sys/rpc/clnt_dg.c

   1 /*      $NetBSD: clnt_dg.c,v 1.4 2000/07/14 08:40:41 fvdl Exp $ */
   2
   3 /*-
   4  * Copyright (c) 2009, Sun Microsystems, Inc.
   5  * All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions are met:
   9  * - Redistributions of source code must retain the above copyright notice,
  10  *   this list of conditions and the following disclaimer.
  11  * - Redistributions in binary form must reproduce the above copyright notice,
  12  *   this list of conditions and the following disclaimer in the documentation
  13  *   and/or other materials provided with the distribution.
  14  * - Neither the name of Sun Microsystems, Inc. nor the names of its
  15  *   contributors may be used to endorse or promote products derived
  16  *   from this software without specific prior written permission.
  17  *
  18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  19  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  21  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
  22  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  23  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  24  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  25  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  26  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  27  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  28  * POSSIBILITY OF SUCH DAMAGE.
  29  */
  30 /*
  31  * Copyright (c) 1986-1991 by Sun Microsystems Inc.
  32  */
  33
  34 #if defined(LIBC_SCCS) && !defined(lint)
  35 #ident  "@(#)clnt_dg.c  1.23    94/04/22 SMI"
  36 static char sccsid[] = "@(#)clnt_dg.c 1.19 89/03/16 Copyr 1988 Sun Micro";
  37 #endif
  38 #include <sys/cdefs.h>
  39 __FBSDID("$FreeBSD$");
  40
  41 /*
  42  * Implements a connectionless client side RPC.
  43  */
  44
  45 #include <sys/param.h>
  46 #include <sys/systm.h>
  47 #include <sys/kernel.h>
  48 #include <sys/lock.h>
  49 #include <sys/malloc.h>
  50 #include <sys/mbuf.h>
  51 #include <sys/mutex.h>
  52 #include <sys/pcpu.h>
  53 #include <sys/proc.h>
  54 #include <sys/socket.h>
  55 #include <sys/socketvar.h>
  56 #include <sys/time.h>
  57 #include <sys/uio.h>
  58
  59 #include <net/vnet.h>
  60
  61 #include <rpc/rpc.h>
  62 #include <rpc/rpc_com.h>
  63
  64
  65 #ifdef _FREEFALL_CONFIG
  66 /*
  67  * Disable RPC exponential back-off for FreeBSD.org systems.
  68  */
  69 #define RPC_MAX_BACKOFF         1 /* second */
  70 #else
  71 #define RPC_MAX_BACKOFF         30 /* seconds */
  72 #endif
  73
  74 static bool_t time_not_ok(struct timeval *);
  75 static enum clnt_stat clnt_dg_call(CLIENT *, struct rpc_callextra *,
  76     rpcproc_t, struct mbuf *, struct mbuf **, struct timeval);
  77 static void clnt_dg_geterr(CLIENT *, struct rpc_err *);
  78 static bool_t clnt_dg_freeres(CLIENT *, xdrproc_t, void *);
  79 static void clnt_dg_abort(CLIENT *);
  80 static bool_t clnt_dg_control(CLIENT *, u_int, void *);
  81 static void clnt_dg_close(CLIENT *);
  82 static void clnt_dg_destroy(CLIENT *);
  83 static int clnt_dg_soupcall(struct socket *so, void *arg, int waitflag);
  84
  85 static struct clnt_ops clnt_dg_ops = {
  86         .cl_call =      clnt_dg_call,
  87         .cl_abort =     clnt_dg_abort,
  88         .cl_geterr =    clnt_dg_geterr,
  89         .cl_freeres =   clnt_dg_freeres,
  90         .cl_close =     clnt_dg_close,
  91         .cl_destroy =   clnt_dg_destroy,
  92         .cl_control =   clnt_dg_control
  93 };
  94
  95 static const char mem_err_clnt_dg[] = "clnt_dg_create: out of memory";
  96
  97 /*
  98  * A pending RPC request which awaits a reply. Requests which have
  99  * received their reply will have cr_xid set to zero and cr_mrep to
 100  * the mbuf chain of the reply.
 101  */
 102 struct cu_request {
 103         TAILQ_ENTRY(cu_request) cr_link;
 104         CLIENT                  *cr_client;     /* owner */
 105         uint32_t                cr_xid;         /* XID of request */
 106         struct mbuf             *cr_mrep;       /* reply received by upcall */
 107         int                     cr_error;       /* any error from upcall */
 108         char                    cr_verf[MAX_AUTH_BYTES]; /* reply verf */
 109 };
 110
 111 TAILQ_HEAD(cu_request_list, cu_request);
 112
 113 #define MCALL_MSG_SIZE 24
 114
 115 /*
 116  * This structure is pointed to by the socket buffer's sb_upcallarg
 117  * member. It is separate from the client private data to facilitate
 118  * multiple clients sharing the same socket. The cs_lock mutex is used
 119  * to protect all fields of this structure, the socket's receive
 120  * buffer SOCKBUF_LOCK is used to ensure that exactly one of these
 121  * structures is installed on the socket.
 122  */
 123 struct cu_socket {
 124         struct mtx              cs_lock;
 125         int                     cs_refs;        /* Count of clients */
 126         struct cu_request_list  cs_pending;     /* Requests awaiting replies */
 127         int                     cs_upcallrefs;  /* Refcnt of upcalls in prog.*/
 128 };
 129
 130 static void clnt_dg_upcallsdone(struct socket *, struct cu_socket *);
 131
 132 /*
 133  * Private data kept per client handle
 134  */
 135 struct cu_data {
 136         int                     cu_threads;     /* # threads in clnt_vc_call */
 137         bool_t                  cu_closing;     /* TRUE if we are closing */
 138         bool_t                  cu_closed;      /* TRUE if we are closed */
 139         struct socket           *cu_socket;     /* connection socket */
 140         bool_t                  cu_closeit;     /* opened by library */
 141         struct sockaddr_storage cu_raddr;       /* remote address */
 142         int                     cu_rlen;
 143         struct timeval          cu_wait;        /* retransmit interval */
 144         struct timeval          cu_total;       /* total time for the call */
 145         struct rpc_err          cu_error;
 146         uint32_t                cu_xid;
 147         char                    cu_mcallc[MCALL_MSG_SIZE]; /* marshalled callmsg */
 148         size_t                  cu_mcalllen;
 149         size_t                  cu_sendsz;      /* send size */
 150         size_t                  cu_recvsz;      /* recv size */
 151         int                     cu_async;
 152         int                     cu_connect;     /* Use connect(). */
 153         int                     cu_connected;   /* Have done connect(). */
 154         const char              *cu_waitchan;
 155         int                     cu_waitflag;
 156         int                     cu_cwnd;        /* congestion window */
 157         int                     cu_sent;        /* number of in-flight RPCs */
 158         bool_t                  cu_cwnd_wait;
 159 };
 160
 161 #define CWNDSCALE       256
 162 #define MAXCWND         (32 * CWNDSCALE)
 163
 164 /*
 165  * Connection less client creation returns with client handle parameters.
 166  * Default options are set, which the user can change using clnt_control().
 167  * fd should be open and bound.
 168  * NB: The rpch->cl_auth is initialized to null authentication.
 169  *      Caller may wish to set this something more useful.
 170  *
 171  * sendsz and recvsz are the maximum allowable packet sizes that can be
 172  * sent and received. Normally they are the same, but they can be
 173  * changed to improve the program efficiency and buffer allocation.
 174  * If they are 0, use the transport default.
 175  *
 176  * If svcaddr is NULL, returns NULL.
 177  */
 178 CLIENT *
 179 clnt_dg_create(
 180         struct socket *so,
 181         struct sockaddr *svcaddr,       /* servers address */
 182         rpcprog_t program,              /* program number */
 183         rpcvers_t version,              /* version number */
 184         size_t sendsz,                  /* buffer recv size */
 185         size_t recvsz)                  /* buffer send size */
 186 {
 187         CLIENT *cl = NULL;              /* client handle */
 188         struct cu_data *cu = NULL;      /* private data */
 189         struct cu_socket *cs = NULL;
 190         struct sockbuf *sb;
 191         struct timeval now;
 192         struct rpc_msg call_msg;
 193         struct __rpc_sockinfo si;
 194         XDR xdrs;
 195         int error;
 196
 197         if (svcaddr == NULL) {
 198                 rpc_createerr.cf_stat = RPC_UNKNOWNADDR;
 199                 return (NULL);
 200         }
 201
 202         if (!__rpc_socket2sockinfo(so, &si)) {
 203                 rpc_createerr.cf_stat = RPC_TLIERROR;
 204                 rpc_createerr.cf_error.re_errno = 0;
 205                 return (NULL);
 206         }
 207
 208         /*
 209          * Find the receive and the send size
 210          */
 211         sendsz = __rpc_get_t_size(si.si_af, si.si_proto, (int)sendsz);
 212         recvsz = __rpc_get_t_size(si.si_af, si.si_proto, (int)recvsz);
 213         if ((sendsz == 0) || (recvsz == 0)) {
 214                 rpc_createerr.cf_stat = RPC_TLIERROR; /* XXX */
 215                 rpc_createerr.cf_error.re_errno = 0;
 216                 return (NULL);
 217         }
 218
 219         cl = mem_alloc(sizeof (CLIENT));
 220
 221         /*
 222          * Should be multiple of 4 for XDR.
 223          */
 224         sendsz = ((sendsz + 3) / 4) * 4;
 225         recvsz = ((recvsz + 3) / 4) * 4;
 226         cu = mem_alloc(sizeof (*cu));
 227         cu->cu_threads = 0;
 228         cu->cu_closing = FALSE;
 229         cu->cu_closed = FALSE;
 230         (void) memcpy(&cu->cu_raddr, svcaddr, (size_t)svcaddr->sa_len);
 231         cu->cu_rlen = svcaddr->sa_len;
 232         /* Other values can also be set through clnt_control() */
 233         cu->cu_wait.tv_sec = 3; /* heuristically chosen */
 234         cu->cu_wait.tv_usec = 0;
 235         cu->cu_total.tv_sec = -1;
 236         cu->cu_total.tv_usec = -1;
 237         cu->cu_sendsz = sendsz;
 238         cu->cu_recvsz = recvsz;
 239         cu->cu_async = FALSE;
 240         cu->cu_connect = FALSE;
 241         cu->cu_connected = FALSE;
 242         cu->cu_waitchan = "rpcrecv";
 243         cu->cu_waitflag = 0;
 244         cu->cu_cwnd = MAXCWND / 2;
 245         cu->cu_sent = 0;
 246         cu->cu_cwnd_wait = FALSE;
 247         (void) getmicrotime(&now);
 248         cu->cu_xid = __RPC_GETXID(&now);
 249         call_msg.rm_xid = cu->cu_xid;
 250         call_msg.rm_call.cb_prog = program;
 251         call_msg.rm_call.cb_vers = version;
 252         xdrmem_create(&xdrs, cu->cu_mcallc, MCALL_MSG_SIZE, XDR_ENCODE);
 253         if (! xdr_callhdr(&xdrs, &call_msg)) {
 254                 rpc_createerr.cf_stat = RPC_CANTENCODEARGS;  /* XXX */
 255                 rpc_createerr.cf_error.re_errno = 0;
 256                 goto err2;
 257         }
 258         cu->cu_mcalllen = XDR_GETPOS(&xdrs);
 259
 260         /*
 261          * By default, closeit is always FALSE. It is users responsibility
 262          * to do a close on it, else the user may use clnt_control
 263          * to let clnt_destroy do it for him/her.
 264          */
 265         cu->cu_closeit = FALSE;
 266         cu->cu_socket = so;
 267         error = soreserve(so, (u_long)sendsz, (u_long)recvsz);
 268         if (error != 0) {
 269                 rpc_createerr.cf_stat = RPC_FAILED;
 270                 rpc_createerr.cf_error.re_errno = error;
 271                 goto err2;
 272         }
 273
 274         sb = &so->so_rcv;
 275         SOCKBUF_LOCK(&so->so_rcv);
 276 recheck_socket:
 277         if (sb->sb_upcall) {
 278                 if (sb->sb_upcall != clnt_dg_soupcall) {
 279                         SOCKBUF_UNLOCK(&so->so_rcv);
 280                         printf("clnt_dg_create(): socket already has an incompatible upcall\n");
 281                         goto err2;
 282                 }
 283                 cs = (struct cu_socket *) sb->sb_upcallarg;
 284                 mtx_lock(&cs->cs_lock);
 285                 cs->cs_refs++;
 286                 mtx_unlock(&cs->cs_lock);
 287         } else {
 288                 /*
 289                  * We are the first on this socket - allocate the
 290                  * structure and install it in the socket.
 291                  */
 292                 SOCKBUF_UNLOCK(&so->so_rcv);
 293                 cs = mem_alloc(sizeof(*cs));
 294                 SOCKBUF_LOCK(&so->so_rcv);
 295                 if (sb->sb_upcall) {
 296                         /*
 297                          * We have lost a race with some other client.
 298                          */
 299                         mem_free(cs, sizeof(*cs));
 300                         goto recheck_socket;
 301                 }
 302                 mtx_init(&cs->cs_lock, "cs->cs_lock", NULL, MTX_DEF);
 303                 cs->cs_refs = 1;
 304                 cs->cs_upcallrefs = 0;
 305                 TAILQ_INIT(&cs->cs_pending);
 306                 soupcall_set(so, SO_RCV, clnt_dg_soupcall, cs);
 307         }
 308         SOCKBUF_UNLOCK(&so->so_rcv);
 309
 310         cl->cl_refs = 1;
 311         cl->cl_ops = &clnt_dg_ops;
 312         cl->cl_private = (caddr_t)(void *)cu;
 313         cl->cl_auth = authnone_create();
 314         cl->cl_tp = NULL;
 315         cl->cl_netid = NULL;
 316         return (cl);
 317 err2:
 318         if (cl) {
 319                 mem_free(cl, sizeof (CLIENT));
 320                 if (cu)
 321                         mem_free(cu, sizeof (*cu));
 322         }
 323         return (NULL);
 324 }
 325
 326 static enum clnt_stat
 327 clnt_dg_call(
 328         CLIENT          *cl,            /* client handle */
 329         struct rpc_callextra *ext,      /* call metadata */
 330         rpcproc_t       proc,           /* procedure number */
 331         struct mbuf     *args,          /* pointer to args */
 332         struct mbuf     **resultsp,     /* pointer to results */
 333         struct timeval  utimeout)       /* seconds to wait before giving up */
 334 {
 335         struct cu_data *cu = (struct cu_data *)cl->cl_private;
 336         struct cu_socket *cs;
 337         struct rpc_timers *rt;
 338         AUTH *auth;
 339         struct rpc_err *errp;
 340         enum clnt_stat stat;
 341         XDR xdrs;
 342         struct rpc_msg reply_msg;
 343         bool_t ok;
 344         int retrans;                    /* number of re-transmits so far */
 345         int nrefreshes = 2;             /* number of times to refresh cred */
 346         struct timeval *tvp;
 347         int timeout;
 348         int retransmit_time;
 349         int next_sendtime, starttime, rtt, time_waited, tv = 0;
 350         struct sockaddr *sa;
 351         socklen_t salen;
 352         uint32_t xid = 0;
 353         struct mbuf *mreq = NULL, *results;
 354         struct cu_request *cr;
 355         int error;
 356
 357         cs = cu->cu_socket->so_rcv.sb_upcallarg;
 358         cr = malloc(sizeof(struct cu_request), M_RPC, M_WAITOK);
 359
 360         mtx_lock(&cs->cs_lock);
 361
 362         if (cu->cu_closing || cu->cu_closed) {
 363                 mtx_unlock(&cs->cs_lock);
 364                 free(cr, M_RPC);
 365                 return (RPC_CANTSEND);
 366         }
 367         cu->cu_threads++;
 368
 369         if (ext) {
 370                 auth = ext->rc_auth;
 371                 errp = &ext->rc_err;
 372         } else {
 373                 auth = cl->cl_auth;
 374                 errp = &cu->cu_error;
 375         }
 376
 377         cr->cr_client = cl;
 378         cr->cr_mrep = NULL;
 379         cr->cr_error = 0;
 380
 381         if (cu->cu_total.tv_usec == -1) {
 382                 tvp = &utimeout; /* use supplied timeout */
 383         } else {
 384                 tvp = &cu->cu_total; /* use default timeout */
 385         }
 386         if (tvp->tv_sec || tvp->tv_usec)
 387                 timeout = tvtohz(tvp);
 388         else
 389                 timeout = 0;
 390
 391         if (cu->cu_connect && !cu->cu_connected) {
 392                 mtx_unlock(&cs->cs_lock);
 393                 error = soconnect(cu->cu_socket,
 394                     (struct sockaddr *)&cu->cu_raddr, curthread);
 395                 mtx_lock(&cs->cs_lock);
 396                 if (error) {
 397                         errp->re_errno = error;
 398                         errp->re_status = stat = RPC_CANTSEND;
 399                         goto out;
 400                 }
 401                 cu->cu_connected = 1;
 402         }
 403         if (cu->cu_connected) {
 404                 sa = NULL;
 405                 salen = 0;
 406         } else {
 407                 sa = (struct sockaddr *)&cu->cu_raddr;
 408                 salen = cu->cu_rlen;
 409         }
 410         time_waited = 0;
 411         retrans = 0;
 412         if (ext && ext->rc_timers) {
 413                 rt = ext->rc_timers;
 414                 if (!rt->rt_rtxcur)
 415                         rt->rt_rtxcur = tvtohz(&cu->cu_wait);
 416                 retransmit_time = next_sendtime = rt->rt_rtxcur;
 417         } else {
 418                 rt = NULL;
 419                 retransmit_time = next_sendtime = tvtohz(&cu->cu_wait);
 420         }
 421
 422         starttime = ticks;
 423
 424 call_again:
 425         mtx_assert(&cs->cs_lock, MA_OWNED);
 426
 427         cu->cu_xid++;
 428         xid = cu->cu_xid;
 429
 430 send_again:
 431         mtx_unlock(&cs->cs_lock);
 432
 433         mreq = m_gethdr(M_WAITOK, MT_DATA);
 434         KASSERT(cu->cu_mcalllen <= MHLEN, ("RPC header too big"));
 435         bcopy(cu->cu_mcallc, mreq->m_data, cu->cu_mcalllen);
 436         mreq->m_len = cu->cu_mcalllen;
 437
 438         /*
 439          * The XID is the first thing in the request.
 440          */
 441         *mtod(mreq, uint32_t *) = htonl(xid);
 442
 443         xdrmbuf_create(&xdrs, mreq, XDR_ENCODE);
 444
 445         if (cu->cu_async == TRUE && args == NULL)
 446                 goto get_reply;
 447
 448         if ((! XDR_PUTINT32(&xdrs, &proc)) ||
 449             (! AUTH_MARSHALL(auth, xid, &xdrs,
 450                 m_copym(args, 0, M_COPYALL, M_WAITOK)))) {
 451                 errp->re_status = stat = RPC_CANTENCODEARGS;
 452                 mtx_lock(&cs->cs_lock);
 453                 goto out;
 454         }
 455         mreq->m_pkthdr.len = m_length(mreq, NULL);
 456
 457         cr->cr_xid = xid;
 458         mtx_lock(&cs->cs_lock);
 459
 460         /*
 461          * Try to get a place in the congestion window.
 462          */
 463         while (cu->cu_sent >= cu->cu_cwnd) {
 464                 cu->cu_cwnd_wait = TRUE;
 465                 error = msleep(&cu->cu_cwnd_wait, &cs->cs_lock,
 466                     cu->cu_waitflag, "rpccwnd", 0);
 467                 if (error) {
 468                         errp->re_errno = error;
 469                         if (error == EINTR || error == ERESTART)
 470                                 errp->re_status = stat = RPC_INTR;
 471                         else
 472                                 errp->re_status = stat = RPC_CANTSEND;
 473                         goto out;
 474                 }
 475         }
 476         cu->cu_sent += CWNDSCALE;
 477
 478         TAILQ_INSERT_TAIL(&cs->cs_pending, cr, cr_link);
 479         mtx_unlock(&cs->cs_lock);
 480
 481         /*
 482          * sosend consumes mreq.
 483          */
 484         error = sosend(cu->cu_socket, sa, NULL, mreq, NULL, 0, curthread);
 485         mreq = NULL;
 486
 487         /*
 488          * sub-optimal code appears here because we have
 489          * some clock time to spare while the packets are in flight.
 490          * (We assume that this is actually only executed once.)
 491          */
 492         reply_msg.acpted_rply.ar_verf.oa_flavor = AUTH_NULL;
 493         reply_msg.acpted_rply.ar_verf.oa_base = cr->cr_verf;
 494         reply_msg.acpted_rply.ar_verf.oa_length = 0;
 495         reply_msg.acpted_rply.ar_results.where = NULL;
 496         reply_msg.acpted_rply.ar_results.proc = (xdrproc_t)xdr_void;
 497
 498         mtx_lock(&cs->cs_lock);
 499         if (error) {
 500                 TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
 501                 errp->re_errno = error;
 502                 errp->re_status = stat = RPC_CANTSEND;
 503                 cu->cu_sent -= CWNDSCALE;
 504                 if (cu->cu_cwnd_wait) {
 505                         cu->cu_cwnd_wait = FALSE;
 506                         wakeup(&cu->cu_cwnd_wait);
 507                 }
 508                 goto out;
 509         }
 510
 511         /*
 512          * Check to see if we got an upcall while waiting for the
 513          * lock.
 514          */
 515         if (cr->cr_error) {
 516                 TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
 517                 errp->re_errno = cr->cr_error;
 518                 errp->re_status = stat = RPC_CANTRECV;
 519                 cu->cu_sent -= CWNDSCALE;
 520                 if (cu->cu_cwnd_wait) {
 521                         cu->cu_cwnd_wait = FALSE;
 522                         wakeup(&cu->cu_cwnd_wait);
 523                 }
 524                 goto out;
 525         }
 526         if (cr->cr_mrep) {
 527                 TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
 528                 cu->cu_sent -= CWNDSCALE;
 529                 if (cu->cu_cwnd_wait) {
 530                         cu->cu_cwnd_wait = FALSE;
 531                         wakeup(&cu->cu_cwnd_wait);
 532                 }
 533                 goto got_reply;
 534         }
 535
 536         /*
 537          * Hack to provide rpc-based message passing
 538          */
 539         if (timeout == 0) {
 540                 TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
 541                 errp->re_status = stat = RPC_TIMEDOUT;
 542                 cu->cu_sent -= CWNDSCALE;
 543                 if (cu->cu_cwnd_wait) {
 544                         cu->cu_cwnd_wait = FALSE;
 545                         wakeup(&cu->cu_cwnd_wait);
 546                 }
 547                 goto out;
 548         }
 549
 550 get_reply:
 551         for (;;) {
 552                 /* Decide how long to wait. */
 553                 if (next_sendtime < timeout)
 554                         tv = next_sendtime;
 555                 else
 556                         tv = timeout;
 557                 tv -= time_waited;
 558
 559                 if (tv > 0) {
 560                         if (cu->cu_closing || cu->cu_closed) {
 561                                 error = 0;
 562                                 cr->cr_error = ESHUTDOWN;
 563                         } else {
 564                                 error = msleep(cr, &cs->cs_lock,
 565                                     cu->cu_waitflag, cu->cu_waitchan, tv);
 566                         }
 567                 } else {
 568                         error = EWOULDBLOCK;
 569                 }
 570
 571                 TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
 572                 cu->cu_sent -= CWNDSCALE;
 573                 if (cu->cu_cwnd_wait) {
 574                         cu->cu_cwnd_wait = FALSE;
 575                         wakeup(&cu->cu_cwnd_wait);
 576                 }
 577
 578                 if (!error) {
 579                         /*
 580                          * We were woken up by the upcall.  If the
 581                          * upcall had a receive error, report that,
 582                          * otherwise we have a reply.
 583                          */
 584                         if (cr->cr_error) {
 585                                 errp->re_errno = cr->cr_error;
 586                                 errp->re_status = stat = RPC_CANTRECV;
 587                                 goto out;
 588                         }
 589
 590                         cu->cu_cwnd += (CWNDSCALE * CWNDSCALE
 591                             + cu->cu_cwnd / 2) / cu->cu_cwnd;
 592                         if (cu->cu_cwnd > MAXCWND)
 593                                 cu->cu_cwnd = MAXCWND;
 594
 595                         if (rt) {
 596                                 /*
 597                                  * Add one to the time since a tick
 598                                  * count of N means that the actual
 599                                  * time taken was somewhere between N
 600                                  * and N+1.
 601                                  */
 602                                 rtt = ticks - starttime + 1;
 603
 604                                 /*
 605                                  * Update our estimate of the round
 606                                  * trip time using roughly the
 607                                  * algorithm described in RFC
 608                                  * 2988. Given an RTT sample R:
 609                                  *
 610                                  * RTTVAR = (1-beta) * RTTVAR + beta * |SRTT-R|
 611                                  * SRTT = (1-alpha) * SRTT + alpha * R
 612                                  *
 613                                  * where alpha = 0.125 and beta = 0.25.
 614                                  *
 615                                  * The initial retransmit timeout is
 616                                  * SRTT + 4*RTTVAR and doubles on each
 617                                  * retransmision.
 618                                  */
 619                                 if (rt->rt_srtt == 0) {
 620                                         rt->rt_srtt = rtt;
 621                                         rt->rt_deviate = rtt / 2;
 622                                 } else {
 623                                         int32_t error = rtt - rt->rt_srtt;
 624                                         rt->rt_srtt += error / 8;
 625                                         error = abs(error) - rt->rt_deviate;
 626                                         rt->rt_deviate += error / 4;
 627                                 }
 628                                 rt->rt_rtxcur = rt->rt_srtt + 4*rt->rt_deviate;
 629                         }
 630
 631                         break;
 632                 }
 633
 634                 /*
 635                  * The sleep returned an error so our request is still
 636                  * on the list. If we got EWOULDBLOCK, we may want to
 637                  * re-send the request.
 638                  */
 639                 if (error != EWOULDBLOCK) {
 640                         errp->re_errno = error;
 641                         if (error == EINTR || error == ERESTART)
 642                                 errp->re_status = stat = RPC_INTR;
 643                         else
 644                                 errp->re_status = stat = RPC_CANTRECV;
 645                         goto out;
 646                 }
 647
 648                 time_waited = ticks - starttime;
 649
 650                 /* Check for timeout. */
 651                 if (time_waited > timeout) {
 652                         errp->re_errno = EWOULDBLOCK;
 653                         errp->re_status = stat = RPC_TIMEDOUT;
 654                         goto out;
 655                 }
 656
 657                 /* Retransmit if necessary. */
 658                 if (time_waited >= next_sendtime) {
 659                         cu->cu_cwnd /= 2;
 660                         if (cu->cu_cwnd < CWNDSCALE)
 661                                 cu->cu_cwnd = CWNDSCALE;
 662                         if (ext && ext->rc_feedback) {
 663                                 mtx_unlock(&cs->cs_lock);
 664                                 if (retrans == 0)
 665                                         ext->rc_feedback(FEEDBACK_REXMIT1,
 666                                             proc, ext->rc_feedback_arg);
 667                                 else
 668                                         ext->rc_feedback(FEEDBACK_REXMIT2,
 669                                             proc, ext->rc_feedback_arg);
 670                                 mtx_lock(&cs->cs_lock);
 671                         }
 672                         if (cu->cu_closing || cu->cu_closed) {
 673                                 errp->re_errno = ESHUTDOWN;
 674                                 errp->re_status = stat = RPC_CANTRECV;
 675                                 goto out;
 676                         }
 677                         retrans++;
 678                         /* update retransmit_time */
 679                         if (retransmit_time < RPC_MAX_BACKOFF * hz)
 680                                 retransmit_time = 2 * retransmit_time;
 681                         next_sendtime += retransmit_time;
 682                         goto send_again;
 683                 }
 684                 cu->cu_sent += CWNDSCALE;
 685                 TAILQ_INSERT_TAIL(&cs->cs_pending, cr, cr_link);
 686         }
 687
 688 got_reply:
 689         /*
 690          * Now decode and validate the response. We need to drop the
 691          * lock since xdr_replymsg may end up sleeping in malloc.
 692          */
 693         mtx_unlock(&cs->cs_lock);
 694
 695         if (ext && ext->rc_feedback)
 696                 ext->rc_feedback(FEEDBACK_OK, proc, ext->rc_feedback_arg);
 697
 698         xdrmbuf_create(&xdrs, cr->cr_mrep, XDR_DECODE);
 699         ok = xdr_replymsg(&xdrs, &reply_msg);
 700         cr->cr_mrep = NULL;
 701
 702         if (ok) {
 703                 if ((reply_msg.rm_reply.rp_stat == MSG_ACCEPTED) &&
 704                     (reply_msg.acpted_rply.ar_stat == SUCCESS))
 705                         errp->re_status = stat = RPC_SUCCESS;
 706                 else
 707                         stat = _seterr_reply(&reply_msg, &(cu->cu_error));
 708
 709                 if (errp->re_status == RPC_SUCCESS) {
 710                         results = xdrmbuf_getall(&xdrs);
 711                         if (! AUTH_VALIDATE(auth, xid,
 712                                 &reply_msg.acpted_rply.ar_verf,
 713                                 &results)) {
 714                                 errp->re_status = stat = RPC_AUTHERROR;
 715                                 errp->re_why = AUTH_INVALIDRESP;
 716                                 if (retrans &&
 717                                     auth->ah_cred.oa_flavor == RPCSEC_GSS) {
 718                                         /*
 719                                          * If we retransmitted, its
 720                                          * possible that we will
 721                                          * receive a reply for one of
 722                                          * the earlier transmissions
 723                                          * (which will use an older
 724                                          * RPCSEC_GSS sequence
 725                                          * number). In this case, just
 726                                          * go back and listen for a
 727                                          * new reply. We could keep a
 728                                          * record of all the seq
 729                                          * numbers we have transmitted
 730                                          * so far so that we could
 731                                          * accept a reply for any of
 732                                          * them here.
 733                                          */
 734                                         XDR_DESTROY(&xdrs);
 735                                         mtx_lock(&cs->cs_lock);
 736                                         cu->cu_sent += CWNDSCALE;
 737                                         TAILQ_INSERT_TAIL(&cs->cs_pending,
 738                                             cr, cr_link);
 739                                         cr->cr_mrep = NULL;
 740                                         goto get_reply;
 741                                 }
 742                         } else {
 743                                 *resultsp = results;
 744                         }
 745                 }               /* end successful completion */
 746                 /*
 747                  * If unsuccesful AND error is an authentication error
 748                  * then refresh credentials and try again, else break
 749                  */
 750                 else if (stat == RPC_AUTHERROR)
 751                         /* maybe our credentials need to be refreshed ... */
 752                         if (nrefreshes > 0 &&
 753                             AUTH_REFRESH(auth, &reply_msg)) {
 754                                 nrefreshes--;
 755                                 XDR_DESTROY(&xdrs);
 756                                 mtx_lock(&cs->cs_lock);
 757                                 goto call_again;
 758                         }
 759                 /* end of unsuccessful completion */
 760         }       /* end of valid reply message */
 761         else {
 762                 errp->re_status = stat = RPC_CANTDECODERES;
 763
 764         }
 765         XDR_DESTROY(&xdrs);
 766         mtx_lock(&cs->cs_lock);
 767 out:
 768         mtx_assert(&cs->cs_lock, MA_OWNED);
 769
 770         if (mreq)
 771                 m_freem(mreq);
 772         if (cr->cr_mrep)
 773                 m_freem(cr->cr_mrep);
 774
 775         cu->cu_threads--;
 776         if (cu->cu_closing)
 777                 wakeup(cu);
 778
 779         mtx_unlock(&cs->cs_lock);
 780
 781         if (auth && stat != RPC_SUCCESS)
 782                 AUTH_VALIDATE(auth, xid, NULL, NULL);
 783
 784         free(cr, M_RPC);
 785
 786         return (stat);
 787 }
 788
 789 static void
 790 clnt_dg_geterr(CLIENT *cl, struct rpc_err *errp)
 791 {
 792         struct cu_data *cu = (struct cu_data *)cl->cl_private;
 793
 794         *errp = cu->cu_error;
 795 }
 796
 797 static bool_t
 798 clnt_dg_freeres(CLIENT *cl, xdrproc_t xdr_res, void *res_ptr)
 799 {
 800         XDR xdrs;
 801         bool_t dummy;
 802
 803         xdrs.x_op = XDR_FREE;
 804         dummy = (*xdr_res)(&xdrs, res_ptr);
 805
 806         return (dummy);
 807 }
 808
 809 /*ARGSUSED*/
 810 static void
 811 clnt_dg_abort(CLIENT *h)
 812 {
 813 }
 814
 815 static bool_t
 816 clnt_dg_control(CLIENT *cl, u_int request, void *info)
 817 {
 818         struct cu_data *cu = (struct cu_data *)cl->cl_private;
 819         struct cu_socket *cs;
 820         struct sockaddr *addr;
 821
 822         cs = cu->cu_socket->so_rcv.sb_upcallarg;
 823         mtx_lock(&cs->cs_lock);
 824
 825         switch (request) {
 826         case CLSET_FD_CLOSE:
 827                 cu->cu_closeit = TRUE;
 828                 mtx_unlock(&cs->cs_lock);
 829                 return (TRUE);
 830         case CLSET_FD_NCLOSE:
 831                 cu->cu_closeit = FALSE;
 832                 mtx_unlock(&cs->cs_lock);
 833                 return (TRUE);
 834         }
 835
 836         /* for other requests which use info */
 837         if (info == NULL) {
 838                 mtx_unlock(&cs->cs_lock);
 839                 return (FALSE);
 840         }
 841         switch (request) {
 842         case CLSET_TIMEOUT:
 843                 if (time_not_ok((struct timeval *)info)) {
 844                         mtx_unlock(&cs->cs_lock);
 845                         return (FALSE);
 846                 }
 847                 cu->cu_total = *(struct timeval *)info;
 848                 break;
 849         case CLGET_TIMEOUT:
 850                 *(struct timeval *)info = cu->cu_total;
 851                 break;
 852         case CLSET_RETRY_TIMEOUT:
 853                 if (time_not_ok((struct timeval *)info)) {
 854                         mtx_unlock(&cs->cs_lock);
 855                         return (FALSE);
 856                 }
 857                 cu->cu_wait = *(struct timeval *)info;
 858                 break;
 859         case CLGET_RETRY_TIMEOUT:
 860                 *(struct timeval *)info = cu->cu_wait;
 861                 break;
 862         case CLGET_SVC_ADDR:
 863                 /*
 864                  * Slightly different semantics to userland - we use
 865                  * sockaddr instead of netbuf.
 866                  */
 867                 memcpy(info, &cu->cu_raddr, cu->cu_raddr.ss_len);
 868                 break;
 869         case CLSET_SVC_ADDR:            /* set to new address */
 870                 addr = (struct sockaddr *)info;
 871                 (void) memcpy(&cu->cu_raddr, addr, addr->sa_len);
 872                 break;
 873         case CLGET_XID:
 874                 *(uint32_t *)info = cu->cu_xid;
 875                 break;
 876
 877         case CLSET_XID:
 878                 /* This will set the xid of the NEXT call */
 879                 /* decrement by 1 as clnt_dg_call() increments once */
 880                 cu->cu_xid = *(uint32_t *)info - 1;
 881                 break;
 882
 883         case CLGET_VERS:
 884                 /*
 885                  * This RELIES on the information that, in the call body,
 886                  * the version number field is the fifth field from the
 887                  * begining of the RPC header. MUST be changed if the
 888                  * call_struct is changed
 889                  */
 890                 *(uint32_t *)info =
 891                     ntohl(*(uint32_t *)(void *)(cu->cu_mcallc +
 892                     4 * BYTES_PER_XDR_UNIT));
 893                 break;
 894
 895         case CLSET_VERS:
 896                 *(uint32_t *)(void *)(cu->cu_mcallc + 4 * BYTES_PER_XDR_UNIT)
 897                         = htonl(*(uint32_t *)info);
 898                 break;
 899
 900         case CLGET_PROG:
 901                 /*
 902                  * This RELIES on the information that, in the call body,
 903                  * the program number field is the fourth field from the
 904                  * begining of the RPC header. MUST be changed if the
 905                  * call_struct is changed
 906                  */
 907                 *(uint32_t *)info =
 908                     ntohl(*(uint32_t *)(void *)(cu->cu_mcallc +
 909                     3 * BYTES_PER_XDR_UNIT));
 910                 break;
 911
 912         case CLSET_PROG:
 913                 *(uint32_t *)(void *)(cu->cu_mcallc + 3 * BYTES_PER_XDR_UNIT)
 914                         = htonl(*(uint32_t *)info);
 915                 break;
 916         case CLSET_ASYNC:
 917                 cu->cu_async = *(int *)info;
 918                 break;
 919         case CLSET_CONNECT:
 920                 cu->cu_connect = *(int *)info;
 921                 break;
 922         case CLSET_WAITCHAN:
 923                 cu->cu_waitchan = (const char *)info;
 924                 break;
 925         case CLGET_WAITCHAN:
 926                 *(const char **) info = cu->cu_waitchan;
 927                 break;
 928         case CLSET_INTERRUPTIBLE:
 929                 if (*(int *) info)
 930                         cu->cu_waitflag = PCATCH;
 931                 else
 932                         cu->cu_waitflag = 0;
 933                 break;
 934         case CLGET_INTERRUPTIBLE:
 935                 if (cu->cu_waitflag)
 936                         *(int *) info = TRUE;
 937                 else
 938                         *(int *) info = FALSE;
 939                 break;
 940         default:
 941                 mtx_unlock(&cs->cs_lock);
 942                 return (FALSE);
 943         }
 944         mtx_unlock(&cs->cs_lock);
 945         return (TRUE);
 946 }
 947
 948 static void
 949 clnt_dg_close(CLIENT *cl)
 950 {
 951         struct cu_data *cu = (struct cu_data *)cl->cl_private;
 952         struct cu_socket *cs;
 953         struct cu_request *cr;
 954
 955         cs = cu->cu_socket->so_rcv.sb_upcallarg;
 956         mtx_lock(&cs->cs_lock);
 957
 958         if (cu->cu_closed) {
 959                 mtx_unlock(&cs->cs_lock);
 960                 return;
 961         }
 962
 963         if (cu->cu_closing) {
 964                 while (cu->cu_closing)
 965                         msleep(cu, &cs->cs_lock, 0, "rpcclose", 0);
 966                 KASSERT(cu->cu_closed, ("client should be closed"));
 967                 mtx_unlock(&cs->cs_lock);
 968                 return;
 969         }
 970
 971         /*
 972          * Abort any pending requests and wait until everyone
 973          * has finished with clnt_vc_call.
 974          */
 975         cu->cu_closing = TRUE;
 976         TAILQ_FOREACH(cr, &cs->cs_pending, cr_link) {
 977                 if (cr->cr_client == cl) {
 978                         cr->cr_xid = 0;
 979                         cr->cr_error = ESHUTDOWN;
 980                         wakeup(cr);
 981                 }
 982         }
 983
 984         while (cu->cu_threads)
 985                 msleep(cu, &cs->cs_lock, 0, "rpcclose", 0);
 986
 987         cu->cu_closing = FALSE;
 988         cu->cu_closed = TRUE;
 989
 990         mtx_unlock(&cs->cs_lock);
 991         wakeup(cu);
 992 }
 993
 994 static void
 995 clnt_dg_destroy(CLIENT *cl)
 996 {
 997         struct cu_data *cu = (struct cu_data *)cl->cl_private;
 998         struct cu_socket *cs;
 999         struct socket *so = NULL;
1000         bool_t lastsocketref;
1001
1002         cs = cu->cu_socket->so_rcv.sb_upcallarg;
1003         clnt_dg_close(cl);
1004
1005         SOCKBUF_LOCK(&cu->cu_socket->so_rcv);
1006         mtx_lock(&cs->cs_lock);
1007
1008         cs->cs_refs--;
1009         if (cs->cs_refs == 0) {
1010                 mtx_unlock(&cs->cs_lock);
1011                 soupcall_clear(cu->cu_socket, SO_RCV);
1012                 clnt_dg_upcallsdone(cu->cu_socket, cs);
1013                 SOCKBUF_UNLOCK(&cu->cu_socket->so_rcv);
1014                 mtx_destroy(&cs->cs_lock);
1015                 mem_free(cs, sizeof(*cs));
1016                 lastsocketref = TRUE;
1017         } else {
1018                 mtx_unlock(&cs->cs_lock);
1019                 SOCKBUF_UNLOCK(&cu->cu_socket->so_rcv);
1020                 lastsocketref = FALSE;
1021         }
1022
1023         if (cu->cu_closeit && lastsocketref) {
1024                 so = cu->cu_socket;
1025                 cu->cu_socket = NULL;
1026         }
1027
1028         if (so)
1029                 soclose(so);
1030
1031         if (cl->cl_netid && cl->cl_netid[0])
1032                 mem_free(cl->cl_netid, strlen(cl->cl_netid) +1);
1033         if (cl->cl_tp && cl->cl_tp[0])
1034                 mem_free(cl->cl_tp, strlen(cl->cl_tp) +1);
1035         mem_free(cu, sizeof (*cu));
1036         mem_free(cl, sizeof (CLIENT));
1037 }
1038
1039 /*
1040  * Make sure that the time is not garbage.  -1 value is allowed.
1041  */
1042 static bool_t
1043 time_not_ok(struct timeval *t)
1044 {
1045         return (t->tv_sec < -1 || t->tv_sec > 100000000 ||
1046                 t->tv_usec < -1 || t->tv_usec > 1000000);
1047 }
1048
1049 int
1050 clnt_dg_soupcall(struct socket *so, void *arg, int waitflag)
1051 {
1052         struct cu_socket *cs = (struct cu_socket *) arg;
1053         struct uio uio;
1054         struct mbuf *m;
1055         struct mbuf *control;
1056         struct cu_request *cr;
1057         int error, rcvflag, foundreq;
1058         uint32_t xid;
1059
1060         cs->cs_upcallrefs++;
1061         uio.uio_resid = 1000000000;
1062         uio.uio_td = curthread;
1063         do {
1064                 SOCKBUF_UNLOCK(&so->so_rcv);
1065                 m = NULL;
1066                 control = NULL;
1067                 rcvflag = MSG_DONTWAIT;
1068                 error = soreceive(so, NULL, &uio, &m, &control, &rcvflag);
1069                 if (control)
1070                         m_freem(control);
1071                 SOCKBUF_LOCK(&so->so_rcv);
1072
1073                 if (error == EWOULDBLOCK)
1074                         break;
1075
1076                 /*
1077                  * If there was an error, wake up all pending
1078                  * requests.
1079                  */
1080                 if (error) {
1081                         mtx_lock(&cs->cs_lock);
1082                         TAILQ_FOREACH(cr, &cs->cs_pending, cr_link) {
1083                                 cr->cr_xid = 0;
1084                                 cr->cr_error = error;
1085                                 wakeup(cr);
1086                         }
1087                         mtx_unlock(&cs->cs_lock);
1088                         break;
1089                 }
1090
1091                 /*
1092                  * The XID is in the first uint32_t of the reply.
1093                  */
1094                 if (m->m_len < sizeof(xid) && m_length(m, NULL) < sizeof(xid)) {
1095                         /*
1096                          * Should never happen.
1097                          */
1098                         m_freem(m);
1099                         continue;
1100                 }
1101
1102                 m_copydata(m, 0, sizeof(xid), (char *)&xid);
1103                 xid = ntohl(xid);
1104
1105                 /*
1106                  * Attempt to match this reply with a pending request.
1107                  */
1108                 mtx_lock(&cs->cs_lock);
1109                 foundreq = 0;
1110                 TAILQ_FOREACH(cr, &cs->cs_pending, cr_link) {
1111                         if (cr->cr_xid == xid) {
1112                                 /*
1113                                  * This one matches. We leave the
1114                                  * reply mbuf in cr->cr_mrep. Set the
1115                                  * XID to zero so that we will ignore
1116                                  * any duplicated replies that arrive
1117                                  * before clnt_dg_call removes it from
1118                                  * the queue.
1119                                  */
1120                                 cr->cr_xid = 0;
1121                                 cr->cr_mrep = m;
1122                                 cr->cr_error = 0;
1123                                 foundreq = 1;
1124                                 wakeup(cr);
1125                                 break;
1126                         }
1127                 }
1128                 mtx_unlock(&cs->cs_lock);
1129
1130                 /*
1131                  * If we didn't find the matching request, just drop
1132                  * it - its probably a repeated reply.
1133                  */
1134                 if (!foundreq)
1135                         m_freem(m);
1136         } while (m);
1137         cs->cs_upcallrefs--;
1138         if (cs->cs_upcallrefs < 0)
1139                 panic("rpcdg upcall refcnt");
1140         if (cs->cs_upcallrefs == 0)
1141                 wakeup(&cs->cs_upcallrefs);
1142         return (SU_OK);
1143 }
1144
1145 /*
1146  * Wait for all upcalls in progress to complete.
1147  */
1148 static void
1149 clnt_dg_upcallsdone(struct socket *so, struct cu_socket *cs)
1150 {
1151
1152         SOCKBUF_LOCK_ASSERT(&so->so_rcv);
1153
1154         while (cs->cs_upcallrefs > 0)
1155                 (void) msleep(&cs->cs_upcallrefs, SOCKBUF_MTX(&so->so_rcv), 0,
1156                     "rpcdgup", 0);
1157 }