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.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2012, 2015, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Seagate, Inc.
32 #define DEBUG_SUBSYSTEM S_LNET
36 #include <linux/net.h>
37 #include <linux/file.h>
38 #include <linux/pagemap.h>
39 /* For sys_open & sys_close */
40 #include <linux/syscalls.h>
43 #include "../../include/linux/libcfs/libcfs.h"
44 #include "../../include/linux/lnet/lib-lnet.h"
47 kernel_sock_unlocked_ioctl(struct file *filp, int cmd, unsigned long arg)
49 mm_segment_t oldfs = get_fs();
53 err = filp->f_op->unlocked_ioctl(filp, cmd, arg);
60 lnet_sock_ioctl(int cmd, unsigned long arg)
62 struct file *sock_filp;
66 rc = sock_create(PF_INET, SOCK_STREAM, 0, &sock);
68 CERROR("Can't create socket: %d\n", rc);
72 sock_filp = sock_alloc_file(sock, 0, NULL);
73 if (IS_ERR(sock_filp)) {
75 rc = PTR_ERR(sock_filp);
79 rc = kernel_sock_unlocked_ioctl(sock_filp, cmd, arg);
87 lnet_ipif_query(char *name, int *up, __u32 *ip, __u32 *mask)
94 nob = strnlen(name, IFNAMSIZ);
95 if (nob == IFNAMSIZ) {
96 CERROR("Interface name %s too long\n", name);
100 CLASSERT(sizeof(ifr.ifr_name) >= IFNAMSIZ);
102 strcpy(ifr.ifr_name, name);
103 rc = lnet_sock_ioctl(SIOCGIFFLAGS, (unsigned long)&ifr);
105 CERROR("Can't get flags for interface %s\n", name);
109 if (!(ifr.ifr_flags & IFF_UP)) {
110 CDEBUG(D_NET, "Interface %s down\n", name);
117 strcpy(ifr.ifr_name, name);
118 ifr.ifr_addr.sa_family = AF_INET;
119 rc = lnet_sock_ioctl(SIOCGIFADDR, (unsigned long)&ifr);
121 CERROR("Can't get IP address for interface %s\n", name);
125 val = ((struct sockaddr_in *)&ifr.ifr_addr)->sin_addr.s_addr;
128 strcpy(ifr.ifr_name, name);
129 ifr.ifr_addr.sa_family = AF_INET;
130 rc = lnet_sock_ioctl(SIOCGIFNETMASK, (unsigned long)&ifr);
132 CERROR("Can't get netmask for interface %s\n", name);
136 val = ((struct sockaddr_in *)&ifr.ifr_netmask)->sin_addr.s_addr;
141 EXPORT_SYMBOL(lnet_ipif_query);
144 lnet_ipif_enumerate(char ***namesp)
146 /* Allocate and fill in 'names', returning # interfaces/error */
157 nalloc = 16; /* first guess at max interfaces */
160 if (nalloc * sizeof(*ifr) > PAGE_CACHE_SIZE) {
162 nalloc = PAGE_CACHE_SIZE / sizeof(*ifr);
163 CWARN("Too many interfaces: only enumerating first %d\n",
167 LIBCFS_ALLOC(ifr, nalloc * sizeof(*ifr));
169 CERROR("ENOMEM enumerating up to %d interfaces\n",
175 ifc.ifc_buf = (char *)ifr;
176 ifc.ifc_len = nalloc * sizeof(*ifr);
178 rc = lnet_sock_ioctl(SIOCGIFCONF, (unsigned long)&ifc);
180 CERROR("Error %d enumerating interfaces\n", rc);
186 nfound = ifc.ifc_len / sizeof(*ifr);
187 LASSERT(nfound <= nalloc);
189 if (nfound < nalloc || toobig)
192 LIBCFS_FREE(ifr, nalloc * sizeof(*ifr));
199 LIBCFS_ALLOC(names, nfound * sizeof(*names));
205 for (i = 0; i < nfound; i++) {
206 nob = strnlen(ifr[i].ifr_name, IFNAMSIZ);
207 if (nob == IFNAMSIZ) {
208 /* no space for terminating NULL */
209 CERROR("interface name %.*s too long (%d max)\n",
210 nob, ifr[i].ifr_name, IFNAMSIZ);
215 LIBCFS_ALLOC(names[i], IFNAMSIZ);
221 memcpy(names[i], ifr[i].ifr_name, nob);
230 lnet_ipif_free_enumeration(names, nfound);
232 LIBCFS_FREE(ifr, nalloc * sizeof(*ifr));
236 EXPORT_SYMBOL(lnet_ipif_enumerate);
239 lnet_ipif_free_enumeration(char **names, int n)
245 for (i = 0; i < n && names[i]; i++)
246 LIBCFS_FREE(names[i], IFNAMSIZ);
248 LIBCFS_FREE(names, n * sizeof(*names));
250 EXPORT_SYMBOL(lnet_ipif_free_enumeration);
253 lnet_sock_write(struct socket *sock, void *buffer, int nob, int timeout)
256 long ticks = timeout * HZ;
262 * Caller may pass a zero timeout if she thinks the socket buffer is
263 * empty enough to take the whole message immediately
270 struct msghdr msg = {
271 .msg_flags = !timeout ? MSG_DONTWAIT : 0
275 /* Set send timeout to remaining time */
276 tv = (struct timeval) {
277 .tv_sec = ticks / HZ,
278 .tv_usec = ((ticks % HZ) * 1000000) / HZ
280 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO,
281 (char *)&tv, sizeof(tv));
283 CERROR("Can't set socket send timeout %ld.%06d: %d\n",
284 (long)tv.tv_sec, (int)tv.tv_usec, rc);
290 rc = kernel_sendmsg(sock, &msg, &iov, 1, nob);
291 ticks -= jiffies - then;
300 CERROR("Unexpected zero rc\n");
301 return -ECONNABORTED;
307 buffer = ((char *)buffer) + rc;
312 EXPORT_SYMBOL(lnet_sock_write);
315 lnet_sock_read(struct socket *sock, void *buffer, int nob, int timeout)
318 long ticks = timeout * HZ;
330 struct msghdr msg = {
334 /* Set receive timeout to remaining time */
335 tv = (struct timeval) {
336 .tv_sec = ticks / HZ,
337 .tv_usec = ((ticks % HZ) * 1000000) / HZ
339 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVTIMEO,
340 (char *)&tv, sizeof(tv));
342 CERROR("Can't set socket recv timeout %ld.%06d: %d\n",
343 (long)tv.tv_sec, (int)tv.tv_usec, rc);
348 rc = kernel_recvmsg(sock, &msg, &iov, 1, nob, 0);
349 ticks -= jiffies - then;
357 buffer = ((char *)buffer) + rc;
367 EXPORT_SYMBOL(lnet_sock_read);
370 lnet_sock_create(struct socket **sockp, int *fatal, __u32 local_ip,
373 struct sockaddr_in locaddr;
378 /* All errors are fatal except bind failure if the port is in use */
381 rc = sock_create(PF_INET, SOCK_STREAM, 0, &sock);
384 CERROR("Can't create socket: %d\n", rc);
389 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
390 (char *)&option, sizeof(option));
392 CERROR("Can't set SO_REUSEADDR for socket: %d\n", rc);
396 if (local_ip || local_port) {
397 memset(&locaddr, 0, sizeof(locaddr));
398 locaddr.sin_family = AF_INET;
399 locaddr.sin_port = htons(local_port);
400 locaddr.sin_addr.s_addr = !local_ip ?
401 INADDR_ANY : htonl(local_ip);
403 rc = kernel_bind(sock, (struct sockaddr *)&locaddr,
405 if (rc == -EADDRINUSE) {
406 CDEBUG(D_NET, "Port %d already in use\n", local_port);
411 CERROR("Error trying to bind to port %d: %d\n",
424 lnet_sock_setbuf(struct socket *sock, int txbufsize, int rxbufsize)
431 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_SNDBUF,
432 (char *)&option, sizeof(option));
434 CERROR("Can't set send buffer %d: %d\n",
442 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVBUF,
443 (char *)&option, sizeof(option));
445 CERROR("Can't set receive buffer %d: %d\n",
452 EXPORT_SYMBOL(lnet_sock_setbuf);
455 lnet_sock_getaddr(struct socket *sock, bool remote, __u32 *ip, int *port)
457 struct sockaddr_in sin;
458 int len = sizeof(sin);
462 rc = kernel_getpeername(sock, (struct sockaddr *)&sin, &len);
464 rc = kernel_getsockname(sock, (struct sockaddr *)&sin, &len);
466 CERROR("Error %d getting sock %s IP/port\n",
467 rc, remote ? "peer" : "local");
472 *ip = ntohl(sin.sin_addr.s_addr);
475 *port = ntohs(sin.sin_port);
479 EXPORT_SYMBOL(lnet_sock_getaddr);
482 lnet_sock_getbuf(struct socket *sock, int *txbufsize, int *rxbufsize)
485 *txbufsize = sock->sk->sk_sndbuf;
488 *rxbufsize = sock->sk->sk_rcvbuf;
492 EXPORT_SYMBOL(lnet_sock_getbuf);
495 lnet_sock_listen(struct socket **sockp, __u32 local_ip, int local_port,
501 rc = lnet_sock_create(sockp, &fatal, local_ip, local_port);
504 CERROR("Can't create socket: port %d already in use\n",
509 rc = kernel_listen(*sockp, backlog);
513 CERROR("Can't set listen backlog %d: %d\n", backlog, rc);
514 sock_release(*sockp);
517 EXPORT_SYMBOL(lnet_sock_listen);
520 lnet_sock_accept(struct socket **newsockp, struct socket *sock)
523 struct socket *newsock;
526 init_waitqueue_entry(&wait, current);
529 * XXX this should add a ref to sock->ops->owner, if
530 * TCP could be a module
532 rc = sock_create_lite(PF_PACKET, sock->type, IPPROTO_TCP, &newsock);
534 CERROR("Can't allocate socket\n");
538 newsock->ops = sock->ops;
540 rc = sock->ops->accept(sock, newsock, O_NONBLOCK);
542 /* Nothing ready, so wait for activity */
543 set_current_state(TASK_INTERRUPTIBLE);
544 add_wait_queue(sk_sleep(sock->sk), &wait);
546 remove_wait_queue(sk_sleep(sock->sk), &wait);
547 set_current_state(TASK_RUNNING);
548 rc = sock->ops->accept(sock, newsock, O_NONBLOCK);
558 sock_release(newsock);
561 EXPORT_SYMBOL(lnet_sock_accept);
564 lnet_sock_connect(struct socket **sockp, int *fatal, __u32 local_ip,
565 int local_port, __u32 peer_ip, int peer_port)
567 struct sockaddr_in srvaddr;
570 rc = lnet_sock_create(sockp, fatal, local_ip, local_port);
574 memset(&srvaddr, 0, sizeof(srvaddr));
575 srvaddr.sin_family = AF_INET;
576 srvaddr.sin_port = htons(peer_port);
577 srvaddr.sin_addr.s_addr = htonl(peer_ip);
579 rc = kernel_connect(*sockp, (struct sockaddr *)&srvaddr,
585 * EADDRNOTAVAIL probably means we're already connected to the same
586 * peer/port on the same local port on a differently typed
587 * connection. Let our caller retry with a different local
590 *fatal = !(rc == -EADDRNOTAVAIL);
592 CDEBUG_LIMIT(*fatal ? D_NETERROR : D_NET,
593 "Error %d connecting %pI4h/%d -> %pI4h/%d\n", rc,
594 &local_ip, local_port, &peer_ip, peer_port);
596 sock_release(*sockp);
599 EXPORT_SYMBOL(lnet_sock_connect);