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/string.h> 23 #include <linux/slab.h> 24 #include <linux/module.h> 25 #include <linux/capability.h> 26 #include <linux/pagemap.h> 27 #include <linux/errno.h> 28 #include <linux/socket.h> 29 #include <linux/in.h> 30 #include <linux/net.h> 31 #include <linux/mm.h> 32 #include <linux/un.h> 33 #include <linux/udp.h> 34 #include <linux/tcp.h> 35 #include <linux/sunrpc/clnt.h> 36 #include <linux/sunrpc/sched.h> 37 #include <linux/sunrpc/svcsock.h> 38 #include <linux/sunrpc/xprtsock.h> 39 #include <linux/file.h> 40 #ifdef CONFIG_SUNRPC_BACKCHANNEL 41 #include <linux/sunrpc/bc_xprt.h> 42 #endif 43 44 #include <net/sock.h> 45 #include <net/checksum.h> 46 #include <net/udp.h> 47 #include <net/tcp.h> 48 49 #include "sunrpc.h" 50 51 static void xs_close(struct rpc_xprt *xprt); 52 53 /* 54 * xprtsock tunables 55 */ 56 static unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE; 57 static unsigned int xprt_tcp_slot_table_entries = RPC_MIN_SLOT_TABLE; 58 static unsigned int xprt_max_tcp_slot_table_entries = RPC_MAX_SLOT_TABLE; 59 60 static unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT; 61 static unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT; 62 63 #define XS_TCP_LINGER_TO (15U * HZ) 64 static unsigned int xs_tcp_fin_timeout __read_mostly = XS_TCP_LINGER_TO; 65 66 /* 67 * We can register our own files under /proc/sys/sunrpc by 68 * calling register_sysctl_table() again. The files in that 69 * directory become the union of all files registered there. 70 * 71 * We simply need to make sure that we don't collide with 72 * someone else's file names! 73 */ 74 75 #ifdef RPC_DEBUG 76 77 static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE; 78 static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE; 79 static unsigned int max_tcp_slot_table_limit = RPC_MAX_SLOT_TABLE_LIMIT; 80 static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT; 81 static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT; 82 83 static struct ctl_table_header *sunrpc_table_header; 84 85 /* 86 * FIXME: changing the UDP slot table size should also resize the UDP 87 * socket buffers for existing UDP transports 88 */ 89 static ctl_table xs_tunables_table[] = { 90 { 91 .procname = "udp_slot_table_entries", 92 .data = &xprt_udp_slot_table_entries, 93 .maxlen = sizeof(unsigned int), 94 .mode = 0644, 95 .proc_handler = proc_dointvec_minmax, 96 .extra1 = &min_slot_table_size, 97 .extra2 = &max_slot_table_size 98 }, 99 { 100 .procname = "tcp_slot_table_entries", 101 .data = &xprt_tcp_slot_table_entries, 102 .maxlen = sizeof(unsigned int), 103 .mode = 0644, 104 .proc_handler = proc_dointvec_minmax, 105 .extra1 = &min_slot_table_size, 106 .extra2 = &max_slot_table_size 107 }, 108 { 109 .procname = "tcp_max_slot_table_entries", 110 .data = &xprt_max_tcp_slot_table_entries, 111 .maxlen = sizeof(unsigned int), 112 .mode = 0644, 113 .proc_handler = proc_dointvec_minmax, 114 .extra1 = &min_slot_table_size, 115 .extra2 = &max_tcp_slot_table_limit 116 }, 117 { 118 .procname = "min_resvport", 119 .data = &xprt_min_resvport, 120 .maxlen = sizeof(unsigned int), 121 .mode = 0644, 122 .proc_handler = proc_dointvec_minmax, 123 .extra1 = &xprt_min_resvport_limit, 124 .extra2 = &xprt_max_resvport_limit 125 }, 126 { 127 .procname = "max_resvport", 128 .data = &xprt_max_resvport, 129 .maxlen = sizeof(unsigned int), 130 .mode = 0644, 131 .proc_handler = proc_dointvec_minmax, 132 .extra1 = &xprt_min_resvport_limit, 133 .extra2 = &xprt_max_resvport_limit 134 }, 135 { 136 .procname = "tcp_fin_timeout", 137 .data = &xs_tcp_fin_timeout, 138 .maxlen = sizeof(xs_tcp_fin_timeout), 139 .mode = 0644, 140 .proc_handler = proc_dointvec_jiffies, 141 }, 142 { }, 143 }; 144 145 static ctl_table sunrpc_table[] = { 146 { 147 .procname = "sunrpc", 148 .mode = 0555, 149 .child = xs_tunables_table 150 }, 151 { }, 152 }; 153 154 #endif 155 156 /* 157 * Wait duration for a reply from the RPC portmapper. 158 */ 159 #define XS_BIND_TO (60U * HZ) 160 161 /* 162 * Delay if a UDP socket connect error occurs. This is most likely some 163 * kind of resource problem on the local host. 164 */ 165 #define XS_UDP_REEST_TO (2U * HZ) 166 167 /* 168 * The reestablish timeout allows clients to delay for a bit before attempting 169 * to reconnect to a server that just dropped our connection. 170 * 171 * We implement an exponential backoff when trying to reestablish a TCP 172 * transport connection with the server. Some servers like to drop a TCP 173 * connection when they are overworked, so we start with a short timeout and 174 * increase over time if the server is down or not responding. 175 */ 176 #define XS_TCP_INIT_REEST_TO (3U * HZ) 177 #define XS_TCP_MAX_REEST_TO (5U * 60 * HZ) 178 179 /* 180 * TCP idle timeout; client drops the transport socket if it is idle 181 * for this long. Note that we also timeout UDP sockets to prevent 182 * holding port numbers when there is no RPC traffic. 183 */ 184 #define XS_IDLE_DISC_TO (5U * 60 * HZ) 185 186 #ifdef RPC_DEBUG 187 # undef RPC_DEBUG_DATA 188 # define RPCDBG_FACILITY RPCDBG_TRANS 189 #endif 190 191 #ifdef RPC_DEBUG_DATA 192 static void xs_pktdump(char *msg, u32 *packet, unsigned int count) 193 { 194 u8 *buf = (u8 *) packet; 195 int j; 196 197 dprintk("RPC: %s\n", msg); 198 for (j = 0; j < count && j < 128; j += 4) { 199 if (!(j & 31)) { 200 if (j) 201 dprintk("\n"); 202 dprintk("0x%04x ", j); 203 } 204 dprintk("%02x%02x%02x%02x ", 205 buf[j], buf[j+1], buf[j+2], buf[j+3]); 206 } 207 dprintk("\n"); 208 } 209 #else 210 static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count) 211 { 212 /* NOP */ 213 } 214 #endif 215 216 struct sock_xprt { 217 struct rpc_xprt xprt; 218 219 /* 220 * Network layer 221 */ 222 struct socket * sock; 223 struct sock * inet; 224 225 /* 226 * State of TCP reply receive 227 */ 228 __be32 tcp_fraghdr, 229 tcp_xid, 230 tcp_calldir; 231 232 u32 tcp_offset, 233 tcp_reclen; 234 235 unsigned long tcp_copied, 236 tcp_flags; 237 238 /* 239 * Connection of transports 240 */ 241 struct delayed_work connect_worker; 242 struct sockaddr_storage srcaddr; 243 unsigned short srcport; 244 245 /* 246 * UDP socket buffer size parameters 247 */ 248 size_t rcvsize, 249 sndsize; 250 251 /* 252 * Saved socket callback addresses 253 */ 254 void (*old_data_ready)(struct sock *, int); 255 void (*old_state_change)(struct sock *); 256 void (*old_write_space)(struct sock *); 257 void (*old_error_report)(struct sock *); 258 }; 259 260 /* 261 * TCP receive state flags 262 */ 263 #define TCP_RCV_LAST_FRAG (1UL << 0) 264 #define TCP_RCV_COPY_FRAGHDR (1UL << 1) 265 #define TCP_RCV_COPY_XID (1UL << 2) 266 #define TCP_RCV_COPY_DATA (1UL << 3) 267 #define TCP_RCV_READ_CALLDIR (1UL << 4) 268 #define TCP_RCV_COPY_CALLDIR (1UL << 5) 269 270 /* 271 * TCP RPC flags 272 */ 273 #define TCP_RPC_REPLY (1UL << 6) 274 275 static inline struct sockaddr *xs_addr(struct rpc_xprt *xprt) 276 { 277 return (struct sockaddr *) &xprt->addr; 278 } 279 280 static inline struct sockaddr_un *xs_addr_un(struct rpc_xprt *xprt) 281 { 282 return (struct sockaddr_un *) &xprt->addr; 283 } 284 285 static inline struct sockaddr_in *xs_addr_in(struct rpc_xprt *xprt) 286 { 287 return (struct sockaddr_in *) &xprt->addr; 288 } 289 290 static inline struct sockaddr_in6 *xs_addr_in6(struct rpc_xprt *xprt) 291 { 292 return (struct sockaddr_in6 *) &xprt->addr; 293 } 294 295 static void xs_format_common_peer_addresses(struct rpc_xprt *xprt) 296 { 297 struct sockaddr *sap = xs_addr(xprt); 298 struct sockaddr_in6 *sin6; 299 struct sockaddr_in *sin; 300 struct sockaddr_un *sun; 301 char buf[128]; 302 303 switch (sap->sa_family) { 304 case AF_LOCAL: 305 sun = xs_addr_un(xprt); 306 strlcpy(buf, sun->sun_path, sizeof(buf)); 307 xprt->address_strings[RPC_DISPLAY_ADDR] = 308 kstrdup(buf, GFP_KERNEL); 309 break; 310 case AF_INET: 311 (void)rpc_ntop(sap, buf, sizeof(buf)); 312 xprt->address_strings[RPC_DISPLAY_ADDR] = 313 kstrdup(buf, GFP_KERNEL); 314 sin = xs_addr_in(xprt); 315 snprintf(buf, sizeof(buf), "%08x", ntohl(sin->sin_addr.s_addr)); 316 break; 317 case AF_INET6: 318 (void)rpc_ntop(sap, buf, sizeof(buf)); 319 xprt->address_strings[RPC_DISPLAY_ADDR] = 320 kstrdup(buf, GFP_KERNEL); 321 sin6 = xs_addr_in6(xprt); 322 snprintf(buf, sizeof(buf), "%pi6", &sin6->sin6_addr); 323 break; 324 default: 325 BUG(); 326 } 327 328 xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = kstrdup(buf, GFP_KERNEL); 329 } 330 331 static void xs_format_common_peer_ports(struct rpc_xprt *xprt) 332 { 333 struct sockaddr *sap = xs_addr(xprt); 334 char buf[128]; 335 336 snprintf(buf, sizeof(buf), "%u", rpc_get_port(sap)); 337 xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL); 338 339 snprintf(buf, sizeof(buf), "%4hx", rpc_get_port(sap)); 340 xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL); 341 } 342 343 static void xs_format_peer_addresses(struct rpc_xprt *xprt, 344 const char *protocol, 345 const char *netid) 346 { 347 xprt->address_strings[RPC_DISPLAY_PROTO] = protocol; 348 xprt->address_strings[RPC_DISPLAY_NETID] = netid; 349 xs_format_common_peer_addresses(xprt); 350 xs_format_common_peer_ports(xprt); 351 } 352 353 static void xs_update_peer_port(struct rpc_xprt *xprt) 354 { 355 kfree(xprt->address_strings[RPC_DISPLAY_HEX_PORT]); 356 kfree(xprt->address_strings[RPC_DISPLAY_PORT]); 357 358 xs_format_common_peer_ports(xprt); 359 } 360 361 static void xs_free_peer_addresses(struct rpc_xprt *xprt) 362 { 363 unsigned int i; 364 365 for (i = 0; i < RPC_DISPLAY_MAX; i++) 366 switch (i) { 367 case RPC_DISPLAY_PROTO: 368 case RPC_DISPLAY_NETID: 369 continue; 370 default: 371 kfree(xprt->address_strings[i]); 372 } 373 } 374 375 #define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL) 376 377 static int xs_send_kvec(struct socket *sock, struct sockaddr *addr, int addrlen, struct kvec *vec, unsigned int base, int more) 378 { 379 struct msghdr msg = { 380 .msg_name = addr, 381 .msg_namelen = addrlen, 382 .msg_flags = XS_SENDMSG_FLAGS | (more ? MSG_MORE : 0), 383 }; 384 struct kvec iov = { 385 .iov_base = vec->iov_base + base, 386 .iov_len = vec->iov_len - base, 387 }; 388 389 if (iov.iov_len != 0) 390 return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len); 391 return kernel_sendmsg(sock, &msg, NULL, 0, 0); 392 } 393 394 static int xs_send_pagedata(struct socket *sock, struct xdr_buf *xdr, unsigned int base, int more) 395 { 396 struct page **ppage; 397 unsigned int remainder; 398 int err, sent = 0; 399 400 remainder = xdr->page_len - base; 401 base += xdr->page_base; 402 ppage = xdr->pages + (base >> PAGE_SHIFT); 403 base &= ~PAGE_MASK; 404 for(;;) { 405 unsigned int len = min_t(unsigned int, PAGE_SIZE - base, remainder); 406 int flags = XS_SENDMSG_FLAGS; 407 408 remainder -= len; 409 if (remainder != 0 || more) 410 flags |= MSG_MORE; 411 err = sock->ops->sendpage(sock, *ppage, base, len, flags); 412 if (remainder == 0 || err != len) 413 break; 414 sent += err; 415 ppage++; 416 base = 0; 417 } 418 if (sent == 0) 419 return err; 420 if (err > 0) 421 sent += err; 422 return sent; 423 } 424 425 /** 426 * xs_sendpages - write pages directly to a socket 427 * @sock: socket to send on 428 * @addr: UDP only -- address of destination 429 * @addrlen: UDP only -- length of destination address 430 * @xdr: buffer containing this request 431 * @base: starting position in the buffer 432 * 433 */ 434 static int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base) 435 { 436 unsigned int remainder = xdr->len - base; 437 int err, sent = 0; 438 439 if (unlikely(!sock)) 440 return -ENOTSOCK; 441 442 clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags); 443 if (base != 0) { 444 addr = NULL; 445 addrlen = 0; 446 } 447 448 if (base < xdr->head[0].iov_len || addr != NULL) { 449 unsigned int len = xdr->head[0].iov_len - base; 450 remainder -= len; 451 err = xs_send_kvec(sock, addr, addrlen, &xdr->head[0], base, remainder != 0); 452 if (remainder == 0 || err != len) 453 goto out; 454 sent += err; 455 base = 0; 456 } else 457 base -= xdr->head[0].iov_len; 458 459 if (base < xdr->page_len) { 460 unsigned int len = xdr->page_len - base; 461 remainder -= len; 462 err = xs_send_pagedata(sock, xdr, base, remainder != 0); 463 if (remainder == 0 || err != len) 464 goto out; 465 sent += err; 466 base = 0; 467 } else 468 base -= xdr->page_len; 469 470 if (base >= xdr->tail[0].iov_len) 471 return sent; 472 err = xs_send_kvec(sock, NULL, 0, &xdr->tail[0], base, 0); 473 out: 474 if (sent == 0) 475 return err; 476 if (err > 0) 477 sent += err; 478 return sent; 479 } 480 481 static void xs_nospace_callback(struct rpc_task *task) 482 { 483 struct sock_xprt *transport = container_of(task->tk_rqstp->rq_xprt, struct sock_xprt, xprt); 484 485 transport->inet->sk_write_pending--; 486 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); 487 } 488 489 /** 490 * xs_nospace - place task on wait queue if transmit was incomplete 491 * @task: task to put to sleep 492 * 493 */ 494 static int xs_nospace(struct rpc_task *task) 495 { 496 struct rpc_rqst *req = task->tk_rqstp; 497 struct rpc_xprt *xprt = req->rq_xprt; 498 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 499 int ret = -EAGAIN; 500 501 dprintk("RPC: %5u xmit incomplete (%u left of %u)\n", 502 task->tk_pid, req->rq_slen - req->rq_bytes_sent, 503 req->rq_slen); 504 505 /* Protect against races with write_space */ 506 spin_lock_bh(&xprt->transport_lock); 507 508 /* Don't race with disconnect */ 509 if (xprt_connected(xprt)) { 510 if (test_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags)) { 511 /* 512 * Notify TCP that we're limited by the application 513 * window size 514 */ 515 set_bit(SOCK_NOSPACE, &transport->sock->flags); 516 transport->inet->sk_write_pending++; 517 /* ...and wait for more buffer space */ 518 xprt_wait_for_buffer_space(task, xs_nospace_callback); 519 } 520 } else { 521 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); 522 ret = -ENOTCONN; 523 } 524 525 spin_unlock_bh(&xprt->transport_lock); 526 return ret; 527 } 528 529 /* 530 * Construct a stream transport record marker in @buf. 531 */ 532 static inline void xs_encode_stream_record_marker(struct xdr_buf *buf) 533 { 534 u32 reclen = buf->len - sizeof(rpc_fraghdr); 535 rpc_fraghdr *base = buf->head[0].iov_base; 536 *base = cpu_to_be32(RPC_LAST_STREAM_FRAGMENT | reclen); 537 } 538 539 /** 540 * xs_local_send_request - write an RPC request to an AF_LOCAL socket 541 * @task: RPC task that manages the state of an RPC request 542 * 543 * Return values: 544 * 0: The request has been sent 545 * EAGAIN: The socket was blocked, please call again later to 546 * complete the request 547 * ENOTCONN: Caller needs to invoke connect logic then call again 548 * other: Some other error occured, the request was not sent 549 */ 550 static int xs_local_send_request(struct rpc_task *task) 551 { 552 struct rpc_rqst *req = task->tk_rqstp; 553 struct rpc_xprt *xprt = req->rq_xprt; 554 struct sock_xprt *transport = 555 container_of(xprt, struct sock_xprt, xprt); 556 struct xdr_buf *xdr = &req->rq_snd_buf; 557 int status; 558 559 xs_encode_stream_record_marker(&req->rq_snd_buf); 560 561 xs_pktdump("packet data:", 562 req->rq_svec->iov_base, req->rq_svec->iov_len); 563 564 status = xs_sendpages(transport->sock, NULL, 0, 565 xdr, req->rq_bytes_sent); 566 dprintk("RPC: %s(%u) = %d\n", 567 __func__, xdr->len - req->rq_bytes_sent, status); 568 if (likely(status >= 0)) { 569 req->rq_bytes_sent += status; 570 req->rq_xmit_bytes_sent += status; 571 if (likely(req->rq_bytes_sent >= req->rq_slen)) { 572 req->rq_bytes_sent = 0; 573 return 0; 574 } 575 status = -EAGAIN; 576 } 577 578 switch (status) { 579 case -EAGAIN: 580 status = xs_nospace(task); 581 break; 582 default: 583 dprintk("RPC: sendmsg returned unrecognized error %d\n", 584 -status); 585 case -EPIPE: 586 xs_close(xprt); 587 status = -ENOTCONN; 588 } 589 590 return status; 591 } 592 593 /** 594 * xs_udp_send_request - write an RPC request to a UDP socket 595 * @task: address of RPC task that manages the state of an RPC request 596 * 597 * Return values: 598 * 0: The request has been sent 599 * EAGAIN: The socket was blocked, please call again later to 600 * complete the request 601 * ENOTCONN: Caller needs to invoke connect logic then call again 602 * other: Some other error occurred, the request was not sent 603 */ 604 static int xs_udp_send_request(struct rpc_task *task) 605 { 606 struct rpc_rqst *req = task->tk_rqstp; 607 struct rpc_xprt *xprt = req->rq_xprt; 608 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 609 struct xdr_buf *xdr = &req->rq_snd_buf; 610 int status; 611 612 xs_pktdump("packet data:", 613 req->rq_svec->iov_base, 614 req->rq_svec->iov_len); 615 616 if (!xprt_bound(xprt)) 617 return -ENOTCONN; 618 status = xs_sendpages(transport->sock, 619 xs_addr(xprt), 620 xprt->addrlen, xdr, 621 req->rq_bytes_sent); 622 623 dprintk("RPC: xs_udp_send_request(%u) = %d\n", 624 xdr->len - req->rq_bytes_sent, status); 625 626 if (status >= 0) { 627 req->rq_xmit_bytes_sent += status; 628 if (status >= req->rq_slen) 629 return 0; 630 /* Still some bytes left; set up for a retry later. */ 631 status = -EAGAIN; 632 } 633 634 switch (status) { 635 case -ENOTSOCK: 636 status = -ENOTCONN; 637 /* Should we call xs_close() here? */ 638 break; 639 case -EAGAIN: 640 status = xs_nospace(task); 641 break; 642 default: 643 dprintk("RPC: sendmsg returned unrecognized error %d\n", 644 -status); 645 case -ENETUNREACH: 646 case -EPIPE: 647 case -ECONNREFUSED: 648 /* When the server has died, an ICMP port unreachable message 649 * prompts ECONNREFUSED. */ 650 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); 651 } 652 653 return status; 654 } 655 656 /** 657 * xs_tcp_shutdown - gracefully shut down a TCP socket 658 * @xprt: transport 659 * 660 * Initiates a graceful shutdown of the TCP socket by calling the 661 * equivalent of shutdown(SHUT_WR); 662 */ 663 static void xs_tcp_shutdown(struct rpc_xprt *xprt) 664 { 665 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 666 struct socket *sock = transport->sock; 667 668 if (sock != NULL) 669 kernel_sock_shutdown(sock, SHUT_WR); 670 } 671 672 /** 673 * xs_tcp_send_request - write an RPC request to a TCP socket 674 * @task: address of RPC task that manages the state of an RPC request 675 * 676 * Return values: 677 * 0: The request has been sent 678 * EAGAIN: The socket was blocked, please call again later to 679 * complete the request 680 * ENOTCONN: Caller needs to invoke connect logic then call again 681 * other: Some other error occurred, the request was not sent 682 * 683 * XXX: In the case of soft timeouts, should we eventually give up 684 * if sendmsg is not able to make progress? 685 */ 686 static int xs_tcp_send_request(struct rpc_task *task) 687 { 688 struct rpc_rqst *req = task->tk_rqstp; 689 struct rpc_xprt *xprt = req->rq_xprt; 690 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 691 struct xdr_buf *xdr = &req->rq_snd_buf; 692 int status; 693 694 xs_encode_stream_record_marker(&req->rq_snd_buf); 695 696 xs_pktdump("packet data:", 697 req->rq_svec->iov_base, 698 req->rq_svec->iov_len); 699 700 /* Continue transmitting the packet/record. We must be careful 701 * to cope with writespace callbacks arriving _after_ we have 702 * called sendmsg(). */ 703 while (1) { 704 status = xs_sendpages(transport->sock, 705 NULL, 0, xdr, req->rq_bytes_sent); 706 707 dprintk("RPC: xs_tcp_send_request(%u) = %d\n", 708 xdr->len - req->rq_bytes_sent, status); 709 710 if (unlikely(status < 0)) 711 break; 712 713 /* If we've sent the entire packet, immediately 714 * reset the count of bytes sent. */ 715 req->rq_bytes_sent += status; 716 req->rq_xmit_bytes_sent += status; 717 if (likely(req->rq_bytes_sent >= req->rq_slen)) { 718 req->rq_bytes_sent = 0; 719 return 0; 720 } 721 722 if (status != 0) 723 continue; 724 status = -EAGAIN; 725 break; 726 } 727 728 switch (status) { 729 case -ENOTSOCK: 730 status = -ENOTCONN; 731 /* Should we call xs_close() here? */ 732 break; 733 case -EAGAIN: 734 status = xs_nospace(task); 735 break; 736 default: 737 dprintk("RPC: sendmsg returned unrecognized error %d\n", 738 -status); 739 case -ECONNRESET: 740 case -EPIPE: 741 xs_tcp_shutdown(xprt); 742 case -ECONNREFUSED: 743 case -ENOTCONN: 744 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); 745 } 746 747 return status; 748 } 749 750 /** 751 * xs_tcp_release_xprt - clean up after a tcp transmission 752 * @xprt: transport 753 * @task: rpc task 754 * 755 * This cleans up if an error causes us to abort the transmission of a request. 756 * In this case, the socket may need to be reset in order to avoid confusing 757 * the server. 758 */ 759 static void xs_tcp_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task) 760 { 761 struct rpc_rqst *req; 762 763 if (task != xprt->snd_task) 764 return; 765 if (task == NULL) 766 goto out_release; 767 req = task->tk_rqstp; 768 if (req == NULL) 769 goto out_release; 770 if (req->rq_bytes_sent == 0) 771 goto out_release; 772 if (req->rq_bytes_sent == req->rq_snd_buf.len) 773 goto out_release; 774 set_bit(XPRT_CLOSE_WAIT, &task->tk_xprt->state); 775 out_release: 776 xprt_release_xprt(xprt, task); 777 } 778 779 static void xs_save_old_callbacks(struct sock_xprt *transport, struct sock *sk) 780 { 781 transport->old_data_ready = sk->sk_data_ready; 782 transport->old_state_change = sk->sk_state_change; 783 transport->old_write_space = sk->sk_write_space; 784 transport->old_error_report = sk->sk_error_report; 785 } 786 787 static void xs_restore_old_callbacks(struct sock_xprt *transport, struct sock *sk) 788 { 789 sk->sk_data_ready = transport->old_data_ready; 790 sk->sk_state_change = transport->old_state_change; 791 sk->sk_write_space = transport->old_write_space; 792 sk->sk_error_report = transport->old_error_report; 793 } 794 795 static void xs_reset_transport(struct sock_xprt *transport) 796 { 797 struct socket *sock = transport->sock; 798 struct sock *sk = transport->inet; 799 800 if (sk == NULL) 801 return; 802 803 transport->srcport = 0; 804 805 write_lock_bh(&sk->sk_callback_lock); 806 transport->inet = NULL; 807 transport->sock = NULL; 808 809 sk->sk_user_data = NULL; 810 811 xs_restore_old_callbacks(transport, sk); 812 write_unlock_bh(&sk->sk_callback_lock); 813 814 sk->sk_no_check = 0; 815 816 sock_release(sock); 817 } 818 819 /** 820 * xs_close - close a socket 821 * @xprt: transport 822 * 823 * This is used when all requests are complete; ie, no DRC state remains 824 * on the server we want to save. 825 * 826 * The caller _must_ be holding XPRT_LOCKED in order to avoid issues with 827 * xs_reset_transport() zeroing the socket from underneath a writer. 828 */ 829 static void xs_close(struct rpc_xprt *xprt) 830 { 831 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 832 833 dprintk("RPC: xs_close xprt %p\n", xprt); 834 835 xs_reset_transport(transport); 836 xprt->reestablish_timeout = 0; 837 838 smp_mb__before_clear_bit(); 839 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state); 840 clear_bit(XPRT_CLOSE_WAIT, &xprt->state); 841 clear_bit(XPRT_CLOSING, &xprt->state); 842 smp_mb__after_clear_bit(); 843 xprt_disconnect_done(xprt); 844 } 845 846 static void xs_tcp_close(struct rpc_xprt *xprt) 847 { 848 if (test_and_clear_bit(XPRT_CONNECTION_CLOSE, &xprt->state)) 849 xs_close(xprt); 850 else 851 xs_tcp_shutdown(xprt); 852 } 853 854 /** 855 * xs_destroy - prepare to shutdown a transport 856 * @xprt: doomed transport 857 * 858 */ 859 static void xs_destroy(struct rpc_xprt *xprt) 860 { 861 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 862 863 dprintk("RPC: xs_destroy xprt %p\n", xprt); 864 865 cancel_delayed_work_sync(&transport->connect_worker); 866 867 xs_close(xprt); 868 xs_free_peer_addresses(xprt); 869 xprt_free(xprt); 870 module_put(THIS_MODULE); 871 } 872 873 static inline struct rpc_xprt *xprt_from_sock(struct sock *sk) 874 { 875 return (struct rpc_xprt *) sk->sk_user_data; 876 } 877 878 static int xs_local_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb) 879 { 880 struct xdr_skb_reader desc = { 881 .skb = skb, 882 .offset = sizeof(rpc_fraghdr), 883 .count = skb->len - sizeof(rpc_fraghdr), 884 }; 885 886 if (xdr_partial_copy_from_skb(xdr, 0, &desc, xdr_skb_read_bits) < 0) 887 return -1; 888 if (desc.count) 889 return -1; 890 return 0; 891 } 892 893 /** 894 * xs_local_data_ready - "data ready" callback for AF_LOCAL sockets 895 * @sk: socket with data to read 896 * @len: how much data to read 897 * 898 * Currently this assumes we can read the whole reply in a single gulp. 899 */ 900 static void xs_local_data_ready(struct sock *sk, int len) 901 { 902 struct rpc_task *task; 903 struct rpc_xprt *xprt; 904 struct rpc_rqst *rovr; 905 struct sk_buff *skb; 906 int err, repsize, copied; 907 u32 _xid; 908 __be32 *xp; 909 910 read_lock_bh(&sk->sk_callback_lock); 911 dprintk("RPC: %s...\n", __func__); 912 xprt = xprt_from_sock(sk); 913 if (xprt == NULL) 914 goto out; 915 916 skb = skb_recv_datagram(sk, 0, 1, &err); 917 if (skb == NULL) 918 goto out; 919 920 if (xprt->shutdown) 921 goto dropit; 922 923 repsize = skb->len - sizeof(rpc_fraghdr); 924 if (repsize < 4) { 925 dprintk("RPC: impossible RPC reply size %d\n", repsize); 926 goto dropit; 927 } 928 929 /* Copy the XID from the skb... */ 930 xp = skb_header_pointer(skb, sizeof(rpc_fraghdr), sizeof(_xid), &_xid); 931 if (xp == NULL) 932 goto dropit; 933 934 /* Look up and lock the request corresponding to the given XID */ 935 spin_lock(&xprt->transport_lock); 936 rovr = xprt_lookup_rqst(xprt, *xp); 937 if (!rovr) 938 goto out_unlock; 939 task = rovr->rq_task; 940 941 copied = rovr->rq_private_buf.buflen; 942 if (copied > repsize) 943 copied = repsize; 944 945 if (xs_local_copy_to_xdr(&rovr->rq_private_buf, skb)) { 946 dprintk("RPC: sk_buff copy failed\n"); 947 goto out_unlock; 948 } 949 950 xprt_complete_rqst(task, copied); 951 952 out_unlock: 953 spin_unlock(&xprt->transport_lock); 954 dropit: 955 skb_free_datagram(sk, skb); 956 out: 957 read_unlock_bh(&sk->sk_callback_lock); 958 } 959 960 /** 961 * xs_udp_data_ready - "data ready" callback for UDP sockets 962 * @sk: socket with data to read 963 * @len: how much data to read 964 * 965 */ 966 static void xs_udp_data_ready(struct sock *sk, int len) 967 { 968 struct rpc_task *task; 969 struct rpc_xprt *xprt; 970 struct rpc_rqst *rovr; 971 struct sk_buff *skb; 972 int err, repsize, copied; 973 u32 _xid; 974 __be32 *xp; 975 976 read_lock_bh(&sk->sk_callback_lock); 977 dprintk("RPC: xs_udp_data_ready...\n"); 978 if (!(xprt = xprt_from_sock(sk))) 979 goto out; 980 981 if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL) 982 goto out; 983 984 if (xprt->shutdown) 985 goto dropit; 986 987 repsize = skb->len - sizeof(struct udphdr); 988 if (repsize < 4) { 989 dprintk("RPC: impossible RPC reply size %d!\n", repsize); 990 goto dropit; 991 } 992 993 /* Copy the XID from the skb... */ 994 xp = skb_header_pointer(skb, sizeof(struct udphdr), 995 sizeof(_xid), &_xid); 996 if (xp == NULL) 997 goto dropit; 998 999 /* Look up and lock the request corresponding to the given XID */ 1000 spin_lock(&xprt->transport_lock); 1001 rovr = xprt_lookup_rqst(xprt, *xp); 1002 if (!rovr) 1003 goto out_unlock; 1004 task = rovr->rq_task; 1005 1006 if ((copied = rovr->rq_private_buf.buflen) > repsize) 1007 copied = repsize; 1008 1009 /* Suck it into the iovec, verify checksum if not done by hw. */ 1010 if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) { 1011 UDPX_INC_STATS_BH(sk, UDP_MIB_INERRORS); 1012 goto out_unlock; 1013 } 1014 1015 UDPX_INC_STATS_BH(sk, UDP_MIB_INDATAGRAMS); 1016 1017 /* Something worked... */ 1018 dst_confirm(skb_dst(skb)); 1019 1020 xprt_adjust_cwnd(task, copied); 1021 xprt_complete_rqst(task, copied); 1022 1023 out_unlock: 1024 spin_unlock(&xprt->transport_lock); 1025 dropit: 1026 skb_free_datagram(sk, skb); 1027 out: 1028 read_unlock_bh(&sk->sk_callback_lock); 1029 } 1030 1031 static inline void xs_tcp_read_fraghdr(struct rpc_xprt *xprt, struct xdr_skb_reader *desc) 1032 { 1033 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1034 size_t len, used; 1035 char *p; 1036 1037 p = ((char *) &transport->tcp_fraghdr) + transport->tcp_offset; 1038 len = sizeof(transport->tcp_fraghdr) - transport->tcp_offset; 1039 used = xdr_skb_read_bits(desc, p, len); 1040 transport->tcp_offset += used; 1041 if (used != len) 1042 return; 1043 1044 transport->tcp_reclen = ntohl(transport->tcp_fraghdr); 1045 if (transport->tcp_reclen & RPC_LAST_STREAM_FRAGMENT) 1046 transport->tcp_flags |= TCP_RCV_LAST_FRAG; 1047 else 1048 transport->tcp_flags &= ~TCP_RCV_LAST_FRAG; 1049 transport->tcp_reclen &= RPC_FRAGMENT_SIZE_MASK; 1050 1051 transport->tcp_flags &= ~TCP_RCV_COPY_FRAGHDR; 1052 transport->tcp_offset = 0; 1053 1054 /* Sanity check of the record length */ 1055 if (unlikely(transport->tcp_reclen < 8)) { 1056 dprintk("RPC: invalid TCP record fragment length\n"); 1057 xprt_force_disconnect(xprt); 1058 return; 1059 } 1060 dprintk("RPC: reading TCP record fragment of length %d\n", 1061 transport->tcp_reclen); 1062 } 1063 1064 static void xs_tcp_check_fraghdr(struct sock_xprt *transport) 1065 { 1066 if (transport->tcp_offset == transport->tcp_reclen) { 1067 transport->tcp_flags |= TCP_RCV_COPY_FRAGHDR; 1068 transport->tcp_offset = 0; 1069 if (transport->tcp_flags & TCP_RCV_LAST_FRAG) { 1070 transport->tcp_flags &= ~TCP_RCV_COPY_DATA; 1071 transport->tcp_flags |= TCP_RCV_COPY_XID; 1072 transport->tcp_copied = 0; 1073 } 1074 } 1075 } 1076 1077 static inline void xs_tcp_read_xid(struct sock_xprt *transport, struct xdr_skb_reader *desc) 1078 { 1079 size_t len, used; 1080 char *p; 1081 1082 len = sizeof(transport->tcp_xid) - transport->tcp_offset; 1083 dprintk("RPC: reading XID (%Zu bytes)\n", len); 1084 p = ((char *) &transport->tcp_xid) + transport->tcp_offset; 1085 used = xdr_skb_read_bits(desc, p, len); 1086 transport->tcp_offset += used; 1087 if (used != len) 1088 return; 1089 transport->tcp_flags &= ~TCP_RCV_COPY_XID; 1090 transport->tcp_flags |= TCP_RCV_READ_CALLDIR; 1091 transport->tcp_copied = 4; 1092 dprintk("RPC: reading %s XID %08x\n", 1093 (transport->tcp_flags & TCP_RPC_REPLY) ? "reply for" 1094 : "request with", 1095 ntohl(transport->tcp_xid)); 1096 xs_tcp_check_fraghdr(transport); 1097 } 1098 1099 static inline void xs_tcp_read_calldir(struct sock_xprt *transport, 1100 struct xdr_skb_reader *desc) 1101 { 1102 size_t len, used; 1103 u32 offset; 1104 char *p; 1105 1106 /* 1107 * We want transport->tcp_offset to be 8 at the end of this routine 1108 * (4 bytes for the xid and 4 bytes for the call/reply flag). 1109 * When this function is called for the first time, 1110 * transport->tcp_offset is 4 (after having already read the xid). 1111 */ 1112 offset = transport->tcp_offset - sizeof(transport->tcp_xid); 1113 len = sizeof(transport->tcp_calldir) - offset; 1114 dprintk("RPC: reading CALL/REPLY flag (%Zu bytes)\n", len); 1115 p = ((char *) &transport->tcp_calldir) + offset; 1116 used = xdr_skb_read_bits(desc, p, len); 1117 transport->tcp_offset += used; 1118 if (used != len) 1119 return; 1120 transport->tcp_flags &= ~TCP_RCV_READ_CALLDIR; 1121 /* 1122 * We don't yet have the XDR buffer, so we will write the calldir 1123 * out after we get the buffer from the 'struct rpc_rqst' 1124 */ 1125 switch (ntohl(transport->tcp_calldir)) { 1126 case RPC_REPLY: 1127 transport->tcp_flags |= TCP_RCV_COPY_CALLDIR; 1128 transport->tcp_flags |= TCP_RCV_COPY_DATA; 1129 transport->tcp_flags |= TCP_RPC_REPLY; 1130 break; 1131 case RPC_CALL: 1132 transport->tcp_flags |= TCP_RCV_COPY_CALLDIR; 1133 transport->tcp_flags |= TCP_RCV_COPY_DATA; 1134 transport->tcp_flags &= ~TCP_RPC_REPLY; 1135 break; 1136 default: 1137 dprintk("RPC: invalid request message type\n"); 1138 xprt_force_disconnect(&transport->xprt); 1139 } 1140 xs_tcp_check_fraghdr(transport); 1141 } 1142 1143 static inline void xs_tcp_read_common(struct rpc_xprt *xprt, 1144 struct xdr_skb_reader *desc, 1145 struct rpc_rqst *req) 1146 { 1147 struct sock_xprt *transport = 1148 container_of(xprt, struct sock_xprt, xprt); 1149 struct xdr_buf *rcvbuf; 1150 size_t len; 1151 ssize_t r; 1152 1153 rcvbuf = &req->rq_private_buf; 1154 1155 if (transport->tcp_flags & TCP_RCV_COPY_CALLDIR) { 1156 /* 1157 * Save the RPC direction in the XDR buffer 1158 */ 1159 memcpy(rcvbuf->head[0].iov_base + transport->tcp_copied, 1160 &transport->tcp_calldir, 1161 sizeof(transport->tcp_calldir)); 1162 transport->tcp_copied += sizeof(transport->tcp_calldir); 1163 transport->tcp_flags &= ~TCP_RCV_COPY_CALLDIR; 1164 } 1165 1166 len = desc->count; 1167 if (len > transport->tcp_reclen - transport->tcp_offset) { 1168 struct xdr_skb_reader my_desc; 1169 1170 len = transport->tcp_reclen - transport->tcp_offset; 1171 memcpy(&my_desc, desc, sizeof(my_desc)); 1172 my_desc.count = len; 1173 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied, 1174 &my_desc, xdr_skb_read_bits); 1175 desc->count -= r; 1176 desc->offset += r; 1177 } else 1178 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied, 1179 desc, xdr_skb_read_bits); 1180 1181 if (r > 0) { 1182 transport->tcp_copied += r; 1183 transport->tcp_offset += r; 1184 } 1185 if (r != len) { 1186 /* Error when copying to the receive buffer, 1187 * usually because we weren't able to allocate 1188 * additional buffer pages. All we can do now 1189 * is turn off TCP_RCV_COPY_DATA, so the request 1190 * will not receive any additional updates, 1191 * and time out. 1192 * Any remaining data from this record will 1193 * be discarded. 1194 */ 1195 transport->tcp_flags &= ~TCP_RCV_COPY_DATA; 1196 dprintk("RPC: XID %08x truncated request\n", 1197 ntohl(transport->tcp_xid)); 1198 dprintk("RPC: xprt = %p, tcp_copied = %lu, " 1199 "tcp_offset = %u, tcp_reclen = %u\n", 1200 xprt, transport->tcp_copied, 1201 transport->tcp_offset, transport->tcp_reclen); 1202 return; 1203 } 1204 1205 dprintk("RPC: XID %08x read %Zd bytes\n", 1206 ntohl(transport->tcp_xid), r); 1207 dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, " 1208 "tcp_reclen = %u\n", xprt, transport->tcp_copied, 1209 transport->tcp_offset, transport->tcp_reclen); 1210 1211 if (transport->tcp_copied == req->rq_private_buf.buflen) 1212 transport->tcp_flags &= ~TCP_RCV_COPY_DATA; 1213 else if (transport->tcp_offset == transport->tcp_reclen) { 1214 if (transport->tcp_flags & TCP_RCV_LAST_FRAG) 1215 transport->tcp_flags &= ~TCP_RCV_COPY_DATA; 1216 } 1217 } 1218 1219 /* 1220 * Finds the request corresponding to the RPC xid and invokes the common 1221 * tcp read code to read the data. 1222 */ 1223 static inline int xs_tcp_read_reply(struct rpc_xprt *xprt, 1224 struct xdr_skb_reader *desc) 1225 { 1226 struct sock_xprt *transport = 1227 container_of(xprt, struct sock_xprt, xprt); 1228 struct rpc_rqst *req; 1229 1230 dprintk("RPC: read reply XID %08x\n", ntohl(transport->tcp_xid)); 1231 1232 /* Find and lock the request corresponding to this xid */ 1233 spin_lock(&xprt->transport_lock); 1234 req = xprt_lookup_rqst(xprt, transport->tcp_xid); 1235 if (!req) { 1236 dprintk("RPC: XID %08x request not found!\n", 1237 ntohl(transport->tcp_xid)); 1238 spin_unlock(&xprt->transport_lock); 1239 return -1; 1240 } 1241 1242 xs_tcp_read_common(xprt, desc, req); 1243 1244 if (!(transport->tcp_flags & TCP_RCV_COPY_DATA)) 1245 xprt_complete_rqst(req->rq_task, transport->tcp_copied); 1246 1247 spin_unlock(&xprt->transport_lock); 1248 return 0; 1249 } 1250 1251 #if defined(CONFIG_SUNRPC_BACKCHANNEL) 1252 /* 1253 * Obtains an rpc_rqst previously allocated and invokes the common 1254 * tcp read code to read the data. The result is placed in the callback 1255 * queue. 1256 * If we're unable to obtain the rpc_rqst we schedule the closing of the 1257 * connection and return -1. 1258 */ 1259 static inline int xs_tcp_read_callback(struct rpc_xprt *xprt, 1260 struct xdr_skb_reader *desc) 1261 { 1262 struct sock_xprt *transport = 1263 container_of(xprt, struct sock_xprt, xprt); 1264 struct rpc_rqst *req; 1265 1266 req = xprt_alloc_bc_request(xprt); 1267 if (req == NULL) { 1268 printk(KERN_WARNING "Callback slot table overflowed\n"); 1269 xprt_force_disconnect(xprt); 1270 return -1; 1271 } 1272 1273 req->rq_xid = transport->tcp_xid; 1274 dprintk("RPC: read callback XID %08x\n", ntohl(req->rq_xid)); 1275 xs_tcp_read_common(xprt, desc, req); 1276 1277 if (!(transport->tcp_flags & TCP_RCV_COPY_DATA)) { 1278 struct svc_serv *bc_serv = xprt->bc_serv; 1279 1280 /* 1281 * Add callback request to callback list. The callback 1282 * service sleeps on the sv_cb_waitq waiting for new 1283 * requests. Wake it up after adding enqueing the 1284 * request. 1285 */ 1286 dprintk("RPC: add callback request to list\n"); 1287 spin_lock(&bc_serv->sv_cb_lock); 1288 list_add(&req->rq_bc_list, &bc_serv->sv_cb_list); 1289 spin_unlock(&bc_serv->sv_cb_lock); 1290 wake_up(&bc_serv->sv_cb_waitq); 1291 } 1292 1293 req->rq_private_buf.len = transport->tcp_copied; 1294 1295 return 0; 1296 } 1297 1298 static inline int _xs_tcp_read_data(struct rpc_xprt *xprt, 1299 struct xdr_skb_reader *desc) 1300 { 1301 struct sock_xprt *transport = 1302 container_of(xprt, struct sock_xprt, xprt); 1303 1304 return (transport->tcp_flags & TCP_RPC_REPLY) ? 1305 xs_tcp_read_reply(xprt, desc) : 1306 xs_tcp_read_callback(xprt, desc); 1307 } 1308 #else 1309 static inline int _xs_tcp_read_data(struct rpc_xprt *xprt, 1310 struct xdr_skb_reader *desc) 1311 { 1312 return xs_tcp_read_reply(xprt, desc); 1313 } 1314 #endif /* CONFIG_SUNRPC_BACKCHANNEL */ 1315 1316 /* 1317 * Read data off the transport. This can be either an RPC_CALL or an 1318 * RPC_REPLY. Relay the processing to helper functions. 1319 */ 1320 static void xs_tcp_read_data(struct rpc_xprt *xprt, 1321 struct xdr_skb_reader *desc) 1322 { 1323 struct sock_xprt *transport = 1324 container_of(xprt, struct sock_xprt, xprt); 1325 1326 if (_xs_tcp_read_data(xprt, desc) == 0) 1327 xs_tcp_check_fraghdr(transport); 1328 else { 1329 /* 1330 * The transport_lock protects the request handling. 1331 * There's no need to hold it to update the tcp_flags. 1332 */ 1333 transport->tcp_flags &= ~TCP_RCV_COPY_DATA; 1334 } 1335 } 1336 1337 static inline void xs_tcp_read_discard(struct sock_xprt *transport, struct xdr_skb_reader *desc) 1338 { 1339 size_t len; 1340 1341 len = transport->tcp_reclen - transport->tcp_offset; 1342 if (len > desc->count) 1343 len = desc->count; 1344 desc->count -= len; 1345 desc->offset += len; 1346 transport->tcp_offset += len; 1347 dprintk("RPC: discarded %Zu bytes\n", len); 1348 xs_tcp_check_fraghdr(transport); 1349 } 1350 1351 static int xs_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, unsigned int offset, size_t len) 1352 { 1353 struct rpc_xprt *xprt = rd_desc->arg.data; 1354 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1355 struct xdr_skb_reader desc = { 1356 .skb = skb, 1357 .offset = offset, 1358 .count = len, 1359 }; 1360 1361 dprintk("RPC: xs_tcp_data_recv started\n"); 1362 do { 1363 /* Read in a new fragment marker if necessary */ 1364 /* Can we ever really expect to get completely empty fragments? */ 1365 if (transport->tcp_flags & TCP_RCV_COPY_FRAGHDR) { 1366 xs_tcp_read_fraghdr(xprt, &desc); 1367 continue; 1368 } 1369 /* Read in the xid if necessary */ 1370 if (transport->tcp_flags & TCP_RCV_COPY_XID) { 1371 xs_tcp_read_xid(transport, &desc); 1372 continue; 1373 } 1374 /* Read in the call/reply flag */ 1375 if (transport->tcp_flags & TCP_RCV_READ_CALLDIR) { 1376 xs_tcp_read_calldir(transport, &desc); 1377 continue; 1378 } 1379 /* Read in the request data */ 1380 if (transport->tcp_flags & TCP_RCV_COPY_DATA) { 1381 xs_tcp_read_data(xprt, &desc); 1382 continue; 1383 } 1384 /* Skip over any trailing bytes on short reads */ 1385 xs_tcp_read_discard(transport, &desc); 1386 } while (desc.count); 1387 dprintk("RPC: xs_tcp_data_recv done\n"); 1388 return len - desc.count; 1389 } 1390 1391 /** 1392 * xs_tcp_data_ready - "data ready" callback for TCP sockets 1393 * @sk: socket with data to read 1394 * @bytes: how much data to read 1395 * 1396 */ 1397 static void xs_tcp_data_ready(struct sock *sk, int bytes) 1398 { 1399 struct rpc_xprt *xprt; 1400 read_descriptor_t rd_desc; 1401 int read; 1402 1403 dprintk("RPC: xs_tcp_data_ready...\n"); 1404 1405 read_lock_bh(&sk->sk_callback_lock); 1406 if (!(xprt = xprt_from_sock(sk))) 1407 goto out; 1408 if (xprt->shutdown) 1409 goto out; 1410 1411 /* Any data means we had a useful conversation, so 1412 * the we don't need to delay the next reconnect 1413 */ 1414 if (xprt->reestablish_timeout) 1415 xprt->reestablish_timeout = 0; 1416 1417 /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */ 1418 rd_desc.arg.data = xprt; 1419 do { 1420 rd_desc.count = 65536; 1421 read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv); 1422 } while (read > 0); 1423 out: 1424 read_unlock_bh(&sk->sk_callback_lock); 1425 } 1426 1427 /* 1428 * Do the equivalent of linger/linger2 handling for dealing with 1429 * broken servers that don't close the socket in a timely 1430 * fashion 1431 */ 1432 static void xs_tcp_schedule_linger_timeout(struct rpc_xprt *xprt, 1433 unsigned long timeout) 1434 { 1435 struct sock_xprt *transport; 1436 1437 if (xprt_test_and_set_connecting(xprt)) 1438 return; 1439 set_bit(XPRT_CONNECTION_ABORT, &xprt->state); 1440 transport = container_of(xprt, struct sock_xprt, xprt); 1441 queue_delayed_work(rpciod_workqueue, &transport->connect_worker, 1442 timeout); 1443 } 1444 1445 static void xs_tcp_cancel_linger_timeout(struct rpc_xprt *xprt) 1446 { 1447 struct sock_xprt *transport; 1448 1449 transport = container_of(xprt, struct sock_xprt, xprt); 1450 1451 if (!test_bit(XPRT_CONNECTION_ABORT, &xprt->state) || 1452 !cancel_delayed_work(&transport->connect_worker)) 1453 return; 1454 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state); 1455 xprt_clear_connecting(xprt); 1456 } 1457 1458 static void xs_sock_mark_closed(struct rpc_xprt *xprt) 1459 { 1460 smp_mb__before_clear_bit(); 1461 clear_bit(XPRT_CLOSE_WAIT, &xprt->state); 1462 clear_bit(XPRT_CLOSING, &xprt->state); 1463 smp_mb__after_clear_bit(); 1464 /* Mark transport as closed and wake up all pending tasks */ 1465 xprt_disconnect_done(xprt); 1466 } 1467 1468 /** 1469 * xs_tcp_state_change - callback to handle TCP socket state changes 1470 * @sk: socket whose state has changed 1471 * 1472 */ 1473 static void xs_tcp_state_change(struct sock *sk) 1474 { 1475 struct rpc_xprt *xprt; 1476 1477 read_lock_bh(&sk->sk_callback_lock); 1478 if (!(xprt = xprt_from_sock(sk))) 1479 goto out; 1480 dprintk("RPC: xs_tcp_state_change client %p...\n", xprt); 1481 dprintk("RPC: state %x conn %d dead %d zapped %d sk_shutdown %d\n", 1482 sk->sk_state, xprt_connected(xprt), 1483 sock_flag(sk, SOCK_DEAD), 1484 sock_flag(sk, SOCK_ZAPPED), 1485 sk->sk_shutdown); 1486 1487 switch (sk->sk_state) { 1488 case TCP_ESTABLISHED: 1489 spin_lock(&xprt->transport_lock); 1490 if (!xprt_test_and_set_connected(xprt)) { 1491 struct sock_xprt *transport = container_of(xprt, 1492 struct sock_xprt, xprt); 1493 1494 /* Reset TCP record info */ 1495 transport->tcp_offset = 0; 1496 transport->tcp_reclen = 0; 1497 transport->tcp_copied = 0; 1498 transport->tcp_flags = 1499 TCP_RCV_COPY_FRAGHDR | TCP_RCV_COPY_XID; 1500 1501 xprt_wake_pending_tasks(xprt, -EAGAIN); 1502 } 1503 spin_unlock(&xprt->transport_lock); 1504 break; 1505 case TCP_FIN_WAIT1: 1506 /* The client initiated a shutdown of the socket */ 1507 xprt->connect_cookie++; 1508 xprt->reestablish_timeout = 0; 1509 set_bit(XPRT_CLOSING, &xprt->state); 1510 smp_mb__before_clear_bit(); 1511 clear_bit(XPRT_CONNECTED, &xprt->state); 1512 clear_bit(XPRT_CLOSE_WAIT, &xprt->state); 1513 smp_mb__after_clear_bit(); 1514 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout); 1515 break; 1516 case TCP_CLOSE_WAIT: 1517 /* The server initiated a shutdown of the socket */ 1518 xprt_force_disconnect(xprt); 1519 xprt->connect_cookie++; 1520 case TCP_CLOSING: 1521 /* 1522 * If the server closed down the connection, make sure that 1523 * we back off before reconnecting 1524 */ 1525 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO) 1526 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; 1527 break; 1528 case TCP_LAST_ACK: 1529 set_bit(XPRT_CLOSING, &xprt->state); 1530 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout); 1531 smp_mb__before_clear_bit(); 1532 clear_bit(XPRT_CONNECTED, &xprt->state); 1533 smp_mb__after_clear_bit(); 1534 break; 1535 case TCP_CLOSE: 1536 xs_tcp_cancel_linger_timeout(xprt); 1537 xs_sock_mark_closed(xprt); 1538 } 1539 out: 1540 read_unlock_bh(&sk->sk_callback_lock); 1541 } 1542 1543 /** 1544 * xs_error_report - callback mainly for catching socket errors 1545 * @sk: socket 1546 */ 1547 static void xs_error_report(struct sock *sk) 1548 { 1549 struct rpc_xprt *xprt; 1550 1551 read_lock_bh(&sk->sk_callback_lock); 1552 if (!(xprt = xprt_from_sock(sk))) 1553 goto out; 1554 dprintk("RPC: %s client %p...\n" 1555 "RPC: error %d\n", 1556 __func__, xprt, sk->sk_err); 1557 xprt_wake_pending_tasks(xprt, -EAGAIN); 1558 out: 1559 read_unlock_bh(&sk->sk_callback_lock); 1560 } 1561 1562 static void xs_write_space(struct sock *sk) 1563 { 1564 struct socket *sock; 1565 struct rpc_xprt *xprt; 1566 1567 if (unlikely(!(sock = sk->sk_socket))) 1568 return; 1569 clear_bit(SOCK_NOSPACE, &sock->flags); 1570 1571 if (unlikely(!(xprt = xprt_from_sock(sk)))) 1572 return; 1573 if (test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags) == 0) 1574 return; 1575 1576 xprt_write_space(xprt); 1577 } 1578 1579 /** 1580 * xs_udp_write_space - callback invoked when socket buffer space 1581 * becomes available 1582 * @sk: socket whose state has changed 1583 * 1584 * Called when more output buffer space is available for this socket. 1585 * We try not to wake our writers until they can make "significant" 1586 * progress, otherwise we'll waste resources thrashing kernel_sendmsg 1587 * with a bunch of small requests. 1588 */ 1589 static void xs_udp_write_space(struct sock *sk) 1590 { 1591 read_lock_bh(&sk->sk_callback_lock); 1592 1593 /* from net/core/sock.c:sock_def_write_space */ 1594 if (sock_writeable(sk)) 1595 xs_write_space(sk); 1596 1597 read_unlock_bh(&sk->sk_callback_lock); 1598 } 1599 1600 /** 1601 * xs_tcp_write_space - callback invoked when socket buffer space 1602 * becomes available 1603 * @sk: socket whose state has changed 1604 * 1605 * Called when more output buffer space is available for this socket. 1606 * We try not to wake our writers until they can make "significant" 1607 * progress, otherwise we'll waste resources thrashing kernel_sendmsg 1608 * with a bunch of small requests. 1609 */ 1610 static void xs_tcp_write_space(struct sock *sk) 1611 { 1612 read_lock_bh(&sk->sk_callback_lock); 1613 1614 /* from net/core/stream.c:sk_stream_write_space */ 1615 if (sk_stream_wspace(sk) >= sk_stream_min_wspace(sk)) 1616 xs_write_space(sk); 1617 1618 read_unlock_bh(&sk->sk_callback_lock); 1619 } 1620 1621 static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt) 1622 { 1623 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1624 struct sock *sk = transport->inet; 1625 1626 if (transport->rcvsize) { 1627 sk->sk_userlocks |= SOCK_RCVBUF_LOCK; 1628 sk->sk_rcvbuf = transport->rcvsize * xprt->max_reqs * 2; 1629 } 1630 if (transport->sndsize) { 1631 sk->sk_userlocks |= SOCK_SNDBUF_LOCK; 1632 sk->sk_sndbuf = transport->sndsize * xprt->max_reqs * 2; 1633 sk->sk_write_space(sk); 1634 } 1635 } 1636 1637 /** 1638 * xs_udp_set_buffer_size - set send and receive limits 1639 * @xprt: generic transport 1640 * @sndsize: requested size of send buffer, in bytes 1641 * @rcvsize: requested size of receive buffer, in bytes 1642 * 1643 * Set socket send and receive buffer size limits. 1644 */ 1645 static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize) 1646 { 1647 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1648 1649 transport->sndsize = 0; 1650 if (sndsize) 1651 transport->sndsize = sndsize + 1024; 1652 transport->rcvsize = 0; 1653 if (rcvsize) 1654 transport->rcvsize = rcvsize + 1024; 1655 1656 xs_udp_do_set_buffer_size(xprt); 1657 } 1658 1659 /** 1660 * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport 1661 * @task: task that timed out 1662 * 1663 * Adjust the congestion window after a retransmit timeout has occurred. 1664 */ 1665 static void xs_udp_timer(struct rpc_task *task) 1666 { 1667 xprt_adjust_cwnd(task, -ETIMEDOUT); 1668 } 1669 1670 static unsigned short xs_get_random_port(void) 1671 { 1672 unsigned short range = xprt_max_resvport - xprt_min_resvport; 1673 unsigned short rand = (unsigned short) net_random() % range; 1674 return rand + xprt_min_resvport; 1675 } 1676 1677 /** 1678 * xs_set_port - reset the port number in the remote endpoint address 1679 * @xprt: generic transport 1680 * @port: new port number 1681 * 1682 */ 1683 static void xs_set_port(struct rpc_xprt *xprt, unsigned short port) 1684 { 1685 dprintk("RPC: setting port for xprt %p to %u\n", xprt, port); 1686 1687 rpc_set_port(xs_addr(xprt), port); 1688 xs_update_peer_port(xprt); 1689 } 1690 1691 static unsigned short xs_get_srcport(struct sock_xprt *transport) 1692 { 1693 unsigned short port = transport->srcport; 1694 1695 if (port == 0 && transport->xprt.resvport) 1696 port = xs_get_random_port(); 1697 return port; 1698 } 1699 1700 static unsigned short xs_next_srcport(struct sock_xprt *transport, unsigned short port) 1701 { 1702 if (transport->srcport != 0) 1703 transport->srcport = 0; 1704 if (!transport->xprt.resvport) 1705 return 0; 1706 if (port <= xprt_min_resvport || port > xprt_max_resvport) 1707 return xprt_max_resvport; 1708 return --port; 1709 } 1710 static int xs_bind(struct sock_xprt *transport, struct socket *sock) 1711 { 1712 struct sockaddr_storage myaddr; 1713 int err, nloop = 0; 1714 unsigned short port = xs_get_srcport(transport); 1715 unsigned short last; 1716 1717 memcpy(&myaddr, &transport->srcaddr, transport->xprt.addrlen); 1718 do { 1719 rpc_set_port((struct sockaddr *)&myaddr, port); 1720 err = kernel_bind(sock, (struct sockaddr *)&myaddr, 1721 transport->xprt.addrlen); 1722 if (port == 0) 1723 break; 1724 if (err == 0) { 1725 transport->srcport = port; 1726 break; 1727 } 1728 last = port; 1729 port = xs_next_srcport(transport, port); 1730 if (port > last) 1731 nloop++; 1732 } while (err == -EADDRINUSE && nloop != 2); 1733 1734 if (myaddr.ss_family == AF_INET) 1735 dprintk("RPC: %s %pI4:%u: %s (%d)\n", __func__, 1736 &((struct sockaddr_in *)&myaddr)->sin_addr, 1737 port, err ? "failed" : "ok", err); 1738 else 1739 dprintk("RPC: %s %pI6:%u: %s (%d)\n", __func__, 1740 &((struct sockaddr_in6 *)&myaddr)->sin6_addr, 1741 port, err ? "failed" : "ok", err); 1742 return err; 1743 } 1744 1745 /* 1746 * We don't support autobind on AF_LOCAL sockets 1747 */ 1748 static void xs_local_rpcbind(struct rpc_task *task) 1749 { 1750 xprt_set_bound(task->tk_xprt); 1751 } 1752 1753 static void xs_local_set_port(struct rpc_xprt *xprt, unsigned short port) 1754 { 1755 } 1756 1757 #ifdef CONFIG_DEBUG_LOCK_ALLOC 1758 static struct lock_class_key xs_key[2]; 1759 static struct lock_class_key xs_slock_key[2]; 1760 1761 static inline void xs_reclassify_socketu(struct socket *sock) 1762 { 1763 struct sock *sk = sock->sk; 1764 1765 BUG_ON(sock_owned_by_user(sk)); 1766 sock_lock_init_class_and_name(sk, "slock-AF_LOCAL-RPC", 1767 &xs_slock_key[1], "sk_lock-AF_LOCAL-RPC", &xs_key[1]); 1768 } 1769 1770 static inline void xs_reclassify_socket4(struct socket *sock) 1771 { 1772 struct sock *sk = sock->sk; 1773 1774 BUG_ON(sock_owned_by_user(sk)); 1775 sock_lock_init_class_and_name(sk, "slock-AF_INET-RPC", 1776 &xs_slock_key[0], "sk_lock-AF_INET-RPC", &xs_key[0]); 1777 } 1778 1779 static inline void xs_reclassify_socket6(struct socket *sock) 1780 { 1781 struct sock *sk = sock->sk; 1782 1783 BUG_ON(sock_owned_by_user(sk)); 1784 sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC", 1785 &xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]); 1786 } 1787 1788 static inline void xs_reclassify_socket(int family, struct socket *sock) 1789 { 1790 switch (family) { 1791 case AF_LOCAL: 1792 xs_reclassify_socketu(sock); 1793 break; 1794 case AF_INET: 1795 xs_reclassify_socket4(sock); 1796 break; 1797 case AF_INET6: 1798 xs_reclassify_socket6(sock); 1799 break; 1800 } 1801 } 1802 #else 1803 static inline void xs_reclassify_socketu(struct socket *sock) 1804 { 1805 } 1806 1807 static inline void xs_reclassify_socket4(struct socket *sock) 1808 { 1809 } 1810 1811 static inline void xs_reclassify_socket6(struct socket *sock) 1812 { 1813 } 1814 1815 static inline void xs_reclassify_socket(int family, struct socket *sock) 1816 { 1817 } 1818 #endif 1819 1820 static struct socket *xs_create_sock(struct rpc_xprt *xprt, 1821 struct sock_xprt *transport, int family, int type, int protocol) 1822 { 1823 struct socket *sock; 1824 int err; 1825 1826 err = __sock_create(xprt->xprt_net, family, type, protocol, &sock, 1); 1827 if (err < 0) { 1828 dprintk("RPC: can't create %d transport socket (%d).\n", 1829 protocol, -err); 1830 goto out; 1831 } 1832 xs_reclassify_socket(family, sock); 1833 1834 err = xs_bind(transport, sock); 1835 if (err) { 1836 sock_release(sock); 1837 goto out; 1838 } 1839 1840 return sock; 1841 out: 1842 return ERR_PTR(err); 1843 } 1844 1845 static int xs_local_finish_connecting(struct rpc_xprt *xprt, 1846 struct socket *sock) 1847 { 1848 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, 1849 xprt); 1850 1851 if (!transport->inet) { 1852 struct sock *sk = sock->sk; 1853 1854 write_lock_bh(&sk->sk_callback_lock); 1855 1856 xs_save_old_callbacks(transport, sk); 1857 1858 sk->sk_user_data = xprt; 1859 sk->sk_data_ready = xs_local_data_ready; 1860 sk->sk_write_space = xs_udp_write_space; 1861 sk->sk_error_report = xs_error_report; 1862 sk->sk_allocation = GFP_ATOMIC; 1863 1864 xprt_clear_connected(xprt); 1865 1866 /* Reset to new socket */ 1867 transport->sock = sock; 1868 transport->inet = sk; 1869 1870 write_unlock_bh(&sk->sk_callback_lock); 1871 } 1872 1873 /* Tell the socket layer to start connecting... */ 1874 xprt->stat.connect_count++; 1875 xprt->stat.connect_start = jiffies; 1876 return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, 0); 1877 } 1878 1879 /** 1880 * xs_local_setup_socket - create AF_LOCAL socket, connect to a local endpoint 1881 * @xprt: RPC transport to connect 1882 * @transport: socket transport to connect 1883 * @create_sock: function to create a socket of the correct type 1884 * 1885 * Invoked by a work queue tasklet. 1886 */ 1887 static void xs_local_setup_socket(struct work_struct *work) 1888 { 1889 struct sock_xprt *transport = 1890 container_of(work, struct sock_xprt, connect_worker.work); 1891 struct rpc_xprt *xprt = &transport->xprt; 1892 struct socket *sock; 1893 int status = -EIO; 1894 1895 if (xprt->shutdown) 1896 goto out; 1897 1898 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state); 1899 status = __sock_create(xprt->xprt_net, AF_LOCAL, 1900 SOCK_STREAM, 0, &sock, 1); 1901 if (status < 0) { 1902 dprintk("RPC: can't create AF_LOCAL " 1903 "transport socket (%d).\n", -status); 1904 goto out; 1905 } 1906 xs_reclassify_socketu(sock); 1907 1908 dprintk("RPC: worker connecting xprt %p via AF_LOCAL to %s\n", 1909 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]); 1910 1911 status = xs_local_finish_connecting(xprt, sock); 1912 switch (status) { 1913 case 0: 1914 dprintk("RPC: xprt %p connected to %s\n", 1915 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]); 1916 xprt_set_connected(xprt); 1917 break; 1918 case -ENOENT: 1919 dprintk("RPC: xprt %p: socket %s does not exist\n", 1920 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]); 1921 break; 1922 default: 1923 printk(KERN_ERR "%s: unhandled error (%d) connecting to %s\n", 1924 __func__, -status, 1925 xprt->address_strings[RPC_DISPLAY_ADDR]); 1926 } 1927 1928 out: 1929 xprt_clear_connecting(xprt); 1930 xprt_wake_pending_tasks(xprt, status); 1931 } 1932 1933 static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock) 1934 { 1935 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1936 1937 if (!transport->inet) { 1938 struct sock *sk = sock->sk; 1939 1940 write_lock_bh(&sk->sk_callback_lock); 1941 1942 xs_save_old_callbacks(transport, sk); 1943 1944 sk->sk_user_data = xprt; 1945 sk->sk_data_ready = xs_udp_data_ready; 1946 sk->sk_write_space = xs_udp_write_space; 1947 sk->sk_error_report = xs_error_report; 1948 sk->sk_no_check = UDP_CSUM_NORCV; 1949 sk->sk_allocation = GFP_ATOMIC; 1950 1951 xprt_set_connected(xprt); 1952 1953 /* Reset to new socket */ 1954 transport->sock = sock; 1955 transport->inet = sk; 1956 1957 write_unlock_bh(&sk->sk_callback_lock); 1958 } 1959 xs_udp_do_set_buffer_size(xprt); 1960 } 1961 1962 static void xs_udp_setup_socket(struct work_struct *work) 1963 { 1964 struct sock_xprt *transport = 1965 container_of(work, struct sock_xprt, connect_worker.work); 1966 struct rpc_xprt *xprt = &transport->xprt; 1967 struct socket *sock = transport->sock; 1968 int status = -EIO; 1969 1970 if (xprt->shutdown) 1971 goto out; 1972 1973 /* Start by resetting any existing state */ 1974 xs_reset_transport(transport); 1975 sock = xs_create_sock(xprt, transport, 1976 xs_addr(xprt)->sa_family, SOCK_DGRAM, IPPROTO_UDP); 1977 if (IS_ERR(sock)) 1978 goto out; 1979 1980 dprintk("RPC: worker connecting xprt %p via %s to " 1981 "%s (port %s)\n", xprt, 1982 xprt->address_strings[RPC_DISPLAY_PROTO], 1983 xprt->address_strings[RPC_DISPLAY_ADDR], 1984 xprt->address_strings[RPC_DISPLAY_PORT]); 1985 1986 xs_udp_finish_connecting(xprt, sock); 1987 status = 0; 1988 out: 1989 xprt_clear_connecting(xprt); 1990 xprt_wake_pending_tasks(xprt, status); 1991 } 1992 1993 /* 1994 * We need to preserve the port number so the reply cache on the server can 1995 * find our cached RPC replies when we get around to reconnecting. 1996 */ 1997 static void xs_abort_connection(struct sock_xprt *transport) 1998 { 1999 int result; 2000 struct sockaddr any; 2001 2002 dprintk("RPC: disconnecting xprt %p to reuse port\n", transport); 2003 2004 /* 2005 * Disconnect the transport socket by doing a connect operation 2006 * with AF_UNSPEC. This should return immediately... 2007 */ 2008 memset(&any, 0, sizeof(any)); 2009 any.sa_family = AF_UNSPEC; 2010 result = kernel_connect(transport->sock, &any, sizeof(any), 0); 2011 if (!result) 2012 xs_sock_mark_closed(&transport->xprt); 2013 else 2014 dprintk("RPC: AF_UNSPEC connect return code %d\n", 2015 result); 2016 } 2017 2018 static void xs_tcp_reuse_connection(struct sock_xprt *transport) 2019 { 2020 unsigned int state = transport->inet->sk_state; 2021 2022 if (state == TCP_CLOSE && transport->sock->state == SS_UNCONNECTED) { 2023 /* we don't need to abort the connection if the socket 2024 * hasn't undergone a shutdown 2025 */ 2026 if (transport->inet->sk_shutdown == 0) 2027 return; 2028 dprintk("RPC: %s: TCP_CLOSEd and sk_shutdown set to %d\n", 2029 __func__, transport->inet->sk_shutdown); 2030 } 2031 if ((1 << state) & (TCPF_ESTABLISHED|TCPF_SYN_SENT)) { 2032 /* we don't need to abort the connection if the socket 2033 * hasn't undergone a shutdown 2034 */ 2035 if (transport->inet->sk_shutdown == 0) 2036 return; 2037 dprintk("RPC: %s: ESTABLISHED/SYN_SENT " 2038 "sk_shutdown set to %d\n", 2039 __func__, transport->inet->sk_shutdown); 2040 } 2041 xs_abort_connection(transport); 2042 } 2043 2044 static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock) 2045 { 2046 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 2047 int ret = -ENOTCONN; 2048 2049 if (!transport->inet) { 2050 struct sock *sk = sock->sk; 2051 2052 write_lock_bh(&sk->sk_callback_lock); 2053 2054 xs_save_old_callbacks(transport, sk); 2055 2056 sk->sk_user_data = xprt; 2057 sk->sk_data_ready = xs_tcp_data_ready; 2058 sk->sk_state_change = xs_tcp_state_change; 2059 sk->sk_write_space = xs_tcp_write_space; 2060 sk->sk_error_report = xs_error_report; 2061 sk->sk_allocation = GFP_ATOMIC; 2062 2063 /* socket options */ 2064 sk->sk_userlocks |= SOCK_BINDPORT_LOCK; 2065 sock_reset_flag(sk, SOCK_LINGER); 2066 tcp_sk(sk)->linger2 = 0; 2067 tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF; 2068 2069 xprt_clear_connected(xprt); 2070 2071 /* Reset to new socket */ 2072 transport->sock = sock; 2073 transport->inet = sk; 2074 2075 write_unlock_bh(&sk->sk_callback_lock); 2076 } 2077 2078 if (!xprt_bound(xprt)) 2079 goto out; 2080 2081 /* Tell the socket layer to start connecting... */ 2082 xprt->stat.connect_count++; 2083 xprt->stat.connect_start = jiffies; 2084 ret = kernel_connect(sock, xs_addr(xprt), xprt->addrlen, O_NONBLOCK); 2085 switch (ret) { 2086 case 0: 2087 case -EINPROGRESS: 2088 /* SYN_SENT! */ 2089 xprt->connect_cookie++; 2090 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO) 2091 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; 2092 } 2093 out: 2094 return ret; 2095 } 2096 2097 /** 2098 * xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint 2099 * @xprt: RPC transport to connect 2100 * @transport: socket transport to connect 2101 * @create_sock: function to create a socket of the correct type 2102 * 2103 * Invoked by a work queue tasklet. 2104 */ 2105 static void xs_tcp_setup_socket(struct work_struct *work) 2106 { 2107 struct sock_xprt *transport = 2108 container_of(work, struct sock_xprt, connect_worker.work); 2109 struct socket *sock = transport->sock; 2110 struct rpc_xprt *xprt = &transport->xprt; 2111 int status = -EIO; 2112 2113 if (xprt->shutdown) 2114 goto out; 2115 2116 if (!sock) { 2117 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state); 2118 sock = xs_create_sock(xprt, transport, 2119 xs_addr(xprt)->sa_family, SOCK_STREAM, IPPROTO_TCP); 2120 if (IS_ERR(sock)) { 2121 status = PTR_ERR(sock); 2122 goto out; 2123 } 2124 } else { 2125 int abort_and_exit; 2126 2127 abort_and_exit = test_and_clear_bit(XPRT_CONNECTION_ABORT, 2128 &xprt->state); 2129 /* "close" the socket, preserving the local port */ 2130 xs_tcp_reuse_connection(transport); 2131 2132 if (abort_and_exit) 2133 goto out_eagain; 2134 } 2135 2136 dprintk("RPC: worker connecting xprt %p via %s to " 2137 "%s (port %s)\n", xprt, 2138 xprt->address_strings[RPC_DISPLAY_PROTO], 2139 xprt->address_strings[RPC_DISPLAY_ADDR], 2140 xprt->address_strings[RPC_DISPLAY_PORT]); 2141 2142 status = xs_tcp_finish_connecting(xprt, sock); 2143 dprintk("RPC: %p connect status %d connected %d sock state %d\n", 2144 xprt, -status, xprt_connected(xprt), 2145 sock->sk->sk_state); 2146 switch (status) { 2147 default: 2148 printk("%s: connect returned unhandled error %d\n", 2149 __func__, status); 2150 case -EADDRNOTAVAIL: 2151 /* We're probably in TIME_WAIT. Get rid of existing socket, 2152 * and retry 2153 */ 2154 set_bit(XPRT_CONNECTION_CLOSE, &xprt->state); 2155 xprt_force_disconnect(xprt); 2156 break; 2157 case -ECONNREFUSED: 2158 case -ECONNRESET: 2159 case -ENETUNREACH: 2160 /* retry with existing socket, after a delay */ 2161 case 0: 2162 case -EINPROGRESS: 2163 case -EALREADY: 2164 xprt_clear_connecting(xprt); 2165 return; 2166 case -EINVAL: 2167 /* Happens, for instance, if the user specified a link 2168 * local IPv6 address without a scope-id. 2169 */ 2170 goto out; 2171 } 2172 out_eagain: 2173 status = -EAGAIN; 2174 out: 2175 xprt_clear_connecting(xprt); 2176 xprt_wake_pending_tasks(xprt, status); 2177 } 2178 2179 /** 2180 * xs_connect - connect a socket to a remote endpoint 2181 * @task: address of RPC task that manages state of connect request 2182 * 2183 * TCP: If the remote end dropped the connection, delay reconnecting. 2184 * 2185 * UDP socket connects are synchronous, but we use a work queue anyway 2186 * to guarantee that even unprivileged user processes can set up a 2187 * socket on a privileged port. 2188 * 2189 * If a UDP socket connect fails, the delay behavior here prevents 2190 * retry floods (hard mounts). 2191 */ 2192 static void xs_connect(struct rpc_task *task) 2193 { 2194 struct rpc_xprt *xprt = task->tk_xprt; 2195 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 2196 2197 if (transport->sock != NULL && !RPC_IS_SOFTCONN(task)) { 2198 dprintk("RPC: xs_connect delayed xprt %p for %lu " 2199 "seconds\n", 2200 xprt, xprt->reestablish_timeout / HZ); 2201 queue_delayed_work(rpciod_workqueue, 2202 &transport->connect_worker, 2203 xprt->reestablish_timeout); 2204 xprt->reestablish_timeout <<= 1; 2205 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO) 2206 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; 2207 if (xprt->reestablish_timeout > XS_TCP_MAX_REEST_TO) 2208 xprt->reestablish_timeout = XS_TCP_MAX_REEST_TO; 2209 } else { 2210 dprintk("RPC: xs_connect scheduled xprt %p\n", xprt); 2211 queue_delayed_work(rpciod_workqueue, 2212 &transport->connect_worker, 0); 2213 } 2214 } 2215 2216 /** 2217 * xs_local_print_stats - display AF_LOCAL socket-specifc stats 2218 * @xprt: rpc_xprt struct containing statistics 2219 * @seq: output file 2220 * 2221 */ 2222 static void xs_local_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) 2223 { 2224 long idle_time = 0; 2225 2226 if (xprt_connected(xprt)) 2227 idle_time = (long)(jiffies - xprt->last_used) / HZ; 2228 2229 seq_printf(seq, "\txprt:\tlocal %lu %lu %lu %ld %lu %lu %lu " 2230 "%llu %llu %lu %llu %llu\n", 2231 xprt->stat.bind_count, 2232 xprt->stat.connect_count, 2233 xprt->stat.connect_time, 2234 idle_time, 2235 xprt->stat.sends, 2236 xprt->stat.recvs, 2237 xprt->stat.bad_xids, 2238 xprt->stat.req_u, 2239 xprt->stat.bklog_u, 2240 xprt->stat.max_slots, 2241 xprt->stat.sending_u, 2242 xprt->stat.pending_u); 2243 } 2244 2245 /** 2246 * xs_udp_print_stats - display UDP socket-specifc stats 2247 * @xprt: rpc_xprt struct containing statistics 2248 * @seq: output file 2249 * 2250 */ 2251 static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) 2252 { 2253 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 2254 2255 seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %llu %llu " 2256 "%lu %llu %llu\n", 2257 transport->srcport, 2258 xprt->stat.bind_count, 2259 xprt->stat.sends, 2260 xprt->stat.recvs, 2261 xprt->stat.bad_xids, 2262 xprt->stat.req_u, 2263 xprt->stat.bklog_u, 2264 xprt->stat.max_slots, 2265 xprt->stat.sending_u, 2266 xprt->stat.pending_u); 2267 } 2268 2269 /** 2270 * xs_tcp_print_stats - display TCP socket-specifc stats 2271 * @xprt: rpc_xprt struct containing statistics 2272 * @seq: output file 2273 * 2274 */ 2275 static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) 2276 { 2277 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 2278 long idle_time = 0; 2279 2280 if (xprt_connected(xprt)) 2281 idle_time = (long)(jiffies - xprt->last_used) / HZ; 2282 2283 seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu " 2284 "%llu %llu %lu %llu %llu\n", 2285 transport->srcport, 2286 xprt->stat.bind_count, 2287 xprt->stat.connect_count, 2288 xprt->stat.connect_time, 2289 idle_time, 2290 xprt->stat.sends, 2291 xprt->stat.recvs, 2292 xprt->stat.bad_xids, 2293 xprt->stat.req_u, 2294 xprt->stat.bklog_u, 2295 xprt->stat.max_slots, 2296 xprt->stat.sending_u, 2297 xprt->stat.pending_u); 2298 } 2299 2300 /* 2301 * Allocate a bunch of pages for a scratch buffer for the rpc code. The reason 2302 * we allocate pages instead doing a kmalloc like rpc_malloc is because we want 2303 * to use the server side send routines. 2304 */ 2305 static void *bc_malloc(struct rpc_task *task, size_t size) 2306 { 2307 struct page *page; 2308 struct rpc_buffer *buf; 2309 2310 BUG_ON(size > PAGE_SIZE - sizeof(struct rpc_buffer)); 2311 page = alloc_page(GFP_KERNEL); 2312 2313 if (!page) 2314 return NULL; 2315 2316 buf = page_address(page); 2317 buf->len = PAGE_SIZE; 2318 2319 return buf->data; 2320 } 2321 2322 /* 2323 * Free the space allocated in the bc_alloc routine 2324 */ 2325 static void bc_free(void *buffer) 2326 { 2327 struct rpc_buffer *buf; 2328 2329 if (!buffer) 2330 return; 2331 2332 buf = container_of(buffer, struct rpc_buffer, data); 2333 free_page((unsigned long)buf); 2334 } 2335 2336 /* 2337 * Use the svc_sock to send the callback. Must be called with svsk->sk_mutex 2338 * held. Borrows heavily from svc_tcp_sendto and xs_tcp_send_request. 2339 */ 2340 static int bc_sendto(struct rpc_rqst *req) 2341 { 2342 int len; 2343 struct xdr_buf *xbufp = &req->rq_snd_buf; 2344 struct rpc_xprt *xprt = req->rq_xprt; 2345 struct sock_xprt *transport = 2346 container_of(xprt, struct sock_xprt, xprt); 2347 struct socket *sock = transport->sock; 2348 unsigned long headoff; 2349 unsigned long tailoff; 2350 2351 xs_encode_stream_record_marker(xbufp); 2352 2353 tailoff = (unsigned long)xbufp->tail[0].iov_base & ~PAGE_MASK; 2354 headoff = (unsigned long)xbufp->head[0].iov_base & ~PAGE_MASK; 2355 len = svc_send_common(sock, xbufp, 2356 virt_to_page(xbufp->head[0].iov_base), headoff, 2357 xbufp->tail[0].iov_base, tailoff); 2358 2359 if (len != xbufp->len) { 2360 printk(KERN_NOTICE "Error sending entire callback!\n"); 2361 len = -EAGAIN; 2362 } 2363 2364 return len; 2365 } 2366 2367 /* 2368 * The send routine. Borrows from svc_send 2369 */ 2370 static int bc_send_request(struct rpc_task *task) 2371 { 2372 struct rpc_rqst *req = task->tk_rqstp; 2373 struct svc_xprt *xprt; 2374 struct svc_sock *svsk; 2375 u32 len; 2376 2377 dprintk("sending request with xid: %08x\n", ntohl(req->rq_xid)); 2378 /* 2379 * Get the server socket associated with this callback xprt 2380 */ 2381 xprt = req->rq_xprt->bc_xprt; 2382 svsk = container_of(xprt, struct svc_sock, sk_xprt); 2383 2384 /* 2385 * Grab the mutex to serialize data as the connection is shared 2386 * with the fore channel 2387 */ 2388 if (!mutex_trylock(&xprt->xpt_mutex)) { 2389 rpc_sleep_on(&xprt->xpt_bc_pending, task, NULL); 2390 if (!mutex_trylock(&xprt->xpt_mutex)) 2391 return -EAGAIN; 2392 rpc_wake_up_queued_task(&xprt->xpt_bc_pending, task); 2393 } 2394 if (test_bit(XPT_DEAD, &xprt->xpt_flags)) 2395 len = -ENOTCONN; 2396 else 2397 len = bc_sendto(req); 2398 mutex_unlock(&xprt->xpt_mutex); 2399 2400 if (len > 0) 2401 len = 0; 2402 2403 return len; 2404 } 2405 2406 /* 2407 * The close routine. Since this is client initiated, we do nothing 2408 */ 2409 2410 static void bc_close(struct rpc_xprt *xprt) 2411 { 2412 } 2413 2414 /* 2415 * The xprt destroy routine. Again, because this connection is client 2416 * initiated, we do nothing 2417 */ 2418 2419 static void bc_destroy(struct rpc_xprt *xprt) 2420 { 2421 } 2422 2423 static struct rpc_xprt_ops xs_local_ops = { 2424 .reserve_xprt = xprt_reserve_xprt, 2425 .release_xprt = xs_tcp_release_xprt, 2426 .rpcbind = xs_local_rpcbind, 2427 .set_port = xs_local_set_port, 2428 .connect = xs_connect, 2429 .buf_alloc = rpc_malloc, 2430 .buf_free = rpc_free, 2431 .send_request = xs_local_send_request, 2432 .set_retrans_timeout = xprt_set_retrans_timeout_def, 2433 .close = xs_close, 2434 .destroy = xs_destroy, 2435 .print_stats = xs_local_print_stats, 2436 }; 2437 2438 static struct rpc_xprt_ops xs_udp_ops = { 2439 .set_buffer_size = xs_udp_set_buffer_size, 2440 .reserve_xprt = xprt_reserve_xprt_cong, 2441 .release_xprt = xprt_release_xprt_cong, 2442 .rpcbind = rpcb_getport_async, 2443 .set_port = xs_set_port, 2444 .connect = xs_connect, 2445 .buf_alloc = rpc_malloc, 2446 .buf_free = rpc_free, 2447 .send_request = xs_udp_send_request, 2448 .set_retrans_timeout = xprt_set_retrans_timeout_rtt, 2449 .timer = xs_udp_timer, 2450 .release_request = xprt_release_rqst_cong, 2451 .close = xs_close, 2452 .destroy = xs_destroy, 2453 .print_stats = xs_udp_print_stats, 2454 }; 2455 2456 static struct rpc_xprt_ops xs_tcp_ops = { 2457 .reserve_xprt = xprt_reserve_xprt, 2458 .release_xprt = xs_tcp_release_xprt, 2459 .rpcbind = rpcb_getport_async, 2460 .set_port = xs_set_port, 2461 .connect = xs_connect, 2462 .buf_alloc = rpc_malloc, 2463 .buf_free = rpc_free, 2464 .send_request = xs_tcp_send_request, 2465 .set_retrans_timeout = xprt_set_retrans_timeout_def, 2466 .close = xs_tcp_close, 2467 .destroy = xs_destroy, 2468 .print_stats = xs_tcp_print_stats, 2469 }; 2470 2471 /* 2472 * The rpc_xprt_ops for the server backchannel 2473 */ 2474 2475 static struct rpc_xprt_ops bc_tcp_ops = { 2476 .reserve_xprt = xprt_reserve_xprt, 2477 .release_xprt = xprt_release_xprt, 2478 .rpcbind = xs_local_rpcbind, 2479 .buf_alloc = bc_malloc, 2480 .buf_free = bc_free, 2481 .send_request = bc_send_request, 2482 .set_retrans_timeout = xprt_set_retrans_timeout_def, 2483 .close = bc_close, 2484 .destroy = bc_destroy, 2485 .print_stats = xs_tcp_print_stats, 2486 }; 2487 2488 static int xs_init_anyaddr(const int family, struct sockaddr *sap) 2489 { 2490 static const struct sockaddr_in sin = { 2491 .sin_family = AF_INET, 2492 .sin_addr.s_addr = htonl(INADDR_ANY), 2493 }; 2494 static const struct sockaddr_in6 sin6 = { 2495 .sin6_family = AF_INET6, 2496 .sin6_addr = IN6ADDR_ANY_INIT, 2497 }; 2498 2499 switch (family) { 2500 case AF_LOCAL: 2501 break; 2502 case AF_INET: 2503 memcpy(sap, &sin, sizeof(sin)); 2504 break; 2505 case AF_INET6: 2506 memcpy(sap, &sin6, sizeof(sin6)); 2507 break; 2508 default: 2509 dprintk("RPC: %s: Bad address family\n", __func__); 2510 return -EAFNOSUPPORT; 2511 } 2512 return 0; 2513 } 2514 2515 static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args, 2516 unsigned int slot_table_size, 2517 unsigned int max_slot_table_size) 2518 { 2519 struct rpc_xprt *xprt; 2520 struct sock_xprt *new; 2521 2522 if (args->addrlen > sizeof(xprt->addr)) { 2523 dprintk("RPC: xs_setup_xprt: address too large\n"); 2524 return ERR_PTR(-EBADF); 2525 } 2526 2527 xprt = xprt_alloc(args->net, sizeof(*new), slot_table_size, 2528 max_slot_table_size); 2529 if (xprt == NULL) { 2530 dprintk("RPC: xs_setup_xprt: couldn't allocate " 2531 "rpc_xprt\n"); 2532 return ERR_PTR(-ENOMEM); 2533 } 2534 2535 new = container_of(xprt, struct sock_xprt, xprt); 2536 memcpy(&xprt->addr, args->dstaddr, args->addrlen); 2537 xprt->addrlen = args->addrlen; 2538 if (args->srcaddr) 2539 memcpy(&new->srcaddr, args->srcaddr, args->addrlen); 2540 else { 2541 int err; 2542 err = xs_init_anyaddr(args->dstaddr->sa_family, 2543 (struct sockaddr *)&new->srcaddr); 2544 if (err != 0) { 2545 xprt_free(xprt); 2546 return ERR_PTR(err); 2547 } 2548 } 2549 2550 return xprt; 2551 } 2552 2553 static const struct rpc_timeout xs_local_default_timeout = { 2554 .to_initval = 10 * HZ, 2555 .to_maxval = 10 * HZ, 2556 .to_retries = 2, 2557 }; 2558 2559 /** 2560 * xs_setup_local - Set up transport to use an AF_LOCAL socket 2561 * @args: rpc transport creation arguments 2562 * 2563 * AF_LOCAL is a "tpi_cots_ord" transport, just like TCP 2564 */ 2565 static struct rpc_xprt *xs_setup_local(struct xprt_create *args) 2566 { 2567 struct sockaddr_un *sun = (struct sockaddr_un *)args->dstaddr; 2568 struct sock_xprt *transport; 2569 struct rpc_xprt *xprt; 2570 struct rpc_xprt *ret; 2571 2572 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries, 2573 xprt_max_tcp_slot_table_entries); 2574 if (IS_ERR(xprt)) 2575 return xprt; 2576 transport = container_of(xprt, struct sock_xprt, xprt); 2577 2578 xprt->prot = 0; 2579 xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32); 2580 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE; 2581 2582 xprt->bind_timeout = XS_BIND_TO; 2583 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; 2584 xprt->idle_timeout = XS_IDLE_DISC_TO; 2585 2586 xprt->ops = &xs_local_ops; 2587 xprt->timeout = &xs_local_default_timeout; 2588 2589 switch (sun->sun_family) { 2590 case AF_LOCAL: 2591 if (sun->sun_path[0] != '/') { 2592 dprintk("RPC: bad AF_LOCAL address: %s\n", 2593 sun->sun_path); 2594 ret = ERR_PTR(-EINVAL); 2595 goto out_err; 2596 } 2597 xprt_set_bound(xprt); 2598 INIT_DELAYED_WORK(&transport->connect_worker, 2599 xs_local_setup_socket); 2600 xs_format_peer_addresses(xprt, "local", RPCBIND_NETID_LOCAL); 2601 break; 2602 default: 2603 ret = ERR_PTR(-EAFNOSUPPORT); 2604 goto out_err; 2605 } 2606 2607 dprintk("RPC: set up xprt to %s via AF_LOCAL\n", 2608 xprt->address_strings[RPC_DISPLAY_ADDR]); 2609 2610 if (try_module_get(THIS_MODULE)) 2611 return xprt; 2612 ret = ERR_PTR(-EINVAL); 2613 out_err: 2614 xprt_free(xprt); 2615 return ret; 2616 } 2617 2618 static const struct rpc_timeout xs_udp_default_timeout = { 2619 .to_initval = 5 * HZ, 2620 .to_maxval = 30 * HZ, 2621 .to_increment = 5 * HZ, 2622 .to_retries = 5, 2623 }; 2624 2625 /** 2626 * xs_setup_udp - Set up transport to use a UDP socket 2627 * @args: rpc transport creation arguments 2628 * 2629 */ 2630 static struct rpc_xprt *xs_setup_udp(struct xprt_create *args) 2631 { 2632 struct sockaddr *addr = args->dstaddr; 2633 struct rpc_xprt *xprt; 2634 struct sock_xprt *transport; 2635 struct rpc_xprt *ret; 2636 2637 xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries, 2638 xprt_udp_slot_table_entries); 2639 if (IS_ERR(xprt)) 2640 return xprt; 2641 transport = container_of(xprt, struct sock_xprt, xprt); 2642 2643 xprt->prot = IPPROTO_UDP; 2644 xprt->tsh_size = 0; 2645 /* XXX: header size can vary due to auth type, IPv6, etc. */ 2646 xprt->max_payload = (1U << 16) - (MAX_HEADER << 3); 2647 2648 xprt->bind_timeout = XS_BIND_TO; 2649 xprt->reestablish_timeout = XS_UDP_REEST_TO; 2650 xprt->idle_timeout = XS_IDLE_DISC_TO; 2651 2652 xprt->ops = &xs_udp_ops; 2653 2654 xprt->timeout = &xs_udp_default_timeout; 2655 2656 switch (addr->sa_family) { 2657 case AF_INET: 2658 if (((struct sockaddr_in *)addr)->sin_port != htons(0)) 2659 xprt_set_bound(xprt); 2660 2661 INIT_DELAYED_WORK(&transport->connect_worker, 2662 xs_udp_setup_socket); 2663 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP); 2664 break; 2665 case AF_INET6: 2666 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0)) 2667 xprt_set_bound(xprt); 2668 2669 INIT_DELAYED_WORK(&transport->connect_worker, 2670 xs_udp_setup_socket); 2671 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6); 2672 break; 2673 default: 2674 ret = ERR_PTR(-EAFNOSUPPORT); 2675 goto out_err; 2676 } 2677 2678 if (xprt_bound(xprt)) 2679 dprintk("RPC: set up xprt to %s (port %s) via %s\n", 2680 xprt->address_strings[RPC_DISPLAY_ADDR], 2681 xprt->address_strings[RPC_DISPLAY_PORT], 2682 xprt->address_strings[RPC_DISPLAY_PROTO]); 2683 else 2684 dprintk("RPC: set up xprt to %s (autobind) via %s\n", 2685 xprt->address_strings[RPC_DISPLAY_ADDR], 2686 xprt->address_strings[RPC_DISPLAY_PROTO]); 2687 2688 if (try_module_get(THIS_MODULE)) 2689 return xprt; 2690 ret = ERR_PTR(-EINVAL); 2691 out_err: 2692 xprt_free(xprt); 2693 return ret; 2694 } 2695 2696 static const struct rpc_timeout xs_tcp_default_timeout = { 2697 .to_initval = 60 * HZ, 2698 .to_maxval = 60 * HZ, 2699 .to_retries = 2, 2700 }; 2701 2702 /** 2703 * xs_setup_tcp - Set up transport to use a TCP socket 2704 * @args: rpc transport creation arguments 2705 * 2706 */ 2707 static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args) 2708 { 2709 struct sockaddr *addr = args->dstaddr; 2710 struct rpc_xprt *xprt; 2711 struct sock_xprt *transport; 2712 struct rpc_xprt *ret; 2713 2714 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries, 2715 xprt_max_tcp_slot_table_entries); 2716 if (IS_ERR(xprt)) 2717 return xprt; 2718 transport = container_of(xprt, struct sock_xprt, xprt); 2719 2720 xprt->prot = IPPROTO_TCP; 2721 xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32); 2722 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE; 2723 2724 xprt->bind_timeout = XS_BIND_TO; 2725 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; 2726 xprt->idle_timeout = XS_IDLE_DISC_TO; 2727 2728 xprt->ops = &xs_tcp_ops; 2729 xprt->timeout = &xs_tcp_default_timeout; 2730 2731 switch (addr->sa_family) { 2732 case AF_INET: 2733 if (((struct sockaddr_in *)addr)->sin_port != htons(0)) 2734 xprt_set_bound(xprt); 2735 2736 INIT_DELAYED_WORK(&transport->connect_worker, 2737 xs_tcp_setup_socket); 2738 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP); 2739 break; 2740 case AF_INET6: 2741 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0)) 2742 xprt_set_bound(xprt); 2743 2744 INIT_DELAYED_WORK(&transport->connect_worker, 2745 xs_tcp_setup_socket); 2746 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6); 2747 break; 2748 default: 2749 ret = ERR_PTR(-EAFNOSUPPORT); 2750 goto out_err; 2751 } 2752 2753 if (xprt_bound(xprt)) 2754 dprintk("RPC: set up xprt to %s (port %s) via %s\n", 2755 xprt->address_strings[RPC_DISPLAY_ADDR], 2756 xprt->address_strings[RPC_DISPLAY_PORT], 2757 xprt->address_strings[RPC_DISPLAY_PROTO]); 2758 else 2759 dprintk("RPC: set up xprt to %s (autobind) via %s\n", 2760 xprt->address_strings[RPC_DISPLAY_ADDR], 2761 xprt->address_strings[RPC_DISPLAY_PROTO]); 2762 2763 2764 if (try_module_get(THIS_MODULE)) 2765 return xprt; 2766 ret = ERR_PTR(-EINVAL); 2767 out_err: 2768 xprt_free(xprt); 2769 return ret; 2770 } 2771 2772 /** 2773 * xs_setup_bc_tcp - Set up transport to use a TCP backchannel socket 2774 * @args: rpc transport creation arguments 2775 * 2776 */ 2777 static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args) 2778 { 2779 struct sockaddr *addr = args->dstaddr; 2780 struct rpc_xprt *xprt; 2781 struct sock_xprt *transport; 2782 struct svc_sock *bc_sock; 2783 struct rpc_xprt *ret; 2784 2785 if (args->bc_xprt->xpt_bc_xprt) { 2786 /* 2787 * This server connection already has a backchannel 2788 * export; we can't create a new one, as we wouldn't be 2789 * able to match replies based on xid any more. So, 2790 * reuse the already-existing one: 2791 */ 2792 return args->bc_xprt->xpt_bc_xprt; 2793 } 2794 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries, 2795 xprt_tcp_slot_table_entries); 2796 if (IS_ERR(xprt)) 2797 return xprt; 2798 transport = container_of(xprt, struct sock_xprt, xprt); 2799 2800 xprt->prot = IPPROTO_TCP; 2801 xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32); 2802 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE; 2803 xprt->timeout = &xs_tcp_default_timeout; 2804 2805 /* backchannel */ 2806 xprt_set_bound(xprt); 2807 xprt->bind_timeout = 0; 2808 xprt->reestablish_timeout = 0; 2809 xprt->idle_timeout = 0; 2810 2811 xprt->ops = &bc_tcp_ops; 2812 2813 switch (addr->sa_family) { 2814 case AF_INET: 2815 xs_format_peer_addresses(xprt, "tcp", 2816 RPCBIND_NETID_TCP); 2817 break; 2818 case AF_INET6: 2819 xs_format_peer_addresses(xprt, "tcp", 2820 RPCBIND_NETID_TCP6); 2821 break; 2822 default: 2823 ret = ERR_PTR(-EAFNOSUPPORT); 2824 goto out_err; 2825 } 2826 2827 dprintk("RPC: set up xprt to %s (port %s) via %s\n", 2828 xprt->address_strings[RPC_DISPLAY_ADDR], 2829 xprt->address_strings[RPC_DISPLAY_PORT], 2830 xprt->address_strings[RPC_DISPLAY_PROTO]); 2831 2832 /* 2833 * Once we've associated a backchannel xprt with a connection, 2834 * we want to keep it around as long as long as the connection 2835 * lasts, in case we need to start using it for a backchannel 2836 * again; this reference won't be dropped until bc_xprt is 2837 * destroyed. 2838 */ 2839 xprt_get(xprt); 2840 args->bc_xprt->xpt_bc_xprt = xprt; 2841 xprt->bc_xprt = args->bc_xprt; 2842 bc_sock = container_of(args->bc_xprt, struct svc_sock, sk_xprt); 2843 transport->sock = bc_sock->sk_sock; 2844 transport->inet = bc_sock->sk_sk; 2845 2846 /* 2847 * Since we don't want connections for the backchannel, we set 2848 * the xprt status to connected 2849 */ 2850 xprt_set_connected(xprt); 2851 2852 2853 if (try_module_get(THIS_MODULE)) 2854 return xprt; 2855 xprt_put(xprt); 2856 ret = ERR_PTR(-EINVAL); 2857 out_err: 2858 xprt_free(xprt); 2859 return ret; 2860 } 2861 2862 static struct xprt_class xs_local_transport = { 2863 .list = LIST_HEAD_INIT(xs_local_transport.list), 2864 .name = "named UNIX socket", 2865 .owner = THIS_MODULE, 2866 .ident = XPRT_TRANSPORT_LOCAL, 2867 .setup = xs_setup_local, 2868 }; 2869 2870 static struct xprt_class xs_udp_transport = { 2871 .list = LIST_HEAD_INIT(xs_udp_transport.list), 2872 .name = "udp", 2873 .owner = THIS_MODULE, 2874 .ident = XPRT_TRANSPORT_UDP, 2875 .setup = xs_setup_udp, 2876 }; 2877 2878 static struct xprt_class xs_tcp_transport = { 2879 .list = LIST_HEAD_INIT(xs_tcp_transport.list), 2880 .name = "tcp", 2881 .owner = THIS_MODULE, 2882 .ident = XPRT_TRANSPORT_TCP, 2883 .setup = xs_setup_tcp, 2884 }; 2885 2886 static struct xprt_class xs_bc_tcp_transport = { 2887 .list = LIST_HEAD_INIT(xs_bc_tcp_transport.list), 2888 .name = "tcp NFSv4.1 backchannel", 2889 .owner = THIS_MODULE, 2890 .ident = XPRT_TRANSPORT_BC_TCP, 2891 .setup = xs_setup_bc_tcp, 2892 }; 2893 2894 /** 2895 * init_socket_xprt - set up xprtsock's sysctls, register with RPC client 2896 * 2897 */ 2898 int init_socket_xprt(void) 2899 { 2900 #ifdef RPC_DEBUG 2901 if (!sunrpc_table_header) 2902 sunrpc_table_header = register_sysctl_table(sunrpc_table); 2903 #endif 2904 2905 xprt_register_transport(&xs_local_transport); 2906 xprt_register_transport(&xs_udp_transport); 2907 xprt_register_transport(&xs_tcp_transport); 2908 xprt_register_transport(&xs_bc_tcp_transport); 2909 2910 return 0; 2911 } 2912 2913 /** 2914 * cleanup_socket_xprt - remove xprtsock's sysctls, unregister 2915 * 2916 */ 2917 void cleanup_socket_xprt(void) 2918 { 2919 #ifdef RPC_DEBUG 2920 if (sunrpc_table_header) { 2921 unregister_sysctl_table(sunrpc_table_header); 2922 sunrpc_table_header = NULL; 2923 } 2924 #endif 2925 2926 xprt_unregister_transport(&xs_local_transport); 2927 xprt_unregister_transport(&xs_udp_transport); 2928 xprt_unregister_transport(&xs_tcp_transport); 2929 xprt_unregister_transport(&xs_bc_tcp_transport); 2930 } 2931 2932 static int param_set_uint_minmax(const char *val, 2933 const struct kernel_param *kp, 2934 unsigned int min, unsigned int max) 2935 { 2936 unsigned long num; 2937 int ret; 2938 2939 if (!val) 2940 return -EINVAL; 2941 ret = strict_strtoul(val, 0, &num); 2942 if (ret == -EINVAL || num < min || num > max) 2943 return -EINVAL; 2944 *((unsigned int *)kp->arg) = num; 2945 return 0; 2946 } 2947 2948 static int param_set_portnr(const char *val, const struct kernel_param *kp) 2949 { 2950 return param_set_uint_minmax(val, kp, 2951 RPC_MIN_RESVPORT, 2952 RPC_MAX_RESVPORT); 2953 } 2954 2955 static struct kernel_param_ops param_ops_portnr = { 2956 .set = param_set_portnr, 2957 .get = param_get_uint, 2958 }; 2959 2960 #define param_check_portnr(name, p) \ 2961 __param_check(name, p, unsigned int); 2962 2963 module_param_named(min_resvport, xprt_min_resvport, portnr, 0644); 2964 module_param_named(max_resvport, xprt_max_resvport, portnr, 0644); 2965 2966 static int param_set_slot_table_size(const char *val, 2967 const struct kernel_param *kp) 2968 { 2969 return param_set_uint_minmax(val, kp, 2970 RPC_MIN_SLOT_TABLE, 2971 RPC_MAX_SLOT_TABLE); 2972 } 2973 2974 static struct kernel_param_ops param_ops_slot_table_size = { 2975 .set = param_set_slot_table_size, 2976 .get = param_get_uint, 2977 }; 2978 2979 #define param_check_slot_table_size(name, p) \ 2980 __param_check(name, p, unsigned int); 2981 2982 static int param_set_max_slot_table_size(const char *val, 2983 const struct kernel_param *kp) 2984 { 2985 return param_set_uint_minmax(val, kp, 2986 RPC_MIN_SLOT_TABLE, 2987 RPC_MAX_SLOT_TABLE_LIMIT); 2988 } 2989 2990 static struct kernel_param_ops param_ops_max_slot_table_size = { 2991 .set = param_set_max_slot_table_size, 2992 .get = param_get_uint, 2993 }; 2994 2995 #define param_check_max_slot_table_size(name, p) \ 2996 __param_check(name, p, unsigned int); 2997 2998 module_param_named(tcp_slot_table_entries, xprt_tcp_slot_table_entries, 2999 slot_table_size, 0644); 3000 module_param_named(tcp_max_slot_table_entries, xprt_max_tcp_slot_table_entries, 3001 max_slot_table_size, 0644); 3002 module_param_named(udp_slot_table_entries, xprt_udp_slot_table_entries, 3003 slot_table_size, 0644); 3004 3005