icl_proxy.c revision 270888
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: stable/10/sys/dev/iscsi/icl_proxy.c 270888 2014-08-31 20:21:08Z trasz $"); 72 73#include <sys/param.h> 74#include <sys/capability.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#include <linux/types.h> 88#include <rdma/rdma_cm.h> 89 90#include "icl.h" 91 92static int debug = 1; 93 94#define ICL_DEBUG(X, ...) \ 95 if (debug > 1) { \ 96 printf("%s: " X "\n", __func__, ## __VA_ARGS__);\ 97 } while (0) 98 99#define ICL_WARN(X, ...) \ 100 if (debug > 0) { \ 101 printf("WARNING: %s: " X "\n", \ 102 __func__, ## __VA_ARGS__); \ 103 } while (0) 104 105static MALLOC_DEFINE(M_ICL_PROXY, "ICL_PROXY", "iSCSI common layer proxy"); 106 107#ifdef ICL_RDMA 108static int icl_conn_connect_rdma(struct icl_conn *ic, int domain, int socktype, 109 int protocol, struct sockaddr *from_sa, struct sockaddr *to_sa); 110static int icl_listen_add_rdma(struct icl_listen *il, int domain, int socktype, int protocol, 111 struct sockaddr *sa); 112#endif /* ICL_RDMA */ 113 114static int 115icl_conn_connect_tcp(struct icl_conn *ic, int domain, int socktype, 116 int protocol, struct sockaddr *from_sa, struct sockaddr *to_sa) 117{ 118 struct socket *so; 119 int error; 120 int interrupted = 0; 121 122 error = socreate(domain, &so, socktype, protocol, 123 curthread->td_ucred, curthread); 124 if (error != 0) 125 return (error); 126 127 if (from_sa != NULL) { 128 error = sobind(so, from_sa, curthread); 129 if (error != 0) { 130 soclose(so); 131 return (error); 132 } 133 } 134 135 error = soconnect(so, to_sa, curthread); 136 if (error != 0) { 137 soclose(so); 138 return (error); 139 } 140 141 SOCK_LOCK(so); 142 while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) { 143 error = msleep(&so->so_timeo, SOCK_MTX(so), PSOCK | PCATCH, 144 "icl_connect", 0); 145 if (error) { 146 if (error == EINTR || error == ERESTART) 147 interrupted = 1; 148 break; 149 } 150 } 151 if (error == 0) { 152 error = so->so_error; 153 so->so_error = 0; 154 } 155 SOCK_UNLOCK(so); 156 157 if (error != 0) { 158 soclose(so); 159 return (error); 160 } 161 162 error = icl_conn_handoff_sock(ic, so); 163 if (error != 0) 164 soclose(so); 165 166 return (error); 167} 168 169int 170icl_conn_connect(struct icl_conn *ic, bool rdma, int domain, int socktype, 171 int protocol, struct sockaddr *from_sa, struct sockaddr *to_sa) 172{ 173 174 if (rdma) { 175#ifdef ICL_RDMA 176 return (icl_conn_connect_rdma(ic, domain, socktype, protocol, from_sa, to_sa)); 177#else 178 ICL_DEBUG("RDMA not supported"); 179 return (EOPNOTSUPP); 180#endif 181 } 182 183 return (icl_conn_connect_tcp(ic, domain, socktype, protocol, from_sa, to_sa)); 184} 185 186struct icl_listen * 187icl_listen_new(void (*accept_cb)(struct socket *, struct sockaddr *, int)) 188{ 189 struct icl_listen *il; 190 191 il = malloc(sizeof(*il), M_ICL_PROXY, M_ZERO | M_WAITOK); 192 TAILQ_INIT(&il->il_sockets); 193 sx_init(&il->il_lock, "icl_listen"); 194 il->il_accept = accept_cb; 195 196 return (il); 197} 198 199void 200icl_listen_free(struct icl_listen *il) 201{ 202 struct icl_listen_sock *ils; 203 204 sx_xlock(&il->il_lock); 205 while (!TAILQ_EMPTY(&il->il_sockets)) { 206 ils = TAILQ_FIRST(&il->il_sockets); 207 while (ils->ils_running) { 208 ICL_DEBUG("waiting for accept thread to terminate"); 209 sx_xunlock(&il->il_lock); 210 ils->ils_disconnecting = true; 211 wakeup(&ils->ils_socket->so_timeo); 212 pause("icl_unlisten", 1 * hz); 213 sx_xlock(&il->il_lock); 214 } 215 216 TAILQ_REMOVE(&il->il_sockets, ils, ils_next); 217 soclose(ils->ils_socket); 218 free(ils, M_ICL_PROXY); 219 } 220 sx_xunlock(&il->il_lock); 221 222 free(il, M_ICL_PROXY); 223} 224 225/* 226 * XXX: Doing accept in a separate thread in each socket might not be the best way 227 * to do stuff, but it's pretty clean and debuggable - and you probably won't 228 * have hundreds of listening sockets anyway. 229 */ 230static void 231icl_accept_thread(void *arg) 232{ 233 struct icl_listen_sock *ils; 234 struct socket *head, *so; 235 struct sockaddr *sa; 236 int error; 237 238 ils = arg; 239 head = ils->ils_socket; 240 241 ils->ils_running = true; 242 243 for (;;) { 244 ACCEPT_LOCK(); 245 while (TAILQ_EMPTY(&head->so_comp) && head->so_error == 0 && ils->ils_disconnecting == false) { 246 if (head->so_rcv.sb_state & SBS_CANTRCVMORE) { 247 head->so_error = ECONNABORTED; 248 break; 249 } 250 error = msleep(&head->so_timeo, &accept_mtx, PSOCK | PCATCH, 251 "accept", 0); 252 if (error) { 253 ACCEPT_UNLOCK(); 254 ICL_WARN("msleep failed with error %d", error); 255 continue; 256 } 257 if (ils->ils_disconnecting) { 258 ACCEPT_UNLOCK(); 259 ICL_DEBUG("terminating"); 260 ils->ils_running = false; 261 kthread_exit(); 262 return; 263 } 264 } 265 if (head->so_error) { 266 error = head->so_error; 267 head->so_error = 0; 268 ACCEPT_UNLOCK(); 269 ICL_WARN("socket error %d", error); 270 continue; 271 } 272 so = TAILQ_FIRST(&head->so_comp); 273 KASSERT(so != NULL, ("NULL so")); 274 KASSERT(!(so->so_qstate & SQ_INCOMP), ("accept1: so SQ_INCOMP")); 275 KASSERT(so->so_qstate & SQ_COMP, ("accept1: so not SQ_COMP")); 276 277 /* 278 * Before changing the flags on the socket, we have to bump the 279 * reference count. Otherwise, if the protocol calls sofree(), 280 * the socket will be released due to a zero refcount. 281 */ 282 SOCK_LOCK(so); /* soref() and so_state update */ 283 soref(so); /* file descriptor reference */ 284 285 TAILQ_REMOVE(&head->so_comp, so, so_list); 286 head->so_qlen--; 287 so->so_state |= (head->so_state & SS_NBIO); 288 so->so_qstate &= ~SQ_COMP; 289 so->so_head = NULL; 290 291 SOCK_UNLOCK(so); 292 ACCEPT_UNLOCK(); 293 294 sa = NULL; 295 error = soaccept(so, &sa); 296 if (error != 0) { 297 ICL_WARN("soaccept error %d", error); 298 if (sa != NULL) 299 free(sa, M_SONAME); 300 soclose(so); 301 continue; 302 } 303 304 (ils->ils_listen->il_accept)(so, sa, ils->ils_id); 305 } 306} 307 308static int 309icl_listen_add_tcp(struct icl_listen *il, int domain, int socktype, 310 int protocol, struct sockaddr *sa, int portal_id) 311{ 312 struct icl_listen_sock *ils; 313 struct socket *so; 314 struct sockopt sopt; 315 int error, one = 1; 316 317 error = socreate(domain, &so, socktype, protocol, 318 curthread->td_ucred, curthread); 319 if (error != 0) { 320 ICL_WARN("socreate failed with error %d", error); 321 return (error); 322 } 323 324 sopt.sopt_dir = SOPT_SET; 325 sopt.sopt_level = SOL_SOCKET; 326 sopt.sopt_name = SO_REUSEADDR; 327 sopt.sopt_val = &one; 328 sopt.sopt_valsize = sizeof(one); 329 sopt.sopt_td = NULL; 330 error = sosetopt(so, &sopt); 331 if (error != 0) { 332 ICL_WARN("failed to set SO_REUSEADDR with error %d", error); 333 soclose(so); 334 return (error); 335 } 336 337 error = sobind(so, sa, curthread); 338 if (error != 0) { 339 ICL_WARN("sobind failed with error %d", error); 340 soclose(so); 341 return (error); 342 } 343 344 error = solisten(so, -1, curthread); 345 if (error != 0) { 346 ICL_WARN("solisten failed with error %d", error); 347 soclose(so); 348 return (error); 349 } 350 351 ils = malloc(sizeof(*ils), M_ICL_PROXY, M_ZERO | M_WAITOK); 352 ils->ils_listen = il; 353 ils->ils_socket = so; 354 ils->ils_id = portal_id; 355 356 error = kthread_add(icl_accept_thread, ils, NULL, NULL, 0, 0, "iclacc"); 357 if (error != 0) { 358 ICL_WARN("kthread_add failed with error %d", error); 359 soclose(so); 360 free(ils, M_ICL_PROXY); 361 362 return (error); 363 } 364 365 sx_xlock(&il->il_lock); 366 TAILQ_INSERT_TAIL(&il->il_sockets, ils, ils_next); 367 sx_xunlock(&il->il_lock); 368 369 return (0); 370} 371 372int 373icl_listen_add(struct icl_listen *il, bool rdma, int domain, int socktype, 374 int protocol, struct sockaddr *sa, int portal_id) 375{ 376 377 if (rdma) { 378#ifndef ICL_RDMA 379 ICL_DEBUG("RDMA not supported"); 380 return (EOPNOTSUPP); 381#else 382 return (icl_listen_add_rdma(il, domain, socktype, protocol, 383 sa, portal_id)); 384#endif 385 } 386 387 388 return (icl_listen_add_tcp(il, domain, socktype, protocol, sa, 389 portal_id)); 390} 391 392int 393icl_listen_remove(struct icl_listen *il, struct sockaddr *sa) 394{ 395 396 /* 397 * XXX 398 */ 399 400 return (EOPNOTSUPP); 401} 402 403#endif /* ICL_KERNEL_PROXY */ 404