git.oblomov.eu Git - linux-2.6/blob - net/sunrpc/xprt.c

   1 /*
   2  *  linux/net/sunrpc/xprt.c
   3  *
   4  *  This is a generic RPC call interface supporting congestion avoidance,
   5  *  and asynchronous calls.
   6  *
   7  *  The interface works like this:
   8  *
   9  *  -   When a process places a call, it allocates a request slot if
  10  *      one is available. Otherwise, it sleeps on the backlog queue
  11  *      (xprt_reserve).
  12  *  -   Next, the caller puts together the RPC message, stuffs it into
  13  *      the request struct, and calls xprt_call().
  14  *  -   xprt_call transmits the message and installs the caller on the
  15  *      socket's wait list. At the same time, it installs a timer that
  16  *      is run after the packet's timeout has expired.
  17  *  -   When a packet arrives, the data_ready handler walks the list of
  18  *      pending requests for that socket. If a matching XID is found, the
  19  *      caller is woken up, and the timer removed.
  20  *  -   When no reply arrives within the timeout interval, the timer is
  21  *      fired by the kernel and runs xprt_timer(). It either adjusts the
  22  *      timeout values (minor timeout) or wakes up the caller with a status
  23  *      of -ETIMEDOUT.
  24  *  -   When the caller receives a notification from RPC that a reply arrived,
  25  *      it should release the RPC slot, and process the reply.
  26  *      If the call timed out, it may choose to retry the operation by
  27  *      adjusting the initial timeout value, and simply calling rpc_call
  28  *      again.
  29  *
  30  *  Support for async RPC is done through a set of RPC-specific scheduling
  31  *  primitives that `transparently' work for processes as well as async
  32  *  tasks that rely on callbacks.
  33  *
  34  *  Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de>
  35  *
  36  *  TCP callback races fixes (C) 1998 Red Hat Software <alan@redhat.com>
  37  *  TCP send fixes (C) 1998 Red Hat Software <alan@redhat.com>
  38  *  TCP NFS related read + write fixes
  39  *   (C) 1999 Dave Airlie, University of Limerick, Ireland <airlied@linux.ie>
  40  *
  41  *  Rewrite of larges part of the code in order to stabilize TCP stuff.
  42  *  Fix behaviour when socket buffer is full.
  43  *   (C) 1999 Trond Myklebust <trond.myklebust@fys.uio.no>
  44  */
  45
  46 #include <linux/types.h>
  47 #include <linux/slab.h>
  48 #include <linux/capability.h>
  49 #include <linux/sched.h>
  50 #include <linux/errno.h>
  51 #include <linux/socket.h>
  52 #include <linux/in.h>
  53 #include <linux/net.h>
  54 #include <linux/mm.h>
  55 #include <linux/udp.h>
  56 #include <linux/tcp.h>
  57 #include <linux/sunrpc/clnt.h>
  58 #include <linux/file.h>
  59 #include <linux/workqueue.h>
  60 #include <linux/random.h>
  61
  62 #include <net/sock.h>
  63 #include <net/checksum.h>
  64 #include <net/udp.h>
  65 #include <net/tcp.h>
  66
  67 /*
  68  * Local variables
  69  */
  70
  71 #ifdef RPC_DEBUG
  72 # undef  RPC_DEBUG_DATA
  73 # define RPCDBG_FACILITY        RPCDBG_XPRT
  74 #endif
  75
  76 #define XPRT_MAX_BACKOFF        (8)
  77 #define XPRT_IDLE_TIMEOUT       (5*60*HZ)
  78 #define XPRT_MAX_RESVPORT       (800)
  79
  80 /*
  81  * Local functions
  82  */
  83 static void     xprt_request_init(struct rpc_task *, struct rpc_xprt *);
  84 static inline void      do_xprt_reserve(struct rpc_task *);
  85 static void     xprt_disconnect(struct rpc_xprt *);
  86 static void     xprt_connect_status(struct rpc_task *task);
  87 static struct rpc_xprt * xprt_setup(int proto, struct sockaddr_in *ap,
  88                                                 struct rpc_timeout *to);
  89 static struct socket *xprt_create_socket(struct rpc_xprt *, int, int);
  90 static void     xprt_bind_socket(struct rpc_xprt *, struct socket *);
  91 static int      __xprt_get_cong(struct rpc_xprt *, struct rpc_task *);
  92
  93 static int      xprt_clear_backlog(struct rpc_xprt *xprt);
  94
  95 #ifdef RPC_DEBUG_DATA
  96 /*
  97  * Print the buffer contents (first 128 bytes only--just enough for
  98  * diropres return).
  99  */
 100 static void
 101 xprt_pktdump(char *msg, u32 *packet, unsigned int count)
 102 {
 103         u8      *buf = (u8 *) packet;
 104         int     j;
 105
 106         dprintk("RPC:      %s\n", msg);
 107         for (j = 0; j < count && j < 128; j += 4) {
 108                 if (!(j & 31)) {
 109                         if (j)
 110                                 dprintk("\n");
 111                         dprintk("0x%04x ", j);
 112                 }
 113                 dprintk("%02x%02x%02x%02x ",
 114                         buf[j], buf[j+1], buf[j+2], buf[j+3]);
 115         }
 116         dprintk("\n");
 117 }
 118 #else
 119 static inline void
 120 xprt_pktdump(char *msg, u32 *packet, unsigned int count)
 121 {
 122         /* NOP */
 123 }
 124 #endif
 125
 126 /*
 127  * Look up RPC transport given an INET socket
 128  */
 129 static inline struct rpc_xprt *
 130 xprt_from_sock(struct sock *sk)
 131 {
 132         return (struct rpc_xprt *) sk->sk_user_data;
 133 }
 134
 135 /*
 136  * Serialize write access to sockets, in order to prevent different
 137  * requests from interfering with each other.
 138  * Also prevents TCP socket connects from colliding with writes.
 139  */
 140 static int
 141 __xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
 142 {
 143         struct rpc_rqst *req = task->tk_rqstp;
 144
 145         if (test_and_set_bit(XPRT_LOCKED, &xprt->sockstate)) {
 146                 if (task == xprt->snd_task)
 147                         return 1;
 148                 if (task == NULL)
 149                         return 0;
 150                 goto out_sleep;
 151         }
 152         if (xprt->nocong || __xprt_get_cong(xprt, task)) {
 153                 xprt->snd_task = task;
 154                 if (req) {
 155                         req->rq_bytes_sent = 0;
 156                         req->rq_ntrans++;
 157                 }
 158                 return 1;
 159         }
 160         smp_mb__before_clear_bit();
 161         clear_bit(XPRT_LOCKED, &xprt->sockstate);
 162         smp_mb__after_clear_bit();
 163 out_sleep:
 164         dprintk("RPC: %4d failed to lock socket %p\n", task->tk_pid, xprt);
 165         task->tk_timeout = 0;
 166         task->tk_status = -EAGAIN;
 167         if (req && req->rq_ntrans)
 168                 rpc_sleep_on(&xprt->resend, task, NULL, NULL);
 169         else
 170                 rpc_sleep_on(&xprt->sending, task, NULL, NULL);
 171         return 0;
 172 }
 173
 174 static inline int
 175 xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
 176 {
 177         int retval;
 178
 179         spin_lock_bh(&xprt->sock_lock);
 180         retval = __xprt_lock_write(xprt, task);
 181         spin_unlock_bh(&xprt->sock_lock);
 182         return retval;
 183 }
 184
 185
 186 static void
 187 __xprt_lock_write_next(struct rpc_xprt *xprt)
 188 {
 189         struct rpc_task *task;
 190
 191         if (test_and_set_bit(XPRT_LOCKED, &xprt->sockstate))
 192                 return;
 193         if (!xprt->nocong && RPCXPRT_CONGESTED(xprt))
 194                 goto out_unlock;
 195         task = rpc_wake_up_next(&xprt->resend);
 196         if (!task) {
 197                 task = rpc_wake_up_next(&xprt->sending);
 198                 if (!task)
 199                         goto out_unlock;
 200         }
 201         if (xprt->nocong || __xprt_get_cong(xprt, task)) {
 202                 struct rpc_rqst *req = task->tk_rqstp;
 203                 xprt->snd_task = task;
 204                 if (req) {
 205                         req->rq_bytes_sent = 0;
 206                         req->rq_ntrans++;
 207                 }
 208                 return;
 209         }
 210 out_unlock:
 211         smp_mb__before_clear_bit();
 212         clear_bit(XPRT_LOCKED, &xprt->sockstate);
 213         smp_mb__after_clear_bit();
 214 }
 215
 216 /*
 217  * Releases the socket for use by other requests.
 218  */
 219 static void
 220 __xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
 221 {
 222         if (xprt->snd_task == task) {
 223                 xprt->snd_task = NULL;
 224                 smp_mb__before_clear_bit();
 225                 clear_bit(XPRT_LOCKED, &xprt->sockstate);
 226                 smp_mb__after_clear_bit();
 227                 __xprt_lock_write_next(xprt);
 228         }
 229 }
 230
 231 static inline void
 232 xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
 233 {
 234         spin_lock_bh(&xprt->sock_lock);
 235         __xprt_release_write(xprt, task);
 236         spin_unlock_bh(&xprt->sock_lock);
 237 }
 238
 239 /*
 240  * Write data to socket.
 241  */
 242 static inline int
 243 xprt_sendmsg(struct rpc_xprt *xprt, struct rpc_rqst *req)
 244 {
 245         struct socket   *sock = xprt->sock;
 246         struct xdr_buf  *xdr = &req->rq_snd_buf;
 247         struct sockaddr *addr = NULL;
 248         int addrlen = 0;
 249         unsigned int    skip;
 250         int             result;
 251
 252         if (!sock)
 253                 return -ENOTCONN;
 254
 255         xprt_pktdump("packet data:",
 256                                 req->rq_svec->iov_base,
 257                                 req->rq_svec->iov_len);
 258
 259         /* For UDP, we need to provide an address */
 260         if (!xprt->stream) {
 261                 addr = (struct sockaddr *) &xprt->addr;
 262                 addrlen = sizeof(xprt->addr);
 263         }
 264         /* Dont repeat bytes */
 265         skip = req->rq_bytes_sent;
 266
 267         clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
 268         result = xdr_sendpages(sock, addr, addrlen, xdr, skip, MSG_DONTWAIT);
 269
 270         dprintk("RPC:      xprt_sendmsg(%d) = %d\n", xdr->len - skip, result);
 271
 272         if (result >= 0)
 273                 return result;
 274
 275         switch (result) {
 276         case -ECONNREFUSED:
 277                 /* When the server has died, an ICMP port unreachable message
 278                  * prompts ECONNREFUSED.
 279                  */
 280         case -EAGAIN:
 281                 break;
 282         case -ECONNRESET:
 283         case -ENOTCONN:
 284         case -EPIPE:
 285                 /* connection broken */
 286                 if (xprt->stream)
 287                         result = -ENOTCONN;
 288                 break;
 289         default:
 290                 printk(KERN_NOTICE "RPC: sendmsg returned error %d\n", -result);
 291         }
 292         return result;
 293 }
 294
 295 /*
 296  * Van Jacobson congestion avoidance. Check if the congestion window
 297  * overflowed. Put the task to sleep if this is the case.
 298  */
 299 static int
 300 __xprt_get_cong(struct rpc_xprt *xprt, struct rpc_task *task)
 301 {
 302         struct rpc_rqst *req = task->tk_rqstp;
 303
 304         if (req->rq_cong)
 305                 return 1;
 306         dprintk("RPC: %4d xprt_cwnd_limited cong = %ld cwnd = %ld\n",
 307                         task->tk_pid, xprt->cong, xprt->cwnd);
 308         if (RPCXPRT_CONGESTED(xprt))
 309                 return 0;
 310         req->rq_cong = 1;
 311         xprt->cong += RPC_CWNDSCALE;
 312         return 1;
 313 }
 314
 315 /*
 316  * Adjust the congestion window, and wake up the next task
 317  * that has been sleeping due to congestion
 318  */
 319 static void
 320 __xprt_put_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
 321 {
 322         if (!req->rq_cong)
 323                 return;
 324         req->rq_cong = 0;
 325         xprt->cong -= RPC_CWNDSCALE;
 326         __xprt_lock_write_next(xprt);
 327 }
 328
 329 /*
 330  * Adjust RPC congestion window
 331  * We use a time-smoothed congestion estimator to avoid heavy oscillation.
 332  */
 333 static void
 334 xprt_adjust_cwnd(struct rpc_xprt *xprt, int result)
 335 {
 336         unsigned long   cwnd;
 337
 338         cwnd = xprt->cwnd;
 339         if (result >= 0 && cwnd <= xprt->cong) {
 340                 /* The (cwnd >> 1) term makes sure
 341                  * the result gets rounded properly. */
 342                 cwnd += (RPC_CWNDSCALE * RPC_CWNDSCALE + (cwnd >> 1)) / cwnd;
 343                 if (cwnd > RPC_MAXCWND(xprt))
 344                         cwnd = RPC_MAXCWND(xprt);
 345                 __xprt_lock_write_next(xprt);
 346         } else if (result == -ETIMEDOUT) {
 347                 cwnd >>= 1;
 348                 if (cwnd < RPC_CWNDSCALE)
 349                         cwnd = RPC_CWNDSCALE;
 350         }
 351         dprintk("RPC:      cong %ld, cwnd was %ld, now %ld\n",
 352                         xprt->cong, xprt->cwnd, cwnd);
 353         xprt->cwnd = cwnd;
 354 }
 355
 356 /*
 357  * Reset the major timeout value
 358  */
 359 static void xprt_reset_majortimeo(struct rpc_rqst *req)
 360 {
 361         struct rpc_timeout *to = &req->rq_xprt->timeout;
 362
 363         req->rq_majortimeo = req->rq_timeout;
 364         if (to->to_exponential)
 365                 req->rq_majortimeo <<= to->to_retries;
 366         else
 367                 req->rq_majortimeo += to->to_increment * to->to_retries;
 368         if (req->rq_majortimeo > to->to_maxval || req->rq_majortimeo == 0)
 369                 req->rq_majortimeo = to->to_maxval;
 370         req->rq_majortimeo += jiffies;
 371 }
 372
 373 /*
 374  * Adjust timeout values etc for next retransmit
 375  */
 376 int xprt_adjust_timeout(struct rpc_rqst *req)
 377 {
 378         struct rpc_xprt *xprt = req->rq_xprt;
 379         struct rpc_timeout *to = &xprt->timeout;
 380         int status = 0;
 381
 382         if (time_before(jiffies, req->rq_majortimeo)) {
 383                 if (to->to_exponential)
 384                         req->rq_timeout <<= 1;
 385                 else
 386                         req->rq_timeout += to->to_increment;
 387                 if (to->to_maxval && req->rq_timeout >= to->to_maxval)
 388                         req->rq_timeout = to->to_maxval;
 389                 req->rq_retries++;
 390                 pprintk("RPC: %lu retrans\n", jiffies);
 391         } else {
 392                 req->rq_timeout = to->to_initval;
 393                 req->rq_retries = 0;
 394                 xprt_reset_majortimeo(req);
 395                 /* Reset the RTT counters == "slow start" */
 396                 spin_lock_bh(&xprt->sock_lock);
 397                 rpc_init_rtt(req->rq_task->tk_client->cl_rtt, to->to_initval);
 398                 spin_unlock_bh(&xprt->sock_lock);
 399                 pprintk("RPC: %lu timeout\n", jiffies);
 400                 status = -ETIMEDOUT;
 401         }
 402
 403         if (req->rq_timeout == 0) {
 404                 printk(KERN_WARNING "xprt_adjust_timeout: rq_timeout = 0!\n");
 405                 req->rq_timeout = 5 * HZ;
 406         }
 407         return status;
 408 }
 409
 410 /*
 411  * Close down a transport socket
 412  */
 413 static void
 414 xprt_close(struct rpc_xprt *xprt)
 415 {
 416         struct socket   *sock = xprt->sock;
 417         struct sock     *sk = xprt->inet;
 418
 419         if (!sk)
 420                 return;
 421
 422         write_lock_bh(&sk->sk_callback_lock);
 423         xprt->inet = NULL;
 424         xprt->sock = NULL;
 425
 426         sk->sk_user_data    = NULL;
 427         sk->sk_data_ready   = xprt->old_data_ready;
 428         sk->sk_state_change = xprt->old_state_change;
 429         sk->sk_write_space  = xprt->old_write_space;
 430         write_unlock_bh(&sk->sk_callback_lock);
 431
 432         sk->sk_no_check  = 0;
 433
 434         sock_release(sock);
 435 }
 436
 437 static void
 438 xprt_socket_autoclose(void *args)
 439 {
 440         struct rpc_xprt *xprt = (struct rpc_xprt *)args;
 441
 442         xprt_disconnect(xprt);
 443         xprt_close(xprt);
 444         xprt_release_write(xprt, NULL);
 445 }
 446
 447 /*
 448  * Mark a transport as disconnected
 449  */
 450 static void
 451 xprt_disconnect(struct rpc_xprt *xprt)
 452 {
 453         dprintk("RPC:      disconnected transport %p\n", xprt);
 454         spin_lock_bh(&xprt->sock_lock);
 455         xprt_clear_connected(xprt);
 456         rpc_wake_up_status(&xprt->pending, -ENOTCONN);
 457         spin_unlock_bh(&xprt->sock_lock);
 458 }
 459
 460 /*
 461  * Used to allow disconnection when we've been idle
 462  */
 463 static void
 464 xprt_init_autodisconnect(unsigned long data)
 465 {
 466         struct rpc_xprt *xprt = (struct rpc_xprt *)data;
 467
 468         spin_lock(&xprt->sock_lock);
 469         if (!list_empty(&xprt->recv) || xprt->shutdown)
 470                 goto out_abort;
 471         if (test_and_set_bit(XPRT_LOCKED, &xprt->sockstate))
 472                 goto out_abort;
 473         spin_unlock(&xprt->sock_lock);
 474         /* Let keventd close the socket */
 475         if (test_bit(XPRT_CONNECTING, &xprt->sockstate) != 0)
 476                 xprt_release_write(xprt, NULL);
 477         else
 478                 schedule_work(&xprt->task_cleanup);
 479         return;
 480 out_abort:
 481         spin_unlock(&xprt->sock_lock);
 482 }
 483
 484 static void xprt_socket_connect(void *args)
 485 {
 486         struct rpc_xprt *xprt = (struct rpc_xprt *)args;
 487         struct socket *sock = xprt->sock;
 488         int status = -EIO;
 489
 490         if (xprt->shutdown || xprt->addr.sin_port == 0)
 491                 goto out;
 492
 493         /*
 494          * Start by resetting any existing state
 495          */
 496         xprt_close(xprt);
 497         sock = xprt_create_socket(xprt, xprt->prot, xprt->resvport);
 498         if (sock == NULL) {
 499                 /* couldn't create socket or bind to reserved port;
 500                  * this is likely a permanent error, so cause an abort */
 501                 goto out;
 502         }
 503         xprt_bind_socket(xprt, sock);
 504         xprt_sock_setbufsize(xprt);
 505
 506         status = 0;
 507         if (!xprt->stream)
 508                 goto out;
 509
 510         /*
 511          * Tell the socket layer to start connecting...
 512          */
 513         status = sock->ops->connect(sock, (struct sockaddr *) &xprt->addr,
 514                         sizeof(xprt->addr), O_NONBLOCK);
 515         dprintk("RPC: %p  connect status %d connected %d sock state %d\n",
 516                         xprt, -status, xprt_connected(xprt), sock->sk->sk_state);
 517         if (status < 0) {
 518                 switch (status) {
 519                         case -EINPROGRESS:
 520                         case -EALREADY:
 521                                 goto out_clear;
 522                 }
 523         }
 524 out:
 525         if (status < 0)
 526                 rpc_wake_up_status(&xprt->pending, status);
 527         else
 528                 rpc_wake_up(&xprt->pending);
 529 out_clear:
 530         smp_mb__before_clear_bit();
 531         clear_bit(XPRT_CONNECTING, &xprt->sockstate);
 532         smp_mb__after_clear_bit();
 533 }
 534
 535 /*
 536  * Attempt to connect a TCP socket.
 537  *
 538  */
 539 void xprt_connect(struct rpc_task *task)
 540 {
 541         struct rpc_xprt *xprt = task->tk_xprt;
 542
 543         dprintk("RPC: %4d xprt_connect xprt %p %s connected\n", task->tk_pid,
 544                         xprt, (xprt_connected(xprt) ? "is" : "is not"));
 545
 546         if (xprt->shutdown) {
 547                 task->tk_status = -EIO;
 548                 return;
 549         }
 550         if (!xprt->addr.sin_port) {
 551                 task->tk_status = -EIO;
 552                 return;
 553         }
 554         if (!xprt_lock_write(xprt, task))
 555                 return;
 556         if (xprt_connected(xprt))
 557                 goto out_write;
 558
 559         if (task->tk_rqstp)
 560                 task->tk_rqstp->rq_bytes_sent = 0;
 561
 562         task->tk_timeout = RPC_CONNECT_TIMEOUT;
 563         rpc_sleep_on(&xprt->pending, task, xprt_connect_status, NULL);
 564         if (!test_and_set_bit(XPRT_CONNECTING, &xprt->sockstate)) {
 565                 /* Note: if we are here due to a dropped connection
 566                  *       we delay reconnecting by RPC_REESTABLISH_TIMEOUT/HZ
 567                  *       seconds
 568                  */
 569                 if (xprt->sock != NULL)
 570                         schedule_delayed_work(&xprt->sock_connect,
 571                                         RPC_REESTABLISH_TIMEOUT);
 572                 else {
 573                         schedule_work(&xprt->sock_connect);
 574                         if (!RPC_IS_ASYNC(task))
 575                                 flush_scheduled_work();
 576                 }
 577         }
 578         return;
 579  out_write:
 580         xprt_release_write(xprt, task);
 581 }
 582
 583 /*
 584  * We arrive here when awoken from waiting on connection establishment.
 585  */
 586 static void
 587 xprt_connect_status(struct rpc_task *task)
 588 {
 589         struct rpc_xprt *xprt = task->tk_xprt;
 590
 591         if (task->tk_status >= 0) {
 592                 dprintk("RPC: %4d xprt_connect_status: connection established\n",
 593                                 task->tk_pid);
 594                 return;
 595         }
 596
 597         /* if soft mounted, just cause this RPC to fail */
 598         if (RPC_IS_SOFT(task))
 599                 task->tk_status = -EIO;
 600
 601         switch (task->tk_status) {
 602         case -ECONNREFUSED:
 603         case -ECONNRESET:
 604         case -ENOTCONN:
 605                 return;
 606         case -ETIMEDOUT:
 607                 dprintk("RPC: %4d xprt_connect_status: timed out\n",
 608                                 task->tk_pid);
 609                 break;
 610         default:
 611                 printk(KERN_ERR "RPC: error %d connecting to server %s\n",
 612                                 -task->tk_status, task->tk_client->cl_server);
 613         }
 614         xprt_release_write(xprt, task);
 615 }
 616
 617 /*
 618  * Look up the RPC request corresponding to a reply, and then lock it.
 619  */
 620 static inline struct rpc_rqst *
 621 xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid)
 622 {
 623         struct list_head *pos;
 624         struct rpc_rqst *req = NULL;
 625
 626         list_for_each(pos, &xprt->recv) {
 627                 struct rpc_rqst *entry = list_entry(pos, struct rpc_rqst, rq_list);
 628                 if (entry->rq_xid == xid) {
 629                         req = entry;
 630                         break;
 631                 }
 632         }
 633         return req;
 634 }
 635
 636 /*
 637  * Complete reply received.
 638  * The TCP code relies on us to remove the request from xprt->pending.
 639  */
 640 static void
 641 xprt_complete_rqst(struct rpc_xprt *xprt, struct rpc_rqst *req, int copied)
 642 {
 643         struct rpc_task *task = req->rq_task;
 644         struct rpc_clnt *clnt = task->tk_client;
 645
 646         /* Adjust congestion window */
 647         if (!xprt->nocong) {
 648                 unsigned timer = task->tk_msg.rpc_proc->p_timer;
 649                 xprt_adjust_cwnd(xprt, copied);
 650                 __xprt_put_cong(xprt, req);
 651                 if (timer) {
 652                         if (req->rq_ntrans == 1)
 653                                 rpc_update_rtt(clnt->cl_rtt, timer,
 654                                                 (long)jiffies - req->rq_xtime);
 655                         rpc_set_timeo(clnt->cl_rtt, timer, req->rq_ntrans - 1);
 656                 }
 657         }
 658
 659 #ifdef RPC_PROFILE
 660         /* Profile only reads for now */
 661         if (copied > 1024) {
 662                 static unsigned long    nextstat;
 663                 static unsigned long    pkt_rtt, pkt_len, pkt_cnt;
 664
 665                 pkt_cnt++;
 666                 pkt_len += req->rq_slen + copied;
 667                 pkt_rtt += jiffies - req->rq_xtime;
 668                 if (time_before(nextstat, jiffies)) {
 669                         printk("RPC: %lu %ld cwnd\n", jiffies, xprt->cwnd);
 670                         printk("RPC: %ld %ld %ld %ld stat\n",
 671                                         jiffies, pkt_cnt, pkt_len, pkt_rtt);
 672                         pkt_rtt = pkt_len = pkt_cnt = 0;
 673                         nextstat = jiffies + 5 * HZ;
 674                 }
 675         }
 676 #endif
 677
 678         dprintk("RPC: %4d has input (%d bytes)\n", task->tk_pid, copied);
 679         list_del_init(&req->rq_list);
 680         req->rq_received = req->rq_private_buf.len = copied;
 681
 682         /* ... and wake up the process. */
 683         rpc_wake_up_task(task);
 684         return;
 685 }
 686
 687 static size_t
 688 skb_read_bits(skb_reader_t *desc, void *to, size_t len)
 689 {
 690         if (len > desc->count)
 691                 len = desc->count;
 692         if (skb_copy_bits(desc->skb, desc->offset, to, len))
 693                 return 0;
 694         desc->count -= len;
 695         desc->offset += len;
 696         return len;
 697 }
 698
 699 static size_t
 700 skb_read_and_csum_bits(skb_reader_t *desc, void *to, size_t len)
 701 {
 702         unsigned int csum2, pos;
 703
 704         if (len > desc->count)
 705                 len = desc->count;
 706         pos = desc->offset;
 707         csum2 = skb_copy_and_csum_bits(desc->skb, pos, to, len, 0);
 708         desc->csum = csum_block_add(desc->csum, csum2, pos);
 709         desc->count -= len;
 710         desc->offset += len;
 711         return len;
 712 }
 713
 714 /*
 715  * We have set things up such that we perform the checksum of the UDP
 716  * packet in parallel with the copies into the RPC client iovec.  -DaveM
 717  */
 718 int
 719 csum_partial_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb)
 720 {
 721         skb_reader_t desc;
 722
 723         desc.skb = skb;
 724         desc.offset = sizeof(struct udphdr);
 725         desc.count = skb->len - desc.offset;
 726
 727         if (skb->ip_summed == CHECKSUM_UNNECESSARY)
 728                 goto no_checksum;
 729
 730         desc.csum = csum_partial(skb->data, desc.offset, skb->csum);
 731         if (xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_and_csum_bits) < 0)
 732                 return -1;
 733         if (desc.offset != skb->len) {
 734                 unsigned int csum2;
 735                 csum2 = skb_checksum(skb, desc.offset, skb->len - desc.offset, 0);
 736                 desc.csum = csum_block_add(desc.csum, csum2, desc.offset);
 737         }
 738         if (desc.count)
 739                 return -1;
 740         if ((unsigned short)csum_fold(desc.csum))
 741                 return -1;
 742         return 0;
 743 no_checksum:
 744         if (xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_bits) < 0)
 745                 return -1;
 746         if (desc.count)
 747                 return -1;
 748         return 0;
 749 }
 750
 751 /*
 752  * Input handler for RPC replies. Called from a bottom half and hence
 753  * atomic.
 754  */
 755 static void
 756 udp_data_ready(struct sock *sk, int len)
 757 {
 758         struct rpc_task *task;
 759         struct rpc_xprt *xprt;
 760         struct rpc_rqst *rovr;
 761         struct sk_buff  *skb;
 762         int err, repsize, copied;
 763         u32 _xid, *xp;
 764
 765         read_lock(&sk->sk_callback_lock);
 766         dprintk("RPC:      udp_data_ready...\n");
 767         if (!(xprt = xprt_from_sock(sk))) {
 768                 printk("RPC:      udp_data_ready request not found!\n");
 769                 goto out;
 770         }
 771
 772         dprintk("RPC:      udp_data_ready client %p\n", xprt);
 773
 774         if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL)
 775                 goto out;
 776
 777         if (xprt->shutdown)
 778                 goto dropit;
 779
 780         repsize = skb->len - sizeof(struct udphdr);
 781         if (repsize < 4) {
 782                 printk("RPC: impossible RPC reply size %d!\n", repsize);
 783                 goto dropit;
 784         }
 785
 786         /* Copy the XID from the skb... */
 787         xp = skb_header_pointer(skb, sizeof(struct udphdr),
 788                                 sizeof(_xid), &_xid);
 789         if (xp == NULL)
 790                 goto dropit;
 791
 792         /* Look up and lock the request corresponding to the given XID */
 793         spin_lock(&xprt->sock_lock);
 794         rovr = xprt_lookup_rqst(xprt, *xp);
 795         if (!rovr)
 796                 goto out_unlock;
 797         task = rovr->rq_task;
 798
 799         dprintk("RPC: %4d received reply\n", task->tk_pid);
 800
 801         if ((copied = rovr->rq_private_buf.buflen) > repsize)
 802                 copied = repsize;
 803
 804         /* Suck it into the iovec, verify checksum if not done by hw. */
 805         if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb))
 806                 goto out_unlock;
 807
 808         /* Something worked... */
 809         dst_confirm(skb->dst);
 810
 811         xprt_complete_rqst(xprt, rovr, copied);
 812
 813  out_unlock:
 814         spin_unlock(&xprt->sock_lock);
 815  dropit:
 816         skb_free_datagram(sk, skb);
 817  out:
 818         read_unlock(&sk->sk_callback_lock);
 819 }
 820
 821 /*
 822  * Copy from an skb into memory and shrink the skb.
 823  */
 824 static inline size_t
 825 tcp_copy_data(skb_reader_t *desc, void *p, size_t len)
 826 {
 827         if (len > desc->count)
 828                 len = desc->count;
 829         if (skb_copy_bits(desc->skb, desc->offset, p, len)) {
 830                 dprintk("RPC:      failed to copy %zu bytes from skb. %zu bytes remain\n",
 831                                 len, desc->count);
 832                 return 0;
 833         }
 834         desc->offset += len;
 835         desc->count -= len;
 836         dprintk("RPC:      copied %zu bytes from skb. %zu bytes remain\n",
 837                         len, desc->count);
 838         return len;
 839 }
 840
 841 /*
 842  * TCP read fragment marker
 843  */
 844 static inline void
 845 tcp_read_fraghdr(struct rpc_xprt *xprt, skb_reader_t *desc)
 846 {
 847         size_t len, used;
 848         char *p;
 849
 850         p = ((char *) &xprt->tcp_recm) + xprt->tcp_offset;
 851         len = sizeof(xprt->tcp_recm) - xprt->tcp_offset;
 852         used = tcp_copy_data(desc, p, len);
 853         xprt->tcp_offset += used;
 854         if (used != len)
 855                 return;
 856         xprt->tcp_reclen = ntohl(xprt->tcp_recm);
 857         if (xprt->tcp_reclen & 0x80000000)
 858                 xprt->tcp_flags |= XPRT_LAST_FRAG;
 859         else
 860                 xprt->tcp_flags &= ~XPRT_LAST_FRAG;
 861         xprt->tcp_reclen &= 0x7fffffff;
 862         xprt->tcp_flags &= ~XPRT_COPY_RECM;
 863         xprt->tcp_offset = 0;
 864         /* Sanity check of the record length */
 865         if (xprt->tcp_reclen < 4) {
 866                 printk(KERN_ERR "RPC: Invalid TCP record fragment length\n");
 867                 xprt_disconnect(xprt);
 868         }
 869         dprintk("RPC:      reading TCP record fragment of length %d\n",
 870                         xprt->tcp_reclen);
 871 }
 872
 873 static void
 874 tcp_check_recm(struct rpc_xprt *xprt)
 875 {
 876         dprintk("RPC:      xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u, tcp_flags = %lx\n",
 877                         xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen, xprt->tcp_flags);
 878         if (xprt->tcp_offset == xprt->tcp_reclen) {
 879                 xprt->tcp_flags |= XPRT_COPY_RECM;
 880                 xprt->tcp_offset = 0;
 881                 if (xprt->tcp_flags & XPRT_LAST_FRAG) {
 882                         xprt->tcp_flags &= ~XPRT_COPY_DATA;
 883                         xprt->tcp_flags |= XPRT_COPY_XID;
 884                         xprt->tcp_copied = 0;
 885                 }
 886         }
 887 }
 888
 889 /*
 890  * TCP read xid
 891  */
 892 static inline void
 893 tcp_read_xid(struct rpc_xprt *xprt, skb_reader_t *desc)
 894 {
 895         size_t len, used;
 896         char *p;
 897
 898         len = sizeof(xprt->tcp_xid) - xprt->tcp_offset;
 899         dprintk("RPC:      reading XID (%Zu bytes)\n", len);
 900         p = ((char *) &xprt->tcp_xid) + xprt->tcp_offset;
 901         used = tcp_copy_data(desc, p, len);
 902         xprt->tcp_offset += used;
 903         if (used != len)
 904                 return;
 905         xprt->tcp_flags &= ~XPRT_COPY_XID;
 906         xprt->tcp_flags |= XPRT_COPY_DATA;
 907         xprt->tcp_copied = 4;
 908         dprintk("RPC:      reading reply for XID %08x\n",
 909                                                 ntohl(xprt->tcp_xid));
 910         tcp_check_recm(xprt);
 911 }
 912
 913 /*
 914  * TCP read and complete request
 915  */
 916 static inline void
 917 tcp_read_request(struct rpc_xprt *xprt, skb_reader_t *desc)
 918 {
 919         struct rpc_rqst *req;
 920         struct xdr_buf *rcvbuf;
 921         size_t len;
 922         ssize_t r;
 923
 924         /* Find and lock the request corresponding to this xid */
 925         spin_lock(&xprt->sock_lock);
 926         req = xprt_lookup_rqst(xprt, xprt->tcp_xid);
 927         if (!req) {
 928                 xprt->tcp_flags &= ~XPRT_COPY_DATA;
 929                 dprintk("RPC:      XID %08x request not found!\n",
 930                                 ntohl(xprt->tcp_xid));
 931                 spin_unlock(&xprt->sock_lock);
 932                 return;
 933         }
 934
 935         rcvbuf = &req->rq_private_buf;
 936         len = desc->count;
 937         if (len > xprt->tcp_reclen - xprt->tcp_offset) {
 938                 skb_reader_t my_desc;
 939
 940                 len = xprt->tcp_reclen - xprt->tcp_offset;
 941                 memcpy(&my_desc, desc, sizeof(my_desc));
 942                 my_desc.count = len;
 943                 r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied,
 944                                           &my_desc, tcp_copy_data);
 945                 desc->count -= r;
 946                 desc->offset += r;
 947         } else
 948                 r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied,
 949                                           desc, tcp_copy_data);
 950
 951         if (r > 0) {
 952                 xprt->tcp_copied += r;
 953                 xprt->tcp_offset += r;
 954         }
 955         if (r != len) {
 956                 /* Error when copying to the receive buffer,
 957                  * usually because we weren't able to allocate
 958                  * additional buffer pages. All we can do now
 959                  * is turn off XPRT_COPY_DATA, so the request
 960                  * will not receive any additional updates,
 961                  * and time out.
 962                  * Any remaining data from this record will
 963                  * be discarded.
 964                  */
 965                 xprt->tcp_flags &= ~XPRT_COPY_DATA;
 966                 dprintk("RPC:      XID %08x truncated request\n",
 967                                 ntohl(xprt->tcp_xid));
 968                 dprintk("RPC:      xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n",
 969                                 xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen);
 970                 goto out;
 971         }
 972
 973         dprintk("RPC:      XID %08x read %Zd bytes\n",
 974                         ntohl(xprt->tcp_xid), r);
 975         dprintk("RPC:      xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n",
 976                         xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen);
 977
 978         if (xprt->tcp_copied == req->rq_private_buf.buflen)
 979                 xprt->tcp_flags &= ~XPRT_COPY_DATA;
 980         else if (xprt->tcp_offset == xprt->tcp_reclen) {
 981                 if (xprt->tcp_flags & XPRT_LAST_FRAG)
 982                         xprt->tcp_flags &= ~XPRT_COPY_DATA;
 983         }
 984
 985 out:
 986         if (!(xprt->tcp_flags & XPRT_COPY_DATA)) {
 987                 dprintk("RPC: %4d received reply complete\n",
 988                                 req->rq_task->tk_pid);
 989                 xprt_complete_rqst(xprt, req, xprt->tcp_copied);
 990         }
 991         spin_unlock(&xprt->sock_lock);
 992         tcp_check_recm(xprt);
 993 }
 994
 995 /*
 996  * TCP discard extra bytes from a short read
 997  */
 998 static inline void
 999 tcp_read_discard(struct rpc_xprt *xprt, skb_reader_t *desc)
1000 {
1001         size_t len;
1002
1003         len = xprt->tcp_reclen - xprt->tcp_offset;
1004         if (len > desc->count)
1005                 len = desc->count;
1006         desc->count -= len;
1007         desc->offset += len;
1008         xprt->tcp_offset += len;
1009         dprintk("RPC:      discarded %Zu bytes\n", len);
1010         tcp_check_recm(xprt);
1011 }
1012
1013 /*
1014  * TCP record receive routine
1015  * We first have to grab the record marker, then the XID, then the data.
1016  */
1017 static int
1018 tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb,
1019                 unsigned int offset, size_t len)
1020 {
1021         struct rpc_xprt *xprt = rd_desc->arg.data;
1022         skb_reader_t desc = {
1023                 .skb    = skb,
1024                 .offset = offset,
1025                 .count  = len,
1026                 .csum   = 0
1027         };
1028
1029         dprintk("RPC:      tcp_data_recv\n");
1030         do {
1031                 /* Read in a new fragment marker if necessary */
1032                 /* Can we ever really expect to get completely empty fragments? */
1033                 if (xprt->tcp_flags & XPRT_COPY_RECM) {
1034                         tcp_read_fraghdr(xprt, &desc);
1035                         continue;
1036                 }
1037                 /* Read in the xid if necessary */
1038                 if (xprt->tcp_flags & XPRT_COPY_XID) {
1039                         tcp_read_xid(xprt, &desc);
1040                         continue;
1041                 }
1042                 /* Read in the request data */
1043                 if (xprt->tcp_flags & XPRT_COPY_DATA) {
1044                         tcp_read_request(xprt, &desc);
1045                         continue;
1046                 }
1047                 /* Skip over any trailing bytes on short reads */
1048                 tcp_read_discard(xprt, &desc);
1049         } while (desc.count);
1050         dprintk("RPC:      tcp_data_recv done\n");
1051         return len - desc.count;
1052 }
1053
1054 static void tcp_data_ready(struct sock *sk, int bytes)
1055 {
1056         struct rpc_xprt *xprt;
1057         read_descriptor_t rd_desc;
1058
1059         read_lock(&sk->sk_callback_lock);
1060         dprintk("RPC:      tcp_data_ready...\n");
1061         if (!(xprt = xprt_from_sock(sk))) {
1062                 printk("RPC:      tcp_data_ready socket info not found!\n");
1063                 goto out;
1064         }
1065         if (xprt->shutdown)
1066                 goto out;
1067
1068         /* We use rd_desc to pass struct xprt to tcp_data_recv */
1069         rd_desc.arg.data = xprt;
1070         rd_desc.count = 65536;
1071         tcp_read_sock(sk, &rd_desc, tcp_data_recv);
1072 out:
1073         read_unlock(&sk->sk_callback_lock);
1074 }
1075
1076 static void
1077 tcp_state_change(struct sock *sk)
1078 {
1079         struct rpc_xprt *xprt;
1080
1081         read_lock(&sk->sk_callback_lock);
1082         if (!(xprt = xprt_from_sock(sk)))
1083                 goto out;
1084         dprintk("RPC:      tcp_state_change client %p...\n", xprt);
1085         dprintk("RPC:      state %x conn %d dead %d zapped %d\n",
1086                                 sk->sk_state, xprt_connected(xprt),
1087                                 sock_flag(sk, SOCK_DEAD),
1088                                 sock_flag(sk, SOCK_ZAPPED));
1089
1090         switch (sk->sk_state) {
1091         case TCP_ESTABLISHED:
1092                 spin_lock_bh(&xprt->sock_lock);
1093                 if (!xprt_test_and_set_connected(xprt)) {
1094                         /* Reset TCP record info */
1095                         xprt->tcp_offset = 0;
1096                         xprt->tcp_reclen = 0;
1097                         xprt->tcp_copied = 0;
1098                         xprt->tcp_flags = XPRT_COPY_RECM | XPRT_COPY_XID;
1099                         rpc_wake_up(&xprt->pending);
1100                 }
1101                 spin_unlock_bh(&xprt->sock_lock);
1102                 break;
1103         case TCP_SYN_SENT:
1104         case TCP_SYN_RECV:
1105                 break;
1106         default:
1107                 xprt_disconnect(xprt);
1108                 break;
1109         }
1110  out:
1111         read_unlock(&sk->sk_callback_lock);
1112 }
1113
1114 /*
1115  * Called when more output buffer space is available for this socket.
1116  * We try not to wake our writers until they can make "significant"
1117  * progress, otherwise we'll waste resources thrashing sock_sendmsg
1118  * with a bunch of small requests.
1119  */
1120 static void
1121 xprt_write_space(struct sock *sk)
1122 {
1123         struct rpc_xprt *xprt;
1124         struct socket   *sock;
1125
1126         read_lock(&sk->sk_callback_lock);
1127         if (!(xprt = xprt_from_sock(sk)) || !(sock = sk->sk_socket))
1128                 goto out;
1129         if (xprt->shutdown)
1130                 goto out;
1131
1132         /* Wait until we have enough socket memory */
1133         if (xprt->stream) {
1134                 /* from net/core/stream.c:sk_stream_write_space */
1135                 if (sk_stream_wspace(sk) < sk_stream_min_wspace(sk))
1136                         goto out;
1137         } else {
1138                 /* from net/core/sock.c:sock_def_write_space */
1139                 if (!sock_writeable(sk))
1140                         goto out;
1141         }
1142
1143         if (!test_and_clear_bit(SOCK_NOSPACE, &sock->flags))
1144                 goto out;
1145
1146         spin_lock_bh(&xprt->sock_lock);
1147         if (xprt->snd_task)
1148                 rpc_wake_up_task(xprt->snd_task);
1149         spin_unlock_bh(&xprt->sock_lock);
1150 out:
1151         read_unlock(&sk->sk_callback_lock);
1152 }
1153
1154 /*
1155  * RPC receive timeout handler.
1156  */
1157 static void
1158 xprt_timer(struct rpc_task *task)
1159 {
1160         struct rpc_rqst *req = task->tk_rqstp;
1161         struct rpc_xprt *xprt = req->rq_xprt;
1162
1163         spin_lock(&xprt->sock_lock);
1164         if (req->rq_received)
1165                 goto out;
1166
1167         xprt_adjust_cwnd(req->rq_xprt, -ETIMEDOUT);
1168         __xprt_put_cong(xprt, req);
1169
1170         dprintk("RPC: %4d xprt_timer (%s request)\n",
1171                 task->tk_pid, req ? "pending" : "backlogged");
1172
1173         task->tk_status  = -ETIMEDOUT;
1174 out:
1175         task->tk_timeout = 0;
1176         rpc_wake_up_task(task);
1177         spin_unlock(&xprt->sock_lock);
1178 }
1179
1180 /*
1181  * Place the actual RPC call.
1182  * We have to copy the iovec because sendmsg fiddles with its contents.
1183  */
1184 int
1185 xprt_prepare_transmit(struct rpc_task *task)
1186 {
1187         struct rpc_rqst *req = task->tk_rqstp;
1188         struct rpc_xprt *xprt = req->rq_xprt;
1189         int err = 0;
1190
1191         dprintk("RPC: %4d xprt_prepare_transmit\n", task->tk_pid);
1192
1193         if (xprt->shutdown)
1194                 return -EIO;
1195
1196         spin_lock_bh(&xprt->sock_lock);
1197         if (req->rq_received && !req->rq_bytes_sent) {
1198                 err = req->rq_received;
1199                 goto out_unlock;
1200         }
1201         if (!__xprt_lock_write(xprt, task)) {
1202                 err = -EAGAIN;
1203                 goto out_unlock;
1204         }
1205
1206         if (!xprt_connected(xprt)) {
1207                 err = -ENOTCONN;
1208                 goto out_unlock;
1209         }
1210 out_unlock:
1211         spin_unlock_bh(&xprt->sock_lock);
1212         return err;
1213 }
1214
1215 void
1216 xprt_transmit(struct rpc_task *task)
1217 {
1218         struct rpc_clnt *clnt = task->tk_client;
1219         struct rpc_rqst *req = task->tk_rqstp;
1220         struct rpc_xprt *xprt = req->rq_xprt;
1221         int status, retry = 0;
1222
1223
1224         dprintk("RPC: %4d xprt_transmit(%u)\n", task->tk_pid, req->rq_slen);
1225
1226         /* set up everything as needed. */
1227         /* Write the record marker */
1228         if (xprt->stream) {
1229                 u32     *marker = req->rq_svec[0].iov_base;
1230
1231                 *marker = htonl(0x80000000|(req->rq_slen-sizeof(*marker)));
1232         }
1233
1234         smp_rmb();
1235         if (!req->rq_received) {
1236                 if (list_empty(&req->rq_list)) {
1237                         spin_lock_bh(&xprt->sock_lock);
1238                         /* Update the softirq receive buffer */
1239                         memcpy(&req->rq_private_buf, &req->rq_rcv_buf,
1240                                         sizeof(req->rq_private_buf));
1241                         /* Add request to the receive list */
1242                         list_add_tail(&req->rq_list, &xprt->recv);
1243                         spin_unlock_bh(&xprt->sock_lock);
1244                         xprt_reset_majortimeo(req);
1245                         /* Turn off autodisconnect */
1246                         del_singleshot_timer_sync(&xprt->timer);
1247                 }
1248         } else if (!req->rq_bytes_sent)
1249                 return;
1250
1251         /* Continue transmitting the packet/record. We must be careful
1252          * to cope with writespace callbacks arriving _after_ we have
1253          * called xprt_sendmsg().
1254          */
1255         while (1) {
1256                 req->rq_xtime = jiffies;
1257                 status = xprt_sendmsg(xprt, req);
1258
1259                 if (status < 0)
1260                         break;
1261
1262                 if (xprt->stream) {
1263                         req->rq_bytes_sent += status;
1264
1265                         /* If we've sent the entire packet, immediately
1266                          * reset the count of bytes sent. */
1267                         if (req->rq_bytes_sent >= req->rq_slen) {
1268                                 req->rq_bytes_sent = 0;
1269                                 goto out_receive;
1270                         }
1271                 } else {
1272                         if (status >= req->rq_slen)
1273                                 goto out_receive;
1274                         status = -EAGAIN;
1275                         break;
1276                 }
1277
1278                 dprintk("RPC: %4d xmit incomplete (%d left of %d)\n",
1279                                 task->tk_pid, req->rq_slen - req->rq_bytes_sent,
1280                                 req->rq_slen);
1281
1282                 status = -EAGAIN;
1283                 if (retry++ > 50)
1284                         break;
1285         }
1286
1287         /* Note: at this point, task->tk_sleeping has not yet been set,
1288          *       hence there is no danger of the waking up task being put on
1289          *       schedq, and being picked up by a parallel run of rpciod().
1290          */
1291         task->tk_status = status;
1292
1293         switch (status) {
1294         case -EAGAIN:
1295                 if (test_bit(SOCK_ASYNC_NOSPACE, &xprt->sock->flags)) {
1296                         /* Protect against races with xprt_write_space */
1297                         spin_lock_bh(&xprt->sock_lock);
1298                         /* Don't race with disconnect */
1299                         if (!xprt_connected(xprt))
1300                                 task->tk_status = -ENOTCONN;
1301                         else if (test_bit(SOCK_NOSPACE, &xprt->sock->flags)) {
1302                                 task->tk_timeout = req->rq_timeout;
1303                                 rpc_sleep_on(&xprt->pending, task, NULL, NULL);
1304                         }
1305                         spin_unlock_bh(&xprt->sock_lock);
1306                         return;
1307                 }
1308                 /* Keep holding the socket if it is blocked */
1309                 rpc_delay(task, HZ>>4);
1310                 return;
1311         case -ECONNREFUSED:
1312                 task->tk_timeout = RPC_REESTABLISH_TIMEOUT;
1313                 rpc_sleep_on(&xprt->sending, task, NULL, NULL);
1314         case -ENOTCONN:
1315                 return;
1316         default:
1317                 if (xprt->stream)
1318                         xprt_disconnect(xprt);
1319         }
1320         xprt_release_write(xprt, task);
1321         return;
1322  out_receive:
1323         dprintk("RPC: %4d xmit complete\n", task->tk_pid);
1324         /* Set the task's receive timeout value */
1325         spin_lock_bh(&xprt->sock_lock);
1326         if (!xprt->nocong) {
1327                 int timer = task->tk_msg.rpc_proc->p_timer;
1328                 task->tk_timeout = rpc_calc_rto(clnt->cl_rtt, timer);
1329                 task->tk_timeout <<= rpc_ntimeo(clnt->cl_rtt, timer) + req->rq_retries;
1330                 if (task->tk_timeout > xprt->timeout.to_maxval || task->tk_timeout == 0)
1331                         task->tk_timeout = xprt->timeout.to_maxval;
1332         } else
1333                 task->tk_timeout = req->rq_timeout;
1334         /* Don't race with disconnect */
1335         if (!xprt_connected(xprt))
1336                 task->tk_status = -ENOTCONN;
1337         else if (!req->rq_received)
1338                 rpc_sleep_on(&xprt->pending, task, NULL, xprt_timer);
1339         __xprt_release_write(xprt, task);
1340         spin_unlock_bh(&xprt->sock_lock);
1341 }
1342
1343 /*
1344  * Reserve an RPC call slot.
1345  */
1346 static inline void
1347 do_xprt_reserve(struct rpc_task *task)
1348 {
1349         struct rpc_xprt *xprt = task->tk_xprt;
1350
1351         task->tk_status = 0;
1352         if (task->tk_rqstp)
1353                 return;
1354         if (!list_empty(&xprt->free)) {
1355                 struct rpc_rqst *req = list_entry(xprt->free.next, struct rpc_rqst, rq_list);
1356                 list_del_init(&req->rq_list);
1357                 task->tk_rqstp = req;
1358                 xprt_request_init(task, xprt);
1359                 return;
1360         }
1361         dprintk("RPC:      waiting for request slot\n");
1362         task->tk_status = -EAGAIN;
1363         task->tk_timeout = 0;
1364         rpc_sleep_on(&xprt->backlog, task, NULL, NULL);
1365 }
1366
1367 void
1368 xprt_reserve(struct rpc_task *task)
1369 {
1370         struct rpc_xprt *xprt = task->tk_xprt;
1371
1372         task->tk_status = -EIO;
1373         if (!xprt->shutdown) {
1374                 spin_lock(&xprt->xprt_lock);
1375                 do_xprt_reserve(task);
1376                 spin_unlock(&xprt->xprt_lock);
1377         }
1378 }
1379
1380 /*
1381  * Allocate a 'unique' XID
1382  */
1383 static inline u32 xprt_alloc_xid(struct rpc_xprt *xprt)
1384 {
1385         return xprt->xid++;
1386 }
1387
1388 static inline void xprt_init_xid(struct rpc_xprt *xprt)
1389 {
1390         get_random_bytes(&xprt->xid, sizeof(xprt->xid));
1391 }
1392
1393 /*
1394  * Initialize RPC request
1395  */
1396 static void
1397 xprt_request_init(struct rpc_task *task, struct rpc_xprt *xprt)
1398 {
1399         struct rpc_rqst *req = task->tk_rqstp;
1400
1401         req->rq_timeout = xprt->timeout.to_initval;
1402         req->rq_task    = task;
1403         req->rq_xprt    = xprt;
1404         req->rq_xid     = xprt_alloc_xid(xprt);
1405         dprintk("RPC: %4d reserved req %p xid %08x\n", task->tk_pid,
1406                         req, ntohl(req->rq_xid));
1407 }
1408
1409 /*
1410  * Release an RPC call slot
1411  */
1412 void
1413 xprt_release(struct rpc_task *task)
1414 {
1415         struct rpc_xprt *xprt = task->tk_xprt;
1416         struct rpc_rqst *req;
1417
1418         if (!(req = task->tk_rqstp))
1419                 return;
1420         spin_lock_bh(&xprt->sock_lock);
1421         __xprt_release_write(xprt, task);
1422         __xprt_put_cong(xprt, req);
1423         if (!list_empty(&req->rq_list))
1424                 list_del(&req->rq_list);
1425         xprt->last_used = jiffies;
1426         if (list_empty(&xprt->recv) && !xprt->shutdown)
1427                 mod_timer(&xprt->timer, xprt->last_used + XPRT_IDLE_TIMEOUT);
1428         spin_unlock_bh(&xprt->sock_lock);
1429         task->tk_rqstp = NULL;
1430         memset(req, 0, sizeof(*req));   /* mark unused */
1431
1432         dprintk("RPC: %4d release request %p\n", task->tk_pid, req);
1433
1434         spin_lock(&xprt->xprt_lock);
1435         list_add(&req->rq_list, &xprt->free);
1436         xprt_clear_backlog(xprt);
1437         spin_unlock(&xprt->xprt_lock);
1438 }
1439
1440 /*
1441  * Set default timeout parameters
1442  */
1443 static void
1444 xprt_default_timeout(struct rpc_timeout *to, int proto)
1445 {
1446         if (proto == IPPROTO_UDP)
1447                 xprt_set_timeout(to, 5,  5 * HZ);
1448         else
1449                 xprt_set_timeout(to, 5, 60 * HZ);
1450 }
1451
1452 /*
1453  * Set constant timeout
1454  */
1455 void
1456 xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long incr)
1457 {
1458         to->to_initval   =
1459         to->to_increment = incr;
1460         to->to_maxval    = incr * retr;
1461         to->to_retries   = retr;
1462         to->to_exponential = 0;
1463 }
1464
1465 unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
1466 unsigned int xprt_tcp_slot_table_entries = RPC_DEF_SLOT_TABLE;
1467
1468 /*
1469  * Initialize an RPC client
1470  */
1471 static struct rpc_xprt *
1472 xprt_setup(int proto, struct sockaddr_in *ap, struct rpc_timeout *to)
1473 {
1474         struct rpc_xprt *xprt;
1475         unsigned int entries;
1476         size_t slot_table_size;
1477         struct rpc_rqst *req;
1478
1479         dprintk("RPC:      setting up %s transport...\n",
1480                                 proto == IPPROTO_UDP? "UDP" : "TCP");
1481
1482         entries = (proto == IPPROTO_TCP)?
1483                 xprt_tcp_slot_table_entries : xprt_udp_slot_table_entries;
1484
1485         if ((xprt = kmalloc(sizeof(struct rpc_xprt), GFP_KERNEL)) == NULL)
1486                 return ERR_PTR(-ENOMEM);
1487         memset(xprt, 0, sizeof(*xprt)); /* Nnnngh! */
1488         xprt->max_reqs = entries;
1489         slot_table_size = entries * sizeof(xprt->slot[0]);
1490         xprt->slot = kmalloc(slot_table_size, GFP_KERNEL);
1491         if (xprt->slot == NULL) {
1492                 kfree(xprt);
1493                 return ERR_PTR(-ENOMEM);
1494         }
1495         memset(xprt->slot, 0, slot_table_size);
1496
1497         xprt->addr = *ap;
1498         xprt->prot = proto;
1499         xprt->stream = (proto == IPPROTO_TCP)? 1 : 0;
1500         if (xprt->stream) {
1501                 xprt->cwnd = RPC_MAXCWND(xprt);
1502                 xprt->nocong = 1;
1503                 xprt->max_payload = (1U << 31) - 1;
1504         } else {
1505                 xprt->cwnd = RPC_INITCWND;
1506                 xprt->max_payload = (1U << 16) - (MAX_HEADER << 3);
1507         }
1508         spin_lock_init(&xprt->sock_lock);
1509         spin_lock_init(&xprt->xprt_lock);
1510         init_waitqueue_head(&xprt->cong_wait);
1511
1512         INIT_LIST_HEAD(&xprt->free);
1513         INIT_LIST_HEAD(&xprt->recv);
1514         INIT_WORK(&xprt->sock_connect, xprt_socket_connect, xprt);
1515         INIT_WORK(&xprt->task_cleanup, xprt_socket_autoclose, xprt);
1516         init_timer(&xprt->timer);
1517         xprt->timer.function = xprt_init_autodisconnect;
1518         xprt->timer.data = (unsigned long) xprt;
1519         xprt->last_used = jiffies;
1520         xprt->port = XPRT_MAX_RESVPORT;
1521
1522         /* Set timeout parameters */
1523         if (to) {
1524                 xprt->timeout = *to;
1525         } else
1526                 xprt_default_timeout(&xprt->timeout, xprt->prot);
1527
1528         rpc_init_wait_queue(&xprt->pending, "xprt_pending");
1529         rpc_init_wait_queue(&xprt->sending, "xprt_sending");
1530         rpc_init_wait_queue(&xprt->resend, "xprt_resend");
1531         rpc_init_priority_wait_queue(&xprt->backlog, "xprt_backlog");
1532
1533         /* initialize free list */
1534         for (req = &xprt->slot[entries-1]; req >= &xprt->slot[0]; req--)
1535                 list_add(&req->rq_list, &xprt->free);
1536
1537         xprt_init_xid(xprt);
1538
1539         /* Check whether we want to use a reserved port */
1540         xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0;
1541
1542         dprintk("RPC:      created transport %p with %u slots\n", xprt,
1543                         xprt->max_reqs);
1544
1545         return xprt;
1546 }
1547
1548 /*
1549  * Bind to a reserved port
1550  */
1551 static inline int xprt_bindresvport(struct rpc_xprt *xprt, struct socket *sock)
1552 {
1553         struct sockaddr_in myaddr = {
1554                 .sin_family = AF_INET,
1555         };
1556         int             err, port;
1557
1558         /* Were we already bound to a given port? Try to reuse it */
1559         port = xprt->port;
1560         do {
1561                 myaddr.sin_port = htons(port);
1562                 err = sock->ops->bind(sock, (struct sockaddr *) &myaddr,
1563                                                 sizeof(myaddr));
1564                 if (err == 0) {
1565                         xprt->port = port;
1566                         return 0;
1567                 }
1568                 if (--port == 0)
1569                         port = XPRT_MAX_RESVPORT;
1570         } while (err == -EADDRINUSE && port != xprt->port);
1571
1572         printk("RPC: Can't bind to reserved port (%d).\n", -err);
1573         return err;
1574 }
1575
1576 static void
1577 xprt_bind_socket(struct rpc_xprt *xprt, struct socket *sock)
1578 {
1579         struct sock     *sk = sock->sk;
1580
1581         if (xprt->inet)
1582                 return;
1583
1584         write_lock_bh(&sk->sk_callback_lock);
1585         sk->sk_user_data = xprt;
1586         xprt->old_data_ready = sk->sk_data_ready;
1587         xprt->old_state_change = sk->sk_state_change;
1588         xprt->old_write_space = sk->sk_write_space;
1589         if (xprt->prot == IPPROTO_UDP) {
1590                 sk->sk_data_ready = udp_data_ready;
1591                 sk->sk_no_check = UDP_CSUM_NORCV;
1592                 xprt_set_connected(xprt);
1593         } else {
1594                 tcp_sk(sk)->nonagle = 1;        /* disable Nagle's algorithm */
1595                 sk->sk_data_ready = tcp_data_ready;
1596                 sk->sk_state_change = tcp_state_change;
1597                 xprt_clear_connected(xprt);
1598         }
1599         sk->sk_write_space = xprt_write_space;
1600
1601         /* Reset to new socket */
1602         xprt->sock = sock;
1603         xprt->inet = sk;
1604         write_unlock_bh(&sk->sk_callback_lock);
1605
1606         return;
1607 }
1608
1609 /*
1610  * Set socket buffer length
1611  */
1612 void
1613 xprt_sock_setbufsize(struct rpc_xprt *xprt)
1614 {
1615         struct sock *sk = xprt->inet;
1616
1617         if (xprt->stream)
1618                 return;
1619         if (xprt->rcvsize) {
1620                 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
1621                 sk->sk_rcvbuf = xprt->rcvsize * xprt->max_reqs *  2;
1622         }
1623         if (xprt->sndsize) {
1624                 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
1625                 sk->sk_sndbuf = xprt->sndsize * xprt->max_reqs * 2;
1626                 sk->sk_write_space(sk);
1627         }
1628 }
1629
1630 /*
1631  * Datastream sockets are created here, but xprt_connect will create
1632  * and connect stream sockets.
1633  */
1634 static struct socket * xprt_create_socket(struct rpc_xprt *xprt, int proto, int resvport)
1635 {
1636         struct socket   *sock;
1637         int             type, err;
1638
1639         dprintk("RPC:      xprt_create_socket(%s %d)\n",
1640                            (proto == IPPROTO_UDP)? "udp" : "tcp", proto);
1641
1642         type = (proto == IPPROTO_UDP)? SOCK_DGRAM : SOCK_STREAM;
1643
1644         if ((err = sock_create_kern(PF_INET, type, proto, &sock)) < 0) {
1645                 printk("RPC: can't create socket (%d).\n", -err);
1646                 return NULL;
1647         }
1648
1649         /* If the caller has the capability, bind to a reserved port */
1650         if (resvport && xprt_bindresvport(xprt, sock) < 0) {
1651                 printk("RPC: can't bind to reserved port.\n");
1652                 goto failed;
1653         }
1654
1655         return sock;
1656
1657 failed:
1658         sock_release(sock);
1659         return NULL;
1660 }
1661
1662 /*
1663  * Create an RPC client transport given the protocol and peer address.
1664  */
1665 struct rpc_xprt *
1666 xprt_create_proto(int proto, struct sockaddr_in *sap, struct rpc_timeout *to)
1667 {
1668         struct rpc_xprt *xprt;
1669
1670         xprt = xprt_setup(proto, sap, to);
1671         if (IS_ERR(xprt))
1672                 dprintk("RPC:      xprt_create_proto failed\n");
1673         else
1674                 dprintk("RPC:      xprt_create_proto created xprt %p\n", xprt);
1675         return xprt;
1676 }
1677
1678 /*
1679  * Prepare for transport shutdown.
1680  */
1681 static void
1682 xprt_shutdown(struct rpc_xprt *xprt)
1683 {
1684         xprt->shutdown = 1;
1685         rpc_wake_up(&xprt->sending);
1686         rpc_wake_up(&xprt->resend);
1687         rpc_wake_up(&xprt->pending);
1688         rpc_wake_up(&xprt->backlog);
1689         wake_up(&xprt->cong_wait);
1690         del_timer_sync(&xprt->timer);
1691
1692         /* synchronously wait for connect worker to finish */
1693         cancel_delayed_work(&xprt->sock_connect);
1694         flush_scheduled_work();
1695 }
1696
1697 /*
1698  * Clear the xprt backlog queue
1699  */
1700 static int
1701 xprt_clear_backlog(struct rpc_xprt *xprt) {
1702         rpc_wake_up_next(&xprt->backlog);
1703         wake_up(&xprt->cong_wait);
1704         return 1;
1705 }
1706
1707 /*
1708  * Destroy an RPC transport, killing off all requests.
1709  */
1710 int
1711 xprt_destroy(struct rpc_xprt *xprt)
1712 {
1713         dprintk("RPC:      destroying transport %p\n", xprt);
1714         xprt_shutdown(xprt);
1715         xprt_disconnect(xprt);
1716         xprt_close(xprt);
1717         kfree(xprt->slot);
1718         kfree(xprt);
1719
1720         return 0;
1721 }