatal,
372 __u32 local_ip, int local_port)
373 {
374 struct sockaddr_in locaddr;
375 struct socket *sock;
376 int rc;
377 int option;
378
379 /* All errors are fatal except bind failure if the port is in use */
380 *fatal = 1;
381
382 rc = sock_create (PF_INET, SOCK_STREAM, 0, &sock);
383 *sockp = sock;
384 if (rc != 0) {
385 CERROR ("Can't create socket: %d\n", rc);
386 return rc;
387 }
388
389 option = 1;
390 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
391 (char *)&option, sizeof (option));
392 if (rc != 0) {
393 CERROR("Can't set SO_REUSEADDR for socket: %d\n", rc);
394 goto failed;
395 }
396
397 if (local_ip != 0 || local_port != 0) {
398 memset(&locaddr, 0, sizeof(locaddr));
399 locaddr.sin_family = AF_INET;
400 locaddr.sin_port = htons(local_port);
401 locaddr.sin_addr.s_addr = (local_ip == 0) ?
402 INADDR_ANY : htonl(local_ip);
403
404 rc = sock->ops->bind(sock, (struct sockaddr *)&locaddr,
405 sizeof(locaddr));
406 if (rc == -EADDRINUSE) {
407 CDEBUG(D_NET, "Port %d already in use\n", local_port);
408 *fatal = 0;
409 goto failed;
410 }
411 if (rc != 0) {
412 CERROR("Error trying to bind to port %d: %d\n",
413 local_port, rc);
414 goto failed;
415 }
416 }
417
418 return 0;
419
420 failed:
421 sock_release(sock);
422 return rc;
423 }
424
425 int
libcfs_sock_setbuf(struct socket * sock,int txbufsize,int rxbufsize)426 libcfs_sock_setbuf (struct socket *sock, int txbufsize, int rxbufsize)
427 {
428 int option;
429 int rc;
430
431 if (txbufsize != 0) {
432 option = txbufsize;
433 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_SNDBUF,
434 (char *)&option, sizeof (option));
435 if (rc != 0) {
436 CERROR ("Can't set send buffer %d: %d\n",
437 option, rc);
438 return rc;
439 }
440 }
441
442 if (rxbufsize != 0) {
443 option = rxbufsize;
444 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVBUF,
445 (char *)&option, sizeof (option));
446 if (rc != 0) {
447 CERROR ("Can't set receive buffer %d: %d\n",
448 option, rc);
449 return rc;
450 }
451 }
452
453 return 0;
454 }
455
456 EXPORT_SYMBOL(libcfs_sock_setbuf);
457
458 int
libcfs_sock_getaddr(struct socket * sock,int remote,__u32 * ip,int * port)459 libcfs_sock_getaddr (struct socket *sock, int remote, __u32 *ip, int *port)
460 {
461 struct sockaddr_in sin;
462 int len = sizeof (sin);
463 int rc;
464
465 rc = sock->ops->getname (sock, (struct sockaddr *)&sin, &len,
466 remote ? 2 : 0);
467 if (rc != 0) {
468 CERROR ("Error %d getting sock %s IP/port\n",
469 rc, remote ? "peer" : "local");
470 return rc;
471 }
472
473 if (ip != NULL)
474 *ip = ntohl (sin.sin_addr.s_addr);
475
476 if (port != NULL)
477 *port = ntohs (sin.sin_port);
478
479 return 0;
480 }
481
482 EXPORT_SYMBOL(libcfs_sock_getaddr);
483
484 int
libcfs_sock_getbuf(struct socket * sock,int * txbufsize,int * rxbufsize)485 libcfs_sock_getbuf (struct socket *sock, int *txbufsize, int *rxbufsize)
486 {
487
488 if (txbufsize != NULL) {
489 *txbufsize = sock->sk->sk_sndbuf;
490 }
491
492 if (rxbufsize != NULL) {
493 *rxbufsize = sock->sk->sk_rcvbuf;
494 }
495
496 return 0;
497 }
498
499 EXPORT_SYMBOL(libcfs_sock_getbuf);
500
501 int
libcfs_sock_listen(struct socket ** sockp,__u32 local_ip,int local_port,int backlog)502 libcfs_sock_listen (struct socket **sockp,
503 __u32 local_ip, int local_port, int backlog)
504 {
505 int fatal;
506 int rc;
507
508 rc = libcfs_sock_create(sockp, &fatal, local_ip, local_port);
509 if (rc != 0) {
510 if (!fatal)
511 CERROR("Can't create socket: port %d already in use\n",
512 local_port);
513 return rc;
514 }
515
516 rc = (*sockp)->ops->listen(*sockp, backlog);
517 if (rc == 0)
518 return 0;
519
520 CERROR("Can't set listen backlog %d: %d\n", backlog, rc);
521 sock_release(*sockp);
522 return rc;
523 }
524
525 EXPORT_SYMBOL(libcfs_sock_listen);
526
527 int
libcfs_sock_accept(struct socket ** newsockp,struct socket * sock)528 libcfs_sock_accept (struct socket **newsockp, struct socket *sock)
529 {
530 wait_queue_t wait;
531 struct socket *newsock;
532 int rc;
533
534 init_waitqueue_entry(&wait, current);
535
536 /* XXX this should add a ref to sock->ops->owner, if
537 * TCP could be a module */
538 rc = sock_create_lite(PF_PACKET, sock->type, IPPROTO_TCP, &newsock);
539 if (rc) {
540 CERROR("Can't allocate socket\n");
541 return rc;
542 }
543
544 newsock->ops = sock->ops;
545
546 rc = sock->ops->accept(sock, newsock, O_NONBLOCK);
547 if (rc == -EAGAIN) {
548 /* Nothing ready, so wait for activity */
549 set_current_state(TASK_INTERRUPTIBLE);
550 add_wait_queue(sk_sleep(sock->sk), &wait);
551 schedule();
552 remove_wait_queue(sk_sleep(sock->sk), &wait);
553 set_current_state(TASK_RUNNING);
554 rc = sock->ops->accept(sock, newsock, O_NONBLOCK);
555 }
556
557 if (rc != 0)
558 goto failed;
559
560 *newsockp = newsock;
561 return 0;
562
563 failed:
564 sock_release(newsock);
565 return rc;
566 }
567
568 EXPORT_SYMBOL(libcfs_sock_accept);
569
570 void
libcfs_sock_abort_accept(struct socket * sock)571 libcfs_sock_abort_accept (struct socket *sock)
572 {
573 wake_up_all(sk_sleep(sock->sk));
574 }
575
576 EXPORT_SYMBOL(libcfs_sock_abort_accept);
577
578 int
libcfs_sock_connect(struct socket ** sockp,int * fatal,__u32 local_ip,int local_port,__u32 peer_ip,int peer_port)579 libcfs_sock_connect (struct socket **sockp, int *fatal,
580 __u32 local_ip, int local_port,
581 __u32 peer_ip, int peer_port)
582 {
583 struct sockaddr_in srvaddr;
584 int rc;
585
586 rc = libcfs_sock_create(sockp, fatal, local_ip, local_port);
587 if (rc != 0)
588 return rc;
589
590 memset (&srvaddr, 0, sizeof (srvaddr));
591 srvaddr.sin_family = AF_INET;
592 srvaddr.sin_port = htons(peer_port);
593 srvaddr.sin_addr.s_addr = htonl(peer_ip);
594
595 rc = (*sockp)->ops->connect(*sockp,
596 (struct sockaddr *)&srvaddr, sizeof(srvaddr),
597 0);
598 if (rc == 0)
599 return 0;
600
601 /* EADDRNOTAVAIL probably means we're already connected to the same
602 * peer/port on the same local port on a differently typed
603 * connection. Let our caller retry with a different local
604 * port... */
605 *fatal = !(rc == -EADDRNOTAVAIL);
606
607 CDEBUG_LIMIT(*fatal ? D_NETERROR : D_NET,
608 "Error %d connecting %pI4h/%d -> %pI4h/%d\n", rc,
609 &local_ip, local_port, &peer_ip, peer_port);
610
611 sock_release(*sockp);
612 return rc;
613 }
614
615 EXPORT_SYMBOL(libcfs_sock_connect);
616
617 void
libcfs_sock_release(struct socket * sock)618 libcfs_sock_release (struct socket *sock)
619 {
620 sock_release(sock);
621 }
622
623 EXPORT_SYMBOL(libcfs_sock_release);
624