1 // SPDX-License-Identifier: GPL-2.0
5 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 only,
9 * as published by the Free Software Foundation.
11 * This program is distributed in the hope that it will be useful, but
12 * WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * General Public License version 2 for more details (a copy is included
15 * in the LICENSE file that accompanied this code).
17 * You should have received a copy of the GNU General Public License
18 * version 2 along with this program; If not, see
19 * http://www.gnu.org/licenses/gpl-2.0.html
24 * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
25 * Use is subject to license terms.
27 * Copyright (c) 2012, 2015, Intel Corporation.
30 * This file is part of Lustre, http://www.lustre.org/
31 * Lustre is a trademark of Seagate, Inc.
33 #define DEBUG_SUBSYSTEM S_LNET
37 #include <linux/net.h>
38 #include <linux/file.h>
39 #include <linux/pagemap.h>
40 /* For sys_open & sys_close */
41 #include <linux/syscalls.h>
44 #include <linux/libcfs/libcfs.h>
45 #include <linux/lnet/lib-lnet.h>
48 kernel_sock_unlocked_ioctl(struct file *filp, int cmd, unsigned long arg)
50 mm_segment_t oldfs = get_fs();
54 err = filp->f_op->unlocked_ioctl(filp, cmd, arg);
61 lnet_sock_ioctl(int cmd, unsigned long arg)
63 struct file *sock_filp;
67 rc = sock_create(PF_INET, SOCK_STREAM, 0, &sock);
69 CERROR("Can't create socket: %d\n", rc);
73 sock_filp = sock_alloc_file(sock, 0, NULL);
74 if (IS_ERR(sock_filp))
75 return PTR_ERR(sock_filp);
77 rc = kernel_sock_unlocked_ioctl(sock_filp, cmd, arg);
84 lnet_ipif_query(char *name, int *up, __u32 *ip, __u32 *mask)
91 nob = strnlen(name, IFNAMSIZ);
92 if (nob == IFNAMSIZ) {
93 CERROR("Interface name %s too long\n", name);
97 BUILD_BUG_ON(sizeof(ifr.ifr_name) < IFNAMSIZ);
99 if (strlen(name) > sizeof(ifr.ifr_name) - 1)
101 strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_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 if (strlen(name) > sizeof(ifr.ifr_name) - 1)
119 strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
121 ifr.ifr_addr.sa_family = AF_INET;
122 rc = lnet_sock_ioctl(SIOCGIFADDR, (unsigned long)&ifr);
124 CERROR("Can't get IP address for interface %s\n", name);
128 val = ((struct sockaddr_in *)&ifr.ifr_addr)->sin_addr.s_addr;
131 if (strlen(name) > sizeof(ifr.ifr_name) - 1)
133 strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
135 ifr.ifr_addr.sa_family = AF_INET;
136 rc = lnet_sock_ioctl(SIOCGIFNETMASK, (unsigned long)&ifr);
138 CERROR("Can't get netmask for interface %s\n", name);
142 val = ((struct sockaddr_in *)&ifr.ifr_netmask)->sin_addr.s_addr;
147 EXPORT_SYMBOL(lnet_ipif_query);
150 lnet_ipif_enumerate(char ***namesp)
152 /* Allocate and fill in 'names', returning # interfaces/error */
163 nalloc = 16; /* first guess at max interfaces */
166 if (nalloc * sizeof(*ifr) > PAGE_SIZE) {
168 nalloc = PAGE_SIZE / sizeof(*ifr);
169 CWARN("Too many interfaces: only enumerating first %d\n",
173 LIBCFS_ALLOC(ifr, nalloc * sizeof(*ifr));
175 CERROR("ENOMEM enumerating up to %d interfaces\n",
181 ifc.ifc_buf = (char *)ifr;
182 ifc.ifc_len = nalloc * sizeof(*ifr);
184 rc = lnet_sock_ioctl(SIOCGIFCONF, (unsigned long)&ifc);
186 CERROR("Error %d enumerating interfaces\n", rc);
192 nfound = ifc.ifc_len / sizeof(*ifr);
193 LASSERT(nfound <= nalloc);
195 if (nfound < nalloc || toobig)
198 LIBCFS_FREE(ifr, nalloc * sizeof(*ifr));
205 LIBCFS_ALLOC(names, nfound * sizeof(*names));
211 for (i = 0; i < nfound; i++) {
212 nob = strnlen(ifr[i].ifr_name, IFNAMSIZ);
213 if (nob == IFNAMSIZ) {
214 /* no space for terminating NULL */
215 CERROR("interface name %.*s too long (%d max)\n",
216 nob, ifr[i].ifr_name, IFNAMSIZ);
221 LIBCFS_ALLOC(names[i], IFNAMSIZ);
227 memcpy(names[i], ifr[i].ifr_name, nob);
236 lnet_ipif_free_enumeration(names, nfound);
238 LIBCFS_FREE(ifr, nalloc * sizeof(*ifr));
242 EXPORT_SYMBOL(lnet_ipif_enumerate);
245 lnet_ipif_free_enumeration(char **names, int n)
251 for (i = 0; i < n && names[i]; i++)
252 LIBCFS_FREE(names[i], IFNAMSIZ);
254 LIBCFS_FREE(names, n * sizeof(*names));
256 EXPORT_SYMBOL(lnet_ipif_free_enumeration);
259 lnet_sock_write(struct socket *sock, void *buffer, int nob, int timeout)
262 long jiffies_left = timeout * msecs_to_jiffies(MSEC_PER_SEC);
265 struct kvec iov = { .iov_base = buffer, .iov_len = nob };
266 struct msghdr msg = {NULL,};
270 * Caller may pass a zero timeout if she thinks the socket buffer is
271 * empty enough to take the whole message immediately
273 iov_iter_kvec(&msg.msg_iter, WRITE | ITER_KVEC, &iov, 1, nob);
275 msg.msg_flags = !timeout ? MSG_DONTWAIT : 0;
277 /* Set send timeout to remaining time */
278 jiffies_to_timeval(jiffies_left, &tv);
279 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO,
280 (char *)&tv, sizeof(tv));
282 CERROR("Can't set socket send timeout %ld.%06d: %d\n",
283 (long)tv.tv_sec, (int)tv.tv_usec, rc);
289 rc = kernel_sendmsg(sock, &msg, &iov, 1, nob);
290 jiffies_left -= jiffies - then;
296 CERROR("Unexpected zero rc\n");
297 return -ECONNABORTED;
300 if (!msg_data_left(&msg))
303 if (jiffies_left <= 0)
308 EXPORT_SYMBOL(lnet_sock_write);
311 lnet_sock_read(struct socket *sock, void *buffer, int nob, int timeout)
314 long jiffies_left = timeout * msecs_to_jiffies(MSEC_PER_SEC);
319 LASSERT(jiffies_left > 0);
326 struct msghdr msg = {
330 /* Set receive timeout to remaining time */
331 jiffies_to_timeval(jiffies_left, &tv);
332 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVTIMEO,
333 (char *)&tv, sizeof(tv));
335 CERROR("Can't set socket recv timeout %ld.%06d: %d\n",
336 (long)tv.tv_sec, (int)tv.tv_usec, rc);
341 rc = kernel_recvmsg(sock, &msg, &iov, 1, nob, 0);
342 jiffies_left -= jiffies - then;
350 buffer = ((char *)buffer) + rc;
356 if (jiffies_left <= 0)
360 EXPORT_SYMBOL(lnet_sock_read);
363 lnet_sock_create(struct socket **sockp, int *fatal, __u32 local_ip,
366 struct sockaddr_in locaddr;
371 /* All errors are fatal except bind failure if the port is in use */
374 rc = sock_create(PF_INET, SOCK_STREAM, 0, &sock);
377 CERROR("Can't create socket: %d\n", rc);
382 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
383 (char *)&option, sizeof(option));
385 CERROR("Can't set SO_REUSEADDR for socket: %d\n", rc);
389 if (local_ip || local_port) {
390 memset(&locaddr, 0, sizeof(locaddr));
391 locaddr.sin_family = AF_INET;
392 locaddr.sin_port = htons(local_port);
394 locaddr.sin_addr.s_addr = htonl(INADDR_ANY);
396 locaddr.sin_addr.s_addr = htonl(local_ip);
398 rc = kernel_bind(sock, (struct sockaddr *)&locaddr,
400 if (rc == -EADDRINUSE) {
401 CDEBUG(D_NET, "Port %d already in use\n", local_port);
406 CERROR("Error trying to bind to port %d: %d\n",
419 lnet_sock_setbuf(struct socket *sock, int txbufsize, int rxbufsize)
426 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_SNDBUF,
427 (char *)&option, sizeof(option));
429 CERROR("Can't set send buffer %d: %d\n",
437 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVBUF,
438 (char *)&option, sizeof(option));
440 CERROR("Can't set receive buffer %d: %d\n",
447 EXPORT_SYMBOL(lnet_sock_setbuf);
450 lnet_sock_getaddr(struct socket *sock, bool remote, __u32 *ip, int *port)
452 struct sockaddr_in sin;
453 int len = sizeof(sin);
457 rc = kernel_getpeername(sock, (struct sockaddr *)&sin, &len);
459 rc = kernel_getsockname(sock, (struct sockaddr *)&sin, &len);
461 CERROR("Error %d getting sock %s IP/port\n",
462 rc, remote ? "peer" : "local");
467 *ip = ntohl(sin.sin_addr.s_addr);
470 *port = ntohs(sin.sin_port);
474 EXPORT_SYMBOL(lnet_sock_getaddr);
477 lnet_sock_getbuf(struct socket *sock, int *txbufsize, int *rxbufsize)
480 *txbufsize = sock->sk->sk_sndbuf;
483 *rxbufsize = sock->sk->sk_rcvbuf;
487 EXPORT_SYMBOL(lnet_sock_getbuf);
490 lnet_sock_listen(struct socket **sockp, __u32 local_ip, int local_port,
496 rc = lnet_sock_create(sockp, &fatal, local_ip, local_port);
499 CERROR("Can't create socket: port %d already in use\n",
504 rc = kernel_listen(*sockp, backlog);
508 CERROR("Can't set listen backlog %d: %d\n", backlog, rc);
509 sock_release(*sockp);
514 lnet_sock_accept(struct socket **newsockp, struct socket *sock)
516 wait_queue_entry_t wait;
517 struct socket *newsock;
521 * XXX this should add a ref to sock->ops->owner, if
522 * TCP could be a module
524 rc = sock_create_lite(PF_PACKET, sock->type, IPPROTO_TCP, &newsock);
526 CERROR("Can't allocate socket\n");
530 newsock->ops = sock->ops;
532 rc = sock->ops->accept(sock, newsock, O_NONBLOCK, false);
534 /* Nothing ready, so wait for activity */
535 init_waitqueue_entry(&wait, current);
536 add_wait_queue(sk_sleep(sock->sk), &wait);
537 set_current_state(TASK_INTERRUPTIBLE);
539 remove_wait_queue(sk_sleep(sock->sk), &wait);
540 rc = sock->ops->accept(sock, newsock, O_NONBLOCK, false);
550 sock_release(newsock);
555 lnet_sock_connect(struct socket **sockp, int *fatal, __u32 local_ip,
556 int local_port, __u32 peer_ip, int peer_port)
558 struct sockaddr_in srvaddr;
561 rc = lnet_sock_create(sockp, fatal, local_ip, local_port);
565 memset(&srvaddr, 0, sizeof(srvaddr));
566 srvaddr.sin_family = AF_INET;
567 srvaddr.sin_port = htons(peer_port);
568 srvaddr.sin_addr.s_addr = htonl(peer_ip);
570 rc = kernel_connect(*sockp, (struct sockaddr *)&srvaddr,
576 * EADDRNOTAVAIL probably means we're already connected to the same
577 * peer/port on the same local port on a differently typed
578 * connection. Let our caller retry with a different local
581 *fatal = !(rc == -EADDRNOTAVAIL);
583 CDEBUG_LIMIT(*fatal ? D_NETERROR : D_NET,
584 "Error %d connecting %pI4h/%d -> %pI4h/%d\n", rc,
585 &local_ip, local_port, &peer_ip, peer_port);
587 sock_release(*sockp);