1 /* 2 * linux/net/sunrpc/xprtsock.c 3 * 4 * Client-side transport implementation for sockets. 5 * 6 * TCP callback races fixes (C) 1998 Red Hat 7 * TCP send fixes (C) 1998 Red Hat 8 * TCP NFS related read + write fixes 9 * (C) 1999 Dave Airlie, University of Limerick, Ireland <airlied@linux.ie> 10 * 11 * Rewrite of larges part of the code in order to stabilize TCP stuff. 12 * Fix behaviour when socket buffer is full. 13 * (C) 1999 Trond Myklebust <trond.myklebust@fys.uio.no> 14 * 15 * IP socket transport implementation, (C) 2005 Chuck Lever <cel@netapp.com> 16 * 17 * IPv6 support contributed by Gilles Quillard, Bull Open Source, 2005. 18 * <gilles.quillard@bull.net> 19 */ 20 21 #include <linux/types.h> 22 #include <linux/slab.h> 23 #include <linux/module.h> 24 #include <linux/capability.h> 25 #include <linux/pagemap.h> 26 #include <linux/errno.h> 27 #include <linux/socket.h> 28 #include <linux/in.h> 29 #include <linux/net.h> 30 #include <linux/mm.h> 31 #include <linux/udp.h> 32 #include <linux/tcp.h> 33 #include <linux/sunrpc/clnt.h> 34 #include <linux/sunrpc/sched.h> 35 #include <linux/sunrpc/xprtsock.h> 36 #include <linux/file.h> 37 38 #include <net/sock.h> 39 #include <net/checksum.h> 40 #include <net/udp.h> 41 #include <net/tcp.h> 42 43 /* 44 * xprtsock tunables 45 */ 46 unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE; 47 unsigned int xprt_tcp_slot_table_entries = RPC_DEF_SLOT_TABLE; 48 49 unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT; 50 unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT; 51 52 #define XS_TCP_LINGER_TO (15U * HZ) 53 static unsigned int xs_tcp_fin_timeout __read_mostly = XS_TCP_LINGER_TO; 54 55 /* 56 * We can register our own files under /proc/sys/sunrpc by 57 * calling register_sysctl_table() again. The files in that 58 * directory become the union of all files registered there. 59 * 60 * We simply need to make sure that we don't collide with 61 * someone else's file names! 62 */ 63 64 #ifdef RPC_DEBUG 65 66 static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE; 67 static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE; 68 static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT; 69 static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT; 70 71 static struct ctl_table_header *sunrpc_table_header; 72 73 /* 74 * FIXME: changing the UDP slot table size should also resize the UDP 75 * socket buffers for existing UDP transports 76 */ 77 static ctl_table xs_tunables_table[] = { 78 { 79 .ctl_name = CTL_SLOTTABLE_UDP, 80 .procname = "udp_slot_table_entries", 81 .data = &xprt_udp_slot_table_entries, 82 .maxlen = sizeof(unsigned int), 83 .mode = 0644, 84 .proc_handler = &proc_dointvec_minmax, 85 .strategy = &sysctl_intvec, 86 .extra1 = &min_slot_table_size, 87 .extra2 = &max_slot_table_size 88 }, 89 { 90 .ctl_name = CTL_SLOTTABLE_TCP, 91 .procname = "tcp_slot_table_entries", 92 .data = &xprt_tcp_slot_table_entries, 93 .maxlen = sizeof(unsigned int), 94 .mode = 0644, 95 .proc_handler = &proc_dointvec_minmax, 96 .strategy = &sysctl_intvec, 97 .extra1 = &min_slot_table_size, 98 .extra2 = &max_slot_table_size 99 }, 100 { 101 .ctl_name = CTL_MIN_RESVPORT, 102 .procname = "min_resvport", 103 .data = &xprt_min_resvport, 104 .maxlen = sizeof(unsigned int), 105 .mode = 0644, 106 .proc_handler = &proc_dointvec_minmax, 107 .strategy = &sysctl_intvec, 108 .extra1 = &xprt_min_resvport_limit, 109 .extra2 = &xprt_max_resvport_limit 110 }, 111 { 112 .ctl_name = CTL_MAX_RESVPORT, 113 .procname = "max_resvport", 114 .data = &xprt_max_resvport, 115 .maxlen = sizeof(unsigned int), 116 .mode = 0644, 117 .proc_handler = &proc_dointvec_minmax, 118 .strategy = &sysctl_intvec, 119 .extra1 = &xprt_min_resvport_limit, 120 .extra2 = &xprt_max_resvport_limit 121 }, 122 { 123 .procname = "tcp_fin_timeout", 124 .data = &xs_tcp_fin_timeout, 125 .maxlen = sizeof(xs_tcp_fin_timeout), 126 .mode = 0644, 127 .proc_handler = &proc_dointvec_jiffies, 128 .strategy = sysctl_jiffies 129 }, 130 { 131 .ctl_name = 0, 132 }, 133 }; 134 135 static ctl_table sunrpc_table[] = { 136 { 137 .ctl_name = CTL_SUNRPC, 138 .procname = "sunrpc", 139 .mode = 0555, 140 .child = xs_tunables_table 141 }, 142 { 143 .ctl_name = 0, 144 }, 145 }; 146 147 #endif 148 149 /* 150 * Time out for an RPC UDP socket connect. UDP socket connects are 151 * synchronous, but we set a timeout anyway in case of resource 152 * exhaustion on the local host. 153 */ 154 #define XS_UDP_CONN_TO (5U * HZ) 155 156 /* 157 * Wait duration for an RPC TCP connection to be established. Solaris 158 * NFS over TCP uses 60 seconds, for example, which is in line with how 159 * long a server takes to reboot. 160 */ 161 #define XS_TCP_CONN_TO (60U * HZ) 162 163 /* 164 * Wait duration for a reply from the RPC portmapper. 165 */ 166 #define XS_BIND_TO (60U * HZ) 167 168 /* 169 * Delay if a UDP socket connect error occurs. This is most likely some 170 * kind of resource problem on the local host. 171 */ 172 #define XS_UDP_REEST_TO (2U * HZ) 173 174 /* 175 * The reestablish timeout allows clients to delay for a bit before attempting 176 * to reconnect to a server that just dropped our connection. 177 * 178 * We implement an exponential backoff when trying to reestablish a TCP 179 * transport connection with the server. Some servers like to drop a TCP 180 * connection when they are overworked, so we start with a short timeout and 181 * increase over time if the server is down or not responding. 182 */ 183 #define XS_TCP_INIT_REEST_TO (3U * HZ) 184 #define XS_TCP_MAX_REEST_TO (5U * 60 * HZ) 185 186 /* 187 * TCP idle timeout; client drops the transport socket if it is idle 188 * for this long. Note that we also timeout UDP sockets to prevent 189 * holding port numbers when there is no RPC traffic. 190 */ 191 #define XS_IDLE_DISC_TO (5U * 60 * HZ) 192 193 #ifdef RPC_DEBUG 194 # undef RPC_DEBUG_DATA 195 # define RPCDBG_FACILITY RPCDBG_TRANS 196 #endif 197 198 #ifdef RPC_DEBUG_DATA 199 static void xs_pktdump(char *msg, u32 *packet, unsigned int count) 200 { 201 u8 *buf = (u8 *) packet; 202 int j; 203 204 dprintk("RPC: %s\n", msg); 205 for (j = 0; j < count && j < 128; j += 4) { 206 if (!(j & 31)) { 207 if (j) 208 dprintk("\n"); 209 dprintk("0x%04x ", j); 210 } 211 dprintk("%02x%02x%02x%02x ", 212 buf[j], buf[j+1], buf[j+2], buf[j+3]); 213 } 214 dprintk("\n"); 215 } 216 #else 217 static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count) 218 { 219 /* NOP */ 220 } 221 #endif 222 223 struct sock_xprt { 224 struct rpc_xprt xprt; 225 226 /* 227 * Network layer 228 */ 229 struct socket * sock; 230 struct sock * inet; 231 232 /* 233 * State of TCP reply receive 234 */ 235 __be32 tcp_fraghdr, 236 tcp_xid; 237 238 u32 tcp_offset, 239 tcp_reclen; 240 241 unsigned long tcp_copied, 242 tcp_flags; 243 244 /* 245 * Connection of transports 246 */ 247 struct delayed_work connect_worker; 248 struct sockaddr_storage addr; 249 unsigned short port; 250 251 /* 252 * UDP socket buffer size parameters 253 */ 254 size_t rcvsize, 255 sndsize; 256 257 /* 258 * Saved socket callback addresses 259 */ 260 void (*old_data_ready)(struct sock *, int); 261 void (*old_state_change)(struct sock *); 262 void (*old_write_space)(struct sock *); 263 void (*old_error_report)(struct sock *); 264 }; 265 266 /* 267 * TCP receive state flags 268 */ 269 #define TCP_RCV_LAST_FRAG (1UL << 0) 270 #define TCP_RCV_COPY_FRAGHDR (1UL << 1) 271 #define TCP_RCV_COPY_XID (1UL << 2) 272 #define TCP_RCV_COPY_DATA (1UL << 3) 273 274 static inline struct sockaddr *xs_addr(struct rpc_xprt *xprt) 275 { 276 return (struct sockaddr *) &xprt->addr; 277 } 278 279 static inline struct sockaddr_in *xs_addr_in(struct rpc_xprt *xprt) 280 { 281 return (struct sockaddr_in *) &xprt->addr; 282 } 283 284 static inline struct sockaddr_in6 *xs_addr_in6(struct rpc_xprt *xprt) 285 { 286 return (struct sockaddr_in6 *) &xprt->addr; 287 } 288 289 static void xs_format_ipv4_peer_addresses(struct rpc_xprt *xprt, 290 const char *protocol, 291 const char *netid) 292 { 293 struct sockaddr_in *addr = xs_addr_in(xprt); 294 char *buf; 295 296 buf = kzalloc(20, GFP_KERNEL); 297 if (buf) { 298 snprintf(buf, 20, "%pI4", &addr->sin_addr.s_addr); 299 } 300 xprt->address_strings[RPC_DISPLAY_ADDR] = buf; 301 302 buf = kzalloc(8, GFP_KERNEL); 303 if (buf) { 304 snprintf(buf, 8, "%u", 305 ntohs(addr->sin_port)); 306 } 307 xprt->address_strings[RPC_DISPLAY_PORT] = buf; 308 309 xprt->address_strings[RPC_DISPLAY_PROTO] = protocol; 310 311 buf = kzalloc(48, GFP_KERNEL); 312 if (buf) { 313 snprintf(buf, 48, "addr=%pI4 port=%u proto=%s", 314 &addr->sin_addr.s_addr, 315 ntohs(addr->sin_port), 316 protocol); 317 } 318 xprt->address_strings[RPC_DISPLAY_ALL] = buf; 319 320 buf = kzalloc(10, GFP_KERNEL); 321 if (buf) { 322 snprintf(buf, 10, "%02x%02x%02x%02x", 323 NIPQUAD(addr->sin_addr.s_addr)); 324 } 325 xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = buf; 326 327 buf = kzalloc(8, GFP_KERNEL); 328 if (buf) { 329 snprintf(buf, 8, "%4hx", 330 ntohs(addr->sin_port)); 331 } 332 xprt->address_strings[RPC_DISPLAY_HEX_PORT] = buf; 333 334 buf = kzalloc(30, GFP_KERNEL); 335 if (buf) { 336 snprintf(buf, 30, "%pI4.%u.%u", 337 &addr->sin_addr.s_addr, 338 ntohs(addr->sin_port) >> 8, 339 ntohs(addr->sin_port) & 0xff); 340 } 341 xprt->address_strings[RPC_DISPLAY_UNIVERSAL_ADDR] = buf; 342 343 xprt->address_strings[RPC_DISPLAY_NETID] = netid; 344 } 345 346 static void xs_format_ipv6_peer_addresses(struct rpc_xprt *xprt, 347 const char *protocol, 348 const char *netid) 349 { 350 struct sockaddr_in6 *addr = xs_addr_in6(xprt); 351 char *buf; 352 353 buf = kzalloc(40, GFP_KERNEL); 354 if (buf) { 355 snprintf(buf, 40, "%pI6",&addr->sin6_addr); 356 } 357 xprt->address_strings[RPC_DISPLAY_ADDR] = buf; 358 359 buf = kzalloc(8, GFP_KERNEL); 360 if (buf) { 361 snprintf(buf, 8, "%u", 362 ntohs(addr->sin6_port)); 363 } 364 xprt->address_strings[RPC_DISPLAY_PORT] = buf; 365 366 xprt->address_strings[RPC_DISPLAY_PROTO] = protocol; 367 368 buf = kzalloc(64, GFP_KERNEL); 369 if (buf) { 370 snprintf(buf, 64, "addr=%pI6 port=%u proto=%s", 371 &addr->sin6_addr, 372 ntohs(addr->sin6_port), 373 protocol); 374 } 375 xprt->address_strings[RPC_DISPLAY_ALL] = buf; 376 377 buf = kzalloc(36, GFP_KERNEL); 378 if (buf) 379 snprintf(buf, 36, "%pi6", &addr->sin6_addr); 380 381 xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = buf; 382 383 buf = kzalloc(8, GFP_KERNEL); 384 if (buf) { 385 snprintf(buf, 8, "%4hx", 386 ntohs(addr->sin6_port)); 387 } 388 xprt->address_strings[RPC_DISPLAY_HEX_PORT] = buf; 389 390 buf = kzalloc(50, GFP_KERNEL); 391 if (buf) { 392 snprintf(buf, 50, "%pI6.%u.%u", 393 &addr->sin6_addr, 394 ntohs(addr->sin6_port) >> 8, 395 ntohs(addr->sin6_port) & 0xff); 396 } 397 xprt->address_strings[RPC_DISPLAY_UNIVERSAL_ADDR] = buf; 398 399 xprt->address_strings[RPC_DISPLAY_NETID] = netid; 400 } 401 402 static void xs_free_peer_addresses(struct rpc_xprt *xprt) 403 { 404 unsigned int i; 405 406 for (i = 0; i < RPC_DISPLAY_MAX; i++) 407 switch (i) { 408 case RPC_DISPLAY_PROTO: 409 case RPC_DISPLAY_NETID: 410 continue; 411 default: 412 kfree(xprt->address_strings[i]); 413 } 414 } 415 416 #define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL) 417 418 static int xs_send_kvec(struct socket *sock, struct sockaddr *addr, int addrlen, struct kvec *vec, unsigned int base, int more) 419 { 420 struct msghdr msg = { 421 .msg_name = addr, 422 .msg_namelen = addrlen, 423 .msg_flags = XS_SENDMSG_FLAGS | (more ? MSG_MORE : 0), 424 }; 425 struct kvec iov = { 426 .iov_base = vec->iov_base + base, 427 .iov_len = vec->iov_len - base, 428 }; 429 430 if (iov.iov_len != 0) 431 return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len); 432 return kernel_sendmsg(sock, &msg, NULL, 0, 0); 433 } 434 435 static int xs_send_pagedata(struct socket *sock, struct xdr_buf *xdr, unsigned int base, int more) 436 { 437 struct page **ppage; 438 unsigned int remainder; 439 int err, sent = 0; 440 441 remainder = xdr->page_len - base; 442 base += xdr->page_base; 443 ppage = xdr->pages + (base >> PAGE_SHIFT); 444 base &= ~PAGE_MASK; 445 for(;;) { 446 unsigned int len = min_t(unsigned int, PAGE_SIZE - base, remainder); 447 int flags = XS_SENDMSG_FLAGS; 448 449 remainder -= len; 450 if (remainder != 0 || more) 451 flags |= MSG_MORE; 452 err = sock->ops->sendpage(sock, *ppage, base, len, flags); 453 if (remainder == 0 || err != len) 454 break; 455 sent += err; 456 ppage++; 457 base = 0; 458 } 459 if (sent == 0) 460 return err; 461 if (err > 0) 462 sent += err; 463 return sent; 464 } 465 466 /** 467 * xs_sendpages - write pages directly to a socket 468 * @sock: socket to send on 469 * @addr: UDP only -- address of destination 470 * @addrlen: UDP only -- length of destination address 471 * @xdr: buffer containing this request 472 * @base: starting position in the buffer 473 * 474 */ 475 static int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base) 476 { 477 unsigned int remainder = xdr->len - base; 478 int err, sent = 0; 479 480 if (unlikely(!sock)) 481 return -ENOTSOCK; 482 483 clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags); 484 if (base != 0) { 485 addr = NULL; 486 addrlen = 0; 487 } 488 489 if (base < xdr->head[0].iov_len || addr != NULL) { 490 unsigned int len = xdr->head[0].iov_len - base; 491 remainder -= len; 492 err = xs_send_kvec(sock, addr, addrlen, &xdr->head[0], base, remainder != 0); 493 if (remainder == 0 || err != len) 494 goto out; 495 sent += err; 496 base = 0; 497 } else 498 base -= xdr->head[0].iov_len; 499 500 if (base < xdr->page_len) { 501 unsigned int len = xdr->page_len - base; 502 remainder -= len; 503 err = xs_send_pagedata(sock, xdr, base, remainder != 0); 504 if (remainder == 0 || err != len) 505 goto out; 506 sent += err; 507 base = 0; 508 } else 509 base -= xdr->page_len; 510 511 if (base >= xdr->tail[0].iov_len) 512 return sent; 513 err = xs_send_kvec(sock, NULL, 0, &xdr->tail[0], base, 0); 514 out: 515 if (sent == 0) 516 return err; 517 if (err > 0) 518 sent += err; 519 return sent; 520 } 521 522 static void xs_nospace_callback(struct rpc_task *task) 523 { 524 struct sock_xprt *transport = container_of(task->tk_rqstp->rq_xprt, struct sock_xprt, xprt); 525 526 transport->inet->sk_write_pending--; 527 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); 528 } 529 530 /** 531 * xs_nospace - place task on wait queue if transmit was incomplete 532 * @task: task to put to sleep 533 * 534 */ 535 static int xs_nospace(struct rpc_task *task) 536 { 537 struct rpc_rqst *req = task->tk_rqstp; 538 struct rpc_xprt *xprt = req->rq_xprt; 539 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 540 int ret = 0; 541 542 dprintk("RPC: %5u xmit incomplete (%u left of %u)\n", 543 task->tk_pid, req->rq_slen - req->rq_bytes_sent, 544 req->rq_slen); 545 546 /* Protect against races with write_space */ 547 spin_lock_bh(&xprt->transport_lock); 548 549 /* Don't race with disconnect */ 550 if (xprt_connected(xprt)) { 551 if (test_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags)) { 552 ret = -EAGAIN; 553 /* 554 * Notify TCP that we're limited by the application 555 * window size 556 */ 557 set_bit(SOCK_NOSPACE, &transport->sock->flags); 558 transport->inet->sk_write_pending++; 559 /* ...and wait for more buffer space */ 560 xprt_wait_for_buffer_space(task, xs_nospace_callback); 561 } 562 } else { 563 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); 564 ret = -ENOTCONN; 565 } 566 567 spin_unlock_bh(&xprt->transport_lock); 568 return ret; 569 } 570 571 /** 572 * xs_udp_send_request - write an RPC request to a UDP socket 573 * @task: address of RPC task that manages the state of an RPC request 574 * 575 * Return values: 576 * 0: The request has been sent 577 * EAGAIN: The socket was blocked, please call again later to 578 * complete the request 579 * ENOTCONN: Caller needs to invoke connect logic then call again 580 * other: Some other error occured, the request was not sent 581 */ 582 static int xs_udp_send_request(struct rpc_task *task) 583 { 584 struct rpc_rqst *req = task->tk_rqstp; 585 struct rpc_xprt *xprt = req->rq_xprt; 586 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 587 struct xdr_buf *xdr = &req->rq_snd_buf; 588 int status; 589 590 xs_pktdump("packet data:", 591 req->rq_svec->iov_base, 592 req->rq_svec->iov_len); 593 594 if (!xprt_bound(xprt)) 595 return -ENOTCONN; 596 status = xs_sendpages(transport->sock, 597 xs_addr(xprt), 598 xprt->addrlen, xdr, 599 req->rq_bytes_sent); 600 601 dprintk("RPC: xs_udp_send_request(%u) = %d\n", 602 xdr->len - req->rq_bytes_sent, status); 603 604 if (status >= 0) { 605 task->tk_bytes_sent += status; 606 if (status >= req->rq_slen) 607 return 0; 608 /* Still some bytes left; set up for a retry later. */ 609 status = -EAGAIN; 610 } 611 if (!transport->sock) 612 goto out; 613 614 switch (status) { 615 case -ENOTSOCK: 616 status = -ENOTCONN; 617 /* Should we call xs_close() here? */ 618 break; 619 case -EAGAIN: 620 status = xs_nospace(task); 621 break; 622 default: 623 dprintk("RPC: sendmsg returned unrecognized error %d\n", 624 -status); 625 case -ENETUNREACH: 626 case -EPIPE: 627 case -ECONNREFUSED: 628 /* When the server has died, an ICMP port unreachable message 629 * prompts ECONNREFUSED. */ 630 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); 631 } 632 out: 633 return status; 634 } 635 636 /** 637 * xs_tcp_shutdown - gracefully shut down a TCP socket 638 * @xprt: transport 639 * 640 * Initiates a graceful shutdown of the TCP socket by calling the 641 * equivalent of shutdown(SHUT_WR); 642 */ 643 static void xs_tcp_shutdown(struct rpc_xprt *xprt) 644 { 645 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 646 struct socket *sock = transport->sock; 647 648 if (sock != NULL) 649 kernel_sock_shutdown(sock, SHUT_WR); 650 } 651 652 static inline void xs_encode_tcp_record_marker(struct xdr_buf *buf) 653 { 654 u32 reclen = buf->len - sizeof(rpc_fraghdr); 655 rpc_fraghdr *base = buf->head[0].iov_base; 656 *base = htonl(RPC_LAST_STREAM_FRAGMENT | reclen); 657 } 658 659 /** 660 * xs_tcp_send_request - write an RPC request to a TCP socket 661 * @task: address of RPC task that manages the state of an RPC request 662 * 663 * Return values: 664 * 0: The request has been sent 665 * EAGAIN: The socket was blocked, please call again later to 666 * complete the request 667 * ENOTCONN: Caller needs to invoke connect logic then call again 668 * other: Some other error occured, the request was not sent 669 * 670 * XXX: In the case of soft timeouts, should we eventually give up 671 * if sendmsg is not able to make progress? 672 */ 673 static int xs_tcp_send_request(struct rpc_task *task) 674 { 675 struct rpc_rqst *req = task->tk_rqstp; 676 struct rpc_xprt *xprt = req->rq_xprt; 677 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 678 struct xdr_buf *xdr = &req->rq_snd_buf; 679 int status; 680 681 xs_encode_tcp_record_marker(&req->rq_snd_buf); 682 683 xs_pktdump("packet data:", 684 req->rq_svec->iov_base, 685 req->rq_svec->iov_len); 686 687 /* Continue transmitting the packet/record. We must be careful 688 * to cope with writespace callbacks arriving _after_ we have 689 * called sendmsg(). */ 690 while (1) { 691 status = xs_sendpages(transport->sock, 692 NULL, 0, xdr, req->rq_bytes_sent); 693 694 dprintk("RPC: xs_tcp_send_request(%u) = %d\n", 695 xdr->len - req->rq_bytes_sent, status); 696 697 if (unlikely(status < 0)) 698 break; 699 700 /* If we've sent the entire packet, immediately 701 * reset the count of bytes sent. */ 702 req->rq_bytes_sent += status; 703 task->tk_bytes_sent += status; 704 if (likely(req->rq_bytes_sent >= req->rq_slen)) { 705 req->rq_bytes_sent = 0; 706 return 0; 707 } 708 709 if (status != 0) 710 continue; 711 status = -EAGAIN; 712 break; 713 } 714 if (!transport->sock) 715 goto out; 716 717 switch (status) { 718 case -ENOTSOCK: 719 status = -ENOTCONN; 720 /* Should we call xs_close() here? */ 721 break; 722 case -EAGAIN: 723 status = xs_nospace(task); 724 break; 725 default: 726 dprintk("RPC: sendmsg returned unrecognized error %d\n", 727 -status); 728 case -ECONNRESET: 729 case -EPIPE: 730 xs_tcp_shutdown(xprt); 731 case -ECONNREFUSED: 732 case -ENOTCONN: 733 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); 734 } 735 out: 736 return status; 737 } 738 739 /** 740 * xs_tcp_release_xprt - clean up after a tcp transmission 741 * @xprt: transport 742 * @task: rpc task 743 * 744 * This cleans up if an error causes us to abort the transmission of a request. 745 * In this case, the socket may need to be reset in order to avoid confusing 746 * the server. 747 */ 748 static void xs_tcp_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task) 749 { 750 struct rpc_rqst *req; 751 752 if (task != xprt->snd_task) 753 return; 754 if (task == NULL) 755 goto out_release; 756 req = task->tk_rqstp; 757 if (req->rq_bytes_sent == 0) 758 goto out_release; 759 if (req->rq_bytes_sent == req->rq_snd_buf.len) 760 goto out_release; 761 set_bit(XPRT_CLOSE_WAIT, &task->tk_xprt->state); 762 out_release: 763 xprt_release_xprt(xprt, task); 764 } 765 766 static void xs_save_old_callbacks(struct sock_xprt *transport, struct sock *sk) 767 { 768 transport->old_data_ready = sk->sk_data_ready; 769 transport->old_state_change = sk->sk_state_change; 770 transport->old_write_space = sk->sk_write_space; 771 transport->old_error_report = sk->sk_error_report; 772 } 773 774 static void xs_restore_old_callbacks(struct sock_xprt *transport, struct sock *sk) 775 { 776 sk->sk_data_ready = transport->old_data_ready; 777 sk->sk_state_change = transport->old_state_change; 778 sk->sk_write_space = transport->old_write_space; 779 sk->sk_error_report = transport->old_error_report; 780 } 781 782 static void xs_reset_transport(struct sock_xprt *transport) 783 { 784 struct socket *sock = transport->sock; 785 struct sock *sk = transport->inet; 786 787 if (sk == NULL) 788 return; 789 790 write_lock_bh(&sk->sk_callback_lock); 791 transport->inet = NULL; 792 transport->sock = NULL; 793 794 sk->sk_user_data = NULL; 795 796 xs_restore_old_callbacks(transport, sk); 797 write_unlock_bh(&sk->sk_callback_lock); 798 799 sk->sk_no_check = 0; 800 801 sock_release(sock); 802 } 803 804 /** 805 * xs_close - close a socket 806 * @xprt: transport 807 * 808 * This is used when all requests are complete; ie, no DRC state remains 809 * on the server we want to save. 810 * 811 * The caller _must_ be holding XPRT_LOCKED in order to avoid issues with 812 * xs_reset_transport() zeroing the socket from underneath a writer. 813 */ 814 static void xs_close(struct rpc_xprt *xprt) 815 { 816 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 817 818 dprintk("RPC: xs_close xprt %p\n", xprt); 819 820 xs_reset_transport(transport); 821 822 smp_mb__before_clear_bit(); 823 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state); 824 clear_bit(XPRT_CLOSE_WAIT, &xprt->state); 825 clear_bit(XPRT_CLOSING, &xprt->state); 826 smp_mb__after_clear_bit(); 827 xprt_disconnect_done(xprt); 828 } 829 830 static void xs_tcp_close(struct rpc_xprt *xprt) 831 { 832 if (test_and_clear_bit(XPRT_CONNECTION_CLOSE, &xprt->state)) 833 xs_close(xprt); 834 else 835 xs_tcp_shutdown(xprt); 836 } 837 838 /** 839 * xs_destroy - prepare to shutdown a transport 840 * @xprt: doomed transport 841 * 842 */ 843 static void xs_destroy(struct rpc_xprt *xprt) 844 { 845 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 846 847 dprintk("RPC: xs_destroy xprt %p\n", xprt); 848 849 cancel_rearming_delayed_work(&transport->connect_worker); 850 851 xs_close(xprt); 852 xs_free_peer_addresses(xprt); 853 kfree(xprt->slot); 854 kfree(xprt); 855 module_put(THIS_MODULE); 856 } 857 858 static inline struct rpc_xprt *xprt_from_sock(struct sock *sk) 859 { 860 return (struct rpc_xprt *) sk->sk_user_data; 861 } 862 863 /** 864 * xs_udp_data_ready - "data ready" callback for UDP sockets 865 * @sk: socket with data to read 866 * @len: how much data to read 867 * 868 */ 869 static void xs_udp_data_ready(struct sock *sk, int len) 870 { 871 struct rpc_task *task; 872 struct rpc_xprt *xprt; 873 struct rpc_rqst *rovr; 874 struct sk_buff *skb; 875 int err, repsize, copied; 876 u32 _xid; 877 __be32 *xp; 878 879 read_lock(&sk->sk_callback_lock); 880 dprintk("RPC: xs_udp_data_ready...\n"); 881 if (!(xprt = xprt_from_sock(sk))) 882 goto out; 883 884 if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL) 885 goto out; 886 887 if (xprt->shutdown) 888 goto dropit; 889 890 repsize = skb->len - sizeof(struct udphdr); 891 if (repsize < 4) { 892 dprintk("RPC: impossible RPC reply size %d!\n", repsize); 893 goto dropit; 894 } 895 896 /* Copy the XID from the skb... */ 897 xp = skb_header_pointer(skb, sizeof(struct udphdr), 898 sizeof(_xid), &_xid); 899 if (xp == NULL) 900 goto dropit; 901 902 /* Look up and lock the request corresponding to the given XID */ 903 spin_lock(&xprt->transport_lock); 904 rovr = xprt_lookup_rqst(xprt, *xp); 905 if (!rovr) 906 goto out_unlock; 907 task = rovr->rq_task; 908 909 if ((copied = rovr->rq_private_buf.buflen) > repsize) 910 copied = repsize; 911 912 /* Suck it into the iovec, verify checksum if not done by hw. */ 913 if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) { 914 UDPX_INC_STATS_BH(sk, UDP_MIB_INERRORS); 915 goto out_unlock; 916 } 917 918 UDPX_INC_STATS_BH(sk, UDP_MIB_INDATAGRAMS); 919 920 /* Something worked... */ 921 dst_confirm(skb->dst); 922 923 xprt_adjust_cwnd(task, copied); 924 xprt_update_rtt(task); 925 xprt_complete_rqst(task, copied); 926 927 out_unlock: 928 spin_unlock(&xprt->transport_lock); 929 dropit: 930 skb_free_datagram(sk, skb); 931 out: 932 read_unlock(&sk->sk_callback_lock); 933 } 934 935 static inline void xs_tcp_read_fraghdr(struct rpc_xprt *xprt, struct xdr_skb_reader *desc) 936 { 937 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 938 size_t len, used; 939 char *p; 940 941 p = ((char *) &transport->tcp_fraghdr) + transport->tcp_offset; 942 len = sizeof(transport->tcp_fraghdr) - transport->tcp_offset; 943 used = xdr_skb_read_bits(desc, p, len); 944 transport->tcp_offset += used; 945 if (used != len) 946 return; 947 948 transport->tcp_reclen = ntohl(transport->tcp_fraghdr); 949 if (transport->tcp_reclen & RPC_LAST_STREAM_FRAGMENT) 950 transport->tcp_flags |= TCP_RCV_LAST_FRAG; 951 else 952 transport->tcp_flags &= ~TCP_RCV_LAST_FRAG; 953 transport->tcp_reclen &= RPC_FRAGMENT_SIZE_MASK; 954 955 transport->tcp_flags &= ~TCP_RCV_COPY_FRAGHDR; 956 transport->tcp_offset = 0; 957 958 /* Sanity check of the record length */ 959 if (unlikely(transport->tcp_reclen < 4)) { 960 dprintk("RPC: invalid TCP record fragment length\n"); 961 xprt_force_disconnect(xprt); 962 return; 963 } 964 dprintk("RPC: reading TCP record fragment of length %d\n", 965 transport->tcp_reclen); 966 } 967 968 static void xs_tcp_check_fraghdr(struct sock_xprt *transport) 969 { 970 if (transport->tcp_offset == transport->tcp_reclen) { 971 transport->tcp_flags |= TCP_RCV_COPY_FRAGHDR; 972 transport->tcp_offset = 0; 973 if (transport->tcp_flags & TCP_RCV_LAST_FRAG) { 974 transport->tcp_flags &= ~TCP_RCV_COPY_DATA; 975 transport->tcp_flags |= TCP_RCV_COPY_XID; 976 transport->tcp_copied = 0; 977 } 978 } 979 } 980 981 static inline void xs_tcp_read_xid(struct sock_xprt *transport, struct xdr_skb_reader *desc) 982 { 983 size_t len, used; 984 char *p; 985 986 len = sizeof(transport->tcp_xid) - transport->tcp_offset; 987 dprintk("RPC: reading XID (%Zu bytes)\n", len); 988 p = ((char *) &transport->tcp_xid) + transport->tcp_offset; 989 used = xdr_skb_read_bits(desc, p, len); 990 transport->tcp_offset += used; 991 if (used != len) 992 return; 993 transport->tcp_flags &= ~TCP_RCV_COPY_XID; 994 transport->tcp_flags |= TCP_RCV_COPY_DATA; 995 transport->tcp_copied = 4; 996 dprintk("RPC: reading reply for XID %08x\n", 997 ntohl(transport->tcp_xid)); 998 xs_tcp_check_fraghdr(transport); 999 } 1000 1001 static inline void xs_tcp_read_request(struct rpc_xprt *xprt, struct xdr_skb_reader *desc) 1002 { 1003 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1004 struct rpc_rqst *req; 1005 struct xdr_buf *rcvbuf; 1006 size_t len; 1007 ssize_t r; 1008 1009 /* Find and lock the request corresponding to this xid */ 1010 spin_lock(&xprt->transport_lock); 1011 req = xprt_lookup_rqst(xprt, transport->tcp_xid); 1012 if (!req) { 1013 transport->tcp_flags &= ~TCP_RCV_COPY_DATA; 1014 dprintk("RPC: XID %08x request not found!\n", 1015 ntohl(transport->tcp_xid)); 1016 spin_unlock(&xprt->transport_lock); 1017 return; 1018 } 1019 1020 rcvbuf = &req->rq_private_buf; 1021 len = desc->count; 1022 if (len > transport->tcp_reclen - transport->tcp_offset) { 1023 struct xdr_skb_reader my_desc; 1024 1025 len = transport->tcp_reclen - transport->tcp_offset; 1026 memcpy(&my_desc, desc, sizeof(my_desc)); 1027 my_desc.count = len; 1028 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied, 1029 &my_desc, xdr_skb_read_bits); 1030 desc->count -= r; 1031 desc->offset += r; 1032 } else 1033 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied, 1034 desc, xdr_skb_read_bits); 1035 1036 if (r > 0) { 1037 transport->tcp_copied += r; 1038 transport->tcp_offset += r; 1039 } 1040 if (r != len) { 1041 /* Error when copying to the receive buffer, 1042 * usually because we weren't able to allocate 1043 * additional buffer pages. All we can do now 1044 * is turn off TCP_RCV_COPY_DATA, so the request 1045 * will not receive any additional updates, 1046 * and time out. 1047 * Any remaining data from this record will 1048 * be discarded. 1049 */ 1050 transport->tcp_flags &= ~TCP_RCV_COPY_DATA; 1051 dprintk("RPC: XID %08x truncated request\n", 1052 ntohl(transport->tcp_xid)); 1053 dprintk("RPC: xprt = %p, tcp_copied = %lu, " 1054 "tcp_offset = %u, tcp_reclen = %u\n", 1055 xprt, transport->tcp_copied, 1056 transport->tcp_offset, transport->tcp_reclen); 1057 goto out; 1058 } 1059 1060 dprintk("RPC: XID %08x read %Zd bytes\n", 1061 ntohl(transport->tcp_xid), r); 1062 dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, " 1063 "tcp_reclen = %u\n", xprt, transport->tcp_copied, 1064 transport->tcp_offset, transport->tcp_reclen); 1065 1066 if (transport->tcp_copied == req->rq_private_buf.buflen) 1067 transport->tcp_flags &= ~TCP_RCV_COPY_DATA; 1068 else if (transport->tcp_offset == transport->tcp_reclen) { 1069 if (transport->tcp_flags & TCP_RCV_LAST_FRAG) 1070 transport->tcp_flags &= ~TCP_RCV_COPY_DATA; 1071 } 1072 1073 out: 1074 if (!(transport->tcp_flags & TCP_RCV_COPY_DATA)) 1075 xprt_complete_rqst(req->rq_task, transport->tcp_copied); 1076 spin_unlock(&xprt->transport_lock); 1077 xs_tcp_check_fraghdr(transport); 1078 } 1079 1080 static inline void xs_tcp_read_discard(struct sock_xprt *transport, struct xdr_skb_reader *desc) 1081 { 1082 size_t len; 1083 1084 len = transport->tcp_reclen - transport->tcp_offset; 1085 if (len > desc->count) 1086 len = desc->count; 1087 desc->count -= len; 1088 desc->offset += len; 1089 transport->tcp_offset += len; 1090 dprintk("RPC: discarded %Zu bytes\n", len); 1091 xs_tcp_check_fraghdr(transport); 1092 } 1093 1094 static int xs_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, unsigned int offset, size_t len) 1095 { 1096 struct rpc_xprt *xprt = rd_desc->arg.data; 1097 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1098 struct xdr_skb_reader desc = { 1099 .skb = skb, 1100 .offset = offset, 1101 .count = len, 1102 }; 1103 1104 dprintk("RPC: xs_tcp_data_recv started\n"); 1105 do { 1106 /* Read in a new fragment marker if necessary */ 1107 /* Can we ever really expect to get completely empty fragments? */ 1108 if (transport->tcp_flags & TCP_RCV_COPY_FRAGHDR) { 1109 xs_tcp_read_fraghdr(xprt, &desc); 1110 continue; 1111 } 1112 /* Read in the xid if necessary */ 1113 if (transport->tcp_flags & TCP_RCV_COPY_XID) { 1114 xs_tcp_read_xid(transport, &desc); 1115 continue; 1116 } 1117 /* Read in the request data */ 1118 if (transport->tcp_flags & TCP_RCV_COPY_DATA) { 1119 xs_tcp_read_request(xprt, &desc); 1120 continue; 1121 } 1122 /* Skip over any trailing bytes on short reads */ 1123 xs_tcp_read_discard(transport, &desc); 1124 } while (desc.count); 1125 dprintk("RPC: xs_tcp_data_recv done\n"); 1126 return len - desc.count; 1127 } 1128 1129 /** 1130 * xs_tcp_data_ready - "data ready" callback for TCP sockets 1131 * @sk: socket with data to read 1132 * @bytes: how much data to read 1133 * 1134 */ 1135 static void xs_tcp_data_ready(struct sock *sk, int bytes) 1136 { 1137 struct rpc_xprt *xprt; 1138 read_descriptor_t rd_desc; 1139 int read; 1140 1141 dprintk("RPC: xs_tcp_data_ready...\n"); 1142 1143 read_lock(&sk->sk_callback_lock); 1144 if (!(xprt = xprt_from_sock(sk))) 1145 goto out; 1146 if (xprt->shutdown) 1147 goto out; 1148 1149 /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */ 1150 rd_desc.arg.data = xprt; 1151 do { 1152 rd_desc.count = 65536; 1153 read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv); 1154 } while (read > 0); 1155 out: 1156 read_unlock(&sk->sk_callback_lock); 1157 } 1158 1159 /* 1160 * Do the equivalent of linger/linger2 handling for dealing with 1161 * broken servers that don't close the socket in a timely 1162 * fashion 1163 */ 1164 static void xs_tcp_schedule_linger_timeout(struct rpc_xprt *xprt, 1165 unsigned long timeout) 1166 { 1167 struct sock_xprt *transport; 1168 1169 if (xprt_test_and_set_connecting(xprt)) 1170 return; 1171 set_bit(XPRT_CONNECTION_ABORT, &xprt->state); 1172 transport = container_of(xprt, struct sock_xprt, xprt); 1173 queue_delayed_work(rpciod_workqueue, &transport->connect_worker, 1174 timeout); 1175 } 1176 1177 static void xs_tcp_cancel_linger_timeout(struct rpc_xprt *xprt) 1178 { 1179 struct sock_xprt *transport; 1180 1181 transport = container_of(xprt, struct sock_xprt, xprt); 1182 1183 if (!test_bit(XPRT_CONNECTION_ABORT, &xprt->state) || 1184 !cancel_delayed_work(&transport->connect_worker)) 1185 return; 1186 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state); 1187 xprt_clear_connecting(xprt); 1188 } 1189 1190 static void xs_sock_mark_closed(struct rpc_xprt *xprt) 1191 { 1192 smp_mb__before_clear_bit(); 1193 clear_bit(XPRT_CLOSE_WAIT, &xprt->state); 1194 clear_bit(XPRT_CLOSING, &xprt->state); 1195 smp_mb__after_clear_bit(); 1196 /* Mark transport as closed and wake up all pending tasks */ 1197 xprt_disconnect_done(xprt); 1198 } 1199 1200 /** 1201 * xs_tcp_state_change - callback to handle TCP socket state changes 1202 * @sk: socket whose state has changed 1203 * 1204 */ 1205 static void xs_tcp_state_change(struct sock *sk) 1206 { 1207 struct rpc_xprt *xprt; 1208 1209 read_lock(&sk->sk_callback_lock); 1210 if (!(xprt = xprt_from_sock(sk))) 1211 goto out; 1212 dprintk("RPC: xs_tcp_state_change client %p...\n", xprt); 1213 dprintk("RPC: state %x conn %d dead %d zapped %d\n", 1214 sk->sk_state, xprt_connected(xprt), 1215 sock_flag(sk, SOCK_DEAD), 1216 sock_flag(sk, SOCK_ZAPPED)); 1217 1218 switch (sk->sk_state) { 1219 case TCP_ESTABLISHED: 1220 spin_lock_bh(&xprt->transport_lock); 1221 if (!xprt_test_and_set_connected(xprt)) { 1222 struct sock_xprt *transport = container_of(xprt, 1223 struct sock_xprt, xprt); 1224 1225 /* Reset TCP record info */ 1226 transport->tcp_offset = 0; 1227 transport->tcp_reclen = 0; 1228 transport->tcp_copied = 0; 1229 transport->tcp_flags = 1230 TCP_RCV_COPY_FRAGHDR | TCP_RCV_COPY_XID; 1231 1232 xprt_wake_pending_tasks(xprt, -EAGAIN); 1233 } 1234 spin_unlock_bh(&xprt->transport_lock); 1235 break; 1236 case TCP_FIN_WAIT1: 1237 /* The client initiated a shutdown of the socket */ 1238 xprt->connect_cookie++; 1239 xprt->reestablish_timeout = 0; 1240 set_bit(XPRT_CLOSING, &xprt->state); 1241 smp_mb__before_clear_bit(); 1242 clear_bit(XPRT_CONNECTED, &xprt->state); 1243 clear_bit(XPRT_CLOSE_WAIT, &xprt->state); 1244 smp_mb__after_clear_bit(); 1245 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout); 1246 break; 1247 case TCP_CLOSE_WAIT: 1248 /* The server initiated a shutdown of the socket */ 1249 xprt_force_disconnect(xprt); 1250 case TCP_SYN_SENT: 1251 xprt->connect_cookie++; 1252 case TCP_CLOSING: 1253 /* 1254 * If the server closed down the connection, make sure that 1255 * we back off before reconnecting 1256 */ 1257 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO) 1258 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; 1259 break; 1260 case TCP_LAST_ACK: 1261 set_bit(XPRT_CLOSING, &xprt->state); 1262 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout); 1263 smp_mb__before_clear_bit(); 1264 clear_bit(XPRT_CONNECTED, &xprt->state); 1265 smp_mb__after_clear_bit(); 1266 break; 1267 case TCP_CLOSE: 1268 xs_tcp_cancel_linger_timeout(xprt); 1269 xs_sock_mark_closed(xprt); 1270 } 1271 out: 1272 read_unlock(&sk->sk_callback_lock); 1273 } 1274 1275 /** 1276 * xs_error_report - callback mainly for catching socket errors 1277 * @sk: socket 1278 */ 1279 static void xs_error_report(struct sock *sk) 1280 { 1281 struct rpc_xprt *xprt; 1282 1283 read_lock(&sk->sk_callback_lock); 1284 if (!(xprt = xprt_from_sock(sk))) 1285 goto out; 1286 dprintk("RPC: %s client %p...\n" 1287 "RPC: error %d\n", 1288 __func__, xprt, sk->sk_err); 1289 xprt_wake_pending_tasks(xprt, -EAGAIN); 1290 out: 1291 read_unlock(&sk->sk_callback_lock); 1292 } 1293 1294 static void xs_write_space(struct sock *sk) 1295 { 1296 struct socket *sock; 1297 struct rpc_xprt *xprt; 1298 1299 if (unlikely(!(sock = sk->sk_socket))) 1300 return; 1301 clear_bit(SOCK_NOSPACE, &sock->flags); 1302 1303 if (unlikely(!(xprt = xprt_from_sock(sk)))) 1304 return; 1305 if (test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags) == 0) 1306 return; 1307 1308 xprt_write_space(xprt); 1309 } 1310 1311 /** 1312 * xs_udp_write_space - callback invoked when socket buffer space 1313 * becomes available 1314 * @sk: socket whose state has changed 1315 * 1316 * Called when more output buffer space is available for this socket. 1317 * We try not to wake our writers until they can make "significant" 1318 * progress, otherwise we'll waste resources thrashing kernel_sendmsg 1319 * with a bunch of small requests. 1320 */ 1321 static void xs_udp_write_space(struct sock *sk) 1322 { 1323 read_lock(&sk->sk_callback_lock); 1324 1325 /* from net/core/sock.c:sock_def_write_space */ 1326 if (sock_writeable(sk)) 1327 xs_write_space(sk); 1328 1329 read_unlock(&sk->sk_callback_lock); 1330 } 1331 1332 /** 1333 * xs_tcp_write_space - callback invoked when socket buffer space 1334 * becomes available 1335 * @sk: socket whose state has changed 1336 * 1337 * Called when more output buffer space is available for this socket. 1338 * We try not to wake our writers until they can make "significant" 1339 * progress, otherwise we'll waste resources thrashing kernel_sendmsg 1340 * with a bunch of small requests. 1341 */ 1342 static void xs_tcp_write_space(struct sock *sk) 1343 { 1344 read_lock(&sk->sk_callback_lock); 1345 1346 /* from net/core/stream.c:sk_stream_write_space */ 1347 if (sk_stream_wspace(sk) >= sk_stream_min_wspace(sk)) 1348 xs_write_space(sk); 1349 1350 read_unlock(&sk->sk_callback_lock); 1351 } 1352 1353 static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt) 1354 { 1355 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1356 struct sock *sk = transport->inet; 1357 1358 if (transport->rcvsize) { 1359 sk->sk_userlocks |= SOCK_RCVBUF_LOCK; 1360 sk->sk_rcvbuf = transport->rcvsize * xprt->max_reqs * 2; 1361 } 1362 if (transport->sndsize) { 1363 sk->sk_userlocks |= SOCK_SNDBUF_LOCK; 1364 sk->sk_sndbuf = transport->sndsize * xprt->max_reqs * 2; 1365 sk->sk_write_space(sk); 1366 } 1367 } 1368 1369 /** 1370 * xs_udp_set_buffer_size - set send and receive limits 1371 * @xprt: generic transport 1372 * @sndsize: requested size of send buffer, in bytes 1373 * @rcvsize: requested size of receive buffer, in bytes 1374 * 1375 * Set socket send and receive buffer size limits. 1376 */ 1377 static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize) 1378 { 1379 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1380 1381 transport->sndsize = 0; 1382 if (sndsize) 1383 transport->sndsize = sndsize + 1024; 1384 transport->rcvsize = 0; 1385 if (rcvsize) 1386 transport->rcvsize = rcvsize + 1024; 1387 1388 xs_udp_do_set_buffer_size(xprt); 1389 } 1390 1391 /** 1392 * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport 1393 * @task: task that timed out 1394 * 1395 * Adjust the congestion window after a retransmit timeout has occurred. 1396 */ 1397 static void xs_udp_timer(struct rpc_task *task) 1398 { 1399 xprt_adjust_cwnd(task, -ETIMEDOUT); 1400 } 1401 1402 static unsigned short xs_get_random_port(void) 1403 { 1404 unsigned short range = xprt_max_resvport - xprt_min_resvport; 1405 unsigned short rand = (unsigned short) net_random() % range; 1406 return rand + xprt_min_resvport; 1407 } 1408 1409 /** 1410 * xs_set_port - reset the port number in the remote endpoint address 1411 * @xprt: generic transport 1412 * @port: new port number 1413 * 1414 */ 1415 static void xs_set_port(struct rpc_xprt *xprt, unsigned short port) 1416 { 1417 struct sockaddr *addr = xs_addr(xprt); 1418 1419 dprintk("RPC: setting port for xprt %p to %u\n", xprt, port); 1420 1421 switch (addr->sa_family) { 1422 case AF_INET: 1423 ((struct sockaddr_in *)addr)->sin_port = htons(port); 1424 break; 1425 case AF_INET6: 1426 ((struct sockaddr_in6 *)addr)->sin6_port = htons(port); 1427 break; 1428 default: 1429 BUG(); 1430 } 1431 } 1432 1433 static unsigned short xs_get_srcport(struct sock_xprt *transport, struct socket *sock) 1434 { 1435 unsigned short port = transport->port; 1436 1437 if (port == 0 && transport->xprt.resvport) 1438 port = xs_get_random_port(); 1439 return port; 1440 } 1441 1442 static unsigned short xs_next_srcport(struct sock_xprt *transport, struct socket *sock, unsigned short port) 1443 { 1444 if (transport->port != 0) 1445 transport->port = 0; 1446 if (!transport->xprt.resvport) 1447 return 0; 1448 if (port <= xprt_min_resvport || port > xprt_max_resvport) 1449 return xprt_max_resvport; 1450 return --port; 1451 } 1452 1453 static int xs_bind4(struct sock_xprt *transport, struct socket *sock) 1454 { 1455 struct sockaddr_in myaddr = { 1456 .sin_family = AF_INET, 1457 }; 1458 struct sockaddr_in *sa; 1459 int err, nloop = 0; 1460 unsigned short port = xs_get_srcport(transport, sock); 1461 unsigned short last; 1462 1463 sa = (struct sockaddr_in *)&transport->addr; 1464 myaddr.sin_addr = sa->sin_addr; 1465 do { 1466 myaddr.sin_port = htons(port); 1467 err = kernel_bind(sock, (struct sockaddr *) &myaddr, 1468 sizeof(myaddr)); 1469 if (port == 0) 1470 break; 1471 if (err == 0) { 1472 transport->port = port; 1473 break; 1474 } 1475 last = port; 1476 port = xs_next_srcport(transport, sock, port); 1477 if (port > last) 1478 nloop++; 1479 } while (err == -EADDRINUSE && nloop != 2); 1480 dprintk("RPC: %s %pI4:%u: %s (%d)\n", 1481 __func__, &myaddr.sin_addr, 1482 port, err ? "failed" : "ok", err); 1483 return err; 1484 } 1485 1486 static int xs_bind6(struct sock_xprt *transport, struct socket *sock) 1487 { 1488 struct sockaddr_in6 myaddr = { 1489 .sin6_family = AF_INET6, 1490 }; 1491 struct sockaddr_in6 *sa; 1492 int err, nloop = 0; 1493 unsigned short port = xs_get_srcport(transport, sock); 1494 unsigned short last; 1495 1496 sa = (struct sockaddr_in6 *)&transport->addr; 1497 myaddr.sin6_addr = sa->sin6_addr; 1498 do { 1499 myaddr.sin6_port = htons(port); 1500 err = kernel_bind(sock, (struct sockaddr *) &myaddr, 1501 sizeof(myaddr)); 1502 if (port == 0) 1503 break; 1504 if (err == 0) { 1505 transport->port = port; 1506 break; 1507 } 1508 last = port; 1509 port = xs_next_srcport(transport, sock, port); 1510 if (port > last) 1511 nloop++; 1512 } while (err == -EADDRINUSE && nloop != 2); 1513 dprintk("RPC: xs_bind6 %pI6:%u: %s (%d)\n", 1514 &myaddr.sin6_addr, port, err ? "failed" : "ok", err); 1515 return err; 1516 } 1517 1518 #ifdef CONFIG_DEBUG_LOCK_ALLOC 1519 static struct lock_class_key xs_key[2]; 1520 static struct lock_class_key xs_slock_key[2]; 1521 1522 static inline void xs_reclassify_socket4(struct socket *sock) 1523 { 1524 struct sock *sk = sock->sk; 1525 1526 BUG_ON(sock_owned_by_user(sk)); 1527 sock_lock_init_class_and_name(sk, "slock-AF_INET-RPC", 1528 &xs_slock_key[0], "sk_lock-AF_INET-RPC", &xs_key[0]); 1529 } 1530 1531 static inline void xs_reclassify_socket6(struct socket *sock) 1532 { 1533 struct sock *sk = sock->sk; 1534 1535 BUG_ON(sock_owned_by_user(sk)); 1536 sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC", 1537 &xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]); 1538 } 1539 #else 1540 static inline void xs_reclassify_socket4(struct socket *sock) 1541 { 1542 } 1543 1544 static inline void xs_reclassify_socket6(struct socket *sock) 1545 { 1546 } 1547 #endif 1548 1549 static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock) 1550 { 1551 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1552 1553 if (!transport->inet) { 1554 struct sock *sk = sock->sk; 1555 1556 write_lock_bh(&sk->sk_callback_lock); 1557 1558 xs_save_old_callbacks(transport, sk); 1559 1560 sk->sk_user_data = xprt; 1561 sk->sk_data_ready = xs_udp_data_ready; 1562 sk->sk_write_space = xs_udp_write_space; 1563 sk->sk_error_report = xs_error_report; 1564 sk->sk_no_check = UDP_CSUM_NORCV; 1565 sk->sk_allocation = GFP_ATOMIC; 1566 1567 xprt_set_connected(xprt); 1568 1569 /* Reset to new socket */ 1570 transport->sock = sock; 1571 transport->inet = sk; 1572 1573 write_unlock_bh(&sk->sk_callback_lock); 1574 } 1575 xs_udp_do_set_buffer_size(xprt); 1576 } 1577 1578 /** 1579 * xs_udp_connect_worker4 - set up a UDP socket 1580 * @work: RPC transport to connect 1581 * 1582 * Invoked by a work queue tasklet. 1583 */ 1584 static void xs_udp_connect_worker4(struct work_struct *work) 1585 { 1586 struct sock_xprt *transport = 1587 container_of(work, struct sock_xprt, connect_worker.work); 1588 struct rpc_xprt *xprt = &transport->xprt; 1589 struct socket *sock = transport->sock; 1590 int err, status = -EIO; 1591 1592 if (xprt->shutdown) 1593 goto out; 1594 1595 /* Start by resetting any existing state */ 1596 xs_reset_transport(transport); 1597 1598 err = sock_create_kern(PF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock); 1599 if (err < 0) { 1600 dprintk("RPC: can't create UDP transport socket (%d).\n", -err); 1601 goto out; 1602 } 1603 xs_reclassify_socket4(sock); 1604 1605 if (xs_bind4(transport, sock)) { 1606 sock_release(sock); 1607 goto out; 1608 } 1609 1610 dprintk("RPC: worker connecting xprt %p to address: %s\n", 1611 xprt, xprt->address_strings[RPC_DISPLAY_ALL]); 1612 1613 xs_udp_finish_connecting(xprt, sock); 1614 status = 0; 1615 out: 1616 xprt_clear_connecting(xprt); 1617 xprt_wake_pending_tasks(xprt, status); 1618 } 1619 1620 /** 1621 * xs_udp_connect_worker6 - set up a UDP socket 1622 * @work: RPC transport to connect 1623 * 1624 * Invoked by a work queue tasklet. 1625 */ 1626 static void xs_udp_connect_worker6(struct work_struct *work) 1627 { 1628 struct sock_xprt *transport = 1629 container_of(work, struct sock_xprt, connect_worker.work); 1630 struct rpc_xprt *xprt = &transport->xprt; 1631 struct socket *sock = transport->sock; 1632 int err, status = -EIO; 1633 1634 if (xprt->shutdown) 1635 goto out; 1636 1637 /* Start by resetting any existing state */ 1638 xs_reset_transport(transport); 1639 1640 err = sock_create_kern(PF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock); 1641 if (err < 0) { 1642 dprintk("RPC: can't create UDP transport socket (%d).\n", -err); 1643 goto out; 1644 } 1645 xs_reclassify_socket6(sock); 1646 1647 if (xs_bind6(transport, sock) < 0) { 1648 sock_release(sock); 1649 goto out; 1650 } 1651 1652 dprintk("RPC: worker connecting xprt %p to address: %s\n", 1653 xprt, xprt->address_strings[RPC_DISPLAY_ALL]); 1654 1655 xs_udp_finish_connecting(xprt, sock); 1656 status = 0; 1657 out: 1658 xprt_clear_connecting(xprt); 1659 xprt_wake_pending_tasks(xprt, status); 1660 } 1661 1662 /* 1663 * We need to preserve the port number so the reply cache on the server can 1664 * find our cached RPC replies when we get around to reconnecting. 1665 */ 1666 static void xs_abort_connection(struct rpc_xprt *xprt, struct sock_xprt *transport) 1667 { 1668 int result; 1669 struct sockaddr any; 1670 1671 dprintk("RPC: disconnecting xprt %p to reuse port\n", xprt); 1672 1673 /* 1674 * Disconnect the transport socket by doing a connect operation 1675 * with AF_UNSPEC. This should return immediately... 1676 */ 1677 memset(&any, 0, sizeof(any)); 1678 any.sa_family = AF_UNSPEC; 1679 result = kernel_connect(transport->sock, &any, sizeof(any), 0); 1680 if (!result) 1681 xs_sock_mark_closed(xprt); 1682 else 1683 dprintk("RPC: AF_UNSPEC connect return code %d\n", 1684 result); 1685 } 1686 1687 static void xs_tcp_reuse_connection(struct rpc_xprt *xprt, struct sock_xprt *transport) 1688 { 1689 unsigned int state = transport->inet->sk_state; 1690 1691 if (state == TCP_CLOSE && transport->sock->state == SS_UNCONNECTED) 1692 return; 1693 if ((1 << state) & (TCPF_ESTABLISHED|TCPF_SYN_SENT)) 1694 return; 1695 xs_abort_connection(xprt, transport); 1696 } 1697 1698 static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock) 1699 { 1700 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1701 1702 if (!transport->inet) { 1703 struct sock *sk = sock->sk; 1704 1705 write_lock_bh(&sk->sk_callback_lock); 1706 1707 xs_save_old_callbacks(transport, sk); 1708 1709 sk->sk_user_data = xprt; 1710 sk->sk_data_ready = xs_tcp_data_ready; 1711 sk->sk_state_change = xs_tcp_state_change; 1712 sk->sk_write_space = xs_tcp_write_space; 1713 sk->sk_error_report = xs_error_report; 1714 sk->sk_allocation = GFP_ATOMIC; 1715 1716 /* socket options */ 1717 sk->sk_userlocks |= SOCK_BINDPORT_LOCK; 1718 sock_reset_flag(sk, SOCK_LINGER); 1719 tcp_sk(sk)->linger2 = 0; 1720 tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF; 1721 1722 xprt_clear_connected(xprt); 1723 1724 /* Reset to new socket */ 1725 transport->sock = sock; 1726 transport->inet = sk; 1727 1728 write_unlock_bh(&sk->sk_callback_lock); 1729 } 1730 1731 if (!xprt_bound(xprt)) 1732 return -ENOTCONN; 1733 1734 /* Tell the socket layer to start connecting... */ 1735 xprt->stat.connect_count++; 1736 xprt->stat.connect_start = jiffies; 1737 return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, O_NONBLOCK); 1738 } 1739 1740 /** 1741 * xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint 1742 * @xprt: RPC transport to connect 1743 * @transport: socket transport to connect 1744 * @create_sock: function to create a socket of the correct type 1745 * 1746 * Invoked by a work queue tasklet. 1747 */ 1748 static void xs_tcp_setup_socket(struct rpc_xprt *xprt, 1749 struct sock_xprt *transport, 1750 struct socket *(*create_sock)(struct rpc_xprt *, 1751 struct sock_xprt *)) 1752 { 1753 struct socket *sock = transport->sock; 1754 int status = -EIO; 1755 1756 if (xprt->shutdown) 1757 goto out; 1758 1759 if (!sock) { 1760 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state); 1761 sock = create_sock(xprt, transport); 1762 if (IS_ERR(sock)) { 1763 status = PTR_ERR(sock); 1764 goto out; 1765 } 1766 } else { 1767 int abort_and_exit; 1768 1769 abort_and_exit = test_and_clear_bit(XPRT_CONNECTION_ABORT, 1770 &xprt->state); 1771 /* "close" the socket, preserving the local port */ 1772 xs_tcp_reuse_connection(xprt, transport); 1773 1774 if (abort_and_exit) 1775 goto out_eagain; 1776 } 1777 1778 dprintk("RPC: worker connecting xprt %p to address: %s\n", 1779 xprt, xprt->address_strings[RPC_DISPLAY_ALL]); 1780 1781 status = xs_tcp_finish_connecting(xprt, sock); 1782 dprintk("RPC: %p connect status %d connected %d sock state %d\n", 1783 xprt, -status, xprt_connected(xprt), 1784 sock->sk->sk_state); 1785 switch (status) { 1786 default: 1787 printk("%s: connect returned unhandled error %d\n", 1788 __func__, status); 1789 case -EADDRNOTAVAIL: 1790 /* We're probably in TIME_WAIT. Get rid of existing socket, 1791 * and retry 1792 */ 1793 set_bit(XPRT_CONNECTION_CLOSE, &xprt->state); 1794 xprt_force_disconnect(xprt); 1795 case -ECONNREFUSED: 1796 case -ECONNRESET: 1797 case -ENETUNREACH: 1798 /* retry with existing socket, after a delay */ 1799 case 0: 1800 case -EINPROGRESS: 1801 case -EALREADY: 1802 xprt_clear_connecting(xprt); 1803 return; 1804 } 1805 out_eagain: 1806 status = -EAGAIN; 1807 out: 1808 xprt_clear_connecting(xprt); 1809 xprt_wake_pending_tasks(xprt, status); 1810 } 1811 1812 static struct socket *xs_create_tcp_sock4(struct rpc_xprt *xprt, 1813 struct sock_xprt *transport) 1814 { 1815 struct socket *sock; 1816 int err; 1817 1818 /* start from scratch */ 1819 err = sock_create_kern(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock); 1820 if (err < 0) { 1821 dprintk("RPC: can't create TCP transport socket (%d).\n", 1822 -err); 1823 goto out_err; 1824 } 1825 xs_reclassify_socket4(sock); 1826 1827 if (xs_bind4(transport, sock) < 0) { 1828 sock_release(sock); 1829 goto out_err; 1830 } 1831 return sock; 1832 out_err: 1833 return ERR_PTR(-EIO); 1834 } 1835 1836 /** 1837 * xs_tcp_connect_worker4 - connect a TCP socket to a remote endpoint 1838 * @work: RPC transport to connect 1839 * 1840 * Invoked by a work queue tasklet. 1841 */ 1842 static void xs_tcp_connect_worker4(struct work_struct *work) 1843 { 1844 struct sock_xprt *transport = 1845 container_of(work, struct sock_xprt, connect_worker.work); 1846 struct rpc_xprt *xprt = &transport->xprt; 1847 1848 xs_tcp_setup_socket(xprt, transport, xs_create_tcp_sock4); 1849 } 1850 1851 static struct socket *xs_create_tcp_sock6(struct rpc_xprt *xprt, 1852 struct sock_xprt *transport) 1853 { 1854 struct socket *sock; 1855 int err; 1856 1857 /* start from scratch */ 1858 err = sock_create_kern(PF_INET6, SOCK_STREAM, IPPROTO_TCP, &sock); 1859 if (err < 0) { 1860 dprintk("RPC: can't create TCP transport socket (%d).\n", 1861 -err); 1862 goto out_err; 1863 } 1864 xs_reclassify_socket6(sock); 1865 1866 if (xs_bind6(transport, sock) < 0) { 1867 sock_release(sock); 1868 goto out_err; 1869 } 1870 return sock; 1871 out_err: 1872 return ERR_PTR(-EIO); 1873 } 1874 1875 /** 1876 * xs_tcp_connect_worker6 - connect a TCP socket to a remote endpoint 1877 * @work: RPC transport to connect 1878 * 1879 * Invoked by a work queue tasklet. 1880 */ 1881 static void xs_tcp_connect_worker6(struct work_struct *work) 1882 { 1883 struct sock_xprt *transport = 1884 container_of(work, struct sock_xprt, connect_worker.work); 1885 struct rpc_xprt *xprt = &transport->xprt; 1886 1887 xs_tcp_setup_socket(xprt, transport, xs_create_tcp_sock6); 1888 } 1889 1890 /** 1891 * xs_connect - connect a socket to a remote endpoint 1892 * @task: address of RPC task that manages state of connect request 1893 * 1894 * TCP: If the remote end dropped the connection, delay reconnecting. 1895 * 1896 * UDP socket connects are synchronous, but we use a work queue anyway 1897 * to guarantee that even unprivileged user processes can set up a 1898 * socket on a privileged port. 1899 * 1900 * If a UDP socket connect fails, the delay behavior here prevents 1901 * retry floods (hard mounts). 1902 */ 1903 static void xs_connect(struct rpc_task *task) 1904 { 1905 struct rpc_xprt *xprt = task->tk_xprt; 1906 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1907 1908 if (xprt_test_and_set_connecting(xprt)) 1909 return; 1910 1911 if (transport->sock != NULL) { 1912 dprintk("RPC: xs_connect delayed xprt %p for %lu " 1913 "seconds\n", 1914 xprt, xprt->reestablish_timeout / HZ); 1915 queue_delayed_work(rpciod_workqueue, 1916 &transport->connect_worker, 1917 xprt->reestablish_timeout); 1918 xprt->reestablish_timeout <<= 1; 1919 if (xprt->reestablish_timeout > XS_TCP_MAX_REEST_TO) 1920 xprt->reestablish_timeout = XS_TCP_MAX_REEST_TO; 1921 } else { 1922 dprintk("RPC: xs_connect scheduled xprt %p\n", xprt); 1923 queue_delayed_work(rpciod_workqueue, 1924 &transport->connect_worker, 0); 1925 } 1926 } 1927 1928 static void xs_tcp_connect(struct rpc_task *task) 1929 { 1930 struct rpc_xprt *xprt = task->tk_xprt; 1931 1932 /* Exit if we need to wait for socket shutdown to complete */ 1933 if (test_bit(XPRT_CLOSING, &xprt->state)) 1934 return; 1935 xs_connect(task); 1936 } 1937 1938 /** 1939 * xs_udp_print_stats - display UDP socket-specifc stats 1940 * @xprt: rpc_xprt struct containing statistics 1941 * @seq: output file 1942 * 1943 */ 1944 static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) 1945 { 1946 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1947 1948 seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %Lu %Lu\n", 1949 transport->port, 1950 xprt->stat.bind_count, 1951 xprt->stat.sends, 1952 xprt->stat.recvs, 1953 xprt->stat.bad_xids, 1954 xprt->stat.req_u, 1955 xprt->stat.bklog_u); 1956 } 1957 1958 /** 1959 * xs_tcp_print_stats - display TCP socket-specifc stats 1960 * @xprt: rpc_xprt struct containing statistics 1961 * @seq: output file 1962 * 1963 */ 1964 static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) 1965 { 1966 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1967 long idle_time = 0; 1968 1969 if (xprt_connected(xprt)) 1970 idle_time = (long)(jiffies - xprt->last_used) / HZ; 1971 1972 seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu %Lu %Lu\n", 1973 transport->port, 1974 xprt->stat.bind_count, 1975 xprt->stat.connect_count, 1976 xprt->stat.connect_time, 1977 idle_time, 1978 xprt->stat.sends, 1979 xprt->stat.recvs, 1980 xprt->stat.bad_xids, 1981 xprt->stat.req_u, 1982 xprt->stat.bklog_u); 1983 } 1984 1985 static struct rpc_xprt_ops xs_udp_ops = { 1986 .set_buffer_size = xs_udp_set_buffer_size, 1987 .reserve_xprt = xprt_reserve_xprt_cong, 1988 .release_xprt = xprt_release_xprt_cong, 1989 .rpcbind = rpcb_getport_async, 1990 .set_port = xs_set_port, 1991 .connect = xs_connect, 1992 .buf_alloc = rpc_malloc, 1993 .buf_free = rpc_free, 1994 .send_request = xs_udp_send_request, 1995 .set_retrans_timeout = xprt_set_retrans_timeout_rtt, 1996 .timer = xs_udp_timer, 1997 .release_request = xprt_release_rqst_cong, 1998 .close = xs_close, 1999 .destroy = xs_destroy, 2000 .print_stats = xs_udp_print_stats, 2001 }; 2002 2003 static struct rpc_xprt_ops xs_tcp_ops = { 2004 .reserve_xprt = xprt_reserve_xprt, 2005 .release_xprt = xs_tcp_release_xprt, 2006 .rpcbind = rpcb_getport_async, 2007 .set_port = xs_set_port, 2008 .connect = xs_tcp_connect, 2009 .buf_alloc = rpc_malloc, 2010 .buf_free = rpc_free, 2011 .send_request = xs_tcp_send_request, 2012 .set_retrans_timeout = xprt_set_retrans_timeout_def, 2013 .close = xs_tcp_close, 2014 .destroy = xs_destroy, 2015 .print_stats = xs_tcp_print_stats, 2016 }; 2017 2018 static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args, 2019 unsigned int slot_table_size) 2020 { 2021 struct rpc_xprt *xprt; 2022 struct sock_xprt *new; 2023 2024 if (args->addrlen > sizeof(xprt->addr)) { 2025 dprintk("RPC: xs_setup_xprt: address too large\n"); 2026 return ERR_PTR(-EBADF); 2027 } 2028 2029 new = kzalloc(sizeof(*new), GFP_KERNEL); 2030 if (new == NULL) { 2031 dprintk("RPC: xs_setup_xprt: couldn't allocate " 2032 "rpc_xprt\n"); 2033 return ERR_PTR(-ENOMEM); 2034 } 2035 xprt = &new->xprt; 2036 2037 xprt->max_reqs = slot_table_size; 2038 xprt->slot = kcalloc(xprt->max_reqs, sizeof(struct rpc_rqst), GFP_KERNEL); 2039 if (xprt->slot == NULL) { 2040 kfree(xprt); 2041 dprintk("RPC: xs_setup_xprt: couldn't allocate slot " 2042 "table\n"); 2043 return ERR_PTR(-ENOMEM); 2044 } 2045 2046 memcpy(&xprt->addr, args->dstaddr, args->addrlen); 2047 xprt->addrlen = args->addrlen; 2048 if (args->srcaddr) 2049 memcpy(&new->addr, args->srcaddr, args->addrlen); 2050 2051 return xprt; 2052 } 2053 2054 static const struct rpc_timeout xs_udp_default_timeout = { 2055 .to_initval = 5 * HZ, 2056 .to_maxval = 30 * HZ, 2057 .to_increment = 5 * HZ, 2058 .to_retries = 5, 2059 }; 2060 2061 /** 2062 * xs_setup_udp - Set up transport to use a UDP socket 2063 * @args: rpc transport creation arguments 2064 * 2065 */ 2066 static struct rpc_xprt *xs_setup_udp(struct xprt_create *args) 2067 { 2068 struct sockaddr *addr = args->dstaddr; 2069 struct rpc_xprt *xprt; 2070 struct sock_xprt *transport; 2071 2072 xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries); 2073 if (IS_ERR(xprt)) 2074 return xprt; 2075 transport = container_of(xprt, struct sock_xprt, xprt); 2076 2077 xprt->prot = IPPROTO_UDP; 2078 xprt->tsh_size = 0; 2079 /* XXX: header size can vary due to auth type, IPv6, etc. */ 2080 xprt->max_payload = (1U << 16) - (MAX_HEADER << 3); 2081 2082 xprt->bind_timeout = XS_BIND_TO; 2083 xprt->connect_timeout = XS_UDP_CONN_TO; 2084 xprt->reestablish_timeout = XS_UDP_REEST_TO; 2085 xprt->idle_timeout = XS_IDLE_DISC_TO; 2086 2087 xprt->ops = &xs_udp_ops; 2088 2089 xprt->timeout = &xs_udp_default_timeout; 2090 2091 switch (addr->sa_family) { 2092 case AF_INET: 2093 if (((struct sockaddr_in *)addr)->sin_port != htons(0)) 2094 xprt_set_bound(xprt); 2095 2096 INIT_DELAYED_WORK(&transport->connect_worker, 2097 xs_udp_connect_worker4); 2098 xs_format_ipv4_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP); 2099 break; 2100 case AF_INET6: 2101 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0)) 2102 xprt_set_bound(xprt); 2103 2104 INIT_DELAYED_WORK(&transport->connect_worker, 2105 xs_udp_connect_worker6); 2106 xs_format_ipv6_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6); 2107 break; 2108 default: 2109 kfree(xprt); 2110 return ERR_PTR(-EAFNOSUPPORT); 2111 } 2112 2113 dprintk("RPC: set up transport to address %s\n", 2114 xprt->address_strings[RPC_DISPLAY_ALL]); 2115 2116 if (try_module_get(THIS_MODULE)) 2117 return xprt; 2118 2119 kfree(xprt->slot); 2120 kfree(xprt); 2121 return ERR_PTR(-EINVAL); 2122 } 2123 2124 static const struct rpc_timeout xs_tcp_default_timeout = { 2125 .to_initval = 60 * HZ, 2126 .to_maxval = 60 * HZ, 2127 .to_retries = 2, 2128 }; 2129 2130 /** 2131 * xs_setup_tcp - Set up transport to use a TCP socket 2132 * @args: rpc transport creation arguments 2133 * 2134 */ 2135 static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args) 2136 { 2137 struct sockaddr *addr = args->dstaddr; 2138 struct rpc_xprt *xprt; 2139 struct sock_xprt *transport; 2140 2141 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries); 2142 if (IS_ERR(xprt)) 2143 return xprt; 2144 transport = container_of(xprt, struct sock_xprt, xprt); 2145 2146 xprt->prot = IPPROTO_TCP; 2147 xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32); 2148 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE; 2149 2150 xprt->bind_timeout = XS_BIND_TO; 2151 xprt->connect_timeout = XS_TCP_CONN_TO; 2152 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; 2153 xprt->idle_timeout = XS_IDLE_DISC_TO; 2154 2155 xprt->ops = &xs_tcp_ops; 2156 xprt->timeout = &xs_tcp_default_timeout; 2157 2158 switch (addr->sa_family) { 2159 case AF_INET: 2160 if (((struct sockaddr_in *)addr)->sin_port != htons(0)) 2161 xprt_set_bound(xprt); 2162 2163 INIT_DELAYED_WORK(&transport->connect_worker, xs_tcp_connect_worker4); 2164 xs_format_ipv4_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP); 2165 break; 2166 case AF_INET6: 2167 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0)) 2168 xprt_set_bound(xprt); 2169 2170 INIT_DELAYED_WORK(&transport->connect_worker, xs_tcp_connect_worker6); 2171 xs_format_ipv6_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6); 2172 break; 2173 default: 2174 kfree(xprt); 2175 return ERR_PTR(-EAFNOSUPPORT); 2176 } 2177 2178 dprintk("RPC: set up transport to address %s\n", 2179 xprt->address_strings[RPC_DISPLAY_ALL]); 2180 2181 if (try_module_get(THIS_MODULE)) 2182 return xprt; 2183 2184 kfree(xprt->slot); 2185 kfree(xprt); 2186 return ERR_PTR(-EINVAL); 2187 } 2188 2189 static struct xprt_class xs_udp_transport = { 2190 .list = LIST_HEAD_INIT(xs_udp_transport.list), 2191 .name = "udp", 2192 .owner = THIS_MODULE, 2193 .ident = IPPROTO_UDP, 2194 .setup = xs_setup_udp, 2195 }; 2196 2197 static struct xprt_class xs_tcp_transport = { 2198 .list = LIST_HEAD_INIT(xs_tcp_transport.list), 2199 .name = "tcp", 2200 .owner = THIS_MODULE, 2201 .ident = IPPROTO_TCP, 2202 .setup = xs_setup_tcp, 2203 }; 2204 2205 /** 2206 * init_socket_xprt - set up xprtsock's sysctls, register with RPC client 2207 * 2208 */ 2209 int init_socket_xprt(void) 2210 { 2211 #ifdef RPC_DEBUG 2212 if (!sunrpc_table_header) 2213 sunrpc_table_header = register_sysctl_table(sunrpc_table); 2214 #endif 2215 2216 xprt_register_transport(&xs_udp_transport); 2217 xprt_register_transport(&xs_tcp_transport); 2218 2219 return 0; 2220 } 2221 2222 /** 2223 * cleanup_socket_xprt - remove xprtsock's sysctls, unregister 2224 * 2225 */ 2226 void cleanup_socket_xprt(void) 2227 { 2228 #ifdef RPC_DEBUG 2229 if (sunrpc_table_header) { 2230 unregister_sysctl_table(sunrpc_table_header); 2231 sunrpc_table_header = NULL; 2232 } 2233 #endif 2234 2235 xprt_unregister_transport(&xs_udp_transport); 2236 xprt_unregister_transport(&xs_tcp_transport); 2237 } 2238