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) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2012, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 #define DEBUG_SUBSYSTEM S_LNET
38 #include <libcfs/libcfs.h>
39 #include <libcfs/libcfs.h>
43 #include <linux/file.h>
44 /* For sys_open & sys_close */
45 #include <linux/syscalls.h>
48 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
55 libcfs_sock_ioctl(int cmd, unsigned long arg)
57 mm_segment_t oldmm = get_fs();
61 struct file *sock_filp;
63 rc = sock_create (PF_INET, SOCK_STREAM, 0, &sock);
65 CERROR ("Can't create socket: %d\n", rc);
69 #if !defined(HAVE_SOCK_ALLOC_FILE) && !defined(HAVE_SOCK_ALLOC_FILE_3ARGS)
70 fd = sock_map_fd(sock, 0);
78 # ifdef HAVE_SOCK_ALLOC_FILE_3ARGS
79 sock_filp = sock_alloc_file(sock, 0, NULL);
81 sock_filp = sock_alloc_file(sock, 0);
91 if (sock_filp->f_op->unlocked_ioctl)
92 rc = sock_filp->f_op->unlocked_ioctl(sock_filp, cmd, arg);
105 libcfs_ipif_query (char *name, int *up, __u32 *ip, __u32 *mask)
112 nob = strnlen(name, IFNAMSIZ);
113 if (nob == IFNAMSIZ) {
114 CERROR("Interface name %s too long\n", name);
118 CLASSERT (sizeof(ifr.ifr_name) >= IFNAMSIZ);
120 if (strlen(name) > sizeof(ifr.ifr_name)-1)
122 strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
124 rc = libcfs_sock_ioctl(SIOCGIFFLAGS, (unsigned long)&ifr);
127 CERROR("Can't get flags for interface %s\n", name);
131 if ((ifr.ifr_flags & IFF_UP) == 0) {
132 CDEBUG(D_NET, "Interface %s down\n", name);
140 if (strlen(name) > sizeof(ifr.ifr_name)-1)
142 strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
144 ifr.ifr_addr.sa_family = AF_INET;
145 rc = libcfs_sock_ioctl(SIOCGIFADDR, (unsigned long)&ifr);
148 CERROR("Can't get IP address for interface %s\n", name);
152 val = ((struct sockaddr_in *)&ifr.ifr_addr)->sin_addr.s_addr;
155 if (strlen(name) > sizeof(ifr.ifr_name)-1)
157 strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
159 ifr.ifr_addr.sa_family = AF_INET;
160 rc = libcfs_sock_ioctl(SIOCGIFNETMASK, (unsigned long)&ifr);
163 CERROR("Can't get netmask for interface %s\n", name);
167 val = ((struct sockaddr_in *)&ifr.ifr_netmask)->sin_addr.s_addr;
173 EXPORT_SYMBOL(libcfs_ipif_query);
176 libcfs_ipif_enumerate (char ***namesp)
178 /* Allocate and fill in 'names', returning # interfaces/error */
190 nalloc = 16; /* first guess at max interfaces */
193 if (nalloc * sizeof(*ifr) > PAGE_CACHE_SIZE) {
195 nalloc = PAGE_CACHE_SIZE/sizeof(*ifr);
196 CWARN("Too many interfaces: only enumerating first %d\n",
200 LIBCFS_ALLOC(ifr, nalloc * sizeof(*ifr));
202 CERROR ("ENOMEM enumerating up to %d interfaces\n", nalloc);
207 ifc.ifc_buf = (char *)ifr;
208 ifc.ifc_len = nalloc * sizeof(*ifr);
210 rc = libcfs_sock_ioctl(SIOCGIFCONF, (unsigned long)&ifc);
213 CERROR ("Error %d enumerating interfaces\n", rc);
219 nfound = ifc.ifc_len/sizeof(*ifr);
220 LASSERT (nfound <= nalloc);
222 if (nfound < nalloc || toobig)
225 LIBCFS_FREE(ifr, nalloc * sizeof(*ifr));
232 LIBCFS_ALLOC(names, nfound * sizeof(*names));
237 /* NULL out all names[i] */
238 memset (names, 0, nfound * sizeof(*names));
240 for (i = 0; i < nfound; i++) {
242 nob = strnlen (ifr[i].ifr_name, IFNAMSIZ);
243 if (nob == IFNAMSIZ) {
244 /* no space for terminating NULL */
245 CERROR("interface name %.*s too long (%d max)\n",
246 nob, ifr[i].ifr_name, IFNAMSIZ);
251 LIBCFS_ALLOC(names[i], IFNAMSIZ);
252 if (names[i] == NULL) {
257 memcpy(names[i], ifr[i].ifr_name, nob);
266 libcfs_ipif_free_enumeration(names, nfound);
268 LIBCFS_FREE(ifr, nalloc * sizeof(*ifr));
273 EXPORT_SYMBOL(libcfs_ipif_enumerate);
276 libcfs_ipif_free_enumeration (char **names, int n)
282 for (i = 0; i < n && names[i] != NULL; i++)
283 LIBCFS_FREE(names[i], IFNAMSIZ);
285 LIBCFS_FREE(names, n * sizeof(*names));
288 EXPORT_SYMBOL(libcfs_ipif_free_enumeration);
291 libcfs_sock_write (struct socket *sock, void *buffer, int nob, int timeout)
294 mm_segment_t oldmm = get_fs();
295 long ticks = timeout * HZ;
300 /* Caller may pass a zero timeout if she thinks the socket buffer is
301 * empty enough to take the whole message immediately */
308 struct msghdr msg = {
315 .msg_flags = (timeout == 0) ? MSG_DONTWAIT : 0
319 /* Set send timeout to remaining time */
320 tv = (struct timeval) {
321 .tv_sec = ticks / HZ,
322 .tv_usec = ((ticks % HZ) * 1000000) / HZ
325 rc = sock_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO,
326 (char *)&tv, sizeof(tv));
329 CERROR("Can't set socket send timeout "
331 (long)tv.tv_sec, (int)tv.tv_usec, rc);
338 rc = sock_sendmsg (sock, &msg, iov.iov_len);
339 ticks -= jiffies - then;
349 CERROR ("Unexpected zero rc\n");
350 return (-ECONNABORTED);
356 buffer = ((char *)buffer) + rc;
362 EXPORT_SYMBOL(libcfs_sock_write);
365 libcfs_sock_read (struct socket *sock, void *buffer, int nob, int timeout)
368 mm_segment_t oldmm = get_fs();
369 long ticks = timeout * HZ;
381 struct msghdr msg = {
391 /* Set receive timeout to remaining time */
392 tv = (struct timeval) {
393 .tv_sec = ticks / HZ,
394 .tv_usec = ((ticks % HZ) * 1000000) / HZ
397 rc = sock_setsockopt(sock, SOL_SOCKET, SO_RCVTIMEO,
398 (char *)&tv, sizeof(tv));
401 CERROR("Can't set socket recv timeout %ld.%06d: %d\n",
402 (long)tv.tv_sec, (int)tv.tv_usec, rc);
408 rc = sock_recvmsg(sock, &msg, iov.iov_len, 0);
409 ticks -= jiffies - then;
418 buffer = ((char *)buffer) + rc;
429 EXPORT_SYMBOL(libcfs_sock_read);
432 libcfs_sock_create (struct socket **sockp, int *fatal,
433 __u32 local_ip, int local_port)
435 struct sockaddr_in locaddr;
439 mm_segment_t oldmm = get_fs();
441 /* All errors are fatal except bind failure if the port is in use */
444 rc = sock_create (PF_INET, SOCK_STREAM, 0, &sock);
447 CERROR ("Can't create socket: %d\n", rc);
453 rc = sock_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
454 (char *)&option, sizeof (option));
457 CERROR("Can't set SO_REUSEADDR for socket: %d\n", rc);
461 if (local_ip != 0 || local_port != 0) {
462 memset(&locaddr, 0, sizeof(locaddr));
463 locaddr.sin_family = AF_INET;
464 locaddr.sin_port = htons(local_port);
465 locaddr.sin_addr.s_addr = (local_ip == 0) ?
466 INADDR_ANY : htonl(local_ip);
468 rc = sock->ops->bind(sock, (struct sockaddr *)&locaddr,
470 if (rc == -EADDRINUSE) {
471 CDEBUG(D_NET, "Port %d already in use\n", local_port);
476 CERROR("Error trying to bind to port %d: %d\n",
490 libcfs_sock_setbuf (struct socket *sock, int txbufsize, int rxbufsize)
492 mm_segment_t oldmm = get_fs();
496 if (txbufsize != 0) {
499 rc = sock_setsockopt(sock, SOL_SOCKET, SO_SNDBUF,
500 (char *)&option, sizeof (option));
503 CERROR ("Can't set send buffer %d: %d\n",
509 if (rxbufsize != 0) {
512 rc = sock_setsockopt (sock, SOL_SOCKET, SO_RCVBUF,
513 (char *)&option, sizeof (option));
516 CERROR ("Can't set receive buffer %d: %d\n",
525 EXPORT_SYMBOL(libcfs_sock_setbuf);
528 libcfs_sock_getaddr (struct socket *sock, int remote, __u32 *ip, int *port)
530 struct sockaddr_in sin;
531 int len = sizeof (sin);
534 rc = sock->ops->getname (sock, (struct sockaddr *)&sin, &len,
537 CERROR ("Error %d getting sock %s IP/port\n",
538 rc, remote ? "peer" : "local");
543 *ip = ntohl (sin.sin_addr.s_addr);
546 *port = ntohs (sin.sin_port);
551 EXPORT_SYMBOL(libcfs_sock_getaddr);
554 libcfs_sock_getbuf (struct socket *sock, int *txbufsize, int *rxbufsize)
557 if (txbufsize != NULL) {
558 *txbufsize = sock->sk->sk_sndbuf;
561 if (rxbufsize != NULL) {
562 *rxbufsize = sock->sk->sk_rcvbuf;
568 EXPORT_SYMBOL(libcfs_sock_getbuf);
571 libcfs_sock_listen (struct socket **sockp,
572 __u32 local_ip, int local_port, int backlog)
577 rc = libcfs_sock_create(sockp, &fatal, local_ip, local_port);
580 CERROR("Can't create socket: port %d already in use\n",
585 rc = (*sockp)->ops->listen(*sockp, backlog);
589 CERROR("Can't set listen backlog %d: %d\n", backlog, rc);
590 sock_release(*sockp);
594 EXPORT_SYMBOL(libcfs_sock_listen);
597 libcfs_sock_accept (struct socket **newsockp, struct socket *sock)
600 struct socket *newsock;
603 init_waitqueue_entry(&wait, current);
605 /* XXX this should add a ref to sock->ops->owner, if
606 * TCP could be a module */
607 rc = sock_create_lite(PF_PACKET, sock->type, IPPROTO_TCP, &newsock);
609 CERROR("Can't allocate socket\n");
613 newsock->ops = sock->ops;
615 set_current_state(TASK_INTERRUPTIBLE);
616 add_wait_queue(sk_sleep(sock->sk), &wait);
618 rc = sock->ops->accept(sock, newsock, O_NONBLOCK);
620 /* Nothing ready, so wait for activity */
622 rc = sock->ops->accept(sock, newsock, O_NONBLOCK);
625 remove_wait_queue(sk_sleep(sock->sk), &wait);
626 set_current_state(TASK_RUNNING);
635 sock_release(newsock);
639 EXPORT_SYMBOL(libcfs_sock_accept);
642 libcfs_sock_abort_accept (struct socket *sock)
644 wake_up_all(sk_sleep(sock->sk));
647 EXPORT_SYMBOL(libcfs_sock_abort_accept);
650 libcfs_sock_connect (struct socket **sockp, int *fatal,
651 __u32 local_ip, int local_port,
652 __u32 peer_ip, int peer_port)
654 struct sockaddr_in srvaddr;
657 rc = libcfs_sock_create(sockp, fatal, local_ip, local_port);
661 memset (&srvaddr, 0, sizeof (srvaddr));
662 srvaddr.sin_family = AF_INET;
663 srvaddr.sin_port = htons(peer_port);
664 srvaddr.sin_addr.s_addr = htonl(peer_ip);
666 rc = (*sockp)->ops->connect(*sockp,
667 (struct sockaddr *)&srvaddr, sizeof(srvaddr),
672 /* EADDRNOTAVAIL probably means we're already connected to the same
673 * peer/port on the same local port on a differently typed
674 * connection. Let our caller retry with a different local
676 *fatal = !(rc == -EADDRNOTAVAIL);
678 CDEBUG_LIMIT(*fatal ? D_NETERROR : D_NET,
679 "Error %d connecting %u.%u.%u.%u/%d -> %u.%u.%u.%u/%d\n", rc,
680 HIPQUAD(local_ip), local_port, HIPQUAD(peer_ip), peer_port);
682 sock_release(*sockp);
686 EXPORT_SYMBOL(libcfs_sock_connect);
689 libcfs_sock_release (struct socket *sock)
694 EXPORT_SYMBOL(libcfs_sock_release);