SAFE public projects git trees. - safe/jmp/linux-2.6/blob - net/sunrpc/xprtsock.c

   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(skb));
 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 }