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/addr.h>
  37 #include <linux/sunrpc/sched.h>
  38 #include <linux/sunrpc/svcsock.h>
  39 #include <linux/sunrpc/xprtsock.h>
  40 #include <linux/file.h>
  41 #ifdef CONFIG_SUNRPC_BACKCHANNEL
  42 #include <linux/sunrpc/bc_xprt.h>
  43 #endif
  44
  45 #include <net/sock.h>
  46 #include <net/checksum.h>
  47 #include <net/udp.h>
  48 #include <net/tcp.h>
  49
  50 #include <trace/events/sunrpc.h>
  51
  52 #include "sunrpc.h"
  53
  54 static void xs_close(struct rpc_xprt *xprt);
  55
  56 /*
  57  * xprtsock tunables
  58  */
  59 static unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
  60 static unsigned int xprt_tcp_slot_table_entries = RPC_MIN_SLOT_TABLE;
  61 static unsigned int xprt_max_tcp_slot_table_entries = RPC_MAX_SLOT_TABLE;
  62
  63 static unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT;
  64 static unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT;
  65
  66 #define XS_TCP_LINGER_TO        (15U * HZ)
  67 static unsigned int xs_tcp_fin_timeout __read_mostly = XS_TCP_LINGER_TO;
  68
  69 /*
  70  * We can register our own files under /proc/sys/sunrpc by
  71  * calling register_sysctl_table() again.  The files in that
  72  * directory become the union of all files registered there.
  73  *
  74  * We simply need to make sure that we don't collide with
  75  * someone else's file names!
  76  */
  77
  78 #ifdef RPC_DEBUG
  79
  80 static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE;
  81 static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE;
  82 static unsigned int max_tcp_slot_table_limit = RPC_MAX_SLOT_TABLE_LIMIT;
  83 static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT;
  84 static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT;
  85
  86 static struct ctl_table_header *sunrpc_table_header;
  87
  88 /*
  89  * FIXME: changing the UDP slot table size should also resize the UDP
  90  *        socket buffers for existing UDP transports
  91  */
  92 static struct ctl_table xs_tunables_table[] = {
  93         {
  94                 .procname       = "udp_slot_table_entries",
  95                 .data           = &xprt_udp_slot_table_entries,
  96                 .maxlen         = sizeof(unsigned int),
  97                 .mode           = 0644,
  98                 .proc_handler   = proc_dointvec_minmax,
  99                 .extra1         = &min_slot_table_size,
 100                 .extra2         = &max_slot_table_size
 101         },
 102         {
 103                 .procname       = "tcp_slot_table_entries",
 104                 .data           = &xprt_tcp_slot_table_entries,
 105                 .maxlen         = sizeof(unsigned int),
 106                 .mode           = 0644,
 107                 .proc_handler   = proc_dointvec_minmax,
 108                 .extra1         = &min_slot_table_size,
 109                 .extra2         = &max_slot_table_size
 110         },
 111         {
 112                 .procname       = "tcp_max_slot_table_entries",
 113                 .data           = &xprt_max_tcp_slot_table_entries,
 114                 .maxlen         = sizeof(unsigned int),
 115                 .mode           = 0644,
 116                 .proc_handler   = proc_dointvec_minmax,
 117                 .extra1         = &min_slot_table_size,
 118                 .extra2         = &max_tcp_slot_table_limit
 119         },
 120         {
 121                 .procname       = "min_resvport",
 122                 .data           = &xprt_min_resvport,
 123                 .maxlen         = sizeof(unsigned int),
 124                 .mode           = 0644,
 125                 .proc_handler   = proc_dointvec_minmax,
 126                 .extra1         = &xprt_min_resvport_limit,
 127                 .extra2         = &xprt_max_resvport_limit
 128         },
 129         {
 130                 .procname       = "max_resvport",
 131                 .data           = &xprt_max_resvport,
 132                 .maxlen         = sizeof(unsigned int),
 133                 .mode           = 0644,
 134                 .proc_handler   = proc_dointvec_minmax,
 135                 .extra1         = &xprt_min_resvport_limit,
 136                 .extra2         = &xprt_max_resvport_limit
 137         },
 138         {
 139                 .procname       = "tcp_fin_timeout",
 140                 .data           = &xs_tcp_fin_timeout,
 141                 .maxlen         = sizeof(xs_tcp_fin_timeout),
 142                 .mode           = 0644,
 143                 .proc_handler   = proc_dointvec_jiffies,
 144         },
 145         { },
 146 };
 147
 148 static struct ctl_table sunrpc_table[] = {
 149         {
 150                 .procname       = "sunrpc",
 151                 .mode           = 0555,
 152                 .child          = xs_tunables_table
 153         },
 154         { },
 155 };
 156
 157 #endif
 158
 159 /*
 160  * Wait duration for a reply from the RPC portmapper.
 161  */
 162 #define XS_BIND_TO              (60U * HZ)
 163
 164 /*
 165  * Delay if a UDP socket connect error occurs.  This is most likely some
 166  * kind of resource problem on the local host.
 167  */
 168 #define XS_UDP_REEST_TO         (2U * HZ)
 169
 170 /*
 171  * The reestablish timeout allows clients to delay for a bit before attempting
 172  * to reconnect to a server that just dropped our connection.
 173  *
 174  * We implement an exponential backoff when trying to reestablish a TCP
 175  * transport connection with the server.  Some servers like to drop a TCP
 176  * connection when they are overworked, so we start with a short timeout and
 177  * increase over time if the server is down or not responding.
 178  */
 179 #define XS_TCP_INIT_REEST_TO    (3U * HZ)
 180 #define XS_TCP_MAX_REEST_TO     (5U * 60 * HZ)
 181
 182 /*
 183  * TCP idle timeout; client drops the transport socket if it is idle
 184  * for this long.  Note that we also timeout UDP sockets to prevent
 185  * holding port numbers when there is no RPC traffic.
 186  */
 187 #define XS_IDLE_DISC_TO         (5U * 60 * HZ)
 188
 189 #ifdef RPC_DEBUG
 190 # undef  RPC_DEBUG_DATA
 191 # define RPCDBG_FACILITY        RPCDBG_TRANS
 192 #endif
 193
 194 #ifdef RPC_DEBUG_DATA
 195 static void xs_pktdump(char *msg, u32 *packet, unsigned int count)
 196 {
 197         u8 *buf = (u8 *) packet;
 198         int j;
 199
 200         dprintk("RPC:       %s\n", msg);
 201         for (j = 0; j < count && j < 128; j += 4) {
 202                 if (!(j & 31)) {
 203                         if (j)
 204                                 dprintk("\n");
 205                         dprintk("0x%04x ", j);
 206                 }
 207                 dprintk("%02x%02x%02x%02x ",
 208                         buf[j], buf[j+1], buf[j+2], buf[j+3]);
 209         }
 210         dprintk("\n");
 211 }
 212 #else
 213 static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count)
 214 {
 215         /* NOP */
 216 }
 217 #endif
 218
 219 struct sock_xprt {
 220         struct rpc_xprt         xprt;
 221
 222         /*
 223          * Network layer
 224          */
 225         struct socket *         sock;
 226         struct sock *           inet;
 227
 228         /*
 229          * State of TCP reply receive
 230          */
 231         __be32                  tcp_fraghdr,
 232                                 tcp_xid,
 233                                 tcp_calldir;
 234
 235         u32                     tcp_offset,
 236                                 tcp_reclen;
 237
 238         unsigned long           tcp_copied,
 239                                 tcp_flags;
 240
 241         /*
 242          * Connection of transports
 243          */
 244         struct delayed_work     connect_worker;
 245         struct sockaddr_storage srcaddr;
 246         unsigned short          srcport;
 247
 248         /*
 249          * UDP socket buffer size parameters
 250          */
 251         size_t                  rcvsize,
 252                                 sndsize;
 253
 254         /*
 255          * Saved socket callback addresses
 256          */
 257         void                    (*old_data_ready)(struct sock *, int);
 258         void                    (*old_state_change)(struct sock *);
 259         void                    (*old_write_space)(struct sock *);
 260 };
 261
 262 /*
 263  * TCP receive state flags
 264  */
 265 #define TCP_RCV_LAST_FRAG       (1UL << 0)
 266 #define TCP_RCV_COPY_FRAGHDR    (1UL << 1)
 267 #define TCP_RCV_COPY_XID        (1UL << 2)
 268 #define TCP_RCV_COPY_DATA       (1UL << 3)
 269 #define TCP_RCV_READ_CALLDIR    (1UL << 4)
 270 #define TCP_RCV_COPY_CALLDIR    (1UL << 5)
 271
 272 /*
 273  * TCP RPC flags
 274  */
 275 #define TCP_RPC_REPLY           (1UL << 6)
 276
 277 static inline struct sockaddr *xs_addr(struct rpc_xprt *xprt)
 278 {
 279         return (struct sockaddr *) &xprt->addr;
 280 }
 281
 282 static inline struct sockaddr_un *xs_addr_un(struct rpc_xprt *xprt)
 283 {
 284         return (struct sockaddr_un *) &xprt->addr;
 285 }
 286
 287 static inline struct sockaddr_in *xs_addr_in(struct rpc_xprt *xprt)
 288 {
 289         return (struct sockaddr_in *) &xprt->addr;
 290 }
 291
 292 static inline struct sockaddr_in6 *xs_addr_in6(struct rpc_xprt *xprt)
 293 {
 294         return (struct sockaddr_in6 *) &xprt->addr;
 295 }
 296
 297 static void xs_format_common_peer_addresses(struct rpc_xprt *xprt)
 298 {
 299         struct sockaddr *sap = xs_addr(xprt);
 300         struct sockaddr_in6 *sin6;
 301         struct sockaddr_in *sin;
 302         struct sockaddr_un *sun;
 303         char buf[128];
 304
 305         switch (sap->sa_family) {
 306         case AF_LOCAL:
 307                 sun = xs_addr_un(xprt);
 308                 strlcpy(buf, sun->sun_path, sizeof(buf));
 309                 xprt->address_strings[RPC_DISPLAY_ADDR] =
 310                                                 kstrdup(buf, GFP_KERNEL);
 311                 break;
 312         case AF_INET:
 313                 (void)rpc_ntop(sap, buf, sizeof(buf));
 314                 xprt->address_strings[RPC_DISPLAY_ADDR] =
 315                                                 kstrdup(buf, GFP_KERNEL);
 316                 sin = xs_addr_in(xprt);
 317                 snprintf(buf, sizeof(buf), "%08x", ntohl(sin->sin_addr.s_addr));
 318                 break;
 319         case AF_INET6:
 320                 (void)rpc_ntop(sap, buf, sizeof(buf));
 321                 xprt->address_strings[RPC_DISPLAY_ADDR] =
 322                                                 kstrdup(buf, GFP_KERNEL);
 323                 sin6 = xs_addr_in6(xprt);
 324                 snprintf(buf, sizeof(buf), "%pi6", &sin6->sin6_addr);
 325                 break;
 326         default:
 327                 BUG();
 328         }
 329
 330         xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = kstrdup(buf, GFP_KERNEL);
 331 }
 332
 333 static void xs_format_common_peer_ports(struct rpc_xprt *xprt)
 334 {
 335         struct sockaddr *sap = xs_addr(xprt);
 336         char buf[128];
 337
 338         snprintf(buf, sizeof(buf), "%u", rpc_get_port(sap));
 339         xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL);
 340
 341         snprintf(buf, sizeof(buf), "%4hx", rpc_get_port(sap));
 342         xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL);
 343 }
 344
 345 static void xs_format_peer_addresses(struct rpc_xprt *xprt,
 346                                      const char *protocol,
 347                                      const char *netid)
 348 {
 349         xprt->address_strings[RPC_DISPLAY_PROTO] = protocol;
 350         xprt->address_strings[RPC_DISPLAY_NETID] = netid;
 351         xs_format_common_peer_addresses(xprt);
 352         xs_format_common_peer_ports(xprt);
 353 }
 354
 355 static void xs_update_peer_port(struct rpc_xprt *xprt)
 356 {
 357         kfree(xprt->address_strings[RPC_DISPLAY_HEX_PORT]);
 358         kfree(xprt->address_strings[RPC_DISPLAY_PORT]);
 359
 360         xs_format_common_peer_ports(xprt);
 361 }
 362
 363 static void xs_free_peer_addresses(struct rpc_xprt *xprt)
 364 {
 365         unsigned int i;
 366
 367         for (i = 0; i < RPC_DISPLAY_MAX; i++)
 368                 switch (i) {
 369                 case RPC_DISPLAY_PROTO:
 370                 case RPC_DISPLAY_NETID:
 371                         continue;
 372                 default:
 373                         kfree(xprt->address_strings[i]);
 374                 }
 375 }
 376
 377 #define XS_SENDMSG_FLAGS        (MSG_DONTWAIT | MSG_NOSIGNAL)
 378
 379 static int xs_send_kvec(struct socket *sock, struct sockaddr *addr, int addrlen, struct kvec *vec, unsigned int base, int more)
 380 {
 381         struct msghdr msg = {
 382                 .msg_name       = addr,
 383                 .msg_namelen    = addrlen,
 384                 .msg_flags      = XS_SENDMSG_FLAGS | (more ? MSG_MORE : 0),
 385         };
 386         struct kvec iov = {
 387                 .iov_base       = vec->iov_base + base,
 388                 .iov_len        = vec->iov_len - base,
 389         };
 390
 391         if (iov.iov_len != 0)
 392                 return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
 393         return kernel_sendmsg(sock, &msg, NULL, 0, 0);
 394 }
 395
 396 static int xs_send_pagedata(struct socket *sock, struct xdr_buf *xdr, unsigned int base, int more)
 397 {
 398         struct page **ppage;
 399         unsigned int remainder;
 400         int err, sent = 0;
 401
 402         remainder = xdr->page_len - base;
 403         base += xdr->page_base;
 404         ppage = xdr->pages + (base >> PAGE_SHIFT);
 405         base &= ~PAGE_MASK;
 406         for(;;) {
 407                 unsigned int len = min_t(unsigned int, PAGE_SIZE - base, remainder);
 408                 int flags = XS_SENDMSG_FLAGS;
 409
 410                 remainder -= len;
 411                 if (remainder != 0 || more)
 412                         flags |= MSG_MORE;
 413                 err = sock->ops->sendpage(sock, *ppage, base, len, flags);
 414                 if (remainder == 0 || err != len)
 415                         break;
 416                 sent += err;
 417                 ppage++;
 418                 base = 0;
 419         }
 420         if (sent == 0)
 421                 return err;
 422         if (err > 0)
 423                 sent += err;
 424         return sent;
 425 }
 426
 427 /**
 428  * xs_sendpages - write pages directly to a socket
 429  * @sock: socket to send on
 430  * @addr: UDP only -- address of destination
 431  * @addrlen: UDP only -- length of destination address
 432  * @xdr: buffer containing this request
 433  * @base: starting position in the buffer
 434  *
 435  */
 436 static int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base)
 437 {
 438         unsigned int remainder = xdr->len - base;
 439         int err, sent = 0;
 440
 441         if (unlikely(!sock))
 442                 return -ENOTSOCK;
 443
 444         clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
 445         if (base != 0) {
 446                 addr = NULL;
 447                 addrlen = 0;
 448         }
 449
 450         if (base < xdr->head[0].iov_len || addr != NULL) {
 451                 unsigned int len = xdr->head[0].iov_len - base;
 452                 remainder -= len;
 453                 err = xs_send_kvec(sock, addr, addrlen, &xdr->head[0], base, remainder != 0);
 454                 if (remainder == 0 || err != len)
 455                         goto out;
 456                 sent += err;
 457                 base = 0;
 458         } else
 459                 base -= xdr->head[0].iov_len;
 460
 461         if (base < xdr->page_len) {
 462                 unsigned int len = xdr->page_len - base;
 463                 remainder -= len;
 464                 err = xs_send_pagedata(sock, xdr, base, remainder != 0);
 465                 if (remainder == 0 || err != len)
 466                         goto out;
 467                 sent += err;
 468                 base = 0;
 469         } else
 470                 base -= xdr->page_len;
 471
 472         if (base >= xdr->tail[0].iov_len)
 473                 return sent;
 474         err = xs_send_kvec(sock, NULL, 0, &xdr->tail[0], base, 0);
 475 out:
 476         if (sent == 0)
 477                 return err;
 478         if (err > 0)
 479                 sent += err;
 480         return sent;
 481 }
 482
 483 static void xs_nospace_callback(struct rpc_task *task)
 484 {
 485         struct sock_xprt *transport = container_of(task->tk_rqstp->rq_xprt, struct sock_xprt, xprt);
 486
 487         transport->inet->sk_write_pending--;
 488         clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
 489 }
 490
 491 /**
 492  * xs_nospace - place task on wait queue if transmit was incomplete
 493  * @task: task to put to sleep
 494  *
 495  */
 496 static int xs_nospace(struct rpc_task *task)
 497 {
 498         struct rpc_rqst *req = task->tk_rqstp;
 499         struct rpc_xprt *xprt = req->rq_xprt;
 500         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
 501         int ret = -EAGAIN;
 502
 503         dprintk("RPC: %5u xmit incomplete (%u left of %u)\n",
 504                         task->tk_pid, req->rq_slen - req->rq_bytes_sent,
 505                         req->rq_slen);
 506
 507         /* Protect against races with write_space */
 508         spin_lock_bh(&xprt->transport_lock);
 509
 510         /* Don't race with disconnect */
 511         if (xprt_connected(xprt)) {
 512                 if (test_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags)) {
 513                         /*
 514                          * Notify TCP that we're limited by the application
 515                          * window size
 516                          */
 517                         set_bit(SOCK_NOSPACE, &transport->sock->flags);
 518                         transport->inet->sk_write_pending++;
 519                         /* ...and wait for more buffer space */
 520                         xprt_wait_for_buffer_space(task, xs_nospace_callback);
 521                 }
 522         } else {
 523                 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
 524                 ret = -ENOTCONN;
 525         }
 526
 527         spin_unlock_bh(&xprt->transport_lock);
 528         return ret;
 529 }
 530
 531 /*
 532  * Construct a stream transport record marker in @buf.
 533  */
 534 static inline void xs_encode_stream_record_marker(struct xdr_buf *buf)
 535 {
 536         u32 reclen = buf->len - sizeof(rpc_fraghdr);
 537         rpc_fraghdr *base = buf->head[0].iov_base;
 538         *base = cpu_to_be32(RPC_LAST_STREAM_FRAGMENT | reclen);
 539 }
 540
 541 /**
 542  * xs_local_send_request - write an RPC request to an AF_LOCAL socket
 543  * @task: RPC task that manages the state of an RPC request
 544  *
 545  * Return values:
 546  *        0:    The request has been sent
 547  *   EAGAIN:    The socket was blocked, please call again later to
 548  *              complete the request
 549  * ENOTCONN:    Caller needs to invoke connect logic then call again
 550  *    other:    Some other error occured, the request was not sent
 551  */
 552 static int xs_local_send_request(struct rpc_task *task)
 553 {
 554         struct rpc_rqst *req = task->tk_rqstp;
 555         struct rpc_xprt *xprt = req->rq_xprt;
 556         struct sock_xprt *transport =
 557                                 container_of(xprt, struct sock_xprt, xprt);
 558         struct xdr_buf *xdr = &req->rq_snd_buf;
 559         int status;
 560
 561         xs_encode_stream_record_marker(&req->rq_snd_buf);
 562
 563         xs_pktdump("packet data:",
 564                         req->rq_svec->iov_base, req->rq_svec->iov_len);
 565
 566         status = xs_sendpages(transport->sock, NULL, 0,
 567                                                 xdr, req->rq_bytes_sent);
 568         dprintk("RPC:       %s(%u) = %d\n",
 569                         __func__, xdr->len - req->rq_bytes_sent, status);
 570         if (likely(status >= 0)) {
 571                 req->rq_bytes_sent += status;
 572                 req->rq_xmit_bytes_sent += status;
 573                 if (likely(req->rq_bytes_sent >= req->rq_slen)) {
 574                         req->rq_bytes_sent = 0;
 575                         return 0;
 576                 }
 577                 status = -EAGAIN;
 578         }
 579
 580         switch (status) {
 581         case -EAGAIN:
 582                 status = xs_nospace(task);
 583                 break;
 584         default:
 585                 dprintk("RPC:       sendmsg returned unrecognized error %d\n",
 586                         -status);
 587         case -EPIPE:
 588                 xs_close(xprt);
 589                 status = -ENOTCONN;
 590         }
 591
 592         return status;
 593 }
 594
 595 /**
 596  * xs_udp_send_request - write an RPC request to a UDP socket
 597  * @task: address of RPC task that manages the state of an RPC request
 598  *
 599  * Return values:
 600  *        0:    The request has been sent
 601  *   EAGAIN:    The socket was blocked, please call again later to
 602  *              complete the request
 603  * ENOTCONN:    Caller needs to invoke connect logic then call again
 604  *    other:    Some other error occurred, the request was not sent
 605  */
 606 static int xs_udp_send_request(struct rpc_task *task)
 607 {
 608         struct rpc_rqst *req = task->tk_rqstp;
 609         struct rpc_xprt *xprt = req->rq_xprt;
 610         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
 611         struct xdr_buf *xdr = &req->rq_snd_buf;
 612         int status;
 613
 614         xs_pktdump("packet data:",
 615                                 req->rq_svec->iov_base,
 616                                 req->rq_svec->iov_len);
 617
 618         if (!xprt_bound(xprt))
 619                 return -ENOTCONN;
 620         status = xs_sendpages(transport->sock,
 621                               xs_addr(xprt),
 622                               xprt->addrlen, xdr,
 623                               req->rq_bytes_sent);
 624
 625         dprintk("RPC:       xs_udp_send_request(%u) = %d\n",
 626                         xdr->len - req->rq_bytes_sent, status);
 627
 628         if (status >= 0) {
 629                 req->rq_xmit_bytes_sent += status;
 630                 if (status >= req->rq_slen)
 631                         return 0;
 632                 /* Still some bytes left; set up for a retry later. */
 633                 status = -EAGAIN;
 634         }
 635
 636         switch (status) {
 637         case -ENOTSOCK:
 638                 status = -ENOTCONN;
 639                 /* Should we call xs_close() here? */
 640                 break;
 641         case -EAGAIN:
 642                 status = xs_nospace(task);
 643                 break;
 644         default:
 645                 dprintk("RPC:       sendmsg returned unrecognized error %d\n",
 646                         -status);
 647         case -ENETUNREACH:
 648         case -EPIPE:
 649         case -ECONNREFUSED:
 650                 /* When the server has died, an ICMP port unreachable message
 651                  * prompts ECONNREFUSED. */
 652                 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
 653         }
 654
 655         return status;
 656 }
 657
 658 /**
 659  * xs_tcp_shutdown - gracefully shut down a TCP socket
 660  * @xprt: transport
 661  *
 662  * Initiates a graceful shutdown of the TCP socket by calling the
 663  * equivalent of shutdown(SHUT_WR);
 664  */
 665 static void xs_tcp_shutdown(struct rpc_xprt *xprt)
 666 {
 667         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
 668         struct socket *sock = transport->sock;
 669
 670         if (sock != NULL) {
 671                 kernel_sock_shutdown(sock, SHUT_WR);
 672                 trace_rpc_socket_shutdown(xprt, sock);
 673         }
 674 }
 675
 676 /**
 677  * xs_tcp_send_request - write an RPC request to a TCP socket
 678  * @task: address of RPC task that manages the state of an RPC request
 679  *
 680  * Return values:
 681  *        0:    The request has been sent
 682  *   EAGAIN:    The socket was blocked, please call again later to
 683  *              complete the request
 684  * ENOTCONN:    Caller needs to invoke connect logic then call again
 685  *    other:    Some other error occurred, the request was not sent
 686  *
 687  * XXX: In the case of soft timeouts, should we eventually give up
 688  *      if sendmsg is not able to make progress?
 689  */
 690 static int xs_tcp_send_request(struct rpc_task *task)
 691 {
 692         struct rpc_rqst *req = task->tk_rqstp;
 693         struct rpc_xprt *xprt = req->rq_xprt;
 694         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
 695         struct xdr_buf *xdr = &req->rq_snd_buf;
 696         int status;
 697
 698         xs_encode_stream_record_marker(&req->rq_snd_buf);
 699
 700         xs_pktdump("packet data:",
 701                                 req->rq_svec->iov_base,
 702                                 req->rq_svec->iov_len);
 703
 704         /* Continue transmitting the packet/record. We must be careful
 705          * to cope with writespace callbacks arriving _after_ we have
 706          * called sendmsg(). */
 707         while (1) {
 708                 status = xs_sendpages(transport->sock,
 709                                         NULL, 0, xdr, req->rq_bytes_sent);
 710
 711                 dprintk("RPC:       xs_tcp_send_request(%u) = %d\n",
 712                                 xdr->len - req->rq_bytes_sent, status);
 713
 714                 if (unlikely(status < 0))
 715                         break;
 716
 717                 /* If we've sent the entire packet, immediately
 718                  * reset the count of bytes sent. */
 719                 req->rq_bytes_sent += status;
 720                 req->rq_xmit_bytes_sent += status;
 721                 if (likely(req->rq_bytes_sent >= req->rq_slen)) {
 722                         req->rq_bytes_sent = 0;
 723                         return 0;
 724                 }
 725
 726                 if (status != 0)
 727                         continue;
 728                 status = -EAGAIN;
 729                 break;
 730         }
 731
 732         switch (status) {
 733         case -ENOTSOCK:
 734                 status = -ENOTCONN;
 735                 /* Should we call xs_close() here? */
 736                 break;
 737         case -EAGAIN:
 738                 status = xs_nospace(task);
 739                 break;
 740         default:
 741                 dprintk("RPC:       sendmsg returned unrecognized error %d\n",
 742                         -status);
 743         case -ECONNRESET:
 744                 xs_tcp_shutdown(xprt);
 745         case -ECONNREFUSED:
 746         case -ENOTCONN:
 747         case -EPIPE:
 748                 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
 749         }
 750
 751         return status;
 752 }
 753
 754 /**
 755  * xs_tcp_release_xprt - clean up after a tcp transmission
 756  * @xprt: transport
 757  * @task: rpc task
 758  *
 759  * This cleans up if an error causes us to abort the transmission of a request.
 760  * In this case, the socket may need to be reset in order to avoid confusing
 761  * the server.
 762  */
 763 static void xs_tcp_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
 764 {
 765         struct rpc_rqst *req;
 766
 767         if (task != xprt->snd_task)
 768                 return;
 769         if (task == NULL)
 770                 goto out_release;
 771         req = task->tk_rqstp;
 772         if (req == NULL)
 773                 goto out_release;
 774         if (req->rq_bytes_sent == 0)
 775                 goto out_release;
 776         if (req->rq_bytes_sent == req->rq_snd_buf.len)
 777                 goto out_release;
 778         set_bit(XPRT_CLOSE_WAIT, &xprt->state);
 779 out_release:
 780         xprt_release_xprt(xprt, task);
 781 }
 782
 783 static void xs_save_old_callbacks(struct sock_xprt *transport, struct sock *sk)
 784 {
 785         transport->old_data_ready = sk->sk_data_ready;
 786         transport->old_state_change = sk->sk_state_change;
 787         transport->old_write_space = sk->sk_write_space;
 788 }
 789
 790 static void xs_restore_old_callbacks(struct sock_xprt *transport, struct sock *sk)
 791 {
 792         sk->sk_data_ready = transport->old_data_ready;
 793         sk->sk_state_change = transport->old_state_change;
 794         sk->sk_write_space = transport->old_write_space;
 795 }
 796
 797 static void xs_reset_transport(struct sock_xprt *transport)
 798 {
 799         struct socket *sock = transport->sock;
 800         struct sock *sk = transport->inet;
 801
 802         if (sk == NULL)
 803                 return;
 804
 805         transport->srcport = 0;
 806
 807         write_lock_bh(&sk->sk_callback_lock);
 808         transport->inet = NULL;
 809         transport->sock = NULL;
 810
 811         sk->sk_user_data = NULL;
 812
 813         xs_restore_old_callbacks(transport, sk);
 814         write_unlock_bh(&sk->sk_callback_lock);
 815
 816         sk->sk_no_check = 0;
 817
 818         trace_rpc_socket_close(&transport->xprt, sock);
 819         sock_release(sock);
 820 }
 821
 822 /**
 823  * xs_close - close a socket
 824  * @xprt: transport
 825  *
 826  * This is used when all requests are complete; ie, no DRC state remains
 827  * on the server we want to save.
 828  *
 829  * The caller _must_ be holding XPRT_LOCKED in order to avoid issues with
 830  * xs_reset_transport() zeroing the socket from underneath a writer.
 831  */
 832 static void xs_close(struct rpc_xprt *xprt)
 833 {
 834         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
 835
 836         dprintk("RPC:       xs_close xprt %p\n", xprt);
 837
 838         cancel_delayed_work_sync(&transport->connect_worker);
 839
 840         xs_reset_transport(transport);
 841         xprt->reestablish_timeout = 0;
 842
 843         smp_mb__before_clear_bit();
 844         clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
 845         clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
 846         clear_bit(XPRT_CLOSING, &xprt->state);
 847         smp_mb__after_clear_bit();
 848         xprt_disconnect_done(xprt);
 849 }
 850
 851 static void xs_tcp_close(struct rpc_xprt *xprt)
 852 {
 853         if (test_and_clear_bit(XPRT_CONNECTION_CLOSE, &xprt->state))
 854                 xs_close(xprt);
 855         else
 856                 xs_tcp_shutdown(xprt);
 857 }
 858
 859 /**
 860  * xs_destroy - prepare to shutdown a transport
 861  * @xprt: doomed transport
 862  *
 863  */
 864 static void xs_destroy(struct rpc_xprt *xprt)
 865 {
 866         dprintk("RPC:       xs_destroy xprt %p\n", xprt);
 867
 868         xs_close(xprt);
 869         xs_free_peer_addresses(xprt);
 870         xprt_free(xprt);
 871         module_put(THIS_MODULE);
 872 }
 873
 874 static inline struct rpc_xprt *xprt_from_sock(struct sock *sk)
 875 {
 876         return (struct rpc_xprt *) sk->sk_user_data;
 877 }
 878
 879 static int xs_local_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb)
 880 {
 881         struct xdr_skb_reader desc = {
 882                 .skb            = skb,
 883                 .offset         = sizeof(rpc_fraghdr),
 884                 .count          = skb->len - sizeof(rpc_fraghdr),
 885         };
 886
 887         if (xdr_partial_copy_from_skb(xdr, 0, &desc, xdr_skb_read_bits) < 0)
 888                 return -1;
 889         if (desc.count)
 890                 return -1;
 891         return 0;
 892 }
 893
 894 /**
 895  * xs_local_data_ready - "data ready" callback for AF_LOCAL sockets
 896  * @sk: socket with data to read
 897  * @len: how much data to read
 898  *
 899  * Currently this assumes we can read the whole reply in a single gulp.
 900  */
 901 static void xs_local_data_ready(struct sock *sk, int len)
 902 {
 903         struct rpc_task *task;
 904         struct rpc_xprt *xprt;
 905         struct rpc_rqst *rovr;
 906         struct sk_buff *skb;
 907         int err, repsize, copied;
 908         u32 _xid;
 909         __be32 *xp;
 910
 911         read_lock_bh(&sk->sk_callback_lock);
 912         dprintk("RPC:       %s...\n", __func__);
 913         xprt = xprt_from_sock(sk);
 914         if (xprt == NULL)
 915                 goto out;
 916
 917         skb = skb_recv_datagram(sk, 0, 1, &err);
 918         if (skb == NULL)
 919                 goto out;
 920
 921         repsize = skb->len - sizeof(rpc_fraghdr);
 922         if (repsize < 4) {
 923                 dprintk("RPC:       impossible RPC reply size %d\n", repsize);
 924                 goto dropit;
 925         }
 926
 927         /* Copy the XID from the skb... */
 928         xp = skb_header_pointer(skb, sizeof(rpc_fraghdr), sizeof(_xid), &_xid);
 929         if (xp == NULL)
 930                 goto dropit;
 931
 932         /* Look up and lock the request corresponding to the given XID */
 933         spin_lock(&xprt->transport_lock);
 934         rovr = xprt_lookup_rqst(xprt, *xp);
 935         if (!rovr)
 936                 goto out_unlock;
 937         task = rovr->rq_task;
 938
 939         copied = rovr->rq_private_buf.buflen;
 940         if (copied > repsize)
 941                 copied = repsize;
 942
 943         if (xs_local_copy_to_xdr(&rovr->rq_private_buf, skb)) {
 944                 dprintk("RPC:       sk_buff copy failed\n");
 945                 goto out_unlock;
 946         }
 947
 948         xprt_complete_rqst(task, copied);
 949
 950  out_unlock:
 951         spin_unlock(&xprt->transport_lock);
 952  dropit:
 953         skb_free_datagram(sk, skb);
 954  out:
 955         read_unlock_bh(&sk->sk_callback_lock);
 956 }
 957
 958 /**
 959  * xs_udp_data_ready - "data ready" callback for UDP sockets
 960  * @sk: socket with data to read
 961  * @len: how much data to read
 962  *
 963  */
 964 static void xs_udp_data_ready(struct sock *sk, int len)
 965 {
 966         struct rpc_task *task;
 967         struct rpc_xprt *xprt;
 968         struct rpc_rqst *rovr;
 969         struct sk_buff *skb;
 970         int err, repsize, copied;
 971         u32 _xid;
 972         __be32 *xp;
 973
 974         read_lock_bh(&sk->sk_callback_lock);
 975         dprintk("RPC:       xs_udp_data_ready...\n");
 976         if (!(xprt = xprt_from_sock(sk)))
 977                 goto out;
 978
 979         if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL)
 980                 goto out;
 981
 982         repsize = skb->len - sizeof(struct udphdr);
 983         if (repsize < 4) {
 984                 dprintk("RPC:       impossible RPC reply size %d!\n", repsize);
 985                 goto dropit;
 986         }
 987
 988         /* Copy the XID from the skb... */
 989         xp = skb_header_pointer(skb, sizeof(struct udphdr),
 990                                 sizeof(_xid), &_xid);
 991         if (xp == NULL)
 992                 goto dropit;
 993
 994         /* Look up and lock the request corresponding to the given XID */
 995         spin_lock(&xprt->transport_lock);
 996         rovr = xprt_lookup_rqst(xprt, *xp);
 997         if (!rovr)
 998                 goto out_unlock;
 999         task = rovr->rq_task;
1000
1001         if ((copied = rovr->rq_private_buf.buflen) > repsize)
1002                 copied = repsize;
1003
1004         /* Suck it into the iovec, verify checksum if not done by hw. */
1005         if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) {
1006                 UDPX_INC_STATS_BH(sk, UDP_MIB_INERRORS);
1007                 goto out_unlock;
1008         }
1009
1010         UDPX_INC_STATS_BH(sk, UDP_MIB_INDATAGRAMS);
1011
1012         xprt_adjust_cwnd(xprt, task, copied);
1013         xprt_complete_rqst(task, copied);
1014
1015  out_unlock:
1016         spin_unlock(&xprt->transport_lock);
1017  dropit:
1018         skb_free_datagram(sk, skb);
1019  out:
1020         read_unlock_bh(&sk->sk_callback_lock);
1021 }
1022
1023 /*
1024  * Helper function to force a TCP close if the server is sending
1025  * junk and/or it has put us in CLOSE_WAIT
1026  */
1027 static void xs_tcp_force_close(struct rpc_xprt *xprt)
1028 {
1029         set_bit(XPRT_CONNECTION_CLOSE, &xprt->state);
1030         xprt_force_disconnect(xprt);
1031 }
1032
1033 static inline void xs_tcp_read_fraghdr(struct rpc_xprt *xprt, struct xdr_skb_reader *desc)
1034 {
1035         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1036         size_t len, used;
1037         char *p;
1038
1039         p = ((char *) &transport->tcp_fraghdr) + transport->tcp_offset;
1040         len = sizeof(transport->tcp_fraghdr) - transport->tcp_offset;
1041         used = xdr_skb_read_bits(desc, p, len);
1042         transport->tcp_offset += used;
1043         if (used != len)
1044                 return;
1045
1046         transport->tcp_reclen = ntohl(transport->tcp_fraghdr);
1047         if (transport->tcp_reclen & RPC_LAST_STREAM_FRAGMENT)
1048                 transport->tcp_flags |= TCP_RCV_LAST_FRAG;
1049         else
1050                 transport->tcp_flags &= ~TCP_RCV_LAST_FRAG;
1051         transport->tcp_reclen &= RPC_FRAGMENT_SIZE_MASK;
1052
1053         transport->tcp_flags &= ~TCP_RCV_COPY_FRAGHDR;
1054         transport->tcp_offset = 0;
1055
1056         /* Sanity check of the record length */
1057         if (unlikely(transport->tcp_reclen < 8)) {
1058                 dprintk("RPC:       invalid TCP record fragment length\n");
1059                 xs_tcp_force_close(xprt);
1060                 return;
1061         }
1062         dprintk("RPC:       reading TCP record fragment of length %d\n",
1063                         transport->tcp_reclen);
1064 }
1065
1066 static void xs_tcp_check_fraghdr(struct sock_xprt *transport)
1067 {
1068         if (transport->tcp_offset == transport->tcp_reclen) {
1069                 transport->tcp_flags |= TCP_RCV_COPY_FRAGHDR;
1070                 transport->tcp_offset = 0;
1071                 if (transport->tcp_flags & TCP_RCV_LAST_FRAG) {
1072                         transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1073                         transport->tcp_flags |= TCP_RCV_COPY_XID;
1074                         transport->tcp_copied = 0;
1075                 }
1076         }
1077 }
1078
1079 static inline void xs_tcp_read_xid(struct sock_xprt *transport, struct xdr_skb_reader *desc)
1080 {
1081         size_t len, used;
1082         char *p;
1083
1084         len = sizeof(transport->tcp_xid) - transport->tcp_offset;
1085         dprintk("RPC:       reading XID (%Zu bytes)\n", len);
1086         p = ((char *) &transport->tcp_xid) + transport->tcp_offset;
1087         used = xdr_skb_read_bits(desc, p, len);
1088         transport->tcp_offset += used;
1089         if (used != len)
1090                 return;
1091         transport->tcp_flags &= ~TCP_RCV_COPY_XID;
1092         transport->tcp_flags |= TCP_RCV_READ_CALLDIR;
1093         transport->tcp_copied = 4;
1094         dprintk("RPC:       reading %s XID %08x\n",
1095                         (transport->tcp_flags & TCP_RPC_REPLY) ? "reply for"
1096                                                               : "request with",
1097                         ntohl(transport->tcp_xid));
1098         xs_tcp_check_fraghdr(transport);
1099 }
1100
1101 static inline void xs_tcp_read_calldir(struct sock_xprt *transport,
1102                                        struct xdr_skb_reader *desc)
1103 {
1104         size_t len, used;
1105         u32 offset;
1106         char *p;
1107
1108         /*
1109          * We want transport->tcp_offset to be 8 at the end of this routine
1110          * (4 bytes for the xid and 4 bytes for the call/reply flag).
1111          * When this function is called for the first time,
1112          * transport->tcp_offset is 4 (after having already read the xid).
1113          */
1114         offset = transport->tcp_offset - sizeof(transport->tcp_xid);
1115         len = sizeof(transport->tcp_calldir) - offset;
1116         dprintk("RPC:       reading CALL/REPLY flag (%Zu bytes)\n", len);
1117         p = ((char *) &transport->tcp_calldir) + offset;
1118         used = xdr_skb_read_bits(desc, p, len);
1119         transport->tcp_offset += used;
1120         if (used != len)
1121                 return;
1122         transport->tcp_flags &= ~TCP_RCV_READ_CALLDIR;
1123         /*
1124          * We don't yet have the XDR buffer, so we will write the calldir
1125          * out after we get the buffer from the 'struct rpc_rqst'
1126          */
1127         switch (ntohl(transport->tcp_calldir)) {
1128         case RPC_REPLY:
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         case RPC_CALL:
1134                 transport->tcp_flags |= TCP_RCV_COPY_CALLDIR;
1135                 transport->tcp_flags |= TCP_RCV_COPY_DATA;
1136                 transport->tcp_flags &= ~TCP_RPC_REPLY;
1137                 break;
1138         default:
1139                 dprintk("RPC:       invalid request message type\n");
1140                 xs_tcp_force_close(&transport->xprt);
1141         }
1142         xs_tcp_check_fraghdr(transport);
1143 }
1144
1145 static inline void xs_tcp_read_common(struct rpc_xprt *xprt,
1146                                      struct xdr_skb_reader *desc,
1147                                      struct rpc_rqst *req)
1148 {
1149         struct sock_xprt *transport =
1150                                 container_of(xprt, struct sock_xprt, xprt);
1151         struct xdr_buf *rcvbuf;
1152         size_t len;
1153         ssize_t r;
1154
1155         rcvbuf = &req->rq_private_buf;
1156
1157         if (transport->tcp_flags & TCP_RCV_COPY_CALLDIR) {
1158                 /*
1159                  * Save the RPC direction in the XDR buffer
1160                  */
1161                 memcpy(rcvbuf->head[0].iov_base + transport->tcp_copied,
1162                         &transport->tcp_calldir,
1163                         sizeof(transport->tcp_calldir));
1164                 transport->tcp_copied += sizeof(transport->tcp_calldir);
1165                 transport->tcp_flags &= ~TCP_RCV_COPY_CALLDIR;
1166         }
1167
1168         len = desc->count;
1169         if (len > transport->tcp_reclen - transport->tcp_offset) {
1170                 struct xdr_skb_reader my_desc;
1171
1172                 len = transport->tcp_reclen - transport->tcp_offset;
1173                 memcpy(&my_desc, desc, sizeof(my_desc));
1174                 my_desc.count = len;
1175                 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
1176                                           &my_desc, xdr_skb_read_bits);
1177                 desc->count -= r;
1178                 desc->offset += r;
1179         } else
1180                 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
1181                                           desc, xdr_skb_read_bits);
1182
1183         if (r > 0) {
1184                 transport->tcp_copied += r;
1185                 transport->tcp_offset += r;
1186         }
1187         if (r != len) {
1188                 /* Error when copying to the receive buffer,
1189                  * usually because we weren't able to allocate
1190                  * additional buffer pages. All we can do now
1191                  * is turn off TCP_RCV_COPY_DATA, so the request
1192                  * will not receive any additional updates,
1193                  * and time out.
1194                  * Any remaining data from this record will
1195                  * be discarded.
1196                  */
1197                 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1198                 dprintk("RPC:       XID %08x truncated request\n",
1199                                 ntohl(transport->tcp_xid));
1200                 dprintk("RPC:       xprt = %p, tcp_copied = %lu, "
1201                                 "tcp_offset = %u, tcp_reclen = %u\n",
1202                                 xprt, transport->tcp_copied,
1203                                 transport->tcp_offset, transport->tcp_reclen);
1204                 return;
1205         }
1206
1207         dprintk("RPC:       XID %08x read %Zd bytes\n",
1208                         ntohl(transport->tcp_xid), r);
1209         dprintk("RPC:       xprt = %p, tcp_copied = %lu, tcp_offset = %u, "
1210                         "tcp_reclen = %u\n", xprt, transport->tcp_copied,
1211                         transport->tcp_offset, transport->tcp_reclen);
1212
1213         if (transport->tcp_copied == req->rq_private_buf.buflen)
1214                 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1215         else if (transport->tcp_offset == transport->tcp_reclen) {
1216                 if (transport->tcp_flags & TCP_RCV_LAST_FRAG)
1217                         transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1218         }
1219 }
1220
1221 /*
1222  * Finds the request corresponding to the RPC xid and invokes the common
1223  * tcp read code to read the data.
1224  */
1225 static inline int xs_tcp_read_reply(struct rpc_xprt *xprt,
1226                                     struct xdr_skb_reader *desc)
1227 {
1228         struct sock_xprt *transport =
1229                                 container_of(xprt, struct sock_xprt, xprt);
1230         struct rpc_rqst *req;
1231
1232         dprintk("RPC:       read reply XID %08x\n", ntohl(transport->tcp_xid));
1233
1234         /* Find and lock the request corresponding to this xid */
1235         spin_lock(&xprt->transport_lock);
1236         req = xprt_lookup_rqst(xprt, transport->tcp_xid);
1237         if (!req) {
1238                 dprintk("RPC:       XID %08x request not found!\n",
1239                                 ntohl(transport->tcp_xid));
1240                 spin_unlock(&xprt->transport_lock);
1241                 return -1;
1242         }
1243
1244         xs_tcp_read_common(xprt, desc, req);
1245
1246         if (!(transport->tcp_flags & TCP_RCV_COPY_DATA))
1247                 xprt_complete_rqst(req->rq_task, transport->tcp_copied);
1248
1249         spin_unlock(&xprt->transport_lock);
1250         return 0;
1251 }
1252
1253 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
1254 /*
1255  * Obtains an rpc_rqst previously allocated and invokes the common
1256  * tcp read code to read the data.  The result is placed in the callback
1257  * queue.
1258  * If we're unable to obtain the rpc_rqst we schedule the closing of the
1259  * connection and return -1.
1260  */
1261 static inline int xs_tcp_read_callback(struct rpc_xprt *xprt,
1262                                        struct xdr_skb_reader *desc)
1263 {
1264         struct sock_xprt *transport =
1265                                 container_of(xprt, struct sock_xprt, xprt);
1266         struct rpc_rqst *req;
1267
1268         req = xprt_alloc_bc_request(xprt);
1269         if (req == NULL) {
1270                 printk(KERN_WARNING "Callback slot table overflowed\n");
1271                 xprt_force_disconnect(xprt);
1272                 return -1;
1273         }
1274
1275         req->rq_xid = transport->tcp_xid;
1276         dprintk("RPC:       read callback  XID %08x\n", ntohl(req->rq_xid));
1277         xs_tcp_read_common(xprt, desc, req);
1278
1279         if (!(transport->tcp_flags & TCP_RCV_COPY_DATA)) {
1280                 struct svc_serv *bc_serv = xprt->bc_serv;
1281
1282                 /*
1283                  * Add callback request to callback list.  The callback
1284                  * service sleeps on the sv_cb_waitq waiting for new
1285                  * requests.  Wake it up after adding enqueing the
1286                  * request.
1287                  */
1288                 dprintk("RPC:       add callback request to list\n");
1289                 spin_lock(&bc_serv->sv_cb_lock);
1290                 list_add(&req->rq_bc_list, &bc_serv->sv_cb_list);
1291                 spin_unlock(&bc_serv->sv_cb_lock);
1292                 wake_up(&bc_serv->sv_cb_waitq);
1293         }
1294
1295         req->rq_private_buf.len = transport->tcp_copied;
1296
1297         return 0;
1298 }
1299
1300 static inline int _xs_tcp_read_data(struct rpc_xprt *xprt,
1301                                         struct xdr_skb_reader *desc)
1302 {
1303         struct sock_xprt *transport =
1304                                 container_of(xprt, struct sock_xprt, xprt);
1305
1306         return (transport->tcp_flags & TCP_RPC_REPLY) ?
1307                 xs_tcp_read_reply(xprt, desc) :
1308                 xs_tcp_read_callback(xprt, desc);
1309 }
1310 #else
1311 static inline int _xs_tcp_read_data(struct rpc_xprt *xprt,
1312                                         struct xdr_skb_reader *desc)
1313 {
1314         return xs_tcp_read_reply(xprt, desc);
1315 }
1316 #endif /* CONFIG_SUNRPC_BACKCHANNEL */
1317
1318 /*
1319  * Read data off the transport.  This can be either an RPC_CALL or an
1320  * RPC_REPLY.  Relay the processing to helper functions.
1321  */
1322 static void xs_tcp_read_data(struct rpc_xprt *xprt,
1323                                     struct xdr_skb_reader *desc)
1324 {
1325         struct sock_xprt *transport =
1326                                 container_of(xprt, struct sock_xprt, xprt);
1327
1328         if (_xs_tcp_read_data(xprt, desc) == 0)
1329                 xs_tcp_check_fraghdr(transport);
1330         else {
1331                 /*
1332                  * The transport_lock protects the request handling.
1333                  * There's no need to hold it to update the tcp_flags.
1334                  */
1335                 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1336         }
1337 }
1338
1339 static inline void xs_tcp_read_discard(struct sock_xprt *transport, struct xdr_skb_reader *desc)
1340 {
1341         size_t len;
1342
1343         len = transport->tcp_reclen - transport->tcp_offset;
1344         if (len > desc->count)
1345                 len = desc->count;
1346         desc->count -= len;
1347         desc->offset += len;
1348         transport->tcp_offset += len;
1349         dprintk("RPC:       discarded %Zu bytes\n", len);
1350         xs_tcp_check_fraghdr(transport);
1351 }
1352
1353 static int xs_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, unsigned int offset, size_t len)
1354 {
1355         struct rpc_xprt *xprt = rd_desc->arg.data;
1356         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1357         struct xdr_skb_reader desc = {
1358                 .skb    = skb,
1359                 .offset = offset,
1360                 .count  = len,
1361         };
1362
1363         dprintk("RPC:       xs_tcp_data_recv started\n");
1364         do {
1365                 /* Read in a new fragment marker if necessary */
1366                 /* Can we ever really expect to get completely empty fragments? */
1367                 if (transport->tcp_flags & TCP_RCV_COPY_FRAGHDR) {
1368                         xs_tcp_read_fraghdr(xprt, &desc);
1369                         continue;
1370                 }
1371                 /* Read in the xid if necessary */
1372                 if (transport->tcp_flags & TCP_RCV_COPY_XID) {
1373                         xs_tcp_read_xid(transport, &desc);
1374                         continue;
1375                 }
1376                 /* Read in the call/reply flag */
1377                 if (transport->tcp_flags & TCP_RCV_READ_CALLDIR) {
1378                         xs_tcp_read_calldir(transport, &desc);
1379                         continue;
1380                 }
1381                 /* Read in the request data */
1382                 if (transport->tcp_flags & TCP_RCV_COPY_DATA) {
1383                         xs_tcp_read_data(xprt, &desc);
1384                         continue;
1385                 }
1386                 /* Skip over any trailing bytes on short reads */
1387                 xs_tcp_read_discard(transport, &desc);
1388         } while (desc.count);
1389         dprintk("RPC:       xs_tcp_data_recv done\n");
1390         return len - desc.count;
1391 }
1392
1393 /**
1394  * xs_tcp_data_ready - "data ready" callback for TCP sockets
1395  * @sk: socket with data to read
1396  * @bytes: how much data to read
1397  *
1398  */
1399 static void xs_tcp_data_ready(struct sock *sk, int bytes)
1400 {
1401         struct rpc_xprt *xprt;
1402         read_descriptor_t rd_desc;
1403         int read;
1404
1405         dprintk("RPC:       xs_tcp_data_ready...\n");
1406
1407         read_lock_bh(&sk->sk_callback_lock);
1408         if (!(xprt = xprt_from_sock(sk)))
1409                 goto out;
1410         /* Any data means we had a useful conversation, so
1411          * the we don't need to delay the next reconnect
1412          */
1413         if (xprt->reestablish_timeout)
1414                 xprt->reestablish_timeout = 0;
1415
1416         /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */
1417         rd_desc.arg.data = xprt;
1418         do {
1419                 rd_desc.count = 65536;
1420                 read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv);
1421         } while (read > 0);
1422 out:
1423         read_unlock_bh(&sk->sk_callback_lock);
1424 }
1425
1426 /*
1427  * Do the equivalent of linger/linger2 handling for dealing with
1428  * broken servers that don't close the socket in a timely
1429  * fashion
1430  */
1431 static void xs_tcp_schedule_linger_timeout(struct rpc_xprt *xprt,
1432                 unsigned long timeout)
1433 {
1434         struct sock_xprt *transport;
1435
1436         if (xprt_test_and_set_connecting(xprt))
1437                 return;
1438         set_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1439         transport = container_of(xprt, struct sock_xprt, xprt);
1440         queue_delayed_work(rpciod_workqueue, &transport->connect_worker,
1441                            timeout);
1442 }
1443
1444 static void xs_tcp_cancel_linger_timeout(struct rpc_xprt *xprt)
1445 {
1446         struct sock_xprt *transport;
1447
1448         transport = container_of(xprt, struct sock_xprt, xprt);
1449
1450         if (!test_bit(XPRT_CONNECTION_ABORT, &xprt->state) ||
1451             !cancel_delayed_work(&transport->connect_worker))
1452                 return;
1453         clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1454         xprt_clear_connecting(xprt);
1455 }
1456
1457 static void xs_sock_reset_connection_flags(struct rpc_xprt *xprt)
1458 {
1459         smp_mb__before_clear_bit();
1460         clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1461         clear_bit(XPRT_CONNECTION_CLOSE, &xprt->state);
1462         clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
1463         clear_bit(XPRT_CLOSING, &xprt->state);
1464         smp_mb__after_clear_bit();
1465 }
1466
1467 static void xs_sock_mark_closed(struct rpc_xprt *xprt)
1468 {
1469         xs_sock_reset_connection_flags(xprt);
1470         /* Mark transport as closed and wake up all pending tasks */
1471         xprt_disconnect_done(xprt);
1472 }
1473
1474 /**
1475  * xs_tcp_state_change - callback to handle TCP socket state changes
1476  * @sk: socket whose state has changed
1477  *
1478  */
1479 static void xs_tcp_state_change(struct sock *sk)
1480 {
1481         struct rpc_xprt *xprt;
1482
1483         read_lock_bh(&sk->sk_callback_lock);
1484         if (!(xprt = xprt_from_sock(sk)))
1485                 goto out;
1486         dprintk("RPC:       xs_tcp_state_change client %p...\n", xprt);
1487         dprintk("RPC:       state %x conn %d dead %d zapped %d sk_shutdown %d\n",
1488                         sk->sk_state, xprt_connected(xprt),
1489                         sock_flag(sk, SOCK_DEAD),
1490                         sock_flag(sk, SOCK_ZAPPED),
1491                         sk->sk_shutdown);
1492
1493         trace_rpc_socket_state_change(xprt, sk->sk_socket);
1494         switch (sk->sk_state) {
1495         case TCP_ESTABLISHED:
1496                 spin_lock(&xprt->transport_lock);
1497                 if (!xprt_test_and_set_connected(xprt)) {
1498                         struct sock_xprt *transport = container_of(xprt,
1499                                         struct sock_xprt, xprt);
1500
1501                         /* Reset TCP record info */
1502                         transport->tcp_offset = 0;
1503                         transport->tcp_reclen = 0;
1504                         transport->tcp_copied = 0;
1505                         transport->tcp_flags =
1506                                 TCP_RCV_COPY_FRAGHDR | TCP_RCV_COPY_XID;
1507                         xprt->connect_cookie++;
1508
1509                         xprt_wake_pending_tasks(xprt, -EAGAIN);
1510                 }
1511                 spin_unlock(&xprt->transport_lock);
1512                 break;
1513         case TCP_FIN_WAIT1:
1514                 /* The client initiated a shutdown of the socket */
1515                 xprt->connect_cookie++;
1516                 xprt->reestablish_timeout = 0;
1517                 set_bit(XPRT_CLOSING, &xprt->state);
1518                 smp_mb__before_clear_bit();
1519                 clear_bit(XPRT_CONNECTED, &xprt->state);
1520                 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
1521                 smp_mb__after_clear_bit();
1522                 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout);
1523                 break;
1524         case TCP_CLOSE_WAIT:
1525                 /* The server initiated a shutdown of the socket */
1526                 xprt->connect_cookie++;
1527                 clear_bit(XPRT_CONNECTED, &xprt->state);
1528                 xs_tcp_force_close(xprt);
1529         case TCP_CLOSING:
1530                 /*
1531                  * If the server closed down the connection, make sure that
1532                  * we back off before reconnecting
1533                  */
1534                 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
1535                         xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
1536                 break;
1537         case TCP_LAST_ACK:
1538                 set_bit(XPRT_CLOSING, &xprt->state);
1539                 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout);
1540                 smp_mb__before_clear_bit();
1541                 clear_bit(XPRT_CONNECTED, &xprt->state);
1542                 smp_mb__after_clear_bit();
1543                 break;
1544         case TCP_CLOSE:
1545                 xs_tcp_cancel_linger_timeout(xprt);
1546                 xs_sock_mark_closed(xprt);
1547         }
1548  out:
1549         read_unlock_bh(&sk->sk_callback_lock);
1550 }
1551
1552 static void xs_write_space(struct sock *sk)
1553 {
1554         struct socket *sock;
1555         struct rpc_xprt *xprt;
1556
1557         if (unlikely(!(sock = sk->sk_socket)))
1558                 return;
1559         clear_bit(SOCK_NOSPACE, &sock->flags);
1560
1561         if (unlikely(!(xprt = xprt_from_sock(sk))))
1562                 return;
1563         if (test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags) == 0)
1564                 return;
1565
1566         xprt_write_space(xprt);
1567 }
1568
1569 /**
1570  * xs_udp_write_space - callback invoked when socket buffer space
1571  *                             becomes available
1572  * @sk: socket whose state has changed
1573  *
1574  * Called when more output buffer space is available for this socket.
1575  * We try not to wake our writers until they can make "significant"
1576  * progress, otherwise we'll waste resources thrashing kernel_sendmsg
1577  * with a bunch of small requests.
1578  */
1579 static void xs_udp_write_space(struct sock *sk)
1580 {
1581         read_lock_bh(&sk->sk_callback_lock);
1582
1583         /* from net/core/sock.c:sock_def_write_space */
1584         if (sock_writeable(sk))
1585                 xs_write_space(sk);
1586
1587         read_unlock_bh(&sk->sk_callback_lock);
1588 }
1589
1590 /**
1591  * xs_tcp_write_space - callback invoked when socket buffer space
1592  *                             becomes available
1593  * @sk: socket whose state has changed
1594  *
1595  * Called when more output buffer space is available for this socket.
1596  * We try not to wake our writers until they can make "significant"
1597  * progress, otherwise we'll waste resources thrashing kernel_sendmsg
1598  * with a bunch of small requests.
1599  */
1600 static void xs_tcp_write_space(struct sock *sk)
1601 {
1602         read_lock_bh(&sk->sk_callback_lock);
1603
1604         /* from net/core/stream.c:sk_stream_write_space */
1605         if (sk_stream_is_writeable(sk))
1606                 xs_write_space(sk);
1607
1608         read_unlock_bh(&sk->sk_callback_lock);
1609 }
1610
1611 static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt)
1612 {
1613         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1614         struct sock *sk = transport->inet;
1615
1616         if (transport->rcvsize) {
1617                 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
1618                 sk->sk_rcvbuf = transport->rcvsize * xprt->max_reqs * 2;
1619         }
1620         if (transport->sndsize) {
1621                 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
1622                 sk->sk_sndbuf = transport->sndsize * xprt->max_reqs * 2;
1623                 sk->sk_write_space(sk);
1624         }
1625 }
1626
1627 /**
1628  * xs_udp_set_buffer_size - set send and receive limits
1629  * @xprt: generic transport
1630  * @sndsize: requested size of send buffer, in bytes
1631  * @rcvsize: requested size of receive buffer, in bytes
1632  *
1633  * Set socket send and receive buffer size limits.
1634  */
1635 static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize)
1636 {
1637         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1638
1639         transport->sndsize = 0;
1640         if (sndsize)
1641                 transport->sndsize = sndsize + 1024;
1642         transport->rcvsize = 0;
1643         if (rcvsize)
1644                 transport->rcvsize = rcvsize + 1024;
1645
1646         xs_udp_do_set_buffer_size(xprt);
1647 }
1648
1649 /**
1650  * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport
1651  * @task: task that timed out
1652  *
1653  * Adjust the congestion window after a retransmit timeout has occurred.
1654  */
1655 static void xs_udp_timer(struct rpc_xprt *xprt, struct rpc_task *task)
1656 {
1657         xprt_adjust_cwnd(xprt, task, -ETIMEDOUT);
1658 }
1659
1660 static unsigned short xs_get_random_port(void)
1661 {
1662         unsigned short range = xprt_max_resvport - xprt_min_resvport;
1663         unsigned short rand = (unsigned short) net_random() % range;
1664         return rand + xprt_min_resvport;
1665 }
1666
1667 /**
1668  * xs_set_port - reset the port number in the remote endpoint address
1669  * @xprt: generic transport
1670  * @port: new port number
1671  *
1672  */
1673 static void xs_set_port(struct rpc_xprt *xprt, unsigned short port)
1674 {
1675         dprintk("RPC:       setting port for xprt %p to %u\n", xprt, port);
1676
1677         rpc_set_port(xs_addr(xprt), port);
1678         xs_update_peer_port(xprt);
1679 }
1680
1681 static unsigned short xs_get_srcport(struct sock_xprt *transport)
1682 {
1683         unsigned short port = transport->srcport;
1684
1685         if (port == 0 && transport->xprt.resvport)
1686                 port = xs_get_random_port();
1687         return port;
1688 }
1689
1690 static unsigned short xs_next_srcport(struct sock_xprt *transport, unsigned short port)
1691 {
1692         if (transport->srcport != 0)
1693                 transport->srcport = 0;
1694         if (!transport->xprt.resvport)
1695                 return 0;
1696         if (port <= xprt_min_resvport || port > xprt_max_resvport)
1697                 return xprt_max_resvport;
1698         return --port;
1699 }
1700 static int xs_bind(struct sock_xprt *transport, struct socket *sock)
1701 {
1702         struct sockaddr_storage myaddr;
1703         int err, nloop = 0;
1704         unsigned short port = xs_get_srcport(transport);
1705         unsigned short last;
1706
1707         memcpy(&myaddr, &transport->srcaddr, transport->xprt.addrlen);
1708         do {
1709                 rpc_set_port((struct sockaddr *)&myaddr, port);
1710                 err = kernel_bind(sock, (struct sockaddr *)&myaddr,
1711                                 transport->xprt.addrlen);
1712                 if (port == 0)
1713                         break;
1714                 if (err == 0) {
1715                         transport->srcport = port;
1716                         break;
1717                 }
1718                 last = port;
1719                 port = xs_next_srcport(transport, port);
1720                 if (port > last)
1721                         nloop++;
1722         } while (err == -EADDRINUSE && nloop != 2);
1723
1724         if (myaddr.ss_family == AF_INET)
1725                 dprintk("RPC:       %s %pI4:%u: %s (%d)\n", __func__,
1726                                 &((struct sockaddr_in *)&myaddr)->sin_addr,
1727                                 port, err ? "failed" : "ok", err);
1728         else
1729                 dprintk("RPC:       %s %pI6:%u: %s (%d)\n", __func__,
1730                                 &((struct sockaddr_in6 *)&myaddr)->sin6_addr,
1731                                 port, err ? "failed" : "ok", err);
1732         return err;
1733 }
1734
1735 /*
1736  * We don't support autobind on AF_LOCAL sockets
1737  */
1738 static void xs_local_rpcbind(struct rpc_task *task)
1739 {
1740         rcu_read_lock();
1741         xprt_set_bound(rcu_dereference(task->tk_client->cl_xprt));
1742         rcu_read_unlock();
1743 }
1744
1745 static void xs_local_set_port(struct rpc_xprt *xprt, unsigned short port)
1746 {
1747 }
1748
1749 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1750 static struct lock_class_key xs_key[2];
1751 static struct lock_class_key xs_slock_key[2];
1752
1753 static inline void xs_reclassify_socketu(struct socket *sock)
1754 {
1755         struct sock *sk = sock->sk;
1756
1757         sock_lock_init_class_and_name(sk, "slock-AF_LOCAL-RPC",
1758                 &xs_slock_key[1], "sk_lock-AF_LOCAL-RPC", &xs_key[1]);
1759 }
1760
1761 static inline void xs_reclassify_socket4(struct socket *sock)
1762 {
1763         struct sock *sk = sock->sk;
1764
1765         sock_lock_init_class_and_name(sk, "slock-AF_INET-RPC",
1766                 &xs_slock_key[0], "sk_lock-AF_INET-RPC", &xs_key[0]);
1767 }
1768
1769 static inline void xs_reclassify_socket6(struct socket *sock)
1770 {
1771         struct sock *sk = sock->sk;
1772
1773         sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC",
1774                 &xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]);
1775 }
1776
1777 static inline void xs_reclassify_socket(int family, struct socket *sock)
1778 {
1779         WARN_ON_ONCE(sock_owned_by_user(sock->sk));
1780         if (sock_owned_by_user(sock->sk))
1781                 return;
1782
1783         switch (family) {
1784         case AF_LOCAL:
1785                 xs_reclassify_socketu(sock);
1786                 break;
1787         case AF_INET:
1788                 xs_reclassify_socket4(sock);
1789                 break;
1790         case AF_INET6:
1791                 xs_reclassify_socket6(sock);
1792                 break;
1793         }
1794 }
1795 #else
1796 static inline void xs_reclassify_socketu(struct socket *sock)
1797 {
1798 }
1799
1800 static inline void xs_reclassify_socket4(struct socket *sock)
1801 {
1802 }
1803
1804 static inline void xs_reclassify_socket6(struct socket *sock)
1805 {
1806 }
1807
1808 static inline void xs_reclassify_socket(int family, struct socket *sock)
1809 {
1810 }
1811 #endif
1812
1813 static void xs_dummy_setup_socket(struct work_struct *work)
1814 {
1815 }
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_allocation = GFP_ATOMIC;
1859
1860                 xprt_clear_connected(xprt);
1861
1862                 /* Reset to new socket */
1863                 transport->sock = sock;
1864                 transport->inet = sk;
1865
1866                 write_unlock_bh(&sk->sk_callback_lock);
1867         }
1868
1869         /* Tell the socket layer to start connecting... */
1870         xprt->stat.connect_count++;
1871         xprt->stat.connect_start = jiffies;
1872         return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, 0);
1873 }
1874
1875 /**
1876  * xs_local_setup_socket - create AF_LOCAL socket, connect to a local endpoint
1877  * @xprt: RPC transport to connect
1878  * @transport: socket transport to connect
1879  * @create_sock: function to create a socket of the correct type
1880  */
1881 static int xs_local_setup_socket(struct sock_xprt *transport)
1882 {
1883         struct rpc_xprt *xprt = &transport->xprt;
1884         struct socket *sock;
1885         int status = -EIO;
1886
1887         current->flags |= PF_FSTRANS;
1888
1889         clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1890         status = __sock_create(xprt->xprt_net, AF_LOCAL,
1891                                         SOCK_STREAM, 0, &sock, 1);
1892         if (status < 0) {
1893                 dprintk("RPC:       can't create AF_LOCAL "
1894                         "transport socket (%d).\n", -status);
1895                 goto out;
1896         }
1897         xs_reclassify_socketu(sock);
1898
1899         dprintk("RPC:       worker connecting xprt %p via AF_LOCAL to %s\n",
1900                         xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1901
1902         status = xs_local_finish_connecting(xprt, sock);
1903         trace_rpc_socket_connect(xprt, sock, status);
1904         switch (status) {
1905         case 0:
1906                 dprintk("RPC:       xprt %p connected to %s\n",
1907                                 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1908                 xprt_set_connected(xprt);
1909                 break;
1910         case -ENOENT:
1911                 dprintk("RPC:       xprt %p: socket %s does not exist\n",
1912                                 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1913                 break;
1914         case -ECONNREFUSED:
1915                 dprintk("RPC:       xprt %p: connection refused for %s\n",
1916                                 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1917                 break;
1918         default:
1919                 printk(KERN_ERR "%s: unhandled error (%d) connecting to %s\n",
1920                                 __func__, -status,
1921                                 xprt->address_strings[RPC_DISPLAY_ADDR]);
1922         }
1923
1924 out:
1925         xprt_clear_connecting(xprt);
1926         xprt_wake_pending_tasks(xprt, status);
1927         current->flags &= ~PF_FSTRANS;
1928         return status;
1929 }
1930
1931 static void xs_local_connect(struct rpc_xprt *xprt, struct rpc_task *task)
1932 {
1933         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1934         int ret;
1935
1936          if (RPC_IS_ASYNC(task)) {
1937                 /*
1938                  * We want the AF_LOCAL connect to be resolved in the
1939                  * filesystem namespace of the process making the rpc
1940                  * call.  Thus we connect synchronously.
1941                  *
1942                  * If we want to support asynchronous AF_LOCAL calls,
1943                  * we'll need to figure out how to pass a namespace to
1944                  * connect.
1945                  */
1946                 rpc_exit(task, -ENOTCONN);
1947                 return;
1948         }
1949         ret = xs_local_setup_socket(transport);
1950         if (ret && !RPC_IS_SOFTCONN(task))
1951                 msleep_interruptible(15000);
1952 }
1953
1954 #ifdef CONFIG_SUNRPC_SWAP
1955 static void xs_set_memalloc(struct rpc_xprt *xprt)
1956 {
1957         struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
1958                         xprt);
1959
1960         if (xprt->swapper)
1961                 sk_set_memalloc(transport->inet);
1962 }
1963
1964 /**
1965  * xs_swapper - Tag this transport as being used for swap.
1966  * @xprt: transport to tag
1967  * @enable: enable/disable
1968  *
1969  */
1970 int xs_swapper(struct rpc_xprt *xprt, int enable)
1971 {
1972         struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
1973                         xprt);
1974         int err = 0;
1975
1976         if (enable) {
1977                 xprt->swapper++;
1978                 xs_set_memalloc(xprt);
1979         } else if (xprt->swapper) {
1980                 xprt->swapper--;
1981                 sk_clear_memalloc(transport->inet);
1982         }
1983
1984         return err;
1985 }
1986 EXPORT_SYMBOL_GPL(xs_swapper);
1987 #else
1988 static void xs_set_memalloc(struct rpc_xprt *xprt)
1989 {
1990 }
1991 #endif
1992
1993 static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
1994 {
1995         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1996
1997         if (!transport->inet) {
1998                 struct sock *sk = sock->sk;
1999
2000                 write_lock_bh(&sk->sk_callback_lock);
2001
2002                 xs_save_old_callbacks(transport, sk);
2003
2004                 sk->sk_user_data = xprt;
2005                 sk->sk_data_ready = xs_udp_data_ready;
2006                 sk->sk_write_space = xs_udp_write_space;
2007                 sk->sk_no_check = UDP_CSUM_NORCV;
2008                 sk->sk_allocation = GFP_ATOMIC;
2009
2010                 xprt_set_connected(xprt);
2011
2012                 /* Reset to new socket */
2013                 transport->sock = sock;
2014                 transport->inet = sk;
2015
2016                 xs_set_memalloc(xprt);
2017
2018                 write_unlock_bh(&sk->sk_callback_lock);
2019         }
2020         xs_udp_do_set_buffer_size(xprt);
2021 }
2022
2023 static void xs_udp_setup_socket(struct work_struct *work)
2024 {
2025         struct sock_xprt *transport =
2026                 container_of(work, struct sock_xprt, connect_worker.work);
2027         struct rpc_xprt *xprt = &transport->xprt;
2028         struct socket *sock = transport->sock;
2029         int status = -EIO;
2030
2031         current->flags |= PF_FSTRANS;
2032
2033         /* Start by resetting any existing state */
2034         xs_reset_transport(transport);
2035         sock = xs_create_sock(xprt, transport,
2036                         xs_addr(xprt)->sa_family, SOCK_DGRAM, IPPROTO_UDP);
2037         if (IS_ERR(sock))
2038                 goto out;
2039
2040         dprintk("RPC:       worker connecting xprt %p via %s to "
2041                                 "%s (port %s)\n", xprt,
2042                         xprt->address_strings[RPC_DISPLAY_PROTO],
2043                         xprt->address_strings[RPC_DISPLAY_ADDR],
2044                         xprt->address_strings[RPC_DISPLAY_PORT]);
2045
2046         xs_udp_finish_connecting(xprt, sock);
2047         trace_rpc_socket_connect(xprt, sock, 0);
2048         status = 0;
2049 out:
2050         xprt_clear_connecting(xprt);
2051         xprt_wake_pending_tasks(xprt, status);
2052         current->flags &= ~PF_FSTRANS;
2053 }
2054
2055 /*
2056  * We need to preserve the port number so the reply cache on the server can
2057  * find our cached RPC replies when we get around to reconnecting.
2058  */
2059 static void xs_abort_connection(struct sock_xprt *transport)
2060 {
2061         int result;
2062         struct sockaddr any;
2063
2064         dprintk("RPC:       disconnecting xprt %p to reuse port\n", transport);
2065
2066         /*
2067          * Disconnect the transport socket by doing a connect operation
2068          * with AF_UNSPEC.  This should return immediately...
2069          */
2070         memset(&any, 0, sizeof(any));
2071         any.sa_family = AF_UNSPEC;
2072         result = kernel_connect(transport->sock, &any, sizeof(any), 0);
2073         trace_rpc_socket_reset_connection(&transport->xprt,
2074                         transport->sock, result);
2075         if (!result)
2076                 xs_sock_reset_connection_flags(&transport->xprt);
2077         dprintk("RPC:       AF_UNSPEC connect return code %d\n", result);
2078 }
2079
2080 static void xs_tcp_reuse_connection(struct sock_xprt *transport)
2081 {
2082         unsigned int state = transport->inet->sk_state;
2083
2084         if (state == TCP_CLOSE && transport->sock->state == SS_UNCONNECTED) {
2085                 /* we don't need to abort the connection if the socket
2086                  * hasn't undergone a shutdown
2087                  */
2088                 if (transport->inet->sk_shutdown == 0)
2089                         return;
2090                 dprintk("RPC:       %s: TCP_CLOSEd and sk_shutdown set to %d\n",
2091                                 __func__, transport->inet->sk_shutdown);
2092         }
2093         if ((1 << state) & (TCPF_ESTABLISHED|TCPF_SYN_SENT)) {
2094                 /* we don't need to abort the connection if the socket
2095                  * hasn't undergone a shutdown
2096                  */
2097                 if (transport->inet->sk_shutdown == 0)
2098                         return;
2099                 dprintk("RPC:       %s: ESTABLISHED/SYN_SENT "
2100                                 "sk_shutdown set to %d\n",
2101                                 __func__, transport->inet->sk_shutdown);
2102         }
2103         xs_abort_connection(transport);
2104 }
2105
2106 static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
2107 {
2108         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2109         int ret = -ENOTCONN;
2110
2111         if (!transport->inet) {
2112                 struct sock *sk = sock->sk;
2113                 unsigned int keepidle = xprt->timeout->to_initval / HZ;
2114                 unsigned int keepcnt = xprt->timeout->to_retries + 1;
2115                 unsigned int opt_on = 1;
2116
2117                 /* TCP Keepalive options */
2118                 kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
2119                                 (char *)&opt_on, sizeof(opt_on));
2120                 kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
2121                                 (char *)&keepidle, sizeof(keepidle));
2122                 kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
2123                                 (char *)&keepidle, sizeof(keepidle));
2124                 kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
2125                                 (char *)&keepcnt, sizeof(keepcnt));
2126
2127                 write_lock_bh(&sk->sk_callback_lock);
2128
2129                 xs_save_old_callbacks(transport, sk);
2130
2131                 sk->sk_user_data = xprt;
2132                 sk->sk_data_ready = xs_tcp_data_ready;
2133                 sk->sk_state_change = xs_tcp_state_change;
2134                 sk->sk_write_space = xs_tcp_write_space;
2135                 sk->sk_allocation = GFP_ATOMIC;
2136
2137                 /* socket options */
2138                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
2139                 sock_reset_flag(sk, SOCK_LINGER);
2140                 tcp_sk(sk)->linger2 = 0;
2141                 tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF;
2142
2143                 xprt_clear_connected(xprt);
2144
2145                 /* Reset to new socket */
2146                 transport->sock = sock;
2147                 transport->inet = sk;
2148
2149                 write_unlock_bh(&sk->sk_callback_lock);
2150         }
2151
2152         if (!xprt_bound(xprt))
2153                 goto out;
2154
2155         xs_set_memalloc(xprt);
2156
2157         /* Tell the socket layer to start connecting... */
2158         xprt->stat.connect_count++;
2159         xprt->stat.connect_start = jiffies;
2160         ret = kernel_connect(sock, xs_addr(xprt), xprt->addrlen, O_NONBLOCK);
2161         switch (ret) {
2162         case 0:
2163         case -EINPROGRESS:
2164                 /* SYN_SENT! */
2165                 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
2166                         xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2167         }
2168 out:
2169         return ret;
2170 }
2171
2172 /**
2173  * xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint
2174  * @xprt: RPC transport to connect
2175  * @transport: socket transport to connect
2176  * @create_sock: function to create a socket of the correct type
2177  *
2178  * Invoked by a work queue tasklet.
2179  */
2180 static void xs_tcp_setup_socket(struct work_struct *work)
2181 {
2182         struct sock_xprt *transport =
2183                 container_of(work, struct sock_xprt, connect_worker.work);
2184         struct socket *sock = transport->sock;
2185         struct rpc_xprt *xprt = &transport->xprt;
2186         int status = -EIO;
2187
2188         current->flags |= PF_FSTRANS;
2189
2190         if (!sock) {
2191                 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
2192                 sock = xs_create_sock(xprt, transport,
2193                                 xs_addr(xprt)->sa_family, SOCK_STREAM, IPPROTO_TCP);
2194                 if (IS_ERR(sock)) {
2195                         status = PTR_ERR(sock);
2196                         goto out;
2197                 }
2198         } else {
2199                 int abort_and_exit;
2200
2201                 abort_and_exit = test_and_clear_bit(XPRT_CONNECTION_ABORT,
2202                                 &xprt->state);
2203                 /* "close" the socket, preserving the local port */
2204                 xs_tcp_reuse_connection(transport);
2205
2206                 if (abort_and_exit)
2207                         goto out_eagain;
2208         }
2209
2210         dprintk("RPC:       worker connecting xprt %p via %s to "
2211                                 "%s (port %s)\n", xprt,
2212                         xprt->address_strings[RPC_DISPLAY_PROTO],
2213                         xprt->address_strings[RPC_DISPLAY_ADDR],
2214                         xprt->address_strings[RPC_DISPLAY_PORT]);
2215
2216         status = xs_tcp_finish_connecting(xprt, sock);
2217         trace_rpc_socket_connect(xprt, sock, status);
2218         dprintk("RPC:       %p connect status %d connected %d sock state %d\n",
2219                         xprt, -status, xprt_connected(xprt),
2220                         sock->sk->sk_state);
2221         switch (status) {
2222         default:
2223                 printk("%s: connect returned unhandled error %d\n",
2224                         __func__, status);
2225         case -EADDRNOTAVAIL:
2226                 /* We're probably in TIME_WAIT. Get rid of existing socket,
2227                  * and retry
2228                  */
2229                 xs_tcp_force_close(xprt);
2230                 break;
2231         case 0:
2232         case -EINPROGRESS:
2233         case -EALREADY:
2234                 xprt_clear_connecting(xprt);
2235                 current->flags &= ~PF_FSTRANS;
2236                 return;
2237         case -EINVAL:
2238                 /* Happens, for instance, if the user specified a link
2239                  * local IPv6 address without a scope-id.
2240                  */
2241         case -ECONNREFUSED:
2242         case -ECONNRESET:
2243         case -ENETUNREACH:
2244                 /* retry with existing socket, after a delay */
2245                 goto out;
2246         }
2247 out_eagain:
2248         status = -EAGAIN;
2249 out:
2250         xprt_clear_connecting(xprt);
2251         xprt_wake_pending_tasks(xprt, status);
2252         current->flags &= ~PF_FSTRANS;
2253 }
2254
2255 /**
2256  * xs_connect - connect a socket to a remote endpoint
2257  * @xprt: pointer to transport structure
2258  * @task: address of RPC task that manages state of connect request
2259  *
2260  * TCP: If the remote end dropped the connection, delay reconnecting.
2261  *
2262  * UDP socket connects are synchronous, but we use a work queue anyway
2263  * to guarantee that even unprivileged user processes can set up a
2264  * socket on a privileged port.
2265  *
2266  * If a UDP socket connect fails, the delay behavior here prevents
2267  * retry floods (hard mounts).
2268  */
2269 static void xs_connect(struct rpc_xprt *xprt, struct rpc_task *task)
2270 {
2271         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2272
2273         if (transport->sock != NULL && !RPC_IS_SOFTCONN(task)) {
2274                 dprintk("RPC:       xs_connect delayed xprt %p for %lu "
2275                                 "seconds\n",
2276                                 xprt, xprt->reestablish_timeout / HZ);
2277                 queue_delayed_work(rpciod_workqueue,
2278                                    &transport->connect_worker,
2279                                    xprt->reestablish_timeout);
2280                 xprt->reestablish_timeout <<= 1;
2281                 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
2282                         xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2283                 if (xprt->reestablish_timeout > XS_TCP_MAX_REEST_TO)
2284                         xprt->reestablish_timeout = XS_TCP_MAX_REEST_TO;
2285         } else {
2286                 dprintk("RPC:       xs_connect scheduled xprt %p\n", xprt);
2287                 queue_delayed_work(rpciod_workqueue,
2288                                    &transport->connect_worker, 0);
2289         }
2290 }
2291
2292 /**
2293  * xs_local_print_stats - display AF_LOCAL socket-specifc stats
2294  * @xprt: rpc_xprt struct containing statistics
2295  * @seq: output file
2296  *
2297  */
2298 static void xs_local_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2299 {
2300         long idle_time = 0;
2301
2302         if (xprt_connected(xprt))
2303                 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2304
2305         seq_printf(seq, "\txprt:\tlocal %lu %lu %lu %ld %lu %lu %lu "
2306                         "%llu %llu %lu %llu %llu\n",
2307                         xprt->stat.bind_count,
2308                         xprt->stat.connect_count,
2309                         xprt->stat.connect_time,
2310                         idle_time,
2311                         xprt->stat.sends,
2312                         xprt->stat.recvs,
2313                         xprt->stat.bad_xids,
2314                         xprt->stat.req_u,
2315                         xprt->stat.bklog_u,
2316                         xprt->stat.max_slots,
2317                         xprt->stat.sending_u,
2318                         xprt->stat.pending_u);
2319 }
2320
2321 /**
2322  * xs_udp_print_stats - display UDP socket-specifc stats
2323  * @xprt: rpc_xprt struct containing statistics
2324  * @seq: output file
2325  *
2326  */
2327 static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2328 {
2329         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2330
2331         seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %llu %llu "
2332                         "%lu %llu %llu\n",
2333                         transport->srcport,
2334                         xprt->stat.bind_count,
2335                         xprt->stat.sends,
2336                         xprt->stat.recvs,
2337                         xprt->stat.bad_xids,
2338                         xprt->stat.req_u,
2339                         xprt->stat.bklog_u,
2340                         xprt->stat.max_slots,
2341                         xprt->stat.sending_u,
2342                         xprt->stat.pending_u);
2343 }
2344
2345 /**
2346  * xs_tcp_print_stats - display TCP socket-specifc stats
2347  * @xprt: rpc_xprt struct containing statistics
2348  * @seq: output file
2349  *
2350  */
2351 static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2352 {
2353         struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2354         long idle_time = 0;
2355
2356         if (xprt_connected(xprt))
2357                 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2358
2359         seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu "
2360                         "%llu %llu %lu %llu %llu\n",
2361                         transport->srcport,
2362                         xprt->stat.bind_count,
2363                         xprt->stat.connect_count,
2364                         xprt->stat.connect_time,
2365                         idle_time,
2366                         xprt->stat.sends,
2367                         xprt->stat.recvs,
2368                         xprt->stat.bad_xids,
2369                         xprt->stat.req_u,
2370                         xprt->stat.bklog_u,
2371                         xprt->stat.max_slots,
2372                         xprt->stat.sending_u,
2373                         xprt->stat.pending_u);
2374 }
2375
2376 /*
2377  * Allocate a bunch of pages for a scratch buffer for the rpc code. The reason
2378  * we allocate pages instead doing a kmalloc like rpc_malloc is because we want
2379  * to use the server side send routines.
2380  */
2381 static void *bc_malloc(struct rpc_task *task, size_t size)
2382 {
2383         struct page *page;
2384         struct rpc_buffer *buf;
2385
2386         WARN_ON_ONCE(size > PAGE_SIZE - sizeof(struct rpc_buffer));
2387         if (size > PAGE_SIZE - sizeof(struct rpc_buffer))
2388                 return NULL;
2389
2390         page = alloc_page(GFP_KERNEL);
2391         if (!page)
2392                 return NULL;
2393
2394         buf = page_address(page);
2395         buf->len = PAGE_SIZE;
2396
2397         return buf->data;
2398 }
2399
2400 /*
2401  * Free the space allocated in the bc_alloc routine
2402  */
2403 static void bc_free(void *buffer)
2404 {
2405         struct rpc_buffer *buf;
2406
2407         if (!buffer)
2408                 return;
2409
2410         buf = container_of(buffer, struct rpc_buffer, data);
2411         free_page((unsigned long)buf);
2412 }
2413
2414 /*
2415  * Use the svc_sock to send the callback. Must be called with svsk->sk_mutex
2416  * held. Borrows heavily from svc_tcp_sendto and xs_tcp_send_request.
2417  */
2418 static int bc_sendto(struct rpc_rqst *req)
2419 {
2420         int len;
2421         struct xdr_buf *xbufp = &req->rq_snd_buf;
2422         struct rpc_xprt *xprt = req->rq_xprt;
2423         struct sock_xprt *transport =
2424                                 container_of(xprt, struct sock_xprt, xprt);
2425         struct socket *sock = transport->sock;
2426         unsigned long headoff;
2427         unsigned long tailoff;
2428
2429         xs_encode_stream_record_marker(xbufp);
2430
2431         tailoff = (unsigned long)xbufp->tail[0].iov_base & ~PAGE_MASK;
2432         headoff = (unsigned long)xbufp->head[0].iov_base & ~PAGE_MASK;
2433         len = svc_send_common(sock, xbufp,
2434                               virt_to_page(xbufp->head[0].iov_base), headoff,
2435                               xbufp->tail[0].iov_base, tailoff);
2436
2437         if (len != xbufp->len) {
2438                 printk(KERN_NOTICE "Error sending entire callback!\n");
2439                 len = -EAGAIN;
2440         }
2441
2442         return len;
2443 }
2444
2445 /*
2446  * The send routine. Borrows from svc_send
2447  */
2448 static int bc_send_request(struct rpc_task *task)
2449 {
2450         struct rpc_rqst *req = task->tk_rqstp;
2451         struct svc_xprt *xprt;
2452         u32                     len;
2453
2454         dprintk("sending request with xid: %08x\n", ntohl(req->rq_xid));
2455         /*
2456          * Get the server socket associated with this callback xprt
2457          */
2458         xprt = req->rq_xprt->bc_xprt;
2459
2460         /*
2461          * Grab the mutex to serialize data as the connection is shared
2462          * with the fore channel
2463          */
2464         if (!mutex_trylock(&xprt->xpt_mutex)) {
2465                 rpc_sleep_on(&xprt->xpt_bc_pending, task, NULL);
2466                 if (!mutex_trylock(&xprt->xpt_mutex))
2467                         return -EAGAIN;
2468                 rpc_wake_up_queued_task(&xprt->xpt_bc_pending, task);
2469         }
2470         if (test_bit(XPT_DEAD, &xprt->xpt_flags))
2471                 len = -ENOTCONN;
2472         else
2473                 len = bc_sendto(req);
2474         mutex_unlock(&xprt->xpt_mutex);
2475
2476         if (len > 0)
2477                 len = 0;
2478
2479         return len;
2480 }
2481
2482 /*
2483  * The close routine. Since this is client initiated, we do nothing
2484  */
2485
2486 static void bc_close(struct rpc_xprt *xprt)
2487 {
2488 }
2489
2490 /*
2491  * The xprt destroy routine. Again, because this connection is client
2492  * initiated, we do nothing
2493  */
2494
2495 static void bc_destroy(struct rpc_xprt *xprt)
2496 {
2497 }
2498
2499 static struct rpc_xprt_ops xs_local_ops = {
2500         .reserve_xprt           = xprt_reserve_xprt,
2501         .release_xprt           = xs_tcp_release_xprt,
2502         .alloc_slot             = xprt_alloc_slot,
2503         .rpcbind                = xs_local_rpcbind,
2504         .set_port               = xs_local_set_port,
2505         .connect                = xs_local_connect,
2506         .buf_alloc              = rpc_malloc,
2507         .buf_free               = rpc_free,
2508         .send_request           = xs_local_send_request,
2509         .set_retrans_timeout    = xprt_set_retrans_timeout_def,
2510         .close                  = xs_close,
2511         .destroy                = xs_destroy,
2512         .print_stats            = xs_local_print_stats,
2513 };
2514
2515 static struct rpc_xprt_ops xs_udp_ops = {
2516         .set_buffer_size        = xs_udp_set_buffer_size,
2517         .reserve_xprt           = xprt_reserve_xprt_cong,
2518         .release_xprt           = xprt_release_xprt_cong,
2519         .alloc_slot             = xprt_alloc_slot,
2520         .rpcbind                = rpcb_getport_async,
2521         .set_port               = xs_set_port,
2522         .connect                = xs_connect,
2523         .buf_alloc              = rpc_malloc,
2524         .buf_free               = rpc_free,
2525         .send_request           = xs_udp_send_request,
2526         .set_retrans_timeout    = xprt_set_retrans_timeout_rtt,
2527         .timer                  = xs_udp_timer,
2528         .release_request        = xprt_release_rqst_cong,
2529         .close                  = xs_close,
2530         .destroy                = xs_destroy,
2531         .print_stats            = xs_udp_print_stats,
2532 };
2533
2534 static struct rpc_xprt_ops xs_tcp_ops = {
2535         .reserve_xprt           = xprt_reserve_xprt,
2536         .release_xprt           = xs_tcp_release_xprt,
2537         .alloc_slot             = xprt_lock_and_alloc_slot,
2538         .rpcbind                = rpcb_getport_async,
2539         .set_port               = xs_set_port,
2540         .connect                = xs_connect,
2541         .buf_alloc              = rpc_malloc,
2542         .buf_free               = rpc_free,
2543         .send_request           = xs_tcp_send_request,
2544         .set_retrans_timeout    = xprt_set_retrans_timeout_def,
2545         .close                  = xs_tcp_close,
2546         .destroy                = xs_destroy,
2547         .print_stats            = xs_tcp_print_stats,
2548 };
2549
2550 /*
2551  * The rpc_xprt_ops for the server backchannel
2552  */
2553
2554 static struct rpc_xprt_ops bc_tcp_ops = {
2555         .reserve_xprt           = xprt_reserve_xprt,
2556         .release_xprt           = xprt_release_xprt,
2557         .alloc_slot             = xprt_alloc_slot,
2558         .buf_alloc              = bc_malloc,
2559         .buf_free               = bc_free,
2560         .send_request           = bc_send_request,
2561         .set_retrans_timeout    = xprt_set_retrans_timeout_def,
2562         .close                  = bc_close,
2563         .destroy                = bc_destroy,
2564         .print_stats            = xs_tcp_print_stats,
2565 };
2566
2567 static int xs_init_anyaddr(const int family, struct sockaddr *sap)
2568 {
2569         static const struct sockaddr_in sin = {
2570                 .sin_family             = AF_INET,
2571                 .sin_addr.s_addr        = htonl(INADDR_ANY),
2572         };
2573         static const struct sockaddr_in6 sin6 = {
2574                 .sin6_family            = AF_INET6,
2575                 .sin6_addr              = IN6ADDR_ANY_INIT,
2576         };
2577
2578         switch (family) {
2579         case AF_LOCAL:
2580                 break;
2581         case AF_INET:
2582                 memcpy(sap, &sin, sizeof(sin));
2583                 break;
2584         case AF_INET6:
2585                 memcpy(sap, &sin6, sizeof(sin6));
2586                 break;
2587         default:
2588                 dprintk("RPC:       %s: Bad address family\n", __func__);
2589                 return -EAFNOSUPPORT;
2590         }
2591         return 0;
2592 }
2593
2594 static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args,
2595                                       unsigned int slot_table_size,
2596                                       unsigned int max_slot_table_size)
2597 {
2598         struct rpc_xprt *xprt;
2599         struct sock_xprt *new;
2600
2601         if (args->addrlen > sizeof(xprt->addr)) {
2602                 dprintk("RPC:       xs_setup_xprt: address too large\n");
2603                 return ERR_PTR(-EBADF);
2604         }
2605
2606         xprt = xprt_alloc(args->net, sizeof(*new), slot_table_size,
2607                         max_slot_table_size);
2608         if (xprt == NULL) {
2609                 dprintk("RPC:       xs_setup_xprt: couldn't allocate "
2610                                 "rpc_xprt\n");
2611                 return ERR_PTR(-ENOMEM);
2612         }
2613
2614         new = container_of(xprt, struct sock_xprt, xprt);
2615         memcpy(&xprt->addr, args->dstaddr, args->addrlen);
2616         xprt->addrlen = args->addrlen;
2617         if (args->srcaddr)
2618                 memcpy(&new->srcaddr, args->srcaddr, args->addrlen);
2619         else {
2620                 int err;
2621                 err = xs_init_anyaddr(args->dstaddr->sa_family,
2622                                         (struct sockaddr *)&new->srcaddr);
2623                 if (err != 0) {
2624                         xprt_free(xprt);
2625                         return ERR_PTR(err);
2626                 }
2627         }
2628
2629         return xprt;
2630 }
2631
2632 static const struct rpc_timeout xs_local_default_timeout = {
2633         .to_initval = 10 * HZ,
2634         .to_maxval = 10 * HZ,
2635         .to_retries = 2,
2636 };
2637
2638 /**
2639  * xs_setup_local - Set up transport to use an AF_LOCAL socket
2640  * @args: rpc transport creation arguments
2641  *
2642  * AF_LOCAL is a "tpi_cots_ord" transport, just like TCP
2643  */
2644 static struct rpc_xprt *xs_setup_local(struct xprt_create *args)
2645 {
2646         struct sockaddr_un *sun = (struct sockaddr_un *)args->dstaddr;
2647         struct sock_xprt *transport;
2648         struct rpc_xprt *xprt;
2649         struct rpc_xprt *ret;
2650
2651         xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2652                         xprt_max_tcp_slot_table_entries);
2653         if (IS_ERR(xprt))
2654                 return xprt;
2655         transport = container_of(xprt, struct sock_xprt, xprt);
2656
2657         xprt->prot = 0;
2658         xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
2659         xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
2660
2661         xprt->bind_timeout = XS_BIND_TO;
2662         xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2663         xprt->idle_timeout = XS_IDLE_DISC_TO;
2664
2665         xprt->ops = &xs_local_ops;
2666         xprt->timeout = &xs_local_default_timeout;
2667
2668         INIT_DELAYED_WORK(&transport->connect_worker,
2669                         xs_dummy_setup_socket);
2670
2671         switch (sun->sun_family) {
2672         case AF_LOCAL:
2673                 if (sun->sun_path[0] != '/') {
2674                         dprintk("RPC:       bad AF_LOCAL address: %s\n",
2675                                         sun->sun_path);
2676                         ret = ERR_PTR(-EINVAL);
2677                         goto out_err;
2678                 }
2679                 xprt_set_bound(xprt);
2680                 xs_format_peer_addresses(xprt, "local", RPCBIND_NETID_LOCAL);
2681                 ret = ERR_PTR(xs_local_setup_socket(transport));
2682                 if (ret)
2683                         goto out_err;
2684                 break;
2685         default:
2686                 ret = ERR_PTR(-EAFNOSUPPORT);
2687                 goto out_err;
2688         }
2689
2690         dprintk("RPC:       set up xprt to %s via AF_LOCAL\n",
2691                         xprt->address_strings[RPC_DISPLAY_ADDR]);
2692
2693         if (try_module_get(THIS_MODULE))
2694                 return xprt;
2695         ret = ERR_PTR(-EINVAL);
2696 out_err:
2697         xprt_free(xprt);
2698         return ret;
2699 }
2700
2701 static const struct rpc_timeout xs_udp_default_timeout = {
2702         .to_initval = 5 * HZ,
2703         .to_maxval = 30 * HZ,
2704         .to_increment = 5 * HZ,
2705         .to_retries = 5,
2706 };
2707
2708 /**
2709  * xs_setup_udp - Set up transport to use a UDP socket
2710  * @args: rpc transport creation arguments
2711  *
2712  */
2713 static struct rpc_xprt *xs_setup_udp(struct xprt_create *args)
2714 {
2715         struct sockaddr *addr = args->dstaddr;
2716         struct rpc_xprt *xprt;
2717         struct sock_xprt *transport;
2718         struct rpc_xprt *ret;
2719
2720         xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries,
2721                         xprt_udp_slot_table_entries);
2722         if (IS_ERR(xprt))
2723                 return xprt;
2724         transport = container_of(xprt, struct sock_xprt, xprt);
2725
2726         xprt->prot = IPPROTO_UDP;
2727         xprt->tsh_size = 0;
2728         /* XXX: header size can vary due to auth type, IPv6, etc. */
2729         xprt->max_payload = (1U << 16) - (MAX_HEADER << 3);
2730
2731         xprt->bind_timeout = XS_BIND_TO;
2732         xprt->reestablish_timeout = XS_UDP_REEST_TO;
2733         xprt->idle_timeout = XS_IDLE_DISC_TO;
2734
2735         xprt->ops = &xs_udp_ops;
2736
2737         xprt->timeout = &xs_udp_default_timeout;
2738
2739         switch (addr->sa_family) {
2740         case AF_INET:
2741                 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
2742                         xprt_set_bound(xprt);
2743
2744                 INIT_DELAYED_WORK(&transport->connect_worker,
2745                                         xs_udp_setup_socket);
2746                 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP);
2747                 break;
2748         case AF_INET6:
2749                 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
2750                         xprt_set_bound(xprt);
2751
2752                 INIT_DELAYED_WORK(&transport->connect_worker,
2753                                         xs_udp_setup_socket);
2754                 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6);
2755                 break;
2756         default:
2757                 ret = ERR_PTR(-EAFNOSUPPORT);
2758                 goto out_err;
2759         }
2760
2761         if (xprt_bound(xprt))
2762                 dprintk("RPC:       set up xprt to %s (port %s) via %s\n",
2763                                 xprt->address_strings[RPC_DISPLAY_ADDR],
2764                                 xprt->address_strings[RPC_DISPLAY_PORT],
2765                                 xprt->address_strings[RPC_DISPLAY_PROTO]);
2766         else
2767                 dprintk("RPC:       set up xprt to %s (autobind) via %s\n",
2768                                 xprt->address_strings[RPC_DISPLAY_ADDR],
2769                                 xprt->address_strings[RPC_DISPLAY_PROTO]);
2770
2771         if (try_module_get(THIS_MODULE))
2772                 return xprt;
2773         ret = ERR_PTR(-EINVAL);
2774 out_err:
2775         xprt_free(xprt);
2776         return ret;
2777 }
2778
2779 static const struct rpc_timeout xs_tcp_default_timeout = {
2780         .to_initval = 60 * HZ,
2781         .to_maxval = 60 * HZ,
2782         .to_retries = 2,
2783 };
2784
2785 /**
2786  * xs_setup_tcp - Set up transport to use a TCP socket
2787  * @args: rpc transport creation arguments
2788  *
2789  */
2790 static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
2791 {
2792         struct sockaddr *addr = args->dstaddr;
2793         struct rpc_xprt *xprt;
2794         struct sock_xprt *transport;
2795         struct rpc_xprt *ret;
2796         unsigned int max_slot_table_size = xprt_max_tcp_slot_table_entries;
2797
2798         if (args->flags & XPRT_CREATE_INFINITE_SLOTS)
2799                 max_slot_table_size = RPC_MAX_SLOT_TABLE_LIMIT;
2800
2801         xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2802                         max_slot_table_size);
2803         if (IS_ERR(xprt))
2804                 return xprt;
2805         transport = container_of(xprt, struct sock_xprt, xprt);
2806
2807         xprt->prot = IPPROTO_TCP;
2808         xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
2809         xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
2810
2811         xprt->bind_timeout = XS_BIND_TO;
2812         xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2813         xprt->idle_timeout = XS_IDLE_DISC_TO;
2814
2815         xprt->ops = &xs_tcp_ops;
2816         xprt->timeout = &xs_tcp_default_timeout;
2817
2818         switch (addr->sa_family) {
2819         case AF_INET:
2820                 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
2821                         xprt_set_bound(xprt);
2822
2823                 INIT_DELAYED_WORK(&transport->connect_worker,
2824                                         xs_tcp_setup_socket);
2825                 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP);
2826                 break;
2827         case AF_INET6:
2828                 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
2829                         xprt_set_bound(xprt);
2830
2831                 INIT_DELAYED_WORK(&transport->connect_worker,
2832                                         xs_tcp_setup_socket);
2833                 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6);
2834                 break;
2835         default:
2836                 ret = ERR_PTR(-EAFNOSUPPORT);
2837                 goto out_err;
2838         }
2839
2840         if (xprt_bound(xprt))
2841                 dprintk("RPC:       set up xprt to %s (port %s) via %s\n",
2842                                 xprt->address_strings[RPC_DISPLAY_ADDR],
2843                                 xprt->address_strings[RPC_DISPLAY_PORT],
2844                                 xprt->address_strings[RPC_DISPLAY_PROTO]);
2845         else
2846                 dprintk("RPC:       set up xprt to %s (autobind) via %s\n",
2847                                 xprt->address_strings[RPC_DISPLAY_ADDR],
2848                                 xprt->address_strings[RPC_DISPLAY_PROTO]);
2849
2850
2851         if (try_module_get(THIS_MODULE))
2852                 return xprt;
2853         ret = ERR_PTR(-EINVAL);
2854 out_err:
2855         xprt_free(xprt);
2856         return ret;
2857 }
2858
2859 /**
2860  * xs_setup_bc_tcp - Set up transport to use a TCP backchannel socket
2861  * @args: rpc transport creation arguments
2862  *
2863  */
2864 static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
2865 {
2866         struct sockaddr *addr = args->dstaddr;
2867         struct rpc_xprt *xprt;
2868         struct sock_xprt *transport;
2869         struct svc_sock *bc_sock;
2870         struct rpc_xprt *ret;
2871
2872         if (args->bc_xprt->xpt_bc_xprt) {
2873                 /*
2874                  * This server connection already has a backchannel
2875                  * transport; we can't create a new one, as we wouldn't
2876                  * be able to match replies based on xid any more.  So,
2877                  * reuse the already-existing one:
2878                  */
2879                  return args->bc_xprt->xpt_bc_xprt;
2880         }
2881         xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2882                         xprt_tcp_slot_table_entries);
2883         if (IS_ERR(xprt))
2884                 return xprt;
2885         transport = container_of(xprt, struct sock_xprt, xprt);
2886
2887         xprt->prot = IPPROTO_TCP;
2888         xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
2889         xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
2890         xprt->timeout = &xs_tcp_default_timeout;
2891
2892         /* backchannel */
2893         xprt_set_bound(xprt);
2894         xprt->bind_timeout = 0;
2895         xprt->reestablish_timeout = 0;
2896         xprt->idle_timeout = 0;
2897
2898         xprt->ops = &bc_tcp_ops;
2899
2900         switch (addr->sa_family) {
2901         case AF_INET:
2902                 xs_format_peer_addresses(xprt, "tcp",
2903                                          RPCBIND_NETID_TCP);
2904                 break;
2905         case AF_INET6:
2906                 xs_format_peer_addresses(xprt, "tcp",
2907                                    RPCBIND_NETID_TCP6);
2908                 break;
2909         default:
2910                 ret = ERR_PTR(-EAFNOSUPPORT);
2911                 goto out_err;
2912         }
2913
2914         dprintk("RPC:       set up xprt to %s (port %s) via %s\n",
2915                         xprt->address_strings[RPC_DISPLAY_ADDR],
2916                         xprt->address_strings[RPC_DISPLAY_PORT],
2917                         xprt->address_strings[RPC_DISPLAY_PROTO]);
2918
2919         /*
2920          * Once we've associated a backchannel xprt with a connection,
2921          * we want to keep it around as long as long as the connection
2922          * lasts, in case we need to start using it for a backchannel
2923          * again; this reference won't be dropped until bc_xprt is
2924          * destroyed.
2925          */
2926         xprt_get(xprt);
2927         args->bc_xprt->xpt_bc_xprt = xprt;
2928         xprt->bc_xprt = args->bc_xprt;
2929         bc_sock = container_of(args->bc_xprt, struct svc_sock, sk_xprt);
2930         transport->sock = bc_sock->sk_sock;
2931         transport->inet = bc_sock->sk_sk;
2932
2933         /*
2934          * Since we don't want connections for the backchannel, we set
2935          * the xprt status to connected
2936          */
2937         xprt_set_connected(xprt);
2938
2939
2940         if (try_module_get(THIS_MODULE))
2941                 return xprt;
2942         xprt_put(xprt);
2943         ret = ERR_PTR(-EINVAL);
2944 out_err:
2945         xprt_free(xprt);
2946         return ret;
2947 }
2948
2949 static struct xprt_class        xs_local_transport = {
2950         .list           = LIST_HEAD_INIT(xs_local_transport.list),
2951         .name           = "named UNIX socket",
2952         .owner          = THIS_MODULE,
2953         .ident          = XPRT_TRANSPORT_LOCAL,
2954         .setup          = xs_setup_local,
2955 };
2956
2957 static struct xprt_class        xs_udp_transport = {
2958         .list           = LIST_HEAD_INIT(xs_udp_transport.list),
2959         .name           = "udp",
2960         .owner          = THIS_MODULE,
2961         .ident          = XPRT_TRANSPORT_UDP,
2962         .setup          = xs_setup_udp,
2963 };
2964
2965 static struct xprt_class        xs_tcp_transport = {
2966         .list           = LIST_HEAD_INIT(xs_tcp_transport.list),
2967         .name           = "tcp",
2968         .owner          = THIS_MODULE,
2969         .ident          = XPRT_TRANSPORT_TCP,
2970         .setup          = xs_setup_tcp,
2971 };
2972
2973 static struct xprt_class        xs_bc_tcp_transport = {
2974         .list           = LIST_HEAD_INIT(xs_bc_tcp_transport.list),
2975         .name           = "tcp NFSv4.1 backchannel",
2976         .owner          = THIS_MODULE,
2977         .ident          = XPRT_TRANSPORT_BC_TCP,
2978         .setup          = xs_setup_bc_tcp,
2979 };
2980
2981 /**
2982  * init_socket_xprt - set up xprtsock's sysctls, register with RPC client
2983  *
2984  */
2985 int init_socket_xprt(void)
2986 {
2987 #ifdef RPC_DEBUG
2988         if (!sunrpc_table_header)
2989                 sunrpc_table_header = register_sysctl_table(sunrpc_table);
2990 #endif
2991
2992         xprt_register_transport(&xs_local_transport);
2993         xprt_register_transport(&xs_udp_transport);
2994         xprt_register_transport(&xs_tcp_transport);
2995         xprt_register_transport(&xs_bc_tcp_transport);
2996
2997         return 0;
2998 }
2999
3000 /**
3001  * cleanup_socket_xprt - remove xprtsock's sysctls, unregister
3002  *
3003  */
3004 void cleanup_socket_xprt(void)
3005 {
3006 #ifdef RPC_DEBUG
3007         if (sunrpc_table_header) {
3008                 unregister_sysctl_table(sunrpc_table_header);
3009                 sunrpc_table_header = NULL;
3010         }
3011 #endif
3012
3013         xprt_unregister_transport(&xs_local_transport);
3014         xprt_unregister_transport(&xs_udp_transport);
3015         xprt_unregister_transport(&xs_tcp_transport);
3016         xprt_unregister_transport(&xs_bc_tcp_transport);
3017 }
3018
3019 static int param_set_uint_minmax(const char *val,
3020                 const struct kernel_param *kp,
3021                 unsigned int min, unsigned int max)
3022 {
3023         unsigned long num;
3024         int ret;
3025
3026         if (!val)
3027                 return -EINVAL;
3028         ret = strict_strtoul(val, 0, &num);
3029         if (ret == -EINVAL || num < min || num > max)
3030                 return -EINVAL;
3031         *((unsigned int *)kp->arg) = num;
3032         return 0;
3033 }
3034
3035 static int param_set_portnr(const char *val, const struct kernel_param *kp)
3036 {
3037         return param_set_uint_minmax(val, kp,
3038                         RPC_MIN_RESVPORT,
3039                         RPC_MAX_RESVPORT);
3040 }
3041
3042 static struct kernel_param_ops param_ops_portnr = {
3043         .set = param_set_portnr,
3044         .get = param_get_uint,
3045 };
3046
3047 #define param_check_portnr(name, p) \
3048         __param_check(name, p, unsigned int);
3049
3050 module_param_named(min_resvport, xprt_min_resvport, portnr, 0644);
3051 module_param_named(max_resvport, xprt_max_resvport, portnr, 0644);
3052
3053 static int param_set_slot_table_size(const char *val,
3054                                      const struct kernel_param *kp)
3055 {
3056         return param_set_uint_minmax(val, kp,
3057                         RPC_MIN_SLOT_TABLE,
3058                         RPC_MAX_SLOT_TABLE);
3059 }
3060
3061 static struct kernel_param_ops param_ops_slot_table_size = {
3062         .set = param_set_slot_table_size,
3063         .get = param_get_uint,
3064 };
3065
3066 #define param_check_slot_table_size(name, p) \
3067         __param_check(name, p, unsigned int);
3068
3069 static int param_set_max_slot_table_size(const char *val,
3070                                      const struct kernel_param *kp)
3071 {
3072         return param_set_uint_minmax(val, kp,
3073                         RPC_MIN_SLOT_TABLE,
3074                         RPC_MAX_SLOT_TABLE_LIMIT);
3075 }
3076
3077 static struct kernel_param_ops param_ops_max_slot_table_size = {
3078         .set = param_set_max_slot_table_size,
3079         .get = param_get_uint,
3080 };
3081
3082 #define param_check_max_slot_table_size(name, p) \
3083         __param_check(name, p, unsigned int);
3084
3085 module_param_named(tcp_slot_table_entries, xprt_tcp_slot_table_entries,
3086                    slot_table_size, 0644);
3087 module_param_named(tcp_max_slot_table_entries, xprt_max_tcp_slot_table_entries,
3088                    max_slot_table_size, 0644);
3089 module_param_named(udp_slot_table_entries, xprt_udp_slot_table_entries,
3090                    slot_table_size, 0644);
3091