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/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         xprt_adjust_cwnd(task, copied);
1018         xprt_complete_rqst(task, copied);
1019
1020  out_unlock:
1021         spin_unlock(&xprt->transport_lock);
1022  dropit:
1023         skb_free_datagram(sk, skb);
1024  out:
1025         read_unlock_bh(&sk->sk_callback_lock);
1026 }
1027
1028 static inline void xs_tcp_read_fraghdr(struct rpc_xprt *xprt, struct xdr_skb_reader *desc)
1029 {
1030         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1031         size_t len, used;
1032         char *p;
1033
1034         p = ((char *) &transport->tcp_fraghdr) + transport->tcp_offset;
1035         len = sizeof(transport->tcp_fraghdr) - transport->tcp_offset;
1036         used = xdr_skb_read_bits(desc, p, len);
1037         transport->tcp_offset += used;
1038         if (used != len)
1039                 return;
1040
1041         transport->tcp_reclen = ntohl(transport->tcp_fraghdr);
1042         if (transport->tcp_reclen & RPC_LAST_STREAM_FRAGMENT)
1043                 transport->tcp_flags |= TCP_RCV_LAST_FRAG;
1044         else
1045                 transport->tcp_flags &= ~TCP_RCV_LAST_FRAG;
1046         transport->tcp_reclen &= RPC_FRAGMENT_SIZE_MASK;
1047
1048         transport->tcp_flags &= ~TCP_RCV_COPY_FRAGHDR;
1049         transport->tcp_offset = 0;
1050
1051         /* Sanity check of the record length */
1052         if (unlikely(transport->tcp_reclen < 8)) {
1053                 dprintk("RPC:       invalid TCP record fragment length\n");
1054                 xprt_force_disconnect(xprt);
1055                 return;
1056         }
1057         dprintk("RPC:       reading TCP record fragment of length %d\n",
1058                         transport->tcp_reclen);
1059 }
1060
1061 static void xs_tcp_check_fraghdr(struct sock_xprt *transport)
1062 {
1063         if (transport->tcp_offset == transport->tcp_reclen) {
1064                 transport->tcp_flags |= TCP_RCV_COPY_FRAGHDR;
1065                 transport->tcp_offset = 0;
1066                 if (transport->tcp_flags & TCP_RCV_LAST_FRAG) {
1067                         transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1068                         transport->tcp_flags |= TCP_RCV_COPY_XID;
1069                         transport->tcp_copied = 0;
1070                 }
1071         }
1072 }
1073
1074 static inline void xs_tcp_read_xid(struct sock_xprt *transport, struct xdr_skb_reader *desc)
1075 {
1076         size_t len, used;
1077         char *p;
1078
1079         len = sizeof(transport->tcp_xid) - transport->tcp_offset;
1080         dprintk("RPC:       reading XID (%Zu bytes)\n", len);
1081         p = ((char *) &transport->tcp_xid) + transport->tcp_offset;
1082         used = xdr_skb_read_bits(desc, p, len);
1083         transport->tcp_offset += used;
1084         if (used != len)
1085                 return;
1086         transport->tcp_flags &= ~TCP_RCV_COPY_XID;
1087         transport->tcp_flags |= TCP_RCV_READ_CALLDIR;
1088         transport->tcp_copied = 4;
1089         dprintk("RPC:       reading %s XID %08x\n",
1090                         (transport->tcp_flags & TCP_RPC_REPLY) ? "reply for"
1091                                                               : "request with",
1092                         ntohl(transport->tcp_xid));
1093         xs_tcp_check_fraghdr(transport);
1094 }
1095
1096 static inline void xs_tcp_read_calldir(struct sock_xprt *transport,
1097                                        struct xdr_skb_reader *desc)
1098 {
1099         size_t len, used;
1100         u32 offset;
1101         char *p;
1102
1103         /*
1104          * We want transport->tcp_offset to be 8 at the end of this routine
1105          * (4 bytes for the xid and 4 bytes for the call/reply flag).
1106          * When this function is called for the first time,
1107          * transport->tcp_offset is 4 (after having already read the xid).
1108          */
1109         offset = transport->tcp_offset - sizeof(transport->tcp_xid);
1110         len = sizeof(transport->tcp_calldir) - offset;
1111         dprintk("RPC:       reading CALL/REPLY flag (%Zu bytes)\n", len);
1112         p = ((char *) &transport->tcp_calldir) + offset;
1113         used = xdr_skb_read_bits(desc, p, len);
1114         transport->tcp_offset += used;
1115         if (used != len)
1116                 return;
1117         transport->tcp_flags &= ~TCP_RCV_READ_CALLDIR;
1118         /*
1119          * We don't yet have the XDR buffer, so we will write the calldir
1120          * out after we get the buffer from the 'struct rpc_rqst'
1121          */
1122         switch (ntohl(transport->tcp_calldir)) {
1123         case RPC_REPLY:
1124                 transport->tcp_flags |= TCP_RCV_COPY_CALLDIR;
1125                 transport->tcp_flags |= TCP_RCV_COPY_DATA;
1126                 transport->tcp_flags |= TCP_RPC_REPLY;
1127                 break;
1128         case RPC_CALL:
1129                 transport->tcp_flags |= TCP_RCV_COPY_CALLDIR;
1130                 transport->tcp_flags |= TCP_RCV_COPY_DATA;
1131                 transport->tcp_flags &= ~TCP_RPC_REPLY;
1132                 break;
1133         default:
1134                 dprintk("RPC:       invalid request message type\n");
1135                 xprt_force_disconnect(&transport->xprt);
1136         }
1137         xs_tcp_check_fraghdr(transport);
1138 }
1139
1140 static inline void xs_tcp_read_common(struct rpc_xprt *xprt,
1141                                      struct xdr_skb_reader *desc,
1142                                      struct rpc_rqst *req)
1143 {
1144         struct sock_xprt *transport =
1145                                 container_of(xprt, struct sock_xprt, xprt);
1146         struct xdr_buf *rcvbuf;
1147         size_t len;
1148         ssize_t r;
1149
1150         rcvbuf = &req->rq_private_buf;
1151
1152         if (transport->tcp_flags & TCP_RCV_COPY_CALLDIR) {
1153                 /*
1154                  * Save the RPC direction in the XDR buffer
1155                  */
1156                 memcpy(rcvbuf->head[0].iov_base + transport->tcp_copied,
1157                         &transport->tcp_calldir,
1158                         sizeof(transport->tcp_calldir));
1159                 transport->tcp_copied += sizeof(transport->tcp_calldir);
1160                 transport->tcp_flags &= ~TCP_RCV_COPY_CALLDIR;
1161         }
1162
1163         len = desc->count;
1164         if (len > transport->tcp_reclen - transport->tcp_offset) {
1165                 struct xdr_skb_reader my_desc;
1166
1167                 len = transport->tcp_reclen - transport->tcp_offset;
1168                 memcpy(&my_desc, desc, sizeof(my_desc));
1169                 my_desc.count = len;
1170                 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
1171                                           &my_desc, xdr_skb_read_bits);
1172                 desc->count -= r;
1173                 desc->offset += r;
1174         } else
1175                 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
1176                                           desc, xdr_skb_read_bits);
1177
1178         if (r > 0) {
1179                 transport->tcp_copied += r;
1180                 transport->tcp_offset += r;
1181         }
1182         if (r != len) {
1183                 /* Error when copying to the receive buffer,
1184                  * usually because we weren't able to allocate
1185                  * additional buffer pages. All we can do now
1186                  * is turn off TCP_RCV_COPY_DATA, so the request
1187                  * will not receive any additional updates,
1188                  * and time out.
1189                  * Any remaining data from this record will
1190                  * be discarded.
1191                  */
1192                 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1193                 dprintk("RPC:       XID %08x truncated request\n",
1194                                 ntohl(transport->tcp_xid));
1195                 dprintk("RPC:       xprt = %p, tcp_copied = %lu, "
1196                                 "tcp_offset = %u, tcp_reclen = %u\n",
1197                                 xprt, transport->tcp_copied,
1198                                 transport->tcp_offset, transport->tcp_reclen);
1199                 return;
1200         }
1201
1202         dprintk("RPC:       XID %08x read %Zd bytes\n",
1203                         ntohl(transport->tcp_xid), r);
1204         dprintk("RPC:       xprt = %p, tcp_copied = %lu, tcp_offset = %u, "
1205                         "tcp_reclen = %u\n", xprt, transport->tcp_copied,
1206                         transport->tcp_offset, transport->tcp_reclen);
1207
1208         if (transport->tcp_copied == req->rq_private_buf.buflen)
1209                 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1210         else if (transport->tcp_offset == transport->tcp_reclen) {
1211                 if (transport->tcp_flags & TCP_RCV_LAST_FRAG)
1212                         transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1213         }
1214 }
1215
1216 /*
1217  * Finds the request corresponding to the RPC xid and invokes the common
1218  * tcp read code to read the data.
1219  */
1220 static inline int xs_tcp_read_reply(struct rpc_xprt *xprt,
1221                                     struct xdr_skb_reader *desc)
1222 {
1223         struct sock_xprt *transport =
1224                                 container_of(xprt, struct sock_xprt, xprt);
1225         struct rpc_rqst *req;
1226
1227         dprintk("RPC:       read reply XID %08x\n", ntohl(transport->tcp_xid));
1228
1229         /* Find and lock the request corresponding to this xid */
1230         spin_lock(&xprt->transport_lock);
1231         req = xprt_lookup_rqst(xprt, transport->tcp_xid);
1232         if (!req) {
1233                 dprintk("RPC:       XID %08x request not found!\n",
1234                                 ntohl(transport->tcp_xid));
1235                 spin_unlock(&xprt->transport_lock);
1236                 return -1;
1237         }
1238
1239         xs_tcp_read_common(xprt, desc, req);
1240
1241         if (!(transport->tcp_flags & TCP_RCV_COPY_DATA))
1242                 xprt_complete_rqst(req->rq_task, transport->tcp_copied);
1243
1244         spin_unlock(&xprt->transport_lock);
1245         return 0;
1246 }
1247
1248 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
1249 /*
1250  * Obtains an rpc_rqst previously allocated and invokes the common
1251  * tcp read code to read the data.  The result is placed in the callback
1252  * queue.
1253  * If we're unable to obtain the rpc_rqst we schedule the closing of the
1254  * connection and return -1.
1255  */
1256 static inline int xs_tcp_read_callback(struct rpc_xprt *xprt,
1257                                        struct xdr_skb_reader *desc)
1258 {
1259         struct sock_xprt *transport =
1260                                 container_of(xprt, struct sock_xprt, xprt);
1261         struct rpc_rqst *req;
1262
1263         req = xprt_alloc_bc_request(xprt);
1264         if (req == NULL) {
1265                 printk(KERN_WARNING "Callback slot table overflowed\n");
1266                 xprt_force_disconnect(xprt);
1267                 return -1;
1268         }
1269
1270         req->rq_xid = transport->tcp_xid;
1271         dprintk("RPC:       read callback  XID %08x\n", ntohl(req->rq_xid));
1272         xs_tcp_read_common(xprt, desc, req);
1273
1274         if (!(transport->tcp_flags & TCP_RCV_COPY_DATA)) {
1275                 struct svc_serv *bc_serv = xprt->bc_serv;
1276
1277                 /*
1278                  * Add callback request to callback list.  The callback
1279                  * service sleeps on the sv_cb_waitq waiting for new
1280                  * requests.  Wake it up after adding enqueing the
1281                  * request.
1282                  */
1283                 dprintk("RPC:       add callback request to list\n");
1284                 spin_lock(&bc_serv->sv_cb_lock);
1285                 list_add(&req->rq_bc_list, &bc_serv->sv_cb_list);
1286                 spin_unlock(&bc_serv->sv_cb_lock);
1287                 wake_up(&bc_serv->sv_cb_waitq);
1288         }
1289
1290         req->rq_private_buf.len = transport->tcp_copied;
1291
1292         return 0;
1293 }
1294
1295 static inline int _xs_tcp_read_data(struct rpc_xprt *xprt,
1296                                         struct xdr_skb_reader *desc)
1297 {
1298         struct sock_xprt *transport =
1299                                 container_of(xprt, struct sock_xprt, xprt);
1300
1301         return (transport->tcp_flags & TCP_RPC_REPLY) ?
1302                 xs_tcp_read_reply(xprt, desc) :
1303                 xs_tcp_read_callback(xprt, desc);
1304 }
1305 #else
1306 static inline int _xs_tcp_read_data(struct rpc_xprt *xprt,
1307                                         struct xdr_skb_reader *desc)
1308 {
1309         return xs_tcp_read_reply(xprt, desc);
1310 }
1311 #endif /* CONFIG_SUNRPC_BACKCHANNEL */
1312
1313 /*
1314  * Read data off the transport.  This can be either an RPC_CALL or an
1315  * RPC_REPLY.  Relay the processing to helper functions.
1316  */
1317 static void xs_tcp_read_data(struct rpc_xprt *xprt,
1318                                     struct xdr_skb_reader *desc)
1319 {
1320         struct sock_xprt *transport =
1321                                 container_of(xprt, struct sock_xprt, xprt);
1322
1323         if (_xs_tcp_read_data(xprt, desc) == 0)
1324                 xs_tcp_check_fraghdr(transport);
1325         else {
1326                 /*
1327                  * The transport_lock protects the request handling.
1328                  * There's no need to hold it to update the tcp_flags.
1329                  */
1330                 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1331         }
1332 }
1333
1334 static inline void xs_tcp_read_discard(struct sock_xprt *transport, struct xdr_skb_reader *desc)
1335 {
1336         size_t len;
1337
1338         len = transport->tcp_reclen - transport->tcp_offset;
1339         if (len > desc->count)
1340                 len = desc->count;
1341         desc->count -= len;
1342         desc->offset += len;
1343         transport->tcp_offset += len;
1344         dprintk("RPC:       discarded %Zu bytes\n", len);
1345         xs_tcp_check_fraghdr(transport);
1346 }
1347
1348 static int xs_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, unsigned int offset, size_t len)
1349 {
1350         struct rpc_xprt *xprt = rd_desc->arg.data;
1351         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1352         struct xdr_skb_reader desc = {
1353                 .skb    = skb,
1354                 .offset = offset,
1355                 .count  = len,
1356         };
1357
1358         dprintk("RPC:       xs_tcp_data_recv started\n");
1359         do {
1360                 /* Read in a new fragment marker if necessary */
1361                 /* Can we ever really expect to get completely empty fragments? */
1362                 if (transport->tcp_flags & TCP_RCV_COPY_FRAGHDR) {
1363                         xs_tcp_read_fraghdr(xprt, &desc);
1364                         continue;
1365                 }
1366                 /* Read in the xid if necessary */
1367                 if (transport->tcp_flags & TCP_RCV_COPY_XID) {
1368                         xs_tcp_read_xid(transport, &desc);
1369                         continue;
1370                 }
1371                 /* Read in the call/reply flag */
1372                 if (transport->tcp_flags & TCP_RCV_READ_CALLDIR) {
1373                         xs_tcp_read_calldir(transport, &desc);
1374                         continue;
1375                 }
1376                 /* Read in the request data */
1377                 if (transport->tcp_flags & TCP_RCV_COPY_DATA) {
1378                         xs_tcp_read_data(xprt, &desc);
1379                         continue;
1380                 }
1381                 /* Skip over any trailing bytes on short reads */
1382                 xs_tcp_read_discard(transport, &desc);
1383         } while (desc.count);
1384         dprintk("RPC:       xs_tcp_data_recv done\n");
1385         return len - desc.count;
1386 }
1387
1388 /**
1389  * xs_tcp_data_ready - "data ready" callback for TCP sockets
1390  * @sk: socket with data to read
1391  * @bytes: how much data to read
1392  *
1393  */
1394 static void xs_tcp_data_ready(struct sock *sk, int bytes)
1395 {
1396         struct rpc_xprt *xprt;
1397         read_descriptor_t rd_desc;
1398         int read;
1399
1400         dprintk("RPC:       xs_tcp_data_ready...\n");
1401
1402         read_lock_bh(&sk->sk_callback_lock);
1403         if (!(xprt = xprt_from_sock(sk)))
1404                 goto out;
1405         if (xprt->shutdown)
1406                 goto out;
1407
1408         /* Any data means we had a useful conversation, so
1409          * the we don't need to delay the next reconnect
1410          */
1411         if (xprt->reestablish_timeout)
1412                 xprt->reestablish_timeout = 0;
1413
1414         /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */
1415         rd_desc.arg.data = xprt;
1416         do {
1417                 rd_desc.count = 65536;
1418                 read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv);
1419         } while (read > 0);
1420 out:
1421         read_unlock_bh(&sk->sk_callback_lock);
1422 }
1423
1424 /*
1425  * Do the equivalent of linger/linger2 handling for dealing with
1426  * broken servers that don't close the socket in a timely
1427  * fashion
1428  */
1429 static void xs_tcp_schedule_linger_timeout(struct rpc_xprt *xprt,
1430                 unsigned long timeout)
1431 {
1432         struct sock_xprt *transport;
1433
1434         if (xprt_test_and_set_connecting(xprt))
1435                 return;
1436         set_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1437         transport = container_of(xprt, struct sock_xprt, xprt);
1438         queue_delayed_work(rpciod_workqueue, &transport->connect_worker,
1439                            timeout);
1440 }
1441
1442 static void xs_tcp_cancel_linger_timeout(struct rpc_xprt *xprt)
1443 {
1444         struct sock_xprt *transport;
1445
1446         transport = container_of(xprt, struct sock_xprt, xprt);
1447
1448         if (!test_bit(XPRT_CONNECTION_ABORT, &xprt->state) ||
1449             !cancel_delayed_work(&transport->connect_worker))
1450                 return;
1451         clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1452         xprt_clear_connecting(xprt);
1453 }
1454
1455 static void xs_sock_mark_closed(struct rpc_xprt *xprt)
1456 {
1457         smp_mb__before_clear_bit();
1458         clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
1459         clear_bit(XPRT_CLOSING, &xprt->state);
1460         smp_mb__after_clear_bit();
1461         /* Mark transport as closed and wake up all pending tasks */
1462         xprt_disconnect_done(xprt);
1463 }
1464
1465 /**
1466  * xs_tcp_state_change - callback to handle TCP socket state changes
1467  * @sk: socket whose state has changed
1468  *
1469  */
1470 static void xs_tcp_state_change(struct sock *sk)
1471 {
1472         struct rpc_xprt *xprt;
1473
1474         read_lock_bh(&sk->sk_callback_lock);
1475         if (!(xprt = xprt_from_sock(sk)))
1476                 goto out;
1477         dprintk("RPC:       xs_tcp_state_change client %p...\n", xprt);
1478         dprintk("RPC:       state %x conn %d dead %d zapped %d sk_shutdown %d\n",
1479                         sk->sk_state, xprt_connected(xprt),
1480                         sock_flag(sk, SOCK_DEAD),
1481                         sock_flag(sk, SOCK_ZAPPED),
1482                         sk->sk_shutdown);
1483
1484         switch (sk->sk_state) {
1485         case TCP_ESTABLISHED:
1486                 spin_lock(&xprt->transport_lock);
1487                 if (!xprt_test_and_set_connected(xprt)) {
1488                         struct sock_xprt *transport = container_of(xprt,
1489                                         struct sock_xprt, xprt);
1490
1491                         /* Reset TCP record info */
1492                         transport->tcp_offset = 0;
1493                         transport->tcp_reclen = 0;
1494                         transport->tcp_copied = 0;
1495                         transport->tcp_flags =
1496                                 TCP_RCV_COPY_FRAGHDR | TCP_RCV_COPY_XID;
1497
1498                         xprt_wake_pending_tasks(xprt, -EAGAIN);
1499                 }
1500                 spin_unlock(&xprt->transport_lock);
1501                 break;
1502         case TCP_FIN_WAIT1:
1503                 /* The client initiated a shutdown of the socket */
1504                 xprt->connect_cookie++;
1505                 xprt->reestablish_timeout = 0;
1506                 set_bit(XPRT_CLOSING, &xprt->state);
1507                 smp_mb__before_clear_bit();
1508                 clear_bit(XPRT_CONNECTED, &xprt->state);
1509                 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
1510                 smp_mb__after_clear_bit();
1511                 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout);
1512                 break;
1513         case TCP_CLOSE_WAIT:
1514                 /* The server initiated a shutdown of the socket */
1515                 xprt_force_disconnect(xprt);
1516                 xprt->connect_cookie++;
1517         case TCP_CLOSING:
1518                 /*
1519                  * If the server closed down the connection, make sure that
1520                  * we back off before reconnecting
1521                  */
1522                 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
1523                         xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
1524                 break;
1525         case TCP_LAST_ACK:
1526                 set_bit(XPRT_CLOSING, &xprt->state);
1527                 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout);
1528                 smp_mb__before_clear_bit();
1529                 clear_bit(XPRT_CONNECTED, &xprt->state);
1530                 smp_mb__after_clear_bit();
1531                 break;
1532         case TCP_CLOSE:
1533                 xs_tcp_cancel_linger_timeout(xprt);
1534                 xs_sock_mark_closed(xprt);
1535         }
1536  out:
1537         read_unlock_bh(&sk->sk_callback_lock);
1538 }
1539
1540 /**
1541  * xs_error_report - callback mainly for catching socket errors
1542  * @sk: socket
1543  */
1544 static void xs_error_report(struct sock *sk)
1545 {
1546         struct rpc_xprt *xprt;
1547
1548         read_lock_bh(&sk->sk_callback_lock);
1549         if (!(xprt = xprt_from_sock(sk)))
1550                 goto out;
1551         dprintk("RPC:       %s client %p...\n"
1552                         "RPC:       error %d\n",
1553                         __func__, xprt, sk->sk_err);
1554         xprt_wake_pending_tasks(xprt, -EAGAIN);
1555 out:
1556         read_unlock_bh(&sk->sk_callback_lock);
1557 }
1558
1559 static void xs_write_space(struct sock *sk)
1560 {
1561         struct socket *sock;
1562         struct rpc_xprt *xprt;
1563
1564         if (unlikely(!(sock = sk->sk_socket)))
1565                 return;
1566         clear_bit(SOCK_NOSPACE, &sock->flags);
1567
1568         if (unlikely(!(xprt = xprt_from_sock(sk))))
1569                 return;
1570         if (test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags) == 0)
1571                 return;
1572
1573         xprt_write_space(xprt);
1574 }
1575
1576 /**
1577  * xs_udp_write_space - callback invoked when socket buffer space
1578  *                             becomes available
1579  * @sk: socket whose state has changed
1580  *
1581  * Called when more output buffer space is available for this socket.
1582  * We try not to wake our writers until they can make "significant"
1583  * progress, otherwise we'll waste resources thrashing kernel_sendmsg
1584  * with a bunch of small requests.
1585  */
1586 static void xs_udp_write_space(struct sock *sk)
1587 {
1588         read_lock_bh(&sk->sk_callback_lock);
1589
1590         /* from net/core/sock.c:sock_def_write_space */
1591         if (sock_writeable(sk))
1592                 xs_write_space(sk);
1593
1594         read_unlock_bh(&sk->sk_callback_lock);
1595 }
1596
1597 /**
1598  * xs_tcp_write_space - callback invoked when socket buffer space
1599  *                             becomes available
1600  * @sk: socket whose state has changed
1601  *
1602  * Called when more output buffer space is available for this socket.
1603  * We try not to wake our writers until they can make "significant"
1604  * progress, otherwise we'll waste resources thrashing kernel_sendmsg
1605  * with a bunch of small requests.
1606  */
1607 static void xs_tcp_write_space(struct sock *sk)
1608 {
1609         read_lock_bh(&sk->sk_callback_lock);
1610
1611         /* from net/core/stream.c:sk_stream_write_space */
1612         if (sk_stream_wspace(sk) >= sk_stream_min_wspace(sk))
1613                 xs_write_space(sk);
1614
1615         read_unlock_bh(&sk->sk_callback_lock);
1616 }
1617
1618 static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt)
1619 {
1620         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1621         struct sock *sk = transport->inet;
1622
1623         if (transport->rcvsize) {
1624                 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
1625                 sk->sk_rcvbuf = transport->rcvsize * xprt->max_reqs * 2;
1626         }
1627         if (transport->sndsize) {
1628                 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
1629                 sk->sk_sndbuf = transport->sndsize * xprt->max_reqs * 2;
1630                 sk->sk_write_space(sk);
1631         }
1632 }
1633
1634 /**
1635  * xs_udp_set_buffer_size - set send and receive limits
1636  * @xprt: generic transport
1637  * @sndsize: requested size of send buffer, in bytes
1638  * @rcvsize: requested size of receive buffer, in bytes
1639  *
1640  * Set socket send and receive buffer size limits.
1641  */
1642 static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize)
1643 {
1644         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1645
1646         transport->sndsize = 0;
1647         if (sndsize)
1648                 transport->sndsize = sndsize + 1024;
1649         transport->rcvsize = 0;
1650         if (rcvsize)
1651                 transport->rcvsize = rcvsize + 1024;
1652
1653         xs_udp_do_set_buffer_size(xprt);
1654 }
1655
1656 /**
1657  * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport
1658  * @task: task that timed out
1659  *
1660  * Adjust the congestion window after a retransmit timeout has occurred.
1661  */
1662 static void xs_udp_timer(struct rpc_task *task)
1663 {
1664         xprt_adjust_cwnd(task, -ETIMEDOUT);
1665 }
1666
1667 static unsigned short xs_get_random_port(void)
1668 {
1669         unsigned short range = xprt_max_resvport - xprt_min_resvport;
1670         unsigned short rand = (unsigned short) net_random() % range;
1671         return rand + xprt_min_resvport;
1672 }
1673
1674 /**
1675  * xs_set_port - reset the port number in the remote endpoint address
1676  * @xprt: generic transport
1677  * @port: new port number
1678  *
1679  */
1680 static void xs_set_port(struct rpc_xprt *xprt, unsigned short port)
1681 {
1682         dprintk("RPC:       setting port for xprt %p to %u\n", xprt, port);
1683
1684         rpc_set_port(xs_addr(xprt), port);
1685         xs_update_peer_port(xprt);
1686 }
1687
1688 static unsigned short xs_get_srcport(struct sock_xprt *transport)
1689 {
1690         unsigned short port = transport->srcport;
1691
1692         if (port == 0 && transport->xprt.resvport)
1693                 port = xs_get_random_port();
1694         return port;
1695 }
1696
1697 static unsigned short xs_next_srcport(struct sock_xprt *transport, unsigned short port)
1698 {
1699         if (transport->srcport != 0)
1700                 transport->srcport = 0;
1701         if (!transport->xprt.resvport)
1702                 return 0;
1703         if (port <= xprt_min_resvport || port > xprt_max_resvport)
1704                 return xprt_max_resvport;
1705         return --port;
1706 }
1707 static int xs_bind(struct sock_xprt *transport, struct socket *sock)
1708 {
1709         struct sockaddr_storage myaddr;
1710         int err, nloop = 0;
1711         unsigned short port = xs_get_srcport(transport);
1712         unsigned short last;
1713
1714         memcpy(&myaddr, &transport->srcaddr, transport->xprt.addrlen);
1715         do {
1716                 rpc_set_port((struct sockaddr *)&myaddr, port);
1717                 err = kernel_bind(sock, (struct sockaddr *)&myaddr,
1718                                 transport->xprt.addrlen);
1719                 if (port == 0)
1720                         break;
1721                 if (err == 0) {
1722                         transport->srcport = port;
1723                         break;
1724                 }
1725                 last = port;
1726                 port = xs_next_srcport(transport, port);
1727                 if (port > last)
1728                         nloop++;
1729         } while (err == -EADDRINUSE && nloop != 2);
1730
1731         if (myaddr.ss_family == AF_INET)
1732                 dprintk("RPC:       %s %pI4:%u: %s (%d)\n", __func__,
1733                                 &((struct sockaddr_in *)&myaddr)->sin_addr,
1734                                 port, err ? "failed" : "ok", err);
1735         else
1736                 dprintk("RPC:       %s %pI6:%u: %s (%d)\n", __func__,
1737                                 &((struct sockaddr_in6 *)&myaddr)->sin6_addr,
1738                                 port, err ? "failed" : "ok", err);
1739         return err;
1740 }
1741
1742 /*
1743  * We don't support autobind on AF_LOCAL sockets
1744  */
1745 static void xs_local_rpcbind(struct rpc_task *task)
1746 {
1747         xprt_set_bound(task->tk_xprt);
1748 }
1749
1750 static void xs_local_set_port(struct rpc_xprt *xprt, unsigned short port)
1751 {
1752 }
1753
1754 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1755 static struct lock_class_key xs_key[2];
1756 static struct lock_class_key xs_slock_key[2];
1757
1758 static inline void xs_reclassify_socketu(struct socket *sock)
1759 {
1760         struct sock *sk = sock->sk;
1761
1762         BUG_ON(sock_owned_by_user(sk));
1763         sock_lock_init_class_and_name(sk, "slock-AF_LOCAL-RPC",
1764                 &xs_slock_key[1], "sk_lock-AF_LOCAL-RPC", &xs_key[1]);
1765 }
1766
1767 static inline void xs_reclassify_socket4(struct socket *sock)
1768 {
1769         struct sock *sk = sock->sk;
1770
1771         BUG_ON(sock_owned_by_user(sk));
1772         sock_lock_init_class_and_name(sk, "slock-AF_INET-RPC",
1773                 &xs_slock_key[0], "sk_lock-AF_INET-RPC", &xs_key[0]);
1774 }
1775
1776 static inline void xs_reclassify_socket6(struct socket *sock)
1777 {
1778         struct sock *sk = sock->sk;
1779
1780         BUG_ON(sock_owned_by_user(sk));
1781         sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC",
1782                 &xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]);
1783 }
1784
1785 static inline void xs_reclassify_socket(int family, struct socket *sock)
1786 {
1787         switch (family) {
1788         case AF_LOCAL:
1789                 xs_reclassify_socketu(sock);
1790                 break;
1791         case AF_INET:
1792                 xs_reclassify_socket4(sock);
1793                 break;
1794         case AF_INET6:
1795                 xs_reclassify_socket6(sock);
1796                 break;
1797         }
1798 }
1799 #else
1800 static inline void xs_reclassify_socketu(struct socket *sock)
1801 {
1802 }
1803
1804 static inline void xs_reclassify_socket4(struct socket *sock)
1805 {
1806 }
1807
1808 static inline void xs_reclassify_socket6(struct socket *sock)
1809 {
1810 }
1811
1812 static inline void xs_reclassify_socket(int family, struct socket *sock)
1813 {
1814 }
1815 #endif
1816
1817 static struct socket *xs_create_sock(struct rpc_xprt *xprt,
1818                 struct sock_xprt *transport, int family, int type, int protocol)
1819 {
1820         struct socket *sock;
1821         int err;
1822
1823         err = __sock_create(xprt->xprt_net, family, type, protocol, &sock, 1);
1824         if (err < 0) {
1825                 dprintk("RPC:       can't create %d transport socket (%d).\n",
1826                                 protocol, -err);
1827                 goto out;
1828         }
1829         xs_reclassify_socket(family, sock);
1830
1831         err = xs_bind(transport, sock);
1832         if (err) {
1833                 sock_release(sock);
1834                 goto out;
1835         }
1836
1837         return sock;
1838 out:
1839         return ERR_PTR(err);
1840 }
1841
1842 static int xs_local_finish_connecting(struct rpc_xprt *xprt,
1843                                       struct socket *sock)
1844 {
1845         struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
1846                                                                         xprt);
1847
1848         if (!transport->inet) {
1849                 struct sock *sk = sock->sk;
1850
1851                 write_lock_bh(&sk->sk_callback_lock);
1852
1853                 xs_save_old_callbacks(transport, sk);
1854
1855                 sk->sk_user_data = xprt;
1856                 sk->sk_data_ready = xs_local_data_ready;
1857                 sk->sk_write_space = xs_udp_write_space;
1858                 sk->sk_error_report = xs_error_report;
1859                 sk->sk_allocation = GFP_ATOMIC;
1860
1861                 xprt_clear_connected(xprt);
1862
1863                 /* Reset to new socket */
1864                 transport->sock = sock;
1865                 transport->inet = sk;
1866
1867                 write_unlock_bh(&sk->sk_callback_lock);
1868         }
1869
1870         /* Tell the socket layer to start connecting... */
1871         xprt->stat.connect_count++;
1872         xprt->stat.connect_start = jiffies;
1873         return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, 0);
1874 }
1875
1876 /**
1877  * xs_local_setup_socket - create AF_LOCAL socket, connect to a local endpoint
1878  * @xprt: RPC transport to connect
1879  * @transport: socket transport to connect
1880  * @create_sock: function to create a socket of the correct type
1881  *
1882  * Invoked by a work queue tasklet.
1883  */
1884 static void xs_local_setup_socket(struct work_struct *work)
1885 {
1886         struct sock_xprt *transport =
1887                 container_of(work, struct sock_xprt, connect_worker.work);
1888         struct rpc_xprt *xprt = &transport->xprt;
1889         struct socket *sock;
1890         int status = -EIO;
1891
1892         if (xprt->shutdown)
1893                 goto out;
1894
1895         current->flags |= PF_FSTRANS;
1896
1897         clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1898         status = __sock_create(xprt->xprt_net, AF_LOCAL,
1899                                         SOCK_STREAM, 0, &sock, 1);
1900         if (status < 0) {
1901                 dprintk("RPC:       can't create AF_LOCAL "
1902                         "transport socket (%d).\n", -status);
1903                 goto out;
1904         }
1905         xs_reclassify_socketu(sock);
1906
1907         dprintk("RPC:       worker connecting xprt %p via AF_LOCAL to %s\n",
1908                         xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1909
1910         status = xs_local_finish_connecting(xprt, sock);
1911         switch (status) {
1912         case 0:
1913                 dprintk("RPC:       xprt %p connected to %s\n",
1914                                 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1915                 xprt_set_connected(xprt);
1916                 break;
1917         case -ENOENT:
1918                 dprintk("RPC:       xprt %p: socket %s does not exist\n",
1919                                 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1920                 break;
1921         default:
1922                 printk(KERN_ERR "%s: unhandled error (%d) connecting to %s\n",
1923                                 __func__, -status,
1924                                 xprt->address_strings[RPC_DISPLAY_ADDR]);
1925         }
1926
1927 out:
1928         xprt_clear_connecting(xprt);
1929         xprt_wake_pending_tasks(xprt, status);
1930         current->flags &= ~PF_FSTRANS;
1931 }
1932
1933 #ifdef CONFIG_SUNRPC_SWAP
1934 static void xs_set_memalloc(struct rpc_xprt *xprt)
1935 {
1936         struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
1937                         xprt);
1938
1939         if (xprt->swapper)
1940                 sk_set_memalloc(transport->inet);
1941 }
1942
1943 /**
1944  * xs_swapper - Tag this transport as being used for swap.
1945  * @xprt: transport to tag
1946  * @enable: enable/disable
1947  *
1948  */
1949 int xs_swapper(struct rpc_xprt *xprt, int enable)
1950 {
1951         struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
1952                         xprt);
1953         int err = 0;
1954
1955         if (enable) {
1956                 xprt->swapper++;
1957                 xs_set_memalloc(xprt);
1958         } else if (xprt->swapper) {
1959                 xprt->swapper--;
1960                 sk_clear_memalloc(transport->inet);
1961         }
1962
1963         return err;
1964 }
1965 EXPORT_SYMBOL_GPL(xs_swapper);
1966 #else
1967 static void xs_set_memalloc(struct rpc_xprt *xprt)
1968 {
1969 }
1970 #endif
1971
1972 static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
1973 {
1974         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1975
1976         if (!transport->inet) {
1977                 struct sock *sk = sock->sk;
1978
1979                 write_lock_bh(&sk->sk_callback_lock);
1980
1981                 xs_save_old_callbacks(transport, sk);
1982
1983                 sk->sk_user_data = xprt;
1984                 sk->sk_data_ready = xs_udp_data_ready;
1985                 sk->sk_write_space = xs_udp_write_space;
1986                 sk->sk_error_report = xs_error_report;
1987                 sk->sk_no_check = UDP_CSUM_NORCV;
1988                 sk->sk_allocation = GFP_ATOMIC;
1989
1990                 xprt_set_connected(xprt);
1991
1992                 /* Reset to new socket */
1993                 transport->sock = sock;
1994                 transport->inet = sk;
1995
1996                 xs_set_memalloc(xprt);
1997
1998                 write_unlock_bh(&sk->sk_callback_lock);
1999         }
2000         xs_udp_do_set_buffer_size(xprt);
2001 }
2002
2003 static void xs_udp_setup_socket(struct work_struct *work)
2004 {
2005         struct sock_xprt *transport =
2006                 container_of(work, struct sock_xprt, connect_worker.work);
2007         struct rpc_xprt *xprt = &transport->xprt;
2008         struct socket *sock = transport->sock;
2009         int status = -EIO;
2010
2011         if (xprt->shutdown)
2012                 goto out;
2013
2014         current->flags |= PF_FSTRANS;
2015
2016         /* Start by resetting any existing state */
2017         xs_reset_transport(transport);
2018         sock = xs_create_sock(xprt, transport,
2019                         xs_addr(xprt)->sa_family, SOCK_DGRAM, IPPROTO_UDP);
2020         if (IS_ERR(sock))
2021                 goto out;
2022
2023         dprintk("RPC:       worker connecting xprt %p via %s to "
2024                                 "%s (port %s)\n", xprt,
2025                         xprt->address_strings[RPC_DISPLAY_PROTO],
2026                         xprt->address_strings[RPC_DISPLAY_ADDR],
2027                         xprt->address_strings[RPC_DISPLAY_PORT]);
2028
2029         xs_udp_finish_connecting(xprt, sock);
2030         status = 0;
2031 out:
2032         xprt_clear_connecting(xprt);
2033         xprt_wake_pending_tasks(xprt, status);
2034         current->flags &= ~PF_FSTRANS;
2035 }
2036
2037 /*
2038  * We need to preserve the port number so the reply cache on the server can
2039  * find our cached RPC replies when we get around to reconnecting.
2040  */
2041 static void xs_abort_connection(struct sock_xprt *transport)
2042 {
2043         int result;
2044         struct sockaddr any;
2045
2046         dprintk("RPC:       disconnecting xprt %p to reuse port\n", transport);
2047
2048         /*
2049          * Disconnect the transport socket by doing a connect operation
2050          * with AF_UNSPEC.  This should return immediately...
2051          */
2052         memset(&any, 0, sizeof(any));
2053         any.sa_family = AF_UNSPEC;
2054         result = kernel_connect(transport->sock, &any, sizeof(any), 0);
2055         if (!result)
2056                 xs_sock_mark_closed(&transport->xprt);
2057         else
2058                 dprintk("RPC:       AF_UNSPEC connect return code %d\n",
2059                                 result);
2060 }
2061
2062 static void xs_tcp_reuse_connection(struct sock_xprt *transport)
2063 {
2064         unsigned int state = transport->inet->sk_state;
2065
2066         if (state == TCP_CLOSE && transport->sock->state == SS_UNCONNECTED) {
2067                 /* we don't need to abort the connection if the socket
2068                  * hasn't undergone a shutdown
2069                  */
2070                 if (transport->inet->sk_shutdown == 0)
2071                         return;
2072                 dprintk("RPC:       %s: TCP_CLOSEd and sk_shutdown set to %d\n",
2073                                 __func__, transport->inet->sk_shutdown);
2074         }
2075         if ((1 << state) & (TCPF_ESTABLISHED|TCPF_SYN_SENT)) {
2076                 /* we don't need to abort the connection if the socket
2077                  * hasn't undergone a shutdown
2078                  */
2079                 if (transport->inet->sk_shutdown == 0)
2080                         return;
2081                 dprintk("RPC:       %s: ESTABLISHED/SYN_SENT "
2082                                 "sk_shutdown set to %d\n",
2083                                 __func__, transport->inet->sk_shutdown);
2084         }
2085         xs_abort_connection(transport);
2086 }
2087
2088 static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
2089 {
2090         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2091         int ret = -ENOTCONN;
2092
2093         if (!transport->inet) {
2094                 struct sock *sk = sock->sk;
2095
2096                 write_lock_bh(&sk->sk_callback_lock);
2097
2098                 xs_save_old_callbacks(transport, sk);
2099
2100                 sk->sk_user_data = xprt;
2101                 sk->sk_data_ready = xs_tcp_data_ready;
2102                 sk->sk_state_change = xs_tcp_state_change;
2103                 sk->sk_write_space = xs_tcp_write_space;
2104                 sk->sk_error_report = xs_error_report;
2105                 sk->sk_allocation = GFP_ATOMIC;
2106
2107                 /* socket options */
2108                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
2109                 sock_reset_flag(sk, SOCK_LINGER);
2110                 tcp_sk(sk)->linger2 = 0;
2111                 tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF;
2112
2113                 xprt_clear_connected(xprt);
2114
2115                 /* Reset to new socket */
2116                 transport->sock = sock;
2117                 transport->inet = sk;
2118
2119                 write_unlock_bh(&sk->sk_callback_lock);
2120         }
2121
2122         if (!xprt_bound(xprt))
2123                 goto out;
2124
2125         xs_set_memalloc(xprt);
2126
2127         /* Tell the socket layer to start connecting... */
2128         xprt->stat.connect_count++;
2129         xprt->stat.connect_start = jiffies;
2130         ret = kernel_connect(sock, xs_addr(xprt), xprt->addrlen, O_NONBLOCK);
2131         switch (ret) {
2132         case 0:
2133         case -EINPROGRESS:
2134                 /* SYN_SENT! */
2135                 xprt->connect_cookie++;
2136                 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
2137                         xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2138         }
2139 out:
2140         return ret;
2141 }
2142
2143 /**
2144  * xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint
2145  * @xprt: RPC transport to connect
2146  * @transport: socket transport to connect
2147  * @create_sock: function to create a socket of the correct type
2148  *
2149  * Invoked by a work queue tasklet.
2150  */
2151 static void xs_tcp_setup_socket(struct work_struct *work)
2152 {
2153         struct sock_xprt *transport =
2154                 container_of(work, struct sock_xprt, connect_worker.work);
2155         struct socket *sock = transport->sock;
2156         struct rpc_xprt *xprt = &transport->xprt;
2157         int status = -EIO;
2158
2159         if (xprt->shutdown)
2160                 goto out;
2161
2162         current->flags |= PF_FSTRANS;
2163
2164         if (!sock) {
2165                 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
2166                 sock = xs_create_sock(xprt, transport,
2167                                 xs_addr(xprt)->sa_family, SOCK_STREAM, IPPROTO_TCP);
2168                 if (IS_ERR(sock)) {
2169                         status = PTR_ERR(sock);
2170                         goto out;
2171                 }
2172         } else {
2173                 int abort_and_exit;
2174
2175                 abort_and_exit = test_and_clear_bit(XPRT_CONNECTION_ABORT,
2176                                 &xprt->state);
2177                 /* "close" the socket, preserving the local port */
2178                 xs_tcp_reuse_connection(transport);
2179
2180                 if (abort_and_exit)
2181                         goto out_eagain;
2182         }
2183
2184         dprintk("RPC:       worker connecting xprt %p via %s to "
2185                                 "%s (port %s)\n", xprt,
2186                         xprt->address_strings[RPC_DISPLAY_PROTO],
2187                         xprt->address_strings[RPC_DISPLAY_ADDR],
2188                         xprt->address_strings[RPC_DISPLAY_PORT]);
2189
2190         status = xs_tcp_finish_connecting(xprt, sock);
2191         dprintk("RPC:       %p connect status %d connected %d sock state %d\n",
2192                         xprt, -status, xprt_connected(xprt),
2193                         sock->sk->sk_state);
2194         switch (status) {
2195         default:
2196                 printk("%s: connect returned unhandled error %d\n",
2197                         __func__, status);
2198         case -EADDRNOTAVAIL:
2199                 /* We're probably in TIME_WAIT. Get rid of existing socket,
2200                  * and retry
2201                  */
2202                 set_bit(XPRT_CONNECTION_CLOSE, &xprt->state);
2203                 xprt_force_disconnect(xprt);
2204                 break;
2205         case -ECONNREFUSED:
2206         case -ECONNRESET:
2207         case -ENETUNREACH:
2208                 /* retry with existing socket, after a delay */
2209         case 0:
2210         case -EINPROGRESS:
2211         case -EALREADY:
2212                 xprt_clear_connecting(xprt);
2213                 current->flags &= ~PF_FSTRANS;
2214                 return;
2215         case -EINVAL:
2216                 /* Happens, for instance, if the user specified a link
2217                  * local IPv6 address without a scope-id.
2218                  */
2219                 goto out;
2220         }
2221 out_eagain:
2222         status = -EAGAIN;
2223 out:
2224         xprt_clear_connecting(xprt);
2225         xprt_wake_pending_tasks(xprt, status);
2226         current->flags &= ~PF_FSTRANS;
2227 }
2228
2229 /**
2230  * xs_connect - connect a socket to a remote endpoint
2231  * @task: address of RPC task that manages state of connect request
2232  *
2233  * TCP: If the remote end dropped the connection, delay reconnecting.
2234  *
2235  * UDP socket connects are synchronous, but we use a work queue anyway
2236  * to guarantee that even unprivileged user processes can set up a
2237  * socket on a privileged port.
2238  *
2239  * If a UDP socket connect fails, the delay behavior here prevents
2240  * retry floods (hard mounts).
2241  */
2242 static void xs_connect(struct rpc_task *task)
2243 {
2244         struct rpc_xprt *xprt = task->tk_xprt;
2245         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2246
2247         if (transport->sock != NULL && !RPC_IS_SOFTCONN(task)) {
2248                 dprintk("RPC:       xs_connect delayed xprt %p for %lu "
2249                                 "seconds\n",
2250                                 xprt, xprt->reestablish_timeout / HZ);
2251                 queue_delayed_work(rpciod_workqueue,
2252                                    &transport->connect_worker,
2253                                    xprt->reestablish_timeout);
2254                 xprt->reestablish_timeout <<= 1;
2255                 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
2256                         xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2257                 if (xprt->reestablish_timeout > XS_TCP_MAX_REEST_TO)
2258                         xprt->reestablish_timeout = XS_TCP_MAX_REEST_TO;
2259         } else {
2260                 dprintk("RPC:       xs_connect scheduled xprt %p\n", xprt);
2261                 queue_delayed_work(rpciod_workqueue,
2262                                    &transport->connect_worker, 0);
2263         }
2264 }
2265
2266 /**
2267  * xs_local_print_stats - display AF_LOCAL socket-specifc stats
2268  * @xprt: rpc_xprt struct containing statistics
2269  * @seq: output file
2270  *
2271  */
2272 static void xs_local_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2273 {
2274         long idle_time = 0;
2275
2276         if (xprt_connected(xprt))
2277                 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2278
2279         seq_printf(seq, "\txprt:\tlocal %lu %lu %lu %ld %lu %lu %lu "
2280                         "%llu %llu %lu %llu %llu\n",
2281                         xprt->stat.bind_count,
2282                         xprt->stat.connect_count,
2283                         xprt->stat.connect_time,
2284                         idle_time,
2285                         xprt->stat.sends,
2286                         xprt->stat.recvs,
2287                         xprt->stat.bad_xids,
2288                         xprt->stat.req_u,
2289                         xprt->stat.bklog_u,
2290                         xprt->stat.max_slots,
2291                         xprt->stat.sending_u,
2292                         xprt->stat.pending_u);
2293 }
2294
2295 /**
2296  * xs_udp_print_stats - display UDP socket-specifc stats
2297  * @xprt: rpc_xprt struct containing statistics
2298  * @seq: output file
2299  *
2300  */
2301 static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2302 {
2303         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2304
2305         seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %llu %llu "
2306                         "%lu %llu %llu\n",
2307                         transport->srcport,
2308                         xprt->stat.bind_count,
2309                         xprt->stat.sends,
2310                         xprt->stat.recvs,
2311                         xprt->stat.bad_xids,
2312                         xprt->stat.req_u,
2313                         xprt->stat.bklog_u,
2314                         xprt->stat.max_slots,
2315                         xprt->stat.sending_u,
2316                         xprt->stat.pending_u);
2317 }
2318
2319 /**
2320  * xs_tcp_print_stats - display TCP socket-specifc stats
2321  * @xprt: rpc_xprt struct containing statistics
2322  * @seq: output file
2323  *
2324  */
2325 static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2326 {
2327         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2328         long idle_time = 0;
2329
2330         if (xprt_connected(xprt))
2331                 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2332
2333         seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu "
2334                         "%llu %llu %lu %llu %llu\n",
2335                         transport->srcport,
2336                         xprt->stat.bind_count,
2337                         xprt->stat.connect_count,
2338                         xprt->stat.connect_time,
2339                         idle_time,
2340                         xprt->stat.sends,
2341                         xprt->stat.recvs,
2342                         xprt->stat.bad_xids,
2343                         xprt->stat.req_u,
2344                         xprt->stat.bklog_u,
2345                         xprt->stat.max_slots,
2346                         xprt->stat.sending_u,
2347                         xprt->stat.pending_u);
2348 }
2349
2350 /*
2351  * Allocate a bunch of pages for a scratch buffer for the rpc code. The reason
2352  * we allocate pages instead doing a kmalloc like rpc_malloc is because we want
2353  * to use the server side send routines.
2354  */
2355 static void *bc_malloc(struct rpc_task *task, size_t size)
2356 {
2357         struct page *page;
2358         struct rpc_buffer *buf;
2359
2360         BUG_ON(size > PAGE_SIZE - sizeof(struct rpc_buffer));
2361         page = alloc_page(GFP_KERNEL);
2362
2363         if (!page)
2364                 return NULL;
2365
2366         buf = page_address(page);
2367         buf->len = PAGE_SIZE;
2368
2369         return buf->data;
2370 }
2371
2372 /*
2373  * Free the space allocated in the bc_alloc routine
2374  */
2375 static void bc_free(void *buffer)
2376 {
2377         struct rpc_buffer *buf;
2378
2379         if (!buffer)
2380                 return;
2381
2382         buf = container_of(buffer, struct rpc_buffer, data);
2383         free_page((unsigned long)buf);
2384 }
2385
2386 /*
2387  * Use the svc_sock to send the callback. Must be called with svsk->sk_mutex
2388  * held. Borrows heavily from svc_tcp_sendto and xs_tcp_send_request.
2389  */
2390 static int bc_sendto(struct rpc_rqst *req)
2391 {
2392         int len;
2393         struct xdr_buf *xbufp = &req->rq_snd_buf;
2394         struct rpc_xprt *xprt = req->rq_xprt;
2395         struct sock_xprt *transport =
2396                                 container_of(xprt, struct sock_xprt, xprt);
2397         struct socket *sock = transport->sock;
2398         unsigned long headoff;
2399         unsigned long tailoff;
2400
2401         xs_encode_stream_record_marker(xbufp);
2402
2403         tailoff = (unsigned long)xbufp->tail[0].iov_base & ~PAGE_MASK;
2404         headoff = (unsigned long)xbufp->head[0].iov_base & ~PAGE_MASK;
2405         len = svc_send_common(sock, xbufp,
2406                               virt_to_page(xbufp->head[0].iov_base), headoff,
2407                               xbufp->tail[0].iov_base, tailoff);
2408
2409         if (len != xbufp->len) {
2410                 printk(KERN_NOTICE "Error sending entire callback!\n");
2411                 len = -EAGAIN;
2412         }
2413
2414         return len;
2415 }
2416
2417 /*
2418  * The send routine. Borrows from svc_send
2419  */
2420 static int bc_send_request(struct rpc_task *task)
2421 {
2422         struct rpc_rqst *req = task->tk_rqstp;
2423         struct svc_xprt *xprt;
2424         struct svc_sock         *svsk;
2425         u32                     len;
2426
2427         dprintk("sending request with xid: %08x\n", ntohl(req->rq_xid));
2428         /*
2429          * Get the server socket associated with this callback xprt
2430          */
2431         xprt = req->rq_xprt->bc_xprt;
2432         svsk = container_of(xprt, struct svc_sock, sk_xprt);
2433
2434         /*
2435          * Grab the mutex to serialize data as the connection is shared
2436          * with the fore channel
2437          */
2438         if (!mutex_trylock(&xprt->xpt_mutex)) {
2439                 rpc_sleep_on(&xprt->xpt_bc_pending, task, NULL);
2440                 if (!mutex_trylock(&xprt->xpt_mutex))
2441                         return -EAGAIN;
2442                 rpc_wake_up_queued_task(&xprt->xpt_bc_pending, task);
2443         }
2444         if (test_bit(XPT_DEAD, &xprt->xpt_flags))
2445                 len = -ENOTCONN;
2446         else
2447                 len = bc_sendto(req);
2448         mutex_unlock(&xprt->xpt_mutex);
2449
2450         if (len > 0)
2451                 len = 0;
2452
2453         return len;
2454 }
2455
2456 /*
2457  * The close routine. Since this is client initiated, we do nothing
2458  */
2459
2460 static void bc_close(struct rpc_xprt *xprt)
2461 {
2462 }
2463
2464 /*
2465  * The xprt destroy routine. Again, because this connection is client
2466  * initiated, we do nothing
2467  */
2468
2469 static void bc_destroy(struct rpc_xprt *xprt)
2470 {
2471 }
2472
2473 static struct rpc_xprt_ops xs_local_ops = {
2474         .reserve_xprt           = xprt_reserve_xprt,
2475         .release_xprt           = xs_tcp_release_xprt,
2476         .alloc_slot             = xprt_alloc_slot,
2477         .rpcbind                = xs_local_rpcbind,
2478         .set_port               = xs_local_set_port,
2479         .connect                = xs_connect,
2480         .buf_alloc              = rpc_malloc,
2481         .buf_free               = rpc_free,
2482         .send_request           = xs_local_send_request,
2483         .set_retrans_timeout    = xprt_set_retrans_timeout_def,
2484         .close                  = xs_close,
2485         .destroy                = xs_destroy,
2486         .print_stats            = xs_local_print_stats,
2487 };
2488
2489 static struct rpc_xprt_ops xs_udp_ops = {
2490         .set_buffer_size        = xs_udp_set_buffer_size,
2491         .reserve_xprt           = xprt_reserve_xprt_cong,
2492         .release_xprt           = xprt_release_xprt_cong,
2493         .alloc_slot             = xprt_alloc_slot,
2494         .rpcbind                = rpcb_getport_async,
2495         .set_port               = xs_set_port,
2496         .connect                = xs_connect,
2497         .buf_alloc              = rpc_malloc,
2498         .buf_free               = rpc_free,
2499         .send_request           = xs_udp_send_request,
2500         .set_retrans_timeout    = xprt_set_retrans_timeout_rtt,
2501         .timer                  = xs_udp_timer,
2502         .release_request        = xprt_release_rqst_cong,
2503         .close                  = xs_close,
2504         .destroy                = xs_destroy,
2505         .print_stats            = xs_udp_print_stats,
2506 };
2507
2508 static struct rpc_xprt_ops xs_tcp_ops = {
2509         .reserve_xprt           = xprt_reserve_xprt,
2510         .release_xprt           = xs_tcp_release_xprt,
2511         .alloc_slot             = xprt_lock_and_alloc_slot,
2512         .rpcbind                = rpcb_getport_async,
2513         .set_port               = xs_set_port,
2514         .connect                = xs_connect,
2515         .buf_alloc              = rpc_malloc,
2516         .buf_free               = rpc_free,
2517         .send_request           = xs_tcp_send_request,
2518         .set_retrans_timeout    = xprt_set_retrans_timeout_def,
2519         .close                  = xs_tcp_close,
2520         .destroy                = xs_destroy,
2521         .print_stats            = xs_tcp_print_stats,
2522 };
2523
2524 /*
2525  * The rpc_xprt_ops for the server backchannel
2526  */
2527
2528 static struct rpc_xprt_ops bc_tcp_ops = {
2529         .reserve_xprt           = xprt_reserve_xprt,
2530         .release_xprt           = xprt_release_xprt,
2531         .rpcbind                = xs_local_rpcbind,
2532         .buf_alloc              = bc_malloc,
2533         .buf_free               = bc_free,
2534         .send_request           = bc_send_request,
2535         .set_retrans_timeout    = xprt_set_retrans_timeout_def,
2536         .close                  = bc_close,
2537         .destroy                = bc_destroy,
2538         .print_stats            = xs_tcp_print_stats,
2539 };
2540
2541 static int xs_init_anyaddr(const int family, struct sockaddr *sap)
2542 {
2543         static const struct sockaddr_in sin = {
2544                 .sin_family             = AF_INET,
2545                 .sin_addr.s_addr        = htonl(INADDR_ANY),
2546         };
2547         static const struct sockaddr_in6 sin6 = {
2548                 .sin6_family            = AF_INET6,
2549                 .sin6_addr              = IN6ADDR_ANY_INIT,
2550         };
2551
2552         switch (family) {
2553         case AF_LOCAL:
2554                 break;
2555         case AF_INET:
2556                 memcpy(sap, &sin, sizeof(sin));
2557                 break;
2558         case AF_INET6:
2559                 memcpy(sap, &sin6, sizeof(sin6));
2560                 break;
2561         default:
2562                 dprintk("RPC:       %s: Bad address family\n", __func__);
2563                 return -EAFNOSUPPORT;
2564         }
2565         return 0;
2566 }
2567
2568 static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args,
2569                                       unsigned int slot_table_size,
2570                                       unsigned int max_slot_table_size)
2571 {
2572         struct rpc_xprt *xprt;
2573         struct sock_xprt *new;
2574
2575         if (args->addrlen > sizeof(xprt->addr)) {
2576                 dprintk("RPC:       xs_setup_xprt: address too large\n");
2577                 return ERR_PTR(-EBADF);
2578         }
2579
2580         xprt = xprt_alloc(args->net, sizeof(*new), slot_table_size,
2581                         max_slot_table_size);
2582         if (xprt == NULL) {
2583                 dprintk("RPC:       xs_setup_xprt: couldn't allocate "
2584                                 "rpc_xprt\n");
2585                 return ERR_PTR(-ENOMEM);
2586         }
2587
2588         new = container_of(xprt, struct sock_xprt, xprt);
2589         memcpy(&xprt->addr, args->dstaddr, args->addrlen);
2590         xprt->addrlen = args->addrlen;
2591         if (args->srcaddr)
2592                 memcpy(&new->srcaddr, args->srcaddr, args->addrlen);
2593         else {
2594                 int err;
2595                 err = xs_init_anyaddr(args->dstaddr->sa_family,
2596                                         (struct sockaddr *)&new->srcaddr);
2597                 if (err != 0) {
2598                         xprt_free(xprt);
2599                         return ERR_PTR(err);
2600                 }
2601         }
2602
2603         return xprt;
2604 }
2605
2606 static const struct rpc_timeout xs_local_default_timeout = {
2607         .to_initval = 10 * HZ,
2608         .to_maxval = 10 * HZ,
2609         .to_retries = 2,
2610 };
2611
2612 /**
2613  * xs_setup_local - Set up transport to use an AF_LOCAL socket
2614  * @args: rpc transport creation arguments
2615  *
2616  * AF_LOCAL is a "tpi_cots_ord" transport, just like TCP
2617  */
2618 static struct rpc_xprt *xs_setup_local(struct xprt_create *args)
2619 {
2620         struct sockaddr_un *sun = (struct sockaddr_un *)args->dstaddr;
2621         struct sock_xprt *transport;
2622         struct rpc_xprt *xprt;
2623         struct rpc_xprt *ret;
2624
2625         xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2626                         xprt_max_tcp_slot_table_entries);
2627         if (IS_ERR(xprt))
2628                 return xprt;
2629         transport = container_of(xprt, struct sock_xprt, xprt);
2630
2631         xprt->prot = 0;
2632         xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
2633         xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
2634
2635         xprt->bind_timeout = XS_BIND_TO;
2636         xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2637         xprt->idle_timeout = XS_IDLE_DISC_TO;
2638
2639         xprt->ops = &xs_local_ops;
2640         xprt->timeout = &xs_local_default_timeout;
2641
2642         switch (sun->sun_family) {
2643         case AF_LOCAL:
2644                 if (sun->sun_path[0] != '/') {
2645                         dprintk("RPC:       bad AF_LOCAL address: %s\n",
2646                                         sun->sun_path);
2647                         ret = ERR_PTR(-EINVAL);
2648                         goto out_err;
2649                 }
2650                 xprt_set_bound(xprt);
2651                 INIT_DELAYED_WORK(&transport->connect_worker,
2652                                         xs_local_setup_socket);
2653                 xs_format_peer_addresses(xprt, "local", RPCBIND_NETID_LOCAL);
2654                 break;
2655         default:
2656                 ret = ERR_PTR(-EAFNOSUPPORT);
2657                 goto out_err;
2658         }
2659
2660         dprintk("RPC:       set up xprt to %s via AF_LOCAL\n",
2661                         xprt->address_strings[RPC_DISPLAY_ADDR]);
2662
2663         if (try_module_get(THIS_MODULE))
2664                 return xprt;
2665         ret = ERR_PTR(-EINVAL);
2666 out_err:
2667         xprt_free(xprt);
2668         return ret;
2669 }
2670
2671 static const struct rpc_timeout xs_udp_default_timeout = {
2672         .to_initval = 5 * HZ,
2673         .to_maxval = 30 * HZ,
2674         .to_increment = 5 * HZ,
2675         .to_retries = 5,
2676 };
2677
2678 /**
2679  * xs_setup_udp - Set up transport to use a UDP socket
2680  * @args: rpc transport creation arguments
2681  *
2682  */
2683 static struct rpc_xprt *xs_setup_udp(struct xprt_create *args)
2684 {
2685         struct sockaddr *addr = args->dstaddr;
2686         struct rpc_xprt *xprt;
2687         struct sock_xprt *transport;
2688         struct rpc_xprt *ret;
2689
2690         xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries,
2691                         xprt_udp_slot_table_entries);
2692         if (IS_ERR(xprt))
2693                 return xprt;
2694         transport = container_of(xprt, struct sock_xprt, xprt);
2695
2696         xprt->prot = IPPROTO_UDP;
2697         xprt->tsh_size = 0;
2698         /* XXX: header size can vary due to auth type, IPv6, etc. */
2699         xprt->max_payload = (1U << 16) - (MAX_HEADER << 3);
2700
2701         xprt->bind_timeout = XS_BIND_TO;
2702         xprt->reestablish_timeout = XS_UDP_REEST_TO;
2703         xprt->idle_timeout = XS_IDLE_DISC_TO;
2704
2705         xprt->ops = &xs_udp_ops;
2706
2707         xprt->timeout = &xs_udp_default_timeout;
2708
2709         switch (addr->sa_family) {
2710         case AF_INET:
2711                 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
2712                         xprt_set_bound(xprt);
2713
2714                 INIT_DELAYED_WORK(&transport->connect_worker,
2715                                         xs_udp_setup_socket);
2716                 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP);
2717                 break;
2718         case AF_INET6:
2719                 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
2720                         xprt_set_bound(xprt);
2721
2722                 INIT_DELAYED_WORK(&transport->connect_worker,
2723                                         xs_udp_setup_socket);
2724                 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6);
2725                 break;
2726         default:
2727                 ret = ERR_PTR(-EAFNOSUPPORT);
2728                 goto out_err;
2729         }
2730
2731         if (xprt_bound(xprt))
2732                 dprintk("RPC:       set up xprt to %s (port %s) via %s\n",
2733                                 xprt->address_strings[RPC_DISPLAY_ADDR],
2734                                 xprt->address_strings[RPC_DISPLAY_PORT],
2735                                 xprt->address_strings[RPC_DISPLAY_PROTO]);
2736         else
2737                 dprintk("RPC:       set up xprt to %s (autobind) via %s\n",
2738                                 xprt->address_strings[RPC_DISPLAY_ADDR],
2739                                 xprt->address_strings[RPC_DISPLAY_PROTO]);
2740
2741         if (try_module_get(THIS_MODULE))
2742                 return xprt;
2743         ret = ERR_PTR(-EINVAL);
2744 out_err:
2745         xprt_free(xprt);
2746         return ret;
2747 }
2748
2749 static const struct rpc_timeout xs_tcp_default_timeout = {
2750         .to_initval = 60 * HZ,
2751         .to_maxval = 60 * HZ,
2752         .to_retries = 2,
2753 };
2754
2755 /**
2756  * xs_setup_tcp - Set up transport to use a TCP socket
2757  * @args: rpc transport creation arguments
2758  *
2759  */
2760 static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
2761 {
2762         struct sockaddr *addr = args->dstaddr;
2763         struct rpc_xprt *xprt;
2764         struct sock_xprt *transport;
2765         struct rpc_xprt *ret;
2766
2767         xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2768                         xprt_max_tcp_slot_table_entries);
2769         if (IS_ERR(xprt))
2770                 return xprt;
2771         transport = container_of(xprt, struct sock_xprt, xprt);
2772
2773         xprt->prot = IPPROTO_TCP;
2774         xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
2775         xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
2776
2777         xprt->bind_timeout = XS_BIND_TO;
2778         xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2779         xprt->idle_timeout = XS_IDLE_DISC_TO;
2780
2781         xprt->ops = &xs_tcp_ops;
2782         xprt->timeout = &xs_tcp_default_timeout;
2783
2784         switch (addr->sa_family) {
2785         case AF_INET:
2786                 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
2787                         xprt_set_bound(xprt);
2788
2789                 INIT_DELAYED_WORK(&transport->connect_worker,
2790                                         xs_tcp_setup_socket);
2791                 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP);
2792                 break;
2793         case AF_INET6:
2794                 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
2795                         xprt_set_bound(xprt);
2796
2797                 INIT_DELAYED_WORK(&transport->connect_worker,
2798                                         xs_tcp_setup_socket);
2799                 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6);
2800                 break;
2801         default:
2802                 ret = ERR_PTR(-EAFNOSUPPORT);
2803                 goto out_err;
2804         }
2805
2806         if (xprt_bound(xprt))
2807                 dprintk("RPC:       set up xprt to %s (port %s) via %s\n",
2808                                 xprt->address_strings[RPC_DISPLAY_ADDR],
2809                                 xprt->address_strings[RPC_DISPLAY_PORT],
2810                                 xprt->address_strings[RPC_DISPLAY_PROTO]);
2811         else
2812                 dprintk("RPC:       set up xprt to %s (autobind) via %s\n",
2813                                 xprt->address_strings[RPC_DISPLAY_ADDR],
2814                                 xprt->address_strings[RPC_DISPLAY_PROTO]);
2815
2816
2817         if (try_module_get(THIS_MODULE))
2818                 return xprt;
2819         ret = ERR_PTR(-EINVAL);
2820 out_err:
2821         xprt_free(xprt);
2822         return ret;
2823 }
2824
2825 /**
2826  * xs_setup_bc_tcp - Set up transport to use a TCP backchannel socket
2827  * @args: rpc transport creation arguments
2828  *
2829  */
2830 static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
2831 {
2832         struct sockaddr *addr = args->dstaddr;
2833         struct rpc_xprt *xprt;
2834         struct sock_xprt *transport;
2835         struct svc_sock *bc_sock;
2836         struct rpc_xprt *ret;
2837
2838         if (args->bc_xprt->xpt_bc_xprt) {
2839                 /*
2840                  * This server connection already has a backchannel
2841                  * export; we can't create a new one, as we wouldn't be
2842                  * able to match replies based on xid any more.  So,
2843                  * reuse the already-existing one:
2844                  */
2845                  return args->bc_xprt->xpt_bc_xprt;
2846         }
2847         xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2848                         xprt_tcp_slot_table_entries);
2849         if (IS_ERR(xprt))
2850                 return xprt;
2851         transport = container_of(xprt, struct sock_xprt, xprt);
2852
2853         xprt->prot = IPPROTO_TCP;
2854         xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
2855         xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
2856         xprt->timeout = &xs_tcp_default_timeout;
2857
2858         /* backchannel */
2859         xprt_set_bound(xprt);
2860         xprt->bind_timeout = 0;
2861         xprt->reestablish_timeout = 0;
2862         xprt->idle_timeout = 0;
2863
2864         xprt->ops = &bc_tcp_ops;
2865
2866         switch (addr->sa_family) {
2867         case AF_INET:
2868                 xs_format_peer_addresses(xprt, "tcp",
2869                                          RPCBIND_NETID_TCP);
2870                 break;
2871         case AF_INET6:
2872                 xs_format_peer_addresses(xprt, "tcp",
2873                                    RPCBIND_NETID_TCP6);
2874                 break;
2875         default:
2876                 ret = ERR_PTR(-EAFNOSUPPORT);
2877                 goto out_err;
2878         }
2879
2880         dprintk("RPC:       set up xprt to %s (port %s) via %s\n",
2881                         xprt->address_strings[RPC_DISPLAY_ADDR],
2882                         xprt->address_strings[RPC_DISPLAY_PORT],
2883                         xprt->address_strings[RPC_DISPLAY_PROTO]);
2884
2885         /*
2886          * Once we've associated a backchannel xprt with a connection,
2887          * we want to keep it around as long as long as the connection
2888          * lasts, in case we need to start using it for a backchannel
2889          * again; this reference won't be dropped until bc_xprt is
2890          * destroyed.
2891          */
2892         xprt_get(xprt);
2893         args->bc_xprt->xpt_bc_xprt = xprt;
2894         xprt->bc_xprt = args->bc_xprt;
2895         bc_sock = container_of(args->bc_xprt, struct svc_sock, sk_xprt);
2896         transport->sock = bc_sock->sk_sock;
2897         transport->inet = bc_sock->sk_sk;
2898
2899         /*
2900          * Since we don't want connections for the backchannel, we set
2901          * the xprt status to connected
2902          */
2903         xprt_set_connected(xprt);
2904
2905
2906         if (try_module_get(THIS_MODULE))
2907                 return xprt;
2908         xprt_put(xprt);
2909         ret = ERR_PTR(-EINVAL);
2910 out_err:
2911         xprt_free(xprt);
2912         return ret;
2913 }
2914
2915 static struct xprt_class        xs_local_transport = {
2916         .list           = LIST_HEAD_INIT(xs_local_transport.list),
2917         .name           = "named UNIX socket",
2918         .owner          = THIS_MODULE,
2919         .ident          = XPRT_TRANSPORT_LOCAL,
2920         .setup          = xs_setup_local,
2921 };
2922
2923 static struct xprt_class        xs_udp_transport = {
2924         .list           = LIST_HEAD_INIT(xs_udp_transport.list),
2925         .name           = "udp",
2926         .owner          = THIS_MODULE,
2927         .ident          = XPRT_TRANSPORT_UDP,
2928         .setup          = xs_setup_udp,
2929 };
2930
2931 static struct xprt_class        xs_tcp_transport = {
2932         .list           = LIST_HEAD_INIT(xs_tcp_transport.list),
2933         .name           = "tcp",
2934         .owner          = THIS_MODULE,
2935         .ident          = XPRT_TRANSPORT_TCP,
2936         .setup          = xs_setup_tcp,
2937 };
2938
2939 static struct xprt_class        xs_bc_tcp_transport = {
2940         .list           = LIST_HEAD_INIT(xs_bc_tcp_transport.list),
2941         .name           = "tcp NFSv4.1 backchannel",
2942         .owner          = THIS_MODULE,
2943         .ident          = XPRT_TRANSPORT_BC_TCP,
2944         .setup          = xs_setup_bc_tcp,
2945 };
2946
2947 /**
2948  * init_socket_xprt - set up xprtsock's sysctls, register with RPC client
2949  *
2950  */
2951 int init_socket_xprt(void)
2952 {
2953 #ifdef RPC_DEBUG
2954         if (!sunrpc_table_header)
2955                 sunrpc_table_header = register_sysctl_table(sunrpc_table);
2956 #endif
2957
2958         xprt_register_transport(&xs_local_transport);
2959         xprt_register_transport(&xs_udp_transport);
2960         xprt_register_transport(&xs_tcp_transport);
2961         xprt_register_transport(&xs_bc_tcp_transport);
2962
2963         return 0;
2964 }
2965
2966 /**
2967  * cleanup_socket_xprt - remove xprtsock's sysctls, unregister
2968  *
2969  */
2970 void cleanup_socket_xprt(void)
2971 {
2972 #ifdef RPC_DEBUG
2973         if (sunrpc_table_header) {
2974                 unregister_sysctl_table(sunrpc_table_header);
2975                 sunrpc_table_header = NULL;
2976         }
2977 #endif
2978
2979         xprt_unregister_transport(&xs_local_transport);
2980         xprt_unregister_transport(&xs_udp_transport);
2981         xprt_unregister_transport(&xs_tcp_transport);
2982         xprt_unregister_transport(&xs_bc_tcp_transport);
2983 }
2984
2985 static int param_set_uint_minmax(const char *val,
2986                 const struct kernel_param *kp,
2987                 unsigned int min, unsigned int max)
2988 {
2989         unsigned long num;
2990         int ret;
2991
2992         if (!val)
2993                 return -EINVAL;
2994         ret = strict_strtoul(val, 0, &num);
2995         if (ret == -EINVAL || num < min || num > max)
2996                 return -EINVAL;
2997         *((unsigned int *)kp->arg) = num;
2998         return 0;
2999 }
3000
3001 static int param_set_portnr(const char *val, const struct kernel_param *kp)
3002 {
3003         return param_set_uint_minmax(val, kp,
3004                         RPC_MIN_RESVPORT,
3005                         RPC_MAX_RESVPORT);
3006 }
3007
3008 static struct kernel_param_ops param_ops_portnr = {
3009         .set = param_set_portnr,
3010         .get = param_get_uint,
3011 };
3012
3013 #define param_check_portnr(name, p) \
3014         __param_check(name, p, unsigned int);
3015
3016 module_param_named(min_resvport, xprt_min_resvport, portnr, 0644);
3017 module_param_named(max_resvport, xprt_max_resvport, portnr, 0644);
3018
3019 static int param_set_slot_table_size(const char *val,
3020                                      const struct kernel_param *kp)
3021 {
3022         return param_set_uint_minmax(val, kp,
3023                         RPC_MIN_SLOT_TABLE,
3024                         RPC_MAX_SLOT_TABLE);
3025 }
3026
3027 static struct kernel_param_ops param_ops_slot_table_size = {
3028         .set = param_set_slot_table_size,
3029         .get = param_get_uint,
3030 };
3031
3032 #define param_check_slot_table_size(name, p) \
3033         __param_check(name, p, unsigned int);
3034
3035 static int param_set_max_slot_table_size(const char *val,
3036                                      const struct kernel_param *kp)
3037 {
3038         return param_set_uint_minmax(val, kp,
3039                         RPC_MIN_SLOT_TABLE,
3040                         RPC_MAX_SLOT_TABLE_LIMIT);
3041 }
3042
3043 static struct kernel_param_ops param_ops_max_slot_table_size = {
3044         .set = param_set_max_slot_table_size,
3045         .get = param_get_uint,
3046 };
3047
3048 #define param_check_max_slot_table_size(name, p) \
3049         __param_check(name, p, unsigned int);
3050
3051 module_param_named(tcp_slot_table_entries, xprt_tcp_slot_table_entries,
3052                    slot_table_size, 0644);
3053 module_param_named(tcp_max_slot_table_entries, xprt_max_tcp_slot_table_entries,
3054                    max_slot_table_size, 0644);
3055 module_param_named(udp_slot_table_entries, xprt_udp_slot_table_entries,
3056                    slot_table_size, 0644);
3057