xref: /freebsd/sys/dev/iscsi/icl_soft_proxy.c (revision 74d9553e43cfafc29448d0bb836916aa21dea0de)
1 /*-
2  * Copyright (c) 2012 The FreeBSD Foundation
3  * All rights reserved.
4  *
5  * This software was developed by Edward Tomasz Napierala under sponsorship
6  * from the FreeBSD Foundation.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  */
30 /*-
31  * Copyright (c) 1982, 1986, 1989, 1990, 1993
32  *	The Regents of the University of California.  All rights reserved.
33  *
34  * sendfile(2) and related extensions:
35  * Copyright (c) 1998, David Greenman. All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 4. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)uipc_syscalls.c	8.4 (Berkeley) 2/21/94
62  */
63 
64 /*
65  * iSCSI Common Layer, kernel proxy part.
66  */
67 
68 #ifdef ICL_KERNEL_PROXY
69 
70 #include <sys/cdefs.h>
71 __FBSDID("$FreeBSD$");
72 
73 #include <sys/param.h>
74 #include <sys/capsicum.h>
75 #include <sys/condvar.h>
76 #include <sys/conf.h>
77 #include <sys/kernel.h>
78 #include <sys/kthread.h>
79 #include <sys/malloc.h>
80 #include <sys/proc.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/sx.h>
84 #include <sys/systm.h>
85 #include <netinet/in.h>
86 #include <netinet/tcp.h>
87 
88 #include <dev/iscsi/icl.h>
89 
90 struct icl_listen_sock {
91 	TAILQ_ENTRY(icl_listen_sock)	ils_next;
92 	struct icl_listen		*ils_listen;
93 	struct socket			*ils_socket;
94 	bool				ils_running;
95 	bool				ils_disconnecting;
96 	int				ils_id;
97 };
98 
99 struct icl_listen	{
100 	TAILQ_HEAD(, icl_listen_sock)	il_sockets;
101 	struct sx			il_lock;
102 	void				(*il_accept)(struct socket *,
103 					    struct sockaddr *, int);
104 };
105 
106 static MALLOC_DEFINE(M_ICL_PROXY, "ICL_PROXY", "iSCSI common layer proxy");
107 
108 int
109 icl_soft_proxy_connect(struct icl_conn *ic, int domain, int socktype,
110     int protocol, struct sockaddr *from_sa, struct sockaddr *to_sa)
111 {
112 	struct socket *so;
113 	int error;
114 	int interrupted = 0;
115 
116 	error = socreate(domain, &so, socktype, protocol,
117 	    curthread->td_ucred, curthread);
118 	if (error != 0)
119 		return (error);
120 
121 	if (from_sa != NULL) {
122 		error = sobind(so, from_sa, curthread);
123 		if (error != 0) {
124 			soclose(so);
125 			return (error);
126 		}
127 	}
128 
129 	error = soconnect(so, to_sa, curthread);
130 	if (error != 0) {
131 		soclose(so);
132 		return (error);
133 	}
134 
135 	SOCK_LOCK(so);
136 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
137 		error = msleep(&so->so_timeo, SOCK_MTX(so), PSOCK | PCATCH,
138 		    "icl_connect", 0);
139 		if (error) {
140 			if (error == EINTR || error == ERESTART)
141 				interrupted = 1;
142 			break;
143 		}
144 	}
145 	if (error == 0) {
146 		error = so->so_error;
147 		so->so_error = 0;
148 	}
149 	SOCK_UNLOCK(so);
150 
151 	if (error != 0) {
152 		soclose(so);
153 		return (error);
154 	}
155 
156 	error = icl_soft_handoff_sock(ic, so);
157 	if (error != 0)
158 		soclose(so);
159 
160 	return (error);
161 }
162 
163 struct icl_listen *
164 icl_listen_new(void (*accept_cb)(struct socket *, struct sockaddr *, int))
165 {
166 	struct icl_listen *il;
167 
168 	il = malloc(sizeof(*il), M_ICL_PROXY, M_ZERO | M_WAITOK);
169 	TAILQ_INIT(&il->il_sockets);
170 	sx_init(&il->il_lock, "icl_listen");
171 	il->il_accept = accept_cb;
172 
173 	return (il);
174 }
175 
176 void
177 icl_listen_free(struct icl_listen *il)
178 {
179 	struct icl_listen_sock *ils;
180 
181 	sx_xlock(&il->il_lock);
182 	while (!TAILQ_EMPTY(&il->il_sockets)) {
183 		ils = TAILQ_FIRST(&il->il_sockets);
184 		while (ils->ils_running) {
185 			ICL_DEBUG("waiting for accept thread to terminate");
186 			sx_xunlock(&il->il_lock);
187 			ils->ils_disconnecting = true;
188 			wakeup(&ils->ils_socket->so_timeo);
189 			pause("icl_unlisten", 1 * hz);
190 			sx_xlock(&il->il_lock);
191 		}
192 
193 		TAILQ_REMOVE(&il->il_sockets, ils, ils_next);
194 		soclose(ils->ils_socket);
195 		free(ils, M_ICL_PROXY);
196 	}
197 	sx_xunlock(&il->il_lock);
198 
199 	free(il, M_ICL_PROXY);
200 }
201 
202 /*
203  * XXX: Doing accept in a separate thread in each socket might not be the best way
204  * 	to do stuff, but it's pretty clean and debuggable - and you probably won't
205  * 	have hundreds of listening sockets anyway.
206  */
207 static void
208 icl_accept_thread(void *arg)
209 {
210 	struct icl_listen_sock *ils;
211 	struct socket *head, *so;
212 	struct sockaddr *sa;
213 	int error;
214 
215 	ils = arg;
216 	head = ils->ils_socket;
217 
218 	ils->ils_running = true;
219 
220 	for (;;) {
221 		ACCEPT_LOCK();
222 		while (TAILQ_EMPTY(&head->so_comp) && head->so_error == 0 && ils->ils_disconnecting == false) {
223 			if (head->so_rcv.sb_state & SBS_CANTRCVMORE) {
224 				head->so_error = ECONNABORTED;
225 				break;
226 			}
227 			error = msleep(&head->so_timeo, &accept_mtx, PSOCK | PCATCH,
228 			    "accept", 0);
229 			if (error) {
230 				ACCEPT_UNLOCK();
231 				ICL_WARN("msleep failed with error %d", error);
232 				continue;
233 			}
234 			if (ils->ils_disconnecting) {
235 				ACCEPT_UNLOCK();
236 				ICL_DEBUG("terminating");
237 				ils->ils_running = false;
238 				kthread_exit();
239 				return;
240 			}
241 		}
242 		if (head->so_error) {
243 			error = head->so_error;
244 			head->so_error = 0;
245 			ACCEPT_UNLOCK();
246 			ICL_WARN("socket error %d", error);
247 			continue;
248 		}
249 		so = TAILQ_FIRST(&head->so_comp);
250 		KASSERT(so != NULL, ("NULL so"));
251 		KASSERT(!(so->so_qstate & SQ_INCOMP), ("accept1: so SQ_INCOMP"));
252 		KASSERT(so->so_qstate & SQ_COMP, ("accept1: so not SQ_COMP"));
253 
254 		/*
255 		 * Before changing the flags on the socket, we have to bump the
256 		 * reference count.  Otherwise, if the protocol calls sofree(),
257 		 * the socket will be released due to a zero refcount.
258 		 */
259 		SOCK_LOCK(so);			/* soref() and so_state update */
260 		soref(so);			/* file descriptor reference */
261 
262 		TAILQ_REMOVE(&head->so_comp, so, so_list);
263 		head->so_qlen--;
264 		so->so_state |= (head->so_state & SS_NBIO);
265 		so->so_qstate &= ~SQ_COMP;
266 		so->so_head = NULL;
267 
268 		SOCK_UNLOCK(so);
269 		ACCEPT_UNLOCK();
270 
271 		sa = NULL;
272 		error = soaccept(so, &sa);
273 		if (error != 0) {
274 			ICL_WARN("soaccept error %d", error);
275 			if (sa != NULL)
276 				free(sa, M_SONAME);
277 			soclose(so);
278 			continue;
279 		}
280 
281 		(ils->ils_listen->il_accept)(so, sa, ils->ils_id);
282 	}
283 }
284 
285 static int
286 icl_listen_add_tcp(struct icl_listen *il, int domain, int socktype,
287     int protocol, struct sockaddr *sa, int portal_id)
288 {
289 	struct icl_listen_sock *ils;
290 	struct socket *so;
291 	struct sockopt sopt;
292 	int error, one = 1;
293 
294 	error = socreate(domain, &so, socktype, protocol,
295 	    curthread->td_ucred, curthread);
296 	if (error != 0) {
297 		ICL_WARN("socreate failed with error %d", error);
298 		return (error);
299 	}
300 
301 	sopt.sopt_dir = SOPT_SET;
302 	sopt.sopt_level = SOL_SOCKET;
303 	sopt.sopt_name = SO_REUSEADDR;
304 	sopt.sopt_val = &one;
305 	sopt.sopt_valsize = sizeof(one);
306 	sopt.sopt_td = NULL;
307 	error = sosetopt(so, &sopt);
308 	if (error != 0) {
309 		ICL_WARN("failed to set SO_REUSEADDR with error %d", error);
310 		soclose(so);
311 		return (error);
312 	}
313 
314 	error = sobind(so, sa, curthread);
315 	if (error != 0) {
316 		ICL_WARN("sobind failed with error %d", error);
317 		soclose(so);
318 		return (error);
319 	}
320 
321 	error = solisten(so, -1, curthread);
322 	if (error != 0) {
323 		ICL_WARN("solisten failed with error %d", error);
324 		soclose(so);
325 		return (error);
326 	}
327 
328 	ils = malloc(sizeof(*ils), M_ICL_PROXY, M_ZERO | M_WAITOK);
329 	ils->ils_listen = il;
330 	ils->ils_socket = so;
331 	ils->ils_id = portal_id;
332 
333 	error = kthread_add(icl_accept_thread, ils, NULL, NULL, 0, 0, "iclacc");
334 	if (error != 0) {
335 		ICL_WARN("kthread_add failed with error %d", error);
336 		soclose(so);
337 		free(ils, M_ICL_PROXY);
338 
339 		return (error);
340 	}
341 
342 	sx_xlock(&il->il_lock);
343 	TAILQ_INSERT_TAIL(&il->il_sockets, ils, ils_next);
344 	sx_xunlock(&il->il_lock);
345 
346 	return (0);
347 }
348 
349 int
350 icl_listen_add(struct icl_listen *il, bool rdma, int domain, int socktype,
351     int protocol, struct sockaddr *sa, int portal_id)
352 {
353 
354 	if (rdma) {
355 		ICL_DEBUG("RDMA not supported");
356 		return (EOPNOTSUPP);
357 	}
358 
359 
360 	return (icl_listen_add_tcp(il, domain, socktype, protocol, sa,
361 	    portal_id));
362 }
363 
364 int
365 icl_listen_remove(struct icl_listen *il, struct sockaddr *sa)
366 {
367 
368 	/*
369 	 * XXX
370 	 */
371 
372 	return (EOPNOTSUPP);
373 }
374 
375 #endif /* ICL_KERNEL_PROXY */
376