drivers/staging/lustre/lnet/selftest/rpc.c

   1 /*
   2  * GPL HEADER START
   3  *
   4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 only,
   8  * as published by the Free Software Foundation.
   9  *
  10  * This program is distributed in the hope that it will be useful, but
  11  * WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13  * General Public License version 2 for more details (a copy is included
  14  * in the LICENSE file that accompanied this code).
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * version 2 along with this program; If not, see
  18  * http://www.gnu.org/licenses/gpl-2.0.html
  19  *
  20  * GPL HEADER END
  21  */
  22 /*
  23  * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
  24  * Use is subject to license terms.
  25  *
  26  * Copyright (c) 2012, 2015, Intel Corporation.
  27  */
  28 /*
  29  * This file is part of Lustre, http://www.lustre.org/
  30  * Lustre is a trademark of Sun Microsystems, Inc.
  31  *
  32  * lnet/selftest/rpc.c
  33  *
  34  * Author: Isaac Huang <isaac@clusterfs.com>
  35  *
  36  * 2012-05-13: Liang Zhen <liang@whamcloud.com>
  37  * - percpt data for service to improve smp performance
  38  * - code cleanup
  39  */
  40
  41 #define DEBUG_SUBSYSTEM S_LNET
  42
  43 #include "selftest.h"
  44
  45 enum srpc_state {
  46         SRPC_STATE_NONE,
  47         SRPC_STATE_NI_INIT,
  48         SRPC_STATE_EQ_INIT,
  49         SRPC_STATE_RUNNING,
  50         SRPC_STATE_STOPPING,
  51 };
  52
  53 static struct smoketest_rpc {
  54         spinlock_t       rpc_glock;     /* global lock */
  55         struct srpc_service     *rpc_services[SRPC_SERVICE_MAX_ID + 1];
  56         lnet_handle_eq_t rpc_lnet_eq;   /* _the_ LNet event queue */
  57         enum srpc_state  rpc_state;
  58         srpc_counters_t  rpc_counters;
  59         __u64            rpc_matchbits; /* matchbits counter */
  60 } srpc_data;
  61
  62 static inline int
  63 srpc_serv_portal(int svc_id)
  64 {
  65         return svc_id < SRPC_FRAMEWORK_SERVICE_MAX_ID ?
  66                SRPC_FRAMEWORK_REQUEST_PORTAL : SRPC_REQUEST_PORTAL;
  67 }
  68
  69 /* forward ref's */
  70 int srpc_handle_rpc(struct swi_workitem *wi);
  71
  72 void srpc_get_counters(srpc_counters_t *cnt)
  73 {
  74         spin_lock(&srpc_data.rpc_glock);
  75         *cnt = srpc_data.rpc_counters;
  76         spin_unlock(&srpc_data.rpc_glock);
  77 }
  78
  79 void srpc_set_counters(const srpc_counters_t *cnt)
  80 {
  81         spin_lock(&srpc_data.rpc_glock);
  82         srpc_data.rpc_counters = *cnt;
  83         spin_unlock(&srpc_data.rpc_glock);
  84 }
  85
  86 static int
  87 srpc_add_bulk_page(struct srpc_bulk *bk, struct page *pg, int i, int nob)
  88 {
  89         nob = min_t(int, nob, PAGE_SIZE);
  90
  91         LASSERT(nob > 0);
  92         LASSERT(i >= 0 && i < bk->bk_niov);
  93
  94         bk->bk_iovs[i].bv_offset = 0;
  95         bk->bk_iovs[i].bv_page = pg;
  96         bk->bk_iovs[i].bv_len = nob;
  97         return nob;
  98 }
  99
 100 void
 101 srpc_free_bulk(struct srpc_bulk *bk)
 102 {
 103         int i;
 104         struct page *pg;
 105
 106         LASSERT(bk);
 107
 108         for (i = 0; i < bk->bk_niov; i++) {
 109                 pg = bk->bk_iovs[i].bv_page;
 110                 if (!pg)
 111                         break;
 112
 113                 __free_page(pg);
 114         }
 115
 116         LIBCFS_FREE(bk, offsetof(struct srpc_bulk, bk_iovs[bk->bk_niov]));
 117 }
 118
 119 struct srpc_bulk *
 120 srpc_alloc_bulk(int cpt, unsigned bulk_npg, unsigned bulk_len, int sink)
 121 {
 122         struct srpc_bulk *bk;
 123         int i;
 124
 125         LASSERT(bulk_npg > 0 && bulk_npg <= LNET_MAX_IOV);
 126
 127         LIBCFS_CPT_ALLOC(bk, lnet_cpt_table(), cpt,
 128                          offsetof(struct srpc_bulk, bk_iovs[bulk_npg]));
 129         if (!bk) {
 130                 CERROR("Can't allocate descriptor for %d pages\n", bulk_npg);
 131                 return NULL;
 132         }
 133
 134         memset(bk, 0, offsetof(struct srpc_bulk, bk_iovs[bulk_npg]));
 135         bk->bk_sink = sink;
 136         bk->bk_len = bulk_len;
 137         bk->bk_niov = bulk_npg;
 138
 139         for (i = 0; i < bulk_npg; i++) {
 140                 struct page *pg;
 141                 int nob;
 142
 143                 pg = alloc_pages_node(cfs_cpt_spread_node(lnet_cpt_table(), cpt),
 144                                       GFP_KERNEL, 0);
 145                 if (!pg) {
 146                         CERROR("Can't allocate page %d of %d\n", i, bulk_npg);
 147                         srpc_free_bulk(bk);
 148                         return NULL;
 149                 }
 150
 151                 nob = srpc_add_bulk_page(bk, pg, i, bulk_len);
 152                 bulk_len -= nob;
 153         }
 154
 155         return bk;
 156 }
 157
 158 static inline __u64
 159 srpc_next_id(void)
 160 {
 161         __u64 id;
 162
 163         spin_lock(&srpc_data.rpc_glock);
 164         id = srpc_data.rpc_matchbits++;
 165         spin_unlock(&srpc_data.rpc_glock);
 166         return id;
 167 }
 168
 169 static void
 170 srpc_init_server_rpc(struct srpc_server_rpc *rpc,
 171                      struct srpc_service_cd *scd,
 172                      struct srpc_buffer *buffer)
 173 {
 174         memset(rpc, 0, sizeof(*rpc));
 175         swi_init_workitem(&rpc->srpc_wi, rpc, srpc_handle_rpc,
 176                           srpc_serv_is_framework(scd->scd_svc) ?
 177                           lst_sched_serial : lst_sched_test[scd->scd_cpt]);
 178
 179         rpc->srpc_ev.ev_fired = 1; /* no event expected now */
 180
 181         rpc->srpc_scd = scd;
 182         rpc->srpc_reqstbuf = buffer;
 183         rpc->srpc_peer = buffer->buf_peer;
 184         rpc->srpc_self = buffer->buf_self;
 185         LNetInvalidateHandle(&rpc->srpc_replymdh);
 186 }
 187
 188 static void
 189 srpc_service_fini(struct srpc_service *svc)
 190 {
 191         struct srpc_service_cd *scd;
 192         struct srpc_server_rpc *rpc;
 193         struct srpc_buffer *buf;
 194         struct list_head *q;
 195         int i;
 196
 197         if (!svc->sv_cpt_data)
 198                 return;
 199
 200         cfs_percpt_for_each(scd, i, svc->sv_cpt_data) {
 201                 while (1) {
 202                         if (!list_empty(&scd->scd_buf_posted))
 203                                 q = &scd->scd_buf_posted;
 204                         else if (!list_empty(&scd->scd_buf_blocked))
 205                                 q = &scd->scd_buf_blocked;
 206                         else
 207                                 break;
 208
 209                         while (!list_empty(q)) {
 210                                 buf = list_entry(q->next, struct srpc_buffer,
 211                                                  buf_list);
 212                                 list_del(&buf->buf_list);
 213                                 LIBCFS_FREE(buf, sizeof(*buf));
 214                         }
 215                 }
 216
 217                 LASSERT(list_empty(&scd->scd_rpc_active));
 218
 219                 while (!list_empty(&scd->scd_rpc_free)) {
 220                         rpc = list_entry(scd->scd_rpc_free.next,
 221                                          struct srpc_server_rpc,
 222                                          srpc_list);
 223                         list_del(&rpc->srpc_list);
 224                         LIBCFS_FREE(rpc, sizeof(*rpc));
 225                 }
 226         }
 227
 228         cfs_percpt_free(svc->sv_cpt_data);
 229         svc->sv_cpt_data = NULL;
 230 }
 231
 232 static int
 233 srpc_service_nrpcs(struct srpc_service *svc)
 234 {
 235         int nrpcs = svc->sv_wi_total / svc->sv_ncpts;
 236
 237         return srpc_serv_is_framework(svc) ?
 238                max(nrpcs, SFW_FRWK_WI_MIN) : max(nrpcs, SFW_TEST_WI_MIN);
 239 }
 240
 241 int srpc_add_buffer(struct swi_workitem *wi);
 242
 243 static int
 244 srpc_service_init(struct srpc_service *svc)
 245 {
 246         struct srpc_service_cd *scd;
 247         struct srpc_server_rpc *rpc;
 248         int nrpcs;
 249         int i;
 250         int j;
 251
 252         svc->sv_shuttingdown = 0;
 253
 254         svc->sv_cpt_data = cfs_percpt_alloc(lnet_cpt_table(),
 255                                             sizeof(**svc->sv_cpt_data));
 256         if (!svc->sv_cpt_data)
 257                 return -ENOMEM;
 258
 259         svc->sv_ncpts = srpc_serv_is_framework(svc) ?
 260                         1 : cfs_cpt_number(lnet_cpt_table());
 261         nrpcs = srpc_service_nrpcs(svc);
 262
 263         cfs_percpt_for_each(scd, i, svc->sv_cpt_data) {
 264                 scd->scd_cpt = i;
 265                 scd->scd_svc = svc;
 266                 spin_lock_init(&scd->scd_lock);
 267                 INIT_LIST_HEAD(&scd->scd_rpc_free);
 268                 INIT_LIST_HEAD(&scd->scd_rpc_active);
 269                 INIT_LIST_HEAD(&scd->scd_buf_posted);
 270                 INIT_LIST_HEAD(&scd->scd_buf_blocked);
 271
 272                 scd->scd_ev.ev_data = scd;
 273                 scd->scd_ev.ev_type = SRPC_REQUEST_RCVD;
 274
 275                 /*
 276                  * NB: don't use lst_sched_serial for adding buffer,
 277                  * see details in srpc_service_add_buffers()
 278                  */
 279                 swi_init_workitem(&scd->scd_buf_wi, scd,
 280                                   srpc_add_buffer, lst_sched_test[i]);
 281
 282                 if (i && srpc_serv_is_framework(svc)) {
 283                         /*
 284                          * NB: framework service only needs srpc_service_cd for
 285                          * one partition, but we allocate for all to make
 286                          * it easier to implement, it will waste a little
 287                          * memory but nobody should care about this
 288                          */
 289                         continue;
 290                 }
 291
 292                 for (j = 0; j < nrpcs; j++) {
 293                         LIBCFS_CPT_ALLOC(rpc, lnet_cpt_table(),
 294                                          i, sizeof(*rpc));
 295                         if (!rpc) {
 296                                 srpc_service_fini(svc);
 297                                 return -ENOMEM;
 298                         }
 299                         list_add(&rpc->srpc_list, &scd->scd_rpc_free);
 300                 }
 301         }
 302
 303         return 0;
 304 }
 305
 306 int
 307 srpc_add_service(struct srpc_service *sv)
 308 {
 309         int id = sv->sv_id;
 310
 311         LASSERT(0 <= id && id <= SRPC_SERVICE_MAX_ID);
 312
 313         if (srpc_service_init(sv))
 314                 return -ENOMEM;
 315
 316         spin_lock(&srpc_data.rpc_glock);
 317
 318         LASSERT(srpc_data.rpc_state == SRPC_STATE_RUNNING);
 319
 320         if (srpc_data.rpc_services[id]) {
 321                 spin_unlock(&srpc_data.rpc_glock);
 322                 goto failed;
 323         }
 324
 325         srpc_data.rpc_services[id] = sv;
 326         spin_unlock(&srpc_data.rpc_glock);
 327
 328         CDEBUG(D_NET, "Adding service: id %d, name %s\n", id, sv->sv_name);
 329         return 0;
 330
 331  failed:
 332         srpc_service_fini(sv);
 333         return -EBUSY;
 334 }
 335
 336 int
 337 srpc_remove_service(struct srpc_service *sv)
 338 {
 339         int id = sv->sv_id;
 340
 341         spin_lock(&srpc_data.rpc_glock);
 342
 343         if (srpc_data.rpc_services[id] != sv) {
 344                 spin_unlock(&srpc_data.rpc_glock);
 345                 return -ENOENT;
 346         }
 347
 348         srpc_data.rpc_services[id] = NULL;
 349         spin_unlock(&srpc_data.rpc_glock);
 350         return 0;
 351 }
 352
 353 static int
 354 srpc_post_passive_rdma(int portal, int local, __u64 matchbits, void *buf,
 355                        int len, int options, lnet_process_id_t peer,
 356                        lnet_handle_md_t *mdh, struct srpc_event *ev)
 357 {
 358         int rc;
 359         lnet_md_t md;
 360         lnet_handle_me_t meh;
 361
 362         rc = LNetMEAttach(portal, peer, matchbits, 0, LNET_UNLINK,
 363                           local ? LNET_INS_LOCAL : LNET_INS_AFTER, &meh);
 364         if (rc) {
 365                 CERROR("LNetMEAttach failed: %d\n", rc);
 366                 LASSERT(rc == -ENOMEM);
 367                 return -ENOMEM;
 368         }
 369
 370         md.threshold = 1;
 371         md.user_ptr = ev;
 372         md.start = buf;
 373         md.length = len;
 374         md.options = options;
 375         md.eq_handle = srpc_data.rpc_lnet_eq;
 376
 377         rc = LNetMDAttach(meh, md, LNET_UNLINK, mdh);
 378         if (rc) {
 379                 CERROR("LNetMDAttach failed: %d\n", rc);
 380                 LASSERT(rc == -ENOMEM);
 381
 382                 rc = LNetMEUnlink(meh);
 383                 LASSERT(!rc);
 384                 return -ENOMEM;
 385         }
 386
 387         CDEBUG(D_NET, "Posted passive RDMA: peer %s, portal %d, matchbits %#llx\n",
 388                libcfs_id2str(peer), portal, matchbits);
 389         return 0;
 390 }
 391
 392 static int
 393 srpc_post_active_rdma(int portal, __u64 matchbits, void *buf, int len,
 394                       int options, lnet_process_id_t peer, lnet_nid_t self,
 395                       lnet_handle_md_t *mdh, struct srpc_event *ev)
 396 {
 397         int rc;
 398         lnet_md_t md;
 399
 400         md.user_ptr = ev;
 401         md.start = buf;
 402         md.length = len;
 403         md.eq_handle = srpc_data.rpc_lnet_eq;
 404         md.threshold = options & LNET_MD_OP_GET ? 2 : 1;
 405         md.options = options & ~(LNET_MD_OP_PUT | LNET_MD_OP_GET);
 406
 407         rc = LNetMDBind(md, LNET_UNLINK, mdh);
 408         if (rc) {
 409                 CERROR("LNetMDBind failed: %d\n", rc);
 410                 LASSERT(rc == -ENOMEM);
 411                 return -ENOMEM;
 412         }
 413
 414         /*
 415          * this is kind of an abuse of the LNET_MD_OP_{PUT,GET} options.
 416          * they're only meaningful for MDs attached to an ME (i.e. passive
 417          * buffers...
 418          */
 419         if (options & LNET_MD_OP_PUT) {
 420                 rc = LNetPut(self, *mdh, LNET_NOACK_REQ, peer,
 421                              portal, matchbits, 0, 0);
 422         } else {
 423                 LASSERT(options & LNET_MD_OP_GET);
 424
 425                 rc = LNetGet(self, *mdh, peer, portal, matchbits, 0);
 426         }
 427
 428         if (rc) {
 429                 CERROR("LNet%s(%s, %d, %lld) failed: %d\n",
 430                        options & LNET_MD_OP_PUT ? "Put" : "Get",
 431                        libcfs_id2str(peer), portal, matchbits, rc);
 432
 433                 /*
 434                  * The forthcoming unlink event will complete this operation
 435                  * with failure, so fall through and return success here.
 436                  */
 437                 rc = LNetMDUnlink(*mdh);
 438                 LASSERT(!rc);
 439         } else {
 440                 CDEBUG(D_NET, "Posted active RDMA: peer %s, portal %u, matchbits %#llx\n",
 441                        libcfs_id2str(peer), portal, matchbits);
 442         }
 443         return 0;
 444 }
 445
 446 static int
 447 srpc_post_passive_rqtbuf(int service, int local, void *buf, int len,
 448                          lnet_handle_md_t *mdh, struct srpc_event *ev)
 449 {
 450         lnet_process_id_t any = { 0 };
 451
 452         any.nid = LNET_NID_ANY;
 453         any.pid = LNET_PID_ANY;
 454
 455         return srpc_post_passive_rdma(srpc_serv_portal(service),
 456                                       local, service, buf, len,
 457                                       LNET_MD_OP_PUT, any, mdh, ev);
 458 }
 459
 460 static int
 461 srpc_service_post_buffer(struct srpc_service_cd *scd, struct srpc_buffer *buf)
 462 __must_hold(&scd->scd_lock)
 463 {
 464         struct srpc_service *sv = scd->scd_svc;
 465         struct srpc_msg *msg = &buf->buf_msg;
 466         int rc;
 467
 468         LNetInvalidateHandle(&buf->buf_mdh);
 469         list_add(&buf->buf_list, &scd->scd_buf_posted);
 470         scd->scd_buf_nposted++;
 471         spin_unlock(&scd->scd_lock);
 472
 473         rc = srpc_post_passive_rqtbuf(sv->sv_id,
 474                                       !srpc_serv_is_framework(sv),
 475                                       msg, sizeof(*msg), &buf->buf_mdh,
 476                                       &scd->scd_ev);
 477
 478         /*
 479          * At this point, a RPC (new or delayed) may have arrived in
 480          * msg and its event handler has been called. So we must add
 481          * buf to scd_buf_posted _before_ dropping scd_lock
 482          */
 483         spin_lock(&scd->scd_lock);
 484
 485         if (!rc) {
 486                 if (!sv->sv_shuttingdown)
 487                         return 0;
 488
 489                 spin_unlock(&scd->scd_lock);
 490                 /*
 491                  * srpc_shutdown_service might have tried to unlink me
 492                  * when my buf_mdh was still invalid
 493                  */
 494                 LNetMDUnlink(buf->buf_mdh);
 495                 spin_lock(&scd->scd_lock);
 496                 return 0;
 497         }
 498
 499         scd->scd_buf_nposted--;
 500         if (sv->sv_shuttingdown)
 501                 return rc; /* don't allow to change scd_buf_posted */
 502
 503         list_del(&buf->buf_list);
 504         spin_unlock(&scd->scd_lock);
 505
 506         LIBCFS_FREE(buf, sizeof(*buf));
 507
 508         spin_lock(&scd->scd_lock);
 509         return rc;
 510 }
 511
 512 int
 513 srpc_add_buffer(struct swi_workitem *wi)
 514 {
 515         struct srpc_service_cd *scd = wi->swi_workitem.wi_data;
 516         struct srpc_buffer *buf;
 517         int rc = 0;
 518
 519         /*
 520          * it's called by workitem scheduler threads, these threads
 521          * should have been set CPT affinity, so buffers will be posted
 522          * on CPT local list of Portal
 523          */
 524         spin_lock(&scd->scd_lock);
 525
 526         while (scd->scd_buf_adjust > 0 &&
 527                !scd->scd_svc->sv_shuttingdown) {
 528                 scd->scd_buf_adjust--; /* consume it */
 529                 scd->scd_buf_posting++;
 530
 531                 spin_unlock(&scd->scd_lock);
 532
 533                 LIBCFS_ALLOC(buf, sizeof(*buf));
 534                 if (!buf) {
 535                         CERROR("Failed to add new buf to service: %s\n",
 536                                scd->scd_svc->sv_name);
 537                         spin_lock(&scd->scd_lock);
 538                         rc = -ENOMEM;
 539                         break;
 540                 }
 541
 542                 spin_lock(&scd->scd_lock);
 543                 if (scd->scd_svc->sv_shuttingdown) {
 544                         spin_unlock(&scd->scd_lock);
 545                         LIBCFS_FREE(buf, sizeof(*buf));
 546
 547                         spin_lock(&scd->scd_lock);
 548                         rc = -ESHUTDOWN;
 549                         break;
 550                 }
 551
 552                 rc = srpc_service_post_buffer(scd, buf);
 553                 if (rc)
 554                         break; /* buf has been freed inside */
 555
 556                 LASSERT(scd->scd_buf_posting > 0);
 557                 scd->scd_buf_posting--;
 558                 scd->scd_buf_total++;
 559                 scd->scd_buf_low = max(2, scd->scd_buf_total / 4);
 560         }
 561
 562         if (rc) {
 563                 scd->scd_buf_err_stamp = ktime_get_real_seconds();
 564                 scd->scd_buf_err = rc;
 565
 566                 LASSERT(scd->scd_buf_posting > 0);
 567                 scd->scd_buf_posting--;
 568         }
 569
 570         spin_unlock(&scd->scd_lock);
 571         return 0;
 572 }
 573
 574 int
 575 srpc_service_add_buffers(struct srpc_service *sv, int nbuffer)
 576 {
 577         struct srpc_service_cd *scd;
 578         int rc = 0;
 579         int i;
 580
 581         LASSERTF(nbuffer > 0, "nbuffer must be positive: %d\n", nbuffer);
 582
 583         cfs_percpt_for_each(scd, i, sv->sv_cpt_data) {
 584                 spin_lock(&scd->scd_lock);
 585
 586                 scd->scd_buf_err = 0;
 587                 scd->scd_buf_err_stamp = 0;
 588                 scd->scd_buf_posting = 0;
 589                 scd->scd_buf_adjust = nbuffer;
 590                 /* start to post buffers */
 591                 swi_schedule_workitem(&scd->scd_buf_wi);
 592                 spin_unlock(&scd->scd_lock);
 593
 594                 /* framework service only post buffer for one partition  */
 595                 if (srpc_serv_is_framework(sv))
 596                         break;
 597         }
 598
 599         cfs_percpt_for_each(scd, i, sv->sv_cpt_data) {
 600                 spin_lock(&scd->scd_lock);
 601                 /*
 602                  * NB: srpc_service_add_buffers() can be called inside
 603                  * thread context of lst_sched_serial, and we don't normally
 604                  * allow to sleep inside thread context of WI scheduler
 605                  * because it will block current scheduler thread from doing
 606                  * anything else, even worse, it could deadlock if it's
 607                  * waiting on result from another WI of the same scheduler.
 608                  * However, it's safe at here because scd_buf_wi is scheduled
 609                  * by thread in a different WI scheduler (lst_sched_test),
 610                  * so we don't have any risk of deadlock, though this could
 611                  * block all WIs pending on lst_sched_serial for a moment
 612                  * which is not good but not fatal.
 613                  */
 614                 lst_wait_until(scd->scd_buf_err ||
 615                                (!scd->scd_buf_adjust &&
 616                                 !scd->scd_buf_posting),
 617                                scd->scd_lock, "waiting for adding buffer\n");
 618
 619                 if (scd->scd_buf_err && !rc)
 620                         rc = scd->scd_buf_err;
 621
 622                 spin_unlock(&scd->scd_lock);
 623         }
 624
 625         return rc;
 626 }
 627
 628 void
 629 srpc_service_remove_buffers(struct srpc_service *sv, int nbuffer)
 630 {
 631         struct srpc_service_cd *scd;
 632         int num;
 633         int i;
 634
 635         LASSERT(!sv->sv_shuttingdown);
 636
 637         cfs_percpt_for_each(scd, i, sv->sv_cpt_data) {
 638                 spin_lock(&scd->scd_lock);
 639
 640                 num = scd->scd_buf_total + scd->scd_buf_posting;
 641                 scd->scd_buf_adjust -= min(nbuffer, num);
 642
 643                 spin_unlock(&scd->scd_lock);
 644         }
 645 }
 646
 647 /* returns 1 if sv has finished, otherwise 0 */
 648 int
 649 srpc_finish_service(struct srpc_service *sv)
 650 {
 651         struct srpc_service_cd *scd;
 652         struct srpc_server_rpc *rpc;
 653         int i;
 654
 655         LASSERT(sv->sv_shuttingdown); /* srpc_shutdown_service called */
 656
 657         cfs_percpt_for_each(scd, i, sv->sv_cpt_data) {
 658                 spin_lock(&scd->scd_lock);
 659                 if (!swi_deschedule_workitem(&scd->scd_buf_wi)) {
 660                         spin_unlock(&scd->scd_lock);
 661                         return 0;
 662                 }
 663
 664                 if (scd->scd_buf_nposted > 0) {
 665                         CDEBUG(D_NET, "waiting for %d posted buffers to unlink\n",
 666                                scd->scd_buf_nposted);
 667                         spin_unlock(&scd->scd_lock);
 668                         return 0;
 669                 }
 670
 671                 if (list_empty(&scd->scd_rpc_active)) {
 672                         spin_unlock(&scd->scd_lock);
 673                         continue;
 674                 }
 675
 676                 rpc = list_entry(scd->scd_rpc_active.next,
 677                                  struct srpc_server_rpc, srpc_list);
 678                 CNETERR("Active RPC %p on shutdown: sv %s, peer %s, wi %s scheduled %d running %d, ev fired %d type %d status %d lnet %d\n",
 679                         rpc, sv->sv_name, libcfs_id2str(rpc->srpc_peer),
 680                         swi_state2str(rpc->srpc_wi.swi_state),
 681                         rpc->srpc_wi.swi_workitem.wi_scheduled,
 682                         rpc->srpc_wi.swi_workitem.wi_running,
 683                         rpc->srpc_ev.ev_fired, rpc->srpc_ev.ev_type,
 684                         rpc->srpc_ev.ev_status, rpc->srpc_ev.ev_lnet);
 685                 spin_unlock(&scd->scd_lock);
 686                 return 0;
 687         }
 688
 689         /* no lock needed from now on */
 690         srpc_service_fini(sv);
 691         return 1;
 692 }
 693
 694 /* called with sv->sv_lock held */
 695 static void
 696 srpc_service_recycle_buffer(struct srpc_service_cd *scd, struct srpc_buffer *buf)
 697 __must_hold(&scd->scd_lock)
 698 {
 699         if (!scd->scd_svc->sv_shuttingdown && scd->scd_buf_adjust >= 0) {
 700                 if (srpc_service_post_buffer(scd, buf)) {
 701                         CWARN("Failed to post %s buffer\n",
 702                               scd->scd_svc->sv_name);
 703                 }
 704                 return;
 705         }
 706
 707         /* service is shutting down, or we want to recycle some buffers */
 708         scd->scd_buf_total--;
 709
 710         if (scd->scd_buf_adjust < 0) {
 711                 scd->scd_buf_adjust++;
 712                 if (scd->scd_buf_adjust < 0 &&
 713                     !scd->scd_buf_total && !scd->scd_buf_posting) {
 714                         CDEBUG(D_INFO,
 715                                "Try to recycle %d buffers but nothing left\n",
 716                                scd->scd_buf_adjust);
 717                         scd->scd_buf_adjust = 0;
 718                 }
 719         }
 720
 721         spin_unlock(&scd->scd_lock);
 722         LIBCFS_FREE(buf, sizeof(*buf));
 723         spin_lock(&scd->scd_lock);
 724 }
 725
 726 void
 727 srpc_abort_service(struct srpc_service *sv)
 728 {
 729         struct srpc_service_cd *scd;
 730         struct srpc_server_rpc *rpc;
 731         int i;
 732
 733         CDEBUG(D_NET, "Aborting service: id %d, name %s\n",
 734                sv->sv_id, sv->sv_name);
 735
 736         cfs_percpt_for_each(scd, i, sv->sv_cpt_data) {
 737                 spin_lock(&scd->scd_lock);
 738
 739                 /*
 740                  * schedule in-flight RPCs to notice the abort, NB:
 741                  * racing with incoming RPCs; complete fix should make test
 742                  * RPCs carry session ID in its headers
 743                  */
 744                 list_for_each_entry(rpc, &scd->scd_rpc_active, srpc_list) {
 745                         rpc->srpc_aborted = 1;
 746                         swi_schedule_workitem(&rpc->srpc_wi);
 747                 }
 748
 749                 spin_unlock(&scd->scd_lock);
 750         }
 751 }
 752
 753 void
 754 srpc_shutdown_service(struct srpc_service *sv)
 755 {
 756         struct srpc_service_cd *scd;
 757         struct srpc_server_rpc *rpc;
 758         struct srpc_buffer *buf;
 759         int i;
 760
 761         CDEBUG(D_NET, "Shutting down service: id %d, name %s\n",
 762                sv->sv_id, sv->sv_name);
 763
 764         cfs_percpt_for_each(scd, i, sv->sv_cpt_data)
 765                 spin_lock(&scd->scd_lock);
 766
 767         sv->sv_shuttingdown = 1; /* i.e. no new active RPC */
 768
 769         cfs_percpt_for_each(scd, i, sv->sv_cpt_data)
 770                 spin_unlock(&scd->scd_lock);
 771
 772         cfs_percpt_for_each(scd, i, sv->sv_cpt_data) {
 773                 spin_lock(&scd->scd_lock);
 774
 775                 /* schedule in-flight RPCs to notice the shutdown */
 776                 list_for_each_entry(rpc, &scd->scd_rpc_active, srpc_list)
 777                         swi_schedule_workitem(&rpc->srpc_wi);
 778
 779                 spin_unlock(&scd->scd_lock);
 780
 781                 /*
 782                  * OK to traverse scd_buf_posted without lock, since no one
 783                  * touches scd_buf_posted now
 784                  */
 785                 list_for_each_entry(buf, &scd->scd_buf_posted, buf_list)
 786                         LNetMDUnlink(buf->buf_mdh);
 787         }
 788 }
 789
 790 static int
 791 srpc_send_request(struct srpc_client_rpc *rpc)
 792 {
 793         struct srpc_event *ev = &rpc->crpc_reqstev;
 794         int rc;
 795
 796         ev->ev_fired = 0;
 797         ev->ev_data = rpc;
 798         ev->ev_type = SRPC_REQUEST_SENT;
 799
 800          rc = srpc_post_active_rdma(srpc_serv_portal(rpc->crpc_service),
 801                                     rpc->crpc_service, &rpc->crpc_reqstmsg,
 802                                     sizeof(struct srpc_msg), LNET_MD_OP_PUT,
 803                                     rpc->crpc_dest, LNET_NID_ANY,
 804                                     &rpc->crpc_reqstmdh, ev);
 805         if (rc) {
 806                 LASSERT(rc == -ENOMEM);
 807                 ev->ev_fired = 1;  /* no more event expected */
 808         }
 809         return rc;
 810 }
 811
 812 static int
 813 srpc_prepare_reply(struct srpc_client_rpc *rpc)
 814 {
 815         struct srpc_event *ev = &rpc->crpc_replyev;
 816         __u64 *id = &rpc->crpc_reqstmsg.msg_body.reqst.rpyid;
 817         int rc;
 818
 819         ev->ev_fired = 0;
 820         ev->ev_data = rpc;
 821         ev->ev_type = SRPC_REPLY_RCVD;
 822
 823         *id = srpc_next_id();
 824
 825         rc = srpc_post_passive_rdma(SRPC_RDMA_PORTAL, 0, *id,
 826                                     &rpc->crpc_replymsg,
 827                                     sizeof(struct srpc_msg),
 828                                     LNET_MD_OP_PUT, rpc->crpc_dest,
 829                                     &rpc->crpc_replymdh, ev);
 830         if (rc) {
 831                 LASSERT(rc == -ENOMEM);
 832                 ev->ev_fired = 1;  /* no more event expected */
 833         }
 834         return rc;
 835 }
 836
 837 static int
 838 srpc_prepare_bulk(struct srpc_client_rpc *rpc)
 839 {
 840         struct srpc_bulk *bk = &rpc->crpc_bulk;
 841         struct srpc_event *ev = &rpc->crpc_bulkev;
 842         __u64 *id = &rpc->crpc_reqstmsg.msg_body.reqst.bulkid;
 843         int rc;
 844         int opt;
 845
 846         LASSERT(bk->bk_niov <= LNET_MAX_IOV);
 847
 848         if (!bk->bk_niov)
 849                 return 0; /* nothing to do */
 850
 851         opt = bk->bk_sink ? LNET_MD_OP_PUT : LNET_MD_OP_GET;
 852         opt |= LNET_MD_KIOV;
 853
 854         ev->ev_fired = 0;
 855         ev->ev_data = rpc;
 856         ev->ev_type = SRPC_BULK_REQ_RCVD;
 857
 858         *id = srpc_next_id();
 859
 860         rc = srpc_post_passive_rdma(SRPC_RDMA_PORTAL, 0, *id,
 861                                     &bk->bk_iovs[0], bk->bk_niov, opt,
 862                                     rpc->crpc_dest, &bk->bk_mdh, ev);
 863         if (rc) {
 864                 LASSERT(rc == -ENOMEM);
 865                 ev->ev_fired = 1;  /* no more event expected */
 866         }
 867         return rc;
 868 }
 869
 870 static int
 871 srpc_do_bulk(struct srpc_server_rpc *rpc)
 872 {
 873         struct srpc_event *ev = &rpc->srpc_ev;
 874         struct srpc_bulk *bk = rpc->srpc_bulk;
 875         __u64 id = rpc->srpc_reqstbuf->buf_msg.msg_body.reqst.bulkid;
 876         int rc;
 877         int opt;
 878
 879         LASSERT(bk);
 880
 881         opt = bk->bk_sink ? LNET_MD_OP_GET : LNET_MD_OP_PUT;
 882         opt |= LNET_MD_KIOV;
 883
 884         ev->ev_fired = 0;
 885         ev->ev_data = rpc;
 886         ev->ev_type = bk->bk_sink ? SRPC_BULK_GET_RPLD : SRPC_BULK_PUT_SENT;
 887
 888         rc = srpc_post_active_rdma(SRPC_RDMA_PORTAL, id,
 889                                    &bk->bk_iovs[0], bk->bk_niov, opt,
 890                                    rpc->srpc_peer, rpc->srpc_self,
 891                                    &bk->bk_mdh, ev);
 892         if (rc)
 893                 ev->ev_fired = 1;  /* no more event expected */
 894         return rc;
 895 }
 896
 897 /* only called from srpc_handle_rpc */
 898 static void
 899 srpc_server_rpc_done(struct srpc_server_rpc *rpc, int status)
 900 {
 901         struct srpc_service_cd *scd = rpc->srpc_scd;
 902         struct srpc_service *sv = scd->scd_svc;
 903         struct srpc_buffer *buffer;
 904
 905         LASSERT(status || rpc->srpc_wi.swi_state == SWI_STATE_DONE);
 906
 907         rpc->srpc_status = status;
 908
 909         CDEBUG_LIMIT(!status ? D_NET : D_NETERROR,
 910                      "Server RPC %p done: service %s, peer %s, status %s:%d\n",
 911                      rpc, sv->sv_name, libcfs_id2str(rpc->srpc_peer),
 912                      swi_state2str(rpc->srpc_wi.swi_state), status);
 913
 914         if (status) {
 915                 spin_lock(&srpc_data.rpc_glock);
 916                 srpc_data.rpc_counters.rpcs_dropped++;
 917                 spin_unlock(&srpc_data.rpc_glock);
 918         }
 919
 920         if (rpc->srpc_done)
 921                 (*rpc->srpc_done) (rpc);
 922         LASSERT(!rpc->srpc_bulk);
 923
 924         spin_lock(&scd->scd_lock);
 925
 926         if (rpc->srpc_reqstbuf) {
 927                 /*
 928                  * NB might drop sv_lock in srpc_service_recycle_buffer, but
 929                  * sv won't go away for scd_rpc_active must not be empty
 930                  */
 931                 srpc_service_recycle_buffer(scd, rpc->srpc_reqstbuf);
 932                 rpc->srpc_reqstbuf = NULL;
 933         }
 934
 935         list_del(&rpc->srpc_list); /* from scd->scd_rpc_active */
 936
 937         /*
 938          * No one can schedule me now since:
 939          * - I'm not on scd_rpc_active.
 940          * - all LNet events have been fired.
 941          * Cancel pending schedules and prevent future schedule attempts:
 942          */
 943         LASSERT(rpc->srpc_ev.ev_fired);
 944         swi_exit_workitem(&rpc->srpc_wi);
 945
 946         if (!sv->sv_shuttingdown && !list_empty(&scd->scd_buf_blocked)) {
 947                 buffer = list_entry(scd->scd_buf_blocked.next,
 948                                     struct srpc_buffer, buf_list);
 949                 list_del(&buffer->buf_list);
 950
 951                 srpc_init_server_rpc(rpc, scd, buffer);
 952                 list_add_tail(&rpc->srpc_list, &scd->scd_rpc_active);
 953                 swi_schedule_workitem(&rpc->srpc_wi);
 954         } else {
 955                 list_add(&rpc->srpc_list, &scd->scd_rpc_free);
 956         }
 957
 958         spin_unlock(&scd->scd_lock);
 959 }
 960
 961 /* handles an incoming RPC */
 962 int
 963 srpc_handle_rpc(struct swi_workitem *wi)
 964 {
 965         struct srpc_server_rpc *rpc = wi->swi_workitem.wi_data;
 966         struct srpc_service_cd *scd = rpc->srpc_scd;
 967         struct srpc_service *sv = scd->scd_svc;
 968         struct srpc_event *ev = &rpc->srpc_ev;
 969         int rc = 0;
 970
 971         LASSERT(wi == &rpc->srpc_wi);
 972
 973         spin_lock(&scd->scd_lock);
 974
 975         if (sv->sv_shuttingdown || rpc->srpc_aborted) {
 976                 spin_unlock(&scd->scd_lock);
 977
 978                 if (rpc->srpc_bulk)
 979                         LNetMDUnlink(rpc->srpc_bulk->bk_mdh);
 980                 LNetMDUnlink(rpc->srpc_replymdh);
 981
 982                 if (ev->ev_fired) { /* no more event, OK to finish */
 983                         srpc_server_rpc_done(rpc, -ESHUTDOWN);
 984                         return 1;
 985                 }
 986                 return 0;
 987         }
 988
 989         spin_unlock(&scd->scd_lock);
 990
 991         switch (wi->swi_state) {
 992         default:
 993                 LBUG();
 994         case SWI_STATE_NEWBORN: {
 995                 struct srpc_msg *msg;
 996                 struct srpc_generic_reply *reply;
 997
 998                 msg = &rpc->srpc_reqstbuf->buf_msg;
 999                 reply = &rpc->srpc_replymsg.msg_body.reply;
1000
1001                 if (!msg->msg_magic) {
1002                         /* moaned already in srpc_lnet_ev_handler */
1003                         srpc_server_rpc_done(rpc, EBADMSG);
1004                         return 1;
1005                 }
1006
1007                 srpc_unpack_msg_hdr(msg);
1008                 if (msg->msg_version != SRPC_MSG_VERSION) {
1009                         CWARN("Version mismatch: %u, %u expected, from %s\n",
1010                               msg->msg_version, SRPC_MSG_VERSION,
1011                               libcfs_id2str(rpc->srpc_peer));
1012                         reply->status = EPROTO;
1013                         /* drop through and send reply */
1014                 } else {
1015                         reply->status = 0;
1016                         rc = (*sv->sv_handler)(rpc);
1017                         LASSERT(!reply->status || !rpc->srpc_bulk);
1018                         if (rc) {
1019                                 srpc_server_rpc_done(rpc, rc);
1020                                 return 1;
1021                         }
1022                 }
1023
1024                 wi->swi_state = SWI_STATE_BULK_STARTED;
1025
1026                 if (rpc->srpc_bulk) {
1027                         rc = srpc_do_bulk(rpc);
1028                         if (!rc)
1029                                 return 0; /* wait for bulk */
1030
1031                         LASSERT(ev->ev_fired);
1032                         ev->ev_status = rc;
1033                 }
1034         }
1035         case SWI_STATE_BULK_STARTED:
1036                 LASSERT(!rpc->srpc_bulk || ev->ev_fired);
1037
1038                 if (rpc->srpc_bulk) {
1039                         rc = ev->ev_status;
1040
1041                         if (sv->sv_bulk_ready)
1042                                 rc = (*sv->sv_bulk_ready) (rpc, rc);
1043
1044                         if (rc) {
1045                                 srpc_server_rpc_done(rpc, rc);
1046                                 return 1;
1047                         }
1048                 }
1049
1050                 wi->swi_state = SWI_STATE_REPLY_SUBMITTED;
1051                 rc = srpc_send_reply(rpc);
1052                 if (!rc)
1053                         return 0; /* wait for reply */
1054                 srpc_server_rpc_done(rpc, rc);
1055                 return 1;
1056
1057         case SWI_STATE_REPLY_SUBMITTED:
1058                 if (!ev->ev_fired) {
1059                         CERROR("RPC %p: bulk %p, service %d\n",
1060                                rpc, rpc->srpc_bulk, sv->sv_id);
1061                         CERROR("Event: status %d, type %d, lnet %d\n",
1062                                ev->ev_status, ev->ev_type, ev->ev_lnet);
1063                         LASSERT(ev->ev_fired);
1064                 }
1065
1066                 wi->swi_state = SWI_STATE_DONE;
1067                 srpc_server_rpc_done(rpc, ev->ev_status);
1068                 return 1;
1069         }
1070
1071         return 0;
1072 }
1073
1074 static void
1075 srpc_client_rpc_expired(void *data)
1076 {
1077         struct srpc_client_rpc *rpc = data;
1078
1079         CWARN("Client RPC expired: service %d, peer %s, timeout %d.\n",
1080               rpc->crpc_service, libcfs_id2str(rpc->crpc_dest),
1081               rpc->crpc_timeout);
1082
1083         spin_lock(&rpc->crpc_lock);
1084
1085         rpc->crpc_timeout = 0;
1086         srpc_abort_rpc(rpc, -ETIMEDOUT);
1087
1088         spin_unlock(&rpc->crpc_lock);
1089
1090         spin_lock(&srpc_data.rpc_glock);
1091         srpc_data.rpc_counters.rpcs_expired++;
1092         spin_unlock(&srpc_data.rpc_glock);
1093 }
1094
1095 static void
1096 srpc_add_client_rpc_timer(struct srpc_client_rpc *rpc)
1097 {
1098         struct stt_timer *timer = &rpc->crpc_timer;
1099
1100         if (!rpc->crpc_timeout)
1101                 return;
1102
1103         INIT_LIST_HEAD(&timer->stt_list);
1104         timer->stt_data = rpc;
1105         timer->stt_func = srpc_client_rpc_expired;
1106         timer->stt_expires = ktime_get_real_seconds() + rpc->crpc_timeout;
1107         stt_add_timer(timer);
1108 }
1109
1110 /*
1111  * Called with rpc->crpc_lock held.
1112  *
1113  * Upon exit the RPC expiry timer is not queued and the handler is not
1114  * running on any CPU.
1115  */
1116 static void
1117 srpc_del_client_rpc_timer(struct srpc_client_rpc *rpc)
1118 {
1119         /* timer not planted or already exploded */
1120         if (!rpc->crpc_timeout)
1121                 return;
1122
1123         /* timer successfully defused */
1124         if (stt_del_timer(&rpc->crpc_timer))
1125                 return;
1126
1127         /* timer detonated, wait for it to explode */
1128         while (rpc->crpc_timeout) {
1129                 spin_unlock(&rpc->crpc_lock);
1130
1131                 schedule();
1132
1133                 spin_lock(&rpc->crpc_lock);
1134         }
1135 }
1136
1137 static void
1138 srpc_client_rpc_done(struct srpc_client_rpc *rpc, int status)
1139 {
1140         struct swi_workitem *wi = &rpc->crpc_wi;
1141
1142         LASSERT(status || wi->swi_state == SWI_STATE_DONE);
1143
1144         spin_lock(&rpc->crpc_lock);
1145
1146         rpc->crpc_closed = 1;
1147         if (!rpc->crpc_status)
1148                 rpc->crpc_status = status;
1149
1150         srpc_del_client_rpc_timer(rpc);
1151
1152         CDEBUG_LIMIT(!status ? D_NET : D_NETERROR,
1153                      "Client RPC done: service %d, peer %s, status %s:%d:%d\n",
1154                      rpc->crpc_service, libcfs_id2str(rpc->crpc_dest),
1155                      swi_state2str(wi->swi_state), rpc->crpc_aborted, status);
1156
1157         /*
1158          * No one can schedule me now since:
1159          * - RPC timer has been defused.
1160          * - all LNet events have been fired.
1161          * - crpc_closed has been set, preventing srpc_abort_rpc from
1162          *   scheduling me.
1163          * Cancel pending schedules and prevent future schedule attempts:
1164          */
1165         LASSERT(!srpc_event_pending(rpc));
1166         swi_exit_workitem(wi);
1167
1168         spin_unlock(&rpc->crpc_lock);
1169
1170         (*rpc->crpc_done)(rpc);
1171 }
1172
1173 /* sends an outgoing RPC */
1174 int
1175 srpc_send_rpc(struct swi_workitem *wi)
1176 {
1177         int rc = 0;
1178         struct srpc_client_rpc *rpc;
1179         struct srpc_msg *reply;
1180         int do_bulk;
1181
1182         LASSERT(wi);
1183
1184         rpc = wi->swi_workitem.wi_data;
1185
1186         LASSERT(rpc);
1187         LASSERT(wi == &rpc->crpc_wi);
1188
1189         reply = &rpc->crpc_replymsg;
1190         do_bulk = rpc->crpc_bulk.bk_niov > 0;
1191
1192         spin_lock(&rpc->crpc_lock);
1193
1194         if (rpc->crpc_aborted) {
1195                 spin_unlock(&rpc->crpc_lock);
1196                 goto abort;
1197         }
1198
1199         spin_unlock(&rpc->crpc_lock);
1200
1201         switch (wi->swi_state) {
1202         default:
1203                 LBUG();
1204         case SWI_STATE_NEWBORN:
1205                 LASSERT(!srpc_event_pending(rpc));
1206
1207                 rc = srpc_prepare_reply(rpc);
1208                 if (rc) {
1209                         srpc_client_rpc_done(rpc, rc);
1210                         return 1;
1211                 }
1212
1213                 rc = srpc_prepare_bulk(rpc);
1214                 if (rc)
1215                         break;
1216
1217                 wi->swi_state = SWI_STATE_REQUEST_SUBMITTED;
1218                 rc = srpc_send_request(rpc);
1219                 break;
1220
1221         case SWI_STATE_REQUEST_SUBMITTED:
1222                 /*
1223                  * CAVEAT EMPTOR: rqtev, rpyev, and bulkev may come in any
1224                  * order; however, they're processed in a strict order:
1225                  * rqt, rpy, and bulk.
1226                  */
1227                 if (!rpc->crpc_reqstev.ev_fired)
1228                         break;
1229
1230                 rc = rpc->crpc_reqstev.ev_status;
1231                 if (rc)
1232                         break;
1233
1234                 wi->swi_state = SWI_STATE_REQUEST_SENT;
1235                 /* perhaps more events, fall thru */
1236         case SWI_STATE_REQUEST_SENT: {
1237                 enum srpc_msg_type type = srpc_service2reply(rpc->crpc_service);
1238
1239                 if (!rpc->crpc_replyev.ev_fired)
1240                         break;
1241
1242                 rc = rpc->crpc_replyev.ev_status;
1243                 if (rc)
1244                         break;
1245
1246                 srpc_unpack_msg_hdr(reply);
1247                 if (reply->msg_type != type ||
1248                     (reply->msg_magic != SRPC_MSG_MAGIC &&
1249                      reply->msg_magic != __swab32(SRPC_MSG_MAGIC))) {
1250                         CWARN("Bad message from %s: type %u (%d expected), magic %u (%d expected).\n",
1251                               libcfs_id2str(rpc->crpc_dest),
1252                               reply->msg_type, type,
1253                               reply->msg_magic, SRPC_MSG_MAGIC);
1254                         rc = -EBADMSG;
1255                         break;
1256                 }
1257
1258                 if (do_bulk && reply->msg_body.reply.status) {
1259                         CWARN("Remote error %d at %s, unlink bulk buffer in case peer didn't initiate bulk transfer\n",
1260                               reply->msg_body.reply.status,
1261                               libcfs_id2str(rpc->crpc_dest));
1262                         LNetMDUnlink(rpc->crpc_bulk.bk_mdh);
1263                 }
1264
1265                 wi->swi_state = SWI_STATE_REPLY_RECEIVED;
1266         }
1267         case SWI_STATE_REPLY_RECEIVED:
1268                 if (do_bulk && !rpc->crpc_bulkev.ev_fired)
1269                         break;
1270
1271                 rc = do_bulk ? rpc->crpc_bulkev.ev_status : 0;
1272
1273                 /*
1274                  * Bulk buffer was unlinked due to remote error. Clear error
1275                  * since reply buffer still contains valid data.
1276                  * NB rpc->crpc_done shouldn't look into bulk data in case of
1277                  * remote error.
1278                  */
1279                 if (do_bulk && rpc->crpc_bulkev.ev_lnet == LNET_EVENT_UNLINK &&
1280                     !rpc->crpc_status && reply->msg_body.reply.status)
1281                         rc = 0;
1282
1283                 wi->swi_state = SWI_STATE_DONE;
1284                 srpc_client_rpc_done(rpc, rc);
1285                 return 1;
1286         }
1287
1288         if (rc) {
1289                 spin_lock(&rpc->crpc_lock);
1290                 srpc_abort_rpc(rpc, rc);
1291                 spin_unlock(&rpc->crpc_lock);
1292         }
1293
1294 abort:
1295         if (rpc->crpc_aborted) {
1296                 LNetMDUnlink(rpc->crpc_reqstmdh);
1297                 LNetMDUnlink(rpc->crpc_replymdh);
1298                 LNetMDUnlink(rpc->crpc_bulk.bk_mdh);
1299
1300                 if (!srpc_event_pending(rpc)) {
1301                         srpc_client_rpc_done(rpc, -EINTR);
1302                         return 1;
1303                 }
1304         }
1305         return 0;
1306 }
1307
1308 struct srpc_client_rpc *
1309 srpc_create_client_rpc(lnet_process_id_t peer, int service,
1310                        int nbulkiov, int bulklen,
1311                        void (*rpc_done)(struct srpc_client_rpc *),
1312                        void (*rpc_fini)(struct srpc_client_rpc *), void *priv)
1313 {
1314         struct srpc_client_rpc *rpc;
1315
1316         LIBCFS_ALLOC(rpc, offsetof(struct srpc_client_rpc,
1317                                    crpc_bulk.bk_iovs[nbulkiov]));
1318         if (!rpc)
1319                 return NULL;
1320
1321         srpc_init_client_rpc(rpc, peer, service, nbulkiov,
1322                              bulklen, rpc_done, rpc_fini, priv);
1323         return rpc;
1324 }
1325
1326 /* called with rpc->crpc_lock held */
1327 void
1328 srpc_abort_rpc(struct srpc_client_rpc *rpc, int why)
1329 {
1330         LASSERT(why);
1331
1332         if (rpc->crpc_aborted ||        /* already aborted */
1333             rpc->crpc_closed)           /* callback imminent */
1334                 return;
1335
1336         CDEBUG(D_NET, "Aborting RPC: service %d, peer %s, state %s, why %d\n",
1337                rpc->crpc_service, libcfs_id2str(rpc->crpc_dest),
1338                swi_state2str(rpc->crpc_wi.swi_state), why);
1339
1340         rpc->crpc_aborted = 1;
1341         rpc->crpc_status = why;
1342         swi_schedule_workitem(&rpc->crpc_wi);
1343 }
1344
1345 /* called with rpc->crpc_lock held */
1346 void
1347 srpc_post_rpc(struct srpc_client_rpc *rpc)
1348 {
1349         LASSERT(!rpc->crpc_aborted);
1350         LASSERT(srpc_data.rpc_state == SRPC_STATE_RUNNING);
1351
1352         CDEBUG(D_NET, "Posting RPC: peer %s, service %d, timeout %d\n",
1353                libcfs_id2str(rpc->crpc_dest), rpc->crpc_service,
1354                rpc->crpc_timeout);
1355
1356         srpc_add_client_rpc_timer(rpc);
1357         swi_schedule_workitem(&rpc->crpc_wi);
1358 }
1359
1360 int
1361 srpc_send_reply(struct srpc_server_rpc *rpc)
1362 {
1363         struct srpc_event *ev = &rpc->srpc_ev;
1364         struct srpc_msg *msg = &rpc->srpc_replymsg;
1365         struct srpc_buffer *buffer = rpc->srpc_reqstbuf;
1366         struct srpc_service_cd *scd = rpc->srpc_scd;
1367         struct srpc_service *sv = scd->scd_svc;
1368         __u64 rpyid;
1369         int rc;
1370
1371         LASSERT(buffer);
1372         rpyid = buffer->buf_msg.msg_body.reqst.rpyid;
1373
1374         spin_lock(&scd->scd_lock);
1375
1376         if (!sv->sv_shuttingdown && !srpc_serv_is_framework(sv)) {
1377                 /*
1378                  * Repost buffer before replying since test client
1379                  * might send me another RPC once it gets the reply
1380                  */
1381                 if (srpc_service_post_buffer(scd, buffer))
1382                         CWARN("Failed to repost %s buffer\n", sv->sv_name);
1383                 rpc->srpc_reqstbuf = NULL;
1384         }
1385
1386         spin_unlock(&scd->scd_lock);
1387
1388         ev->ev_fired = 0;
1389         ev->ev_data = rpc;
1390         ev->ev_type = SRPC_REPLY_SENT;
1391
1392         msg->msg_magic = SRPC_MSG_MAGIC;
1393         msg->msg_version = SRPC_MSG_VERSION;
1394         msg->msg_type = srpc_service2reply(sv->sv_id);
1395
1396         rc = srpc_post_active_rdma(SRPC_RDMA_PORTAL, rpyid, msg,
1397                                    sizeof(*msg), LNET_MD_OP_PUT,
1398                                    rpc->srpc_peer, rpc->srpc_self,
1399                                    &rpc->srpc_replymdh, ev);
1400         if (rc)
1401                 ev->ev_fired = 1; /* no more event expected */
1402         return rc;
1403 }
1404
1405 /* when in kernel always called with LNET_LOCK() held, and in thread context */
1406 static void
1407 srpc_lnet_ev_handler(lnet_event_t *ev)
1408 {
1409         struct srpc_service_cd *scd;
1410         struct srpc_event *rpcev = ev->md.user_ptr;
1411         struct srpc_client_rpc *crpc;
1412         struct srpc_server_rpc *srpc;
1413         struct srpc_buffer *buffer;
1414         struct srpc_service *sv;
1415         struct srpc_msg *msg;
1416         enum srpc_msg_type type;
1417
1418         LASSERT(!in_interrupt());
1419
1420         if (ev->status) {
1421                 __u32 errors;
1422
1423                 spin_lock(&srpc_data.rpc_glock);
1424                 if (ev->status != -ECANCELED) /* cancellation is not error */
1425                         srpc_data.rpc_counters.errors++;
1426                 errors = srpc_data.rpc_counters.errors;
1427                 spin_unlock(&srpc_data.rpc_glock);
1428
1429                 CNETERR("LNet event status %d type %d, RPC errors %u\n",
1430                         ev->status, ev->type, errors);
1431         }
1432
1433         rpcev->ev_lnet = ev->type;
1434
1435         switch (rpcev->ev_type) {
1436         default:
1437                 CERROR("Unknown event: status %d, type %d, lnet %d\n",
1438                        rpcev->ev_status, rpcev->ev_type, rpcev->ev_lnet);
1439                 LBUG();
1440         case SRPC_REQUEST_SENT:
1441                 if (!ev->status && ev->type != LNET_EVENT_UNLINK) {
1442                         spin_lock(&srpc_data.rpc_glock);
1443                         srpc_data.rpc_counters.rpcs_sent++;
1444                         spin_unlock(&srpc_data.rpc_glock);
1445                 }
1446         case SRPC_REPLY_RCVD:
1447         case SRPC_BULK_REQ_RCVD:
1448                 crpc = rpcev->ev_data;
1449
1450                 if (rpcev != &crpc->crpc_reqstev &&
1451                     rpcev != &crpc->crpc_replyev &&
1452                     rpcev != &crpc->crpc_bulkev) {
1453                         CERROR("rpcev %p, crpc %p, reqstev %p, replyev %p, bulkev %p\n",
1454                                rpcev, crpc, &crpc->crpc_reqstev,
1455                                &crpc->crpc_replyev, &crpc->crpc_bulkev);
1456                         CERROR("Bad event: status %d, type %d, lnet %d\n",
1457                                rpcev->ev_status, rpcev->ev_type, rpcev->ev_lnet);
1458                         LBUG();
1459                 }
1460
1461                 spin_lock(&crpc->crpc_lock);
1462
1463                 LASSERT(!rpcev->ev_fired);
1464                 rpcev->ev_fired = 1;
1465                 rpcev->ev_status = (ev->type == LNET_EVENT_UNLINK) ?
1466                                                 -EINTR : ev->status;
1467                 swi_schedule_workitem(&crpc->crpc_wi);
1468
1469                 spin_unlock(&crpc->crpc_lock);
1470                 break;
1471
1472         case SRPC_REQUEST_RCVD:
1473                 scd = rpcev->ev_data;
1474                 sv = scd->scd_svc;
1475
1476                 LASSERT(rpcev == &scd->scd_ev);
1477
1478                 spin_lock(&scd->scd_lock);
1479
1480                 LASSERT(ev->unlinked);
1481                 LASSERT(ev->type == LNET_EVENT_PUT ||
1482                         ev->type == LNET_EVENT_UNLINK);
1483                 LASSERT(ev->type != LNET_EVENT_UNLINK ||
1484                         sv->sv_shuttingdown);
1485
1486                 buffer = container_of(ev->md.start, struct srpc_buffer, buf_msg);
1487                 buffer->buf_peer = ev->initiator;
1488                 buffer->buf_self = ev->target.nid;
1489
1490                 LASSERT(scd->scd_buf_nposted > 0);
1491                 scd->scd_buf_nposted--;
1492
1493                 if (sv->sv_shuttingdown) {
1494                         /*
1495                          * Leave buffer on scd->scd_buf_nposted since
1496                          * srpc_finish_service needs to traverse it.
1497                          */
1498                         spin_unlock(&scd->scd_lock);
1499                         break;
1500                 }
1501
1502                 if (scd->scd_buf_err_stamp &&
1503                     scd->scd_buf_err_stamp < ktime_get_real_seconds()) {
1504                         /* re-enable adding buffer */
1505                         scd->scd_buf_err_stamp = 0;
1506                         scd->scd_buf_err = 0;
1507                 }
1508
1509                 if (!scd->scd_buf_err &&        /* adding buffer is enabled */
1510                     !scd->scd_buf_adjust &&
1511                     scd->scd_buf_nposted < scd->scd_buf_low) {
1512                         scd->scd_buf_adjust = max(scd->scd_buf_total / 2,
1513                                                   SFW_TEST_WI_MIN);
1514                         swi_schedule_workitem(&scd->scd_buf_wi);
1515                 }
1516
1517                 list_del(&buffer->buf_list); /* from scd->scd_buf_posted */
1518                 msg = &buffer->buf_msg;
1519                 type = srpc_service2request(sv->sv_id);
1520
1521                 if (ev->status || ev->mlength != sizeof(*msg) ||
1522                     (msg->msg_type != type &&
1523                      msg->msg_type != __swab32(type)) ||
1524                     (msg->msg_magic != SRPC_MSG_MAGIC &&
1525                      msg->msg_magic != __swab32(SRPC_MSG_MAGIC))) {
1526                         CERROR("Dropping RPC (%s) from %s: status %d mlength %d type %u magic %u.\n",
1527                                sv->sv_name, libcfs_id2str(ev->initiator),
1528                                ev->status, ev->mlength,
1529                                msg->msg_type, msg->msg_magic);
1530
1531                         /*
1532                          * NB can't call srpc_service_recycle_buffer here since
1533                          * it may call LNetM[DE]Attach. The invalid magic tells
1534                          * srpc_handle_rpc to drop this RPC
1535                          */
1536                         msg->msg_magic = 0;
1537                 }
1538
1539                 if (!list_empty(&scd->scd_rpc_free)) {
1540                         srpc = list_entry(scd->scd_rpc_free.next,
1541                                           struct srpc_server_rpc,
1542                                           srpc_list);
1543                         list_del(&srpc->srpc_list);
1544
1545                         srpc_init_server_rpc(srpc, scd, buffer);
1546                         list_add_tail(&srpc->srpc_list,
1547                                       &scd->scd_rpc_active);
1548                         swi_schedule_workitem(&srpc->srpc_wi);
1549                 } else {
1550                         list_add_tail(&buffer->buf_list,
1551                                       &scd->scd_buf_blocked);
1552                 }
1553
1554                 spin_unlock(&scd->scd_lock);
1555
1556                 spin_lock(&srpc_data.rpc_glock);
1557                 srpc_data.rpc_counters.rpcs_rcvd++;
1558                 spin_unlock(&srpc_data.rpc_glock);
1559                 break;
1560
1561         case SRPC_BULK_GET_RPLD:
1562                 LASSERT(ev->type == LNET_EVENT_SEND ||
1563                         ev->type == LNET_EVENT_REPLY ||
1564                         ev->type == LNET_EVENT_UNLINK);
1565
1566                 if (!ev->unlinked)
1567                         break; /* wait for final event */
1568
1569         case SRPC_BULK_PUT_SENT:
1570                 if (!ev->status && ev->type != LNET_EVENT_UNLINK) {
1571                         spin_lock(&srpc_data.rpc_glock);
1572
1573                         if (rpcev->ev_type == SRPC_BULK_GET_RPLD)
1574                                 srpc_data.rpc_counters.bulk_get += ev->mlength;
1575                         else
1576                                 srpc_data.rpc_counters.bulk_put += ev->mlength;
1577
1578                         spin_unlock(&srpc_data.rpc_glock);
1579                 }
1580         case SRPC_REPLY_SENT:
1581                 srpc = rpcev->ev_data;
1582                 scd = srpc->srpc_scd;
1583
1584                 LASSERT(rpcev == &srpc->srpc_ev);
1585
1586                 spin_lock(&scd->scd_lock);
1587
1588                 rpcev->ev_fired = 1;
1589                 rpcev->ev_status = (ev->type == LNET_EVENT_UNLINK) ?
1590                                    -EINTR : ev->status;
1591                 swi_schedule_workitem(&srpc->srpc_wi);
1592
1593                 spin_unlock(&scd->scd_lock);
1594                 break;
1595         }
1596 }
1597
1598 int
1599 srpc_startup(void)
1600 {
1601         int rc;
1602
1603         memset(&srpc_data, 0, sizeof(struct smoketest_rpc));
1604         spin_lock_init(&srpc_data.rpc_glock);
1605
1606         /* 1 second pause to avoid timestamp reuse */
1607         set_current_state(TASK_UNINTERRUPTIBLE);
1608         schedule_timeout(cfs_time_seconds(1));
1609         srpc_data.rpc_matchbits = ((__u64)ktime_get_real_seconds()) << 48;
1610
1611         srpc_data.rpc_state = SRPC_STATE_NONE;
1612
1613         rc = LNetNIInit(LNET_PID_LUSTRE);
1614         if (rc < 0) {
1615                 CERROR("LNetNIInit() has failed: %d\n", rc);
1616                 return rc;
1617         }
1618
1619         srpc_data.rpc_state = SRPC_STATE_NI_INIT;
1620
1621         LNetInvalidateHandle(&srpc_data.rpc_lnet_eq);
1622         rc = LNetEQAlloc(0, srpc_lnet_ev_handler, &srpc_data.rpc_lnet_eq);
1623         if (rc) {
1624                 CERROR("LNetEQAlloc() has failed: %d\n", rc);
1625                 goto bail;
1626         }
1627
1628         rc = LNetSetLazyPortal(SRPC_FRAMEWORK_REQUEST_PORTAL);
1629         LASSERT(!rc);
1630         rc = LNetSetLazyPortal(SRPC_REQUEST_PORTAL);
1631         LASSERT(!rc);
1632
1633         srpc_data.rpc_state = SRPC_STATE_EQ_INIT;
1634
1635         rc = stt_startup();
1636
1637 bail:
1638         if (rc)
1639                 srpc_shutdown();
1640         else
1641                 srpc_data.rpc_state = SRPC_STATE_RUNNING;
1642
1643         return rc;
1644 }
1645
1646 void
1647 srpc_shutdown(void)
1648 {
1649         int i;
1650         int rc;
1651         int state;
1652
1653         state = srpc_data.rpc_state;
1654         srpc_data.rpc_state = SRPC_STATE_STOPPING;
1655
1656         switch (state) {
1657         default:
1658                 LBUG();
1659         case SRPC_STATE_RUNNING:
1660                 spin_lock(&srpc_data.rpc_glock);
1661
1662                 for (i = 0; i <= SRPC_SERVICE_MAX_ID; i++) {
1663                         struct srpc_service *sv = srpc_data.rpc_services[i];
1664
1665                         LASSERTF(!sv, "service not empty: id %d, name %s\n",
1666                                  i, sv->sv_name);
1667                 }
1668
1669                 spin_unlock(&srpc_data.rpc_glock);
1670
1671                 stt_shutdown();
1672
1673         case SRPC_STATE_EQ_INIT:
1674                 rc = LNetClearLazyPortal(SRPC_FRAMEWORK_REQUEST_PORTAL);
1675                 rc = LNetClearLazyPortal(SRPC_REQUEST_PORTAL);
1676                 LASSERT(!rc);
1677                 rc = LNetEQFree(srpc_data.rpc_lnet_eq);
1678                 LASSERT(!rc); /* the EQ should have no user by now */
1679
1680         case SRPC_STATE_NI_INIT:
1681                 LNetNIFini();
1682         }
1683 }