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 BUILD_BUG_ON(sizeof(ifr.ifr_name) < IFNAMSIZ);
102 if (strlen(name) > sizeof(ifr.ifr_name) - 1)
104 strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
106 rc = lnet_sock_ioctl(SIOCGIFFLAGS, (unsigned long)&ifr);
108 CERROR("Can't get flags for interface %s\n", name);
112 if (!(ifr.ifr_flags & IFF_UP)) {
113 CDEBUG(D_NET, "Interface %s down\n", name);
120 if (strlen(name) > sizeof(ifr.ifr_name) - 1)
122 strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
124 ifr.ifr_addr.sa_family = AF_INET;
125 rc = lnet_sock_ioctl(SIOCGIFADDR, (unsigned long)&ifr);
127 CERROR("Can't get IP address for interface %s\n", name);
131 val = ((struct sockaddr_in *)&ifr.ifr_addr)->sin_addr.s_addr;
134 if (strlen(name) > sizeof(ifr.ifr_name) - 1)
136 strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
138 ifr.ifr_addr.sa_family = AF_INET;
139 rc = lnet_sock_ioctl(SIOCGIFNETMASK, (unsigned long)&ifr);
141 CERROR("Can't get netmask for interface %s\n", name);
145 val = ((struct sockaddr_in *)&ifr.ifr_netmask)->sin_addr.s_addr;
150 EXPORT_SYMBOL(lnet_ipif_query);
153 lnet_ipif_enumerate(char ***namesp)
155 /* Allocate and fill in 'names', returning # interfaces/error */
166 nalloc = 16; /* first guess at max interfaces */
169 if (nalloc * sizeof(*ifr) > PAGE_SIZE) {
171 nalloc = PAGE_SIZE / sizeof(*ifr);
172 CWARN("Too many interfaces: only enumerating first %d\n",
176 LIBCFS_ALLOC(ifr, nalloc * sizeof(*ifr));
178 CERROR("ENOMEM enumerating up to %d interfaces\n",
184 ifc.ifc_buf = (char *)ifr;
185 ifc.ifc_len = nalloc * sizeof(*ifr);
187 rc = lnet_sock_ioctl(SIOCGIFCONF, (unsigned long)&ifc);
189 CERROR("Error %d enumerating interfaces\n", rc);
195 nfound = ifc.ifc_len / sizeof(*ifr);
196 LASSERT(nfound <= nalloc);
198 if (nfound < nalloc || toobig)
201 LIBCFS_FREE(ifr, nalloc * sizeof(*ifr));
208 LIBCFS_ALLOC(names, nfound * sizeof(*names));
214 for (i = 0; i < nfound; i++) {
215 nob = strnlen(ifr[i].ifr_name, IFNAMSIZ);
216 if (nob == IFNAMSIZ) {
217 /* no space for terminating NULL */
218 CERROR("interface name %.*s too long (%d max)\n",
219 nob, ifr[i].ifr_name, IFNAMSIZ);
224 LIBCFS_ALLOC(names[i], IFNAMSIZ);
230 memcpy(names[i], ifr[i].ifr_name, nob);
239 lnet_ipif_free_enumeration(names, nfound);
241 LIBCFS_FREE(ifr, nalloc * sizeof(*ifr));
245 EXPORT_SYMBOL(lnet_ipif_enumerate);
248 lnet_ipif_free_enumeration(char **names, int n)
254 for (i = 0; i < n && names[i]; i++)
255 LIBCFS_FREE(names[i], IFNAMSIZ);
257 LIBCFS_FREE(names, n * sizeof(*names));
259 EXPORT_SYMBOL(lnet_ipif_free_enumeration);
262 lnet_sock_write(struct socket *sock, void *buffer, int nob, int timeout)
265 long jiffies_left = timeout * msecs_to_jiffies(MSEC_PER_SEC);
268 struct kvec iov = { .iov_base = buffer, .iov_len = nob };
269 struct msghdr msg = {NULL,};
273 * Caller may pass a zero timeout if she thinks the socket buffer is
274 * empty enough to take the whole message immediately
276 iov_iter_kvec(&msg.msg_iter, WRITE | ITER_KVEC, &iov, 1, nob);
278 msg.msg_flags = !timeout ? MSG_DONTWAIT : 0;
280 /* Set send timeout to remaining time */
281 jiffies_to_timeval(jiffies_left, &tv);
282 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO,
283 (char *)&tv, sizeof(tv));
285 CERROR("Can't set socket send timeout %ld.%06d: %d\n",
286 (long)tv.tv_sec, (int)tv.tv_usec, rc);
292 rc = kernel_sendmsg(sock, &msg, &iov, 1, nob);
293 jiffies_left -= jiffies - then;
299 CERROR("Unexpected zero rc\n");
300 return -ECONNABORTED;
303 if (!msg_data_left(&msg))
306 if (jiffies_left <= 0)
311 EXPORT_SYMBOL(lnet_sock_write);
314 lnet_sock_read(struct socket *sock, void *buffer, int nob, int timeout)
317 long jiffies_left = timeout * msecs_to_jiffies(MSEC_PER_SEC);
322 LASSERT(jiffies_left > 0);
329 struct msghdr msg = {
333 /* Set receive timeout to remaining time */
334 jiffies_to_timeval(jiffies_left, &tv);
335 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVTIMEO,
336 (char *)&tv, sizeof(tv));
338 CERROR("Can't set socket recv timeout %ld.%06d: %d\n",
339 (long)tv.tv_sec, (int)tv.tv_usec, rc);
344 rc = kernel_recvmsg(sock, &msg, &iov, 1, nob, 0);
345 jiffies_left -= jiffies - then;
353 buffer = ((char *)buffer) + rc;
359 if (jiffies_left <= 0)
363 EXPORT_SYMBOL(lnet_sock_read);
366 lnet_sock_create(struct socket **sockp, int *fatal, __u32 local_ip,
369 struct sockaddr_in locaddr;
374 /* All errors are fatal except bind failure if the port is in use */
377 rc = sock_create(PF_INET, SOCK_STREAM, 0, &sock);
380 CERROR("Can't create socket: %d\n", rc);
385 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
386 (char *)&option, sizeof(option));
388 CERROR("Can't set SO_REUSEADDR for socket: %d\n", rc);
392 if (local_ip || local_port) {
393 memset(&locaddr, 0, sizeof(locaddr));
394 locaddr.sin_family = AF_INET;
395 locaddr.sin_port = htons(local_port);
397 locaddr.sin_addr.s_addr = htonl(INADDR_ANY);
399 locaddr.sin_addr.s_addr = htonl(local_ip);
401 rc = kernel_bind(sock, (struct sockaddr *)&locaddr,
403 if (rc == -EADDRINUSE) {
404 CDEBUG(D_NET, "Port %d already in use\n", local_port);
409 CERROR("Error trying to bind to port %d: %d\n",
422 lnet_sock_setbuf(struct socket *sock, int txbufsize, int rxbufsize)
429 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_SNDBUF,
430 (char *)&option, sizeof(option));
432 CERROR("Can't set send buffer %d: %d\n",
440 rc = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVBUF,
441 (char *)&option, sizeof(option));
443 CERROR("Can't set receive buffer %d: %d\n",
450 EXPORT_SYMBOL(lnet_sock_setbuf);
453 lnet_sock_getaddr(struct socket *sock, bool remote, __u32 *ip, int *port)
455 struct sockaddr_in sin;
456 int len = sizeof(sin);
460 rc = kernel_getpeername(sock, (struct sockaddr *)&sin, &len);
462 rc = kernel_getsockname(sock, (struct sockaddr *)&sin, &len);
464 CERROR("Error %d getting sock %s IP/port\n",
465 rc, remote ? "peer" : "local");
470 *ip = ntohl(sin.sin_addr.s_addr);
473 *port = ntohs(sin.sin_port);
477 EXPORT_SYMBOL(lnet_sock_getaddr);
480 lnet_sock_getbuf(struct socket *sock, int *txbufsize, int *rxbufsize)
483 *txbufsize = sock->sk->sk_sndbuf;
486 *rxbufsize = sock->sk->sk_rcvbuf;
490 EXPORT_SYMBOL(lnet_sock_getbuf);
493 lnet_sock_listen(struct socket **sockp, __u32 local_ip, int local_port,
499 rc = lnet_sock_create(sockp, &fatal, local_ip, local_port);
502 CERROR("Can't create socket: port %d already in use\n",
507 rc = kernel_listen(*sockp, backlog);
511 CERROR("Can't set listen backlog %d: %d\n", backlog, rc);
512 sock_release(*sockp);
517 lnet_sock_accept(struct socket **newsockp, struct socket *sock)
520 struct socket *newsock;
524 * XXX this should add a ref to sock->ops->owner, if
525 * TCP could be a module
527 rc = sock_create_lite(PF_PACKET, sock->type, IPPROTO_TCP, &newsock);
529 CERROR("Can't allocate socket\n");
533 newsock->ops = sock->ops;
535 rc = sock->ops->accept(sock, newsock, O_NONBLOCK);
537 /* Nothing ready, so wait for activity */
538 init_waitqueue_entry(&wait, current);
539 add_wait_queue(sk_sleep(sock->sk), &wait);
540 set_current_state(TASK_INTERRUPTIBLE);
542 remove_wait_queue(sk_sleep(sock->sk), &wait);
543 rc = sock->ops->accept(sock, newsock, O_NONBLOCK);
553 sock_release(newsock);
558 lnet_sock_connect(struct socket **sockp, int *fatal, __u32 local_ip,
559 int local_port, __u32 peer_ip, int peer_port)
561 struct sockaddr_in srvaddr;
564 rc = lnet_sock_create(sockp, fatal, local_ip, local_port);
568 memset(&srvaddr, 0, sizeof(srvaddr));
569 srvaddr.sin_family = AF_INET;
570 srvaddr.sin_port = htons(peer_port);
571 srvaddr.sin_addr.s_addr = htonl(peer_ip);
573 rc = kernel_connect(*sockp, (struct sockaddr *)&srvaddr,
579 * EADDRNOTAVAIL probably means we're already connected to the same
580 * peer/port on the same local port on a differently typed
581 * connection. Let our caller retry with a different local
584 *fatal = !(rc == -EADDRNOTAVAIL);
586 CDEBUG_LIMIT(*fatal ? D_NETERROR : D_NET,
587 "Error %d connecting %pI4h/%d -> %pI4h/%d\n", rc,
588 &local_ip, local_port, &peer_ip, peer_port);
590 sock_release(*sockp);