lustre/osc/osc_request.c

   1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
   2  * vim:expandtab:shiftwidth=8:tabstop=8:
   3  *
   4  * GPL HEADER START
   5  *
   6  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2 only,
  10  * as published by the Free Software Foundation.
  11  *
  12  * This program is distributed in the hope that it will be useful, but
  13  * WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * General Public License version 2 for more details (a copy is included
  16  * in the LICENSE file that accompanied this code).
  17  *
  18  * You should have received a copy of the GNU General Public License
  19  * version 2 along with this program; If not, see
  20  * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
  21  *
  22  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
  23  * CA 95054 USA or visit www.sun.com if you need additional information or
  24  * have any questions.
  25  *
  26  * GPL HEADER END
  27  */
  28 /*
  29  * Copyright  2008 Sun Microsystems, Inc. All rights reserved
  30  * Use is subject to license terms.
  31  */
  32 /*
  33  * This file is part of Lustre, http://www.lustre.org/
  34  * Lustre is a trademark of Sun Microsystems, Inc.
  35  */
  36
  37 #ifndef EXPORT_SYMTAB
  38 # define EXPORT_SYMTAB
  39 #endif
  40 #define DEBUG_SUBSYSTEM S_OSC
  41
  42 #include <libcfs/libcfs.h>
  43
  44 #ifndef __KERNEL__
  45 # include <liblustre.h>
  46 #endif
  47
  48 #include <lustre_dlm.h>
  49 #include <lustre_net.h>
  50 #include <lustre/lustre_user.h>
  51 #include <obd_cksum.h>
  52 #include <obd_ost.h>
  53 #include <obd_lov.h>
  54
  55 #ifdef  __CYGWIN__
  56 # include <ctype.h>
  57 #endif
  58
  59 #include <lustre_ha.h>
  60 #include <lprocfs_status.h>
  61 #include <lustre_log.h>
  62 #include <lustre_debug.h>
  63 #include <lustre_param.h>
  64 #include "osc_internal.h"
  65
  66 static quota_interface_t *quota_interface = NULL;
  67 extern quota_interface_t osc_quota_interface;
  68
  69 static void osc_release_ppga(struct brw_page **ppga, obd_count count);
  70 static int brw_interpret(const struct lu_env *env,
  71                          struct ptlrpc_request *req, void *data, int rc);
  72 int osc_cleanup(struct obd_device *obd);
  73
  74 /* Pack OSC object metadata for disk storage (LE byte order). */
  75 static int osc_packmd(struct obd_export *exp, struct lov_mds_md **lmmp,
  76                       struct lov_stripe_md *lsm)
  77 {
  78         int lmm_size;
  79         ENTRY;
  80
  81         lmm_size = sizeof(**lmmp);
  82         if (!lmmp)
  83                 RETURN(lmm_size);
  84
  85         if (*lmmp && !lsm) {
  86                 OBD_FREE(*lmmp, lmm_size);
  87                 *lmmp = NULL;
  88                 RETURN(0);
  89         }
  90
  91         if (!*lmmp) {
  92                 OBD_ALLOC(*lmmp, lmm_size);
  93                 if (!*lmmp)
  94                         RETURN(-ENOMEM);
  95         }
  96
  97         if (lsm) {
  98                 LASSERT(lsm->lsm_object_id);
  99                 LASSERT_MDS_GROUP(lsm->lsm_object_gr);
 100                 (*lmmp)->lmm_object_id = cpu_to_le64(lsm->lsm_object_id);
 101                 (*lmmp)->lmm_object_gr = cpu_to_le64(lsm->lsm_object_gr);
 102         }
 103
 104         RETURN(lmm_size);
 105 }
 106
 107 /* Unpack OSC object metadata from disk storage (LE byte order). */
 108 static int osc_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp,
 109                         struct lov_mds_md *lmm, int lmm_bytes)
 110 {
 111         int lsm_size;
 112         ENTRY;
 113
 114         if (lmm != NULL) {
 115                 if (lmm_bytes < sizeof (*lmm)) {
 116                         CERROR("lov_mds_md too small: %d, need %d\n",
 117                                lmm_bytes, (int)sizeof(*lmm));
 118                         RETURN(-EINVAL);
 119                 }
 120                 /* XXX LOV_MAGIC etc check? */
 121
 122                 if (lmm->lmm_object_id == 0) {
 123                         CERROR("lov_mds_md: zero lmm_object_id\n");
 124                         RETURN(-EINVAL);
 125                 }
 126         }
 127
 128         lsm_size = lov_stripe_md_size(1);
 129         if (lsmp == NULL)
 130                 RETURN(lsm_size);
 131
 132         if (*lsmp != NULL && lmm == NULL) {
 133                 OBD_FREE((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
 134                 OBD_FREE(*lsmp, lsm_size);
 135                 *lsmp = NULL;
 136                 RETURN(0);
 137         }
 138
 139         if (*lsmp == NULL) {
 140                 OBD_ALLOC(*lsmp, lsm_size);
 141                 if (*lsmp == NULL)
 142                         RETURN(-ENOMEM);
 143                 OBD_ALLOC((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
 144                 if ((*lsmp)->lsm_oinfo[0] == NULL) {
 145                         OBD_FREE(*lsmp, lsm_size);
 146                         RETURN(-ENOMEM);
 147                 }
 148                 loi_init((*lsmp)->lsm_oinfo[0]);
 149         }
 150
 151         if (lmm != NULL) {
 152                 /* XXX zero *lsmp? */
 153                 (*lsmp)->lsm_object_id = le64_to_cpu (lmm->lmm_object_id);
 154                 (*lsmp)->lsm_object_gr = le64_to_cpu (lmm->lmm_object_gr);
 155                 LASSERT((*lsmp)->lsm_object_id);
 156                 LASSERT_MDS_GROUP((*lsmp)->lsm_object_gr);
 157         }
 158
 159         (*lsmp)->lsm_maxbytes = LUSTRE_STRIPE_MAXBYTES;
 160
 161         RETURN(lsm_size);
 162 }
 163
 164 static inline void osc_pack_capa(struct ptlrpc_request *req,
 165                                  struct ost_body *body, void *capa)
 166 {
 167         struct obd_capa *oc = (struct obd_capa *)capa;
 168         struct lustre_capa *c;
 169
 170         if (!capa)
 171                 return;
 172
 173         c = req_capsule_client_get(&req->rq_pill, &RMF_CAPA1);
 174         LASSERT(c);
 175         capa_cpy(c, oc);
 176         body->oa.o_valid |= OBD_MD_FLOSSCAPA;
 177         DEBUG_CAPA(D_SEC, c, "pack");
 178 }
 179
 180 static inline void osc_pack_req_body(struct ptlrpc_request *req,
 181                                      struct obd_info *oinfo)
 182 {
 183         struct ost_body *body;
 184
 185         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
 186         LASSERT(body);
 187
 188         body->oa = *oinfo->oi_oa;
 189         osc_pack_capa(req, body, oinfo->oi_capa);
 190 }
 191
 192 static inline void osc_set_capa_size(struct ptlrpc_request *req,
 193                                      const struct req_msg_field *field,
 194                                      struct obd_capa *oc)
 195 {
 196         if (oc == NULL)
 197                 req_capsule_set_size(&req->rq_pill, field, RCL_CLIENT, 0);
 198         else
 199                 /* it is already calculated as sizeof struct obd_capa */
 200                 ;
 201 }
 202
 203 static int osc_getattr_interpret(const struct lu_env *env,
 204                                  struct ptlrpc_request *req,
 205                                  struct osc_async_args *aa, int rc)
 206 {
 207         struct ost_body *body;
 208         ENTRY;
 209
 210         if (rc != 0)
 211                 GOTO(out, rc);
 212
 213         body = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*body),
 214                                   lustre_swab_ost_body);
 215         if (body) {
 216                 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
 217                 memcpy(aa->aa_oi->oi_oa, &body->oa, sizeof(*aa->aa_oi->oi_oa));
 218
 219                 /* This should really be sent by the OST */
 220                 aa->aa_oi->oi_oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
 221                 aa->aa_oi->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
 222         } else {
 223                 CDEBUG(D_INFO, "can't unpack ost_body\n");
 224                 rc = -EPROTO;
 225                 aa->aa_oi->oi_oa->o_valid = 0;
 226         }
 227 out:
 228         rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
 229         RETURN(rc);
 230 }
 231
 232 static int osc_getattr_async(struct obd_export *exp, struct obd_info *oinfo,
 233                              struct ptlrpc_request_set *set)
 234 {
 235         struct ptlrpc_request *req;
 236         struct osc_async_args *aa;
 237         int                    rc;
 238         ENTRY;
 239
 240         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
 241         if (req == NULL)
 242                 RETURN(-ENOMEM);
 243
 244         osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
 245         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
 246         if (rc) {
 247                 ptlrpc_request_free(req);
 248                 RETURN(rc);
 249         }
 250
 251         osc_pack_req_body(req, oinfo);
 252
 253         ptlrpc_request_set_replen(req);
 254         req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_getattr_interpret;
 255
 256         CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
 257         aa = ptlrpc_req_async_args(req);
 258         aa->aa_oi = oinfo;
 259
 260         ptlrpc_set_add_req(set, req);
 261         RETURN(0);
 262 }
 263
 264 static int osc_getattr(struct obd_export *exp, struct obd_info *oinfo)
 265 {
 266         struct ptlrpc_request *req;
 267         struct ost_body       *body;
 268         int                    rc;
 269         ENTRY;
 270
 271         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
 272         if (req == NULL)
 273                 RETURN(-ENOMEM);
 274
 275         osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
 276         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
 277         if (rc) {
 278                 ptlrpc_request_free(req);
 279                 RETURN(rc);
 280         }
 281
 282         osc_pack_req_body(req, oinfo);
 283
 284         ptlrpc_request_set_replen(req);
 285
 286         rc = ptlrpc_queue_wait(req);
 287         if (rc)
 288                 GOTO(out, rc);
 289
 290         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
 291         if (body == NULL)
 292                 GOTO(out, rc = -EPROTO);
 293
 294         CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
 295         *oinfo->oi_oa = body->oa;
 296
 297         /* This should really be sent by the OST */
 298         oinfo->oi_oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
 299         oinfo->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
 300
 301         EXIT;
 302  out:
 303         ptlrpc_req_finished(req);
 304         return rc;
 305 }
 306
 307 static int osc_setattr(struct obd_export *exp, struct obd_info *oinfo,
 308                        struct obd_trans_info *oti)
 309 {
 310         struct ptlrpc_request *req;
 311         struct ost_body       *body;
 312         int                    rc;
 313         ENTRY;
 314
 315         LASSERTF(!(oinfo->oi_oa->o_valid & OBD_MD_FLGROUP) ||
 316                  CHECK_MDS_GROUP(oinfo->oi_oa->o_gr),
 317                  "oinfo->oi_oa->o_valid="LPU64" oinfo->oi_oa->o_gr="LPU64"\n",
 318                  oinfo->oi_oa->o_valid, oinfo->oi_oa->o_gr);
 319
 320         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
 321         if (req == NULL)
 322                 RETURN(-ENOMEM);
 323
 324         osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
 325         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
 326         if (rc) {
 327                 ptlrpc_request_free(req);
 328                 RETURN(rc);
 329         }
 330
 331         osc_pack_req_body(req, oinfo);
 332
 333         ptlrpc_request_set_replen(req);
 334
 335         rc = ptlrpc_queue_wait(req);
 336         if (rc)
 337                 GOTO(out, rc);
 338
 339         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
 340         if (body == NULL)
 341                 GOTO(out, rc = -EPROTO);
 342
 343         *oinfo->oi_oa = body->oa;
 344
 345         EXIT;
 346 out:
 347         ptlrpc_req_finished(req);
 348         RETURN(rc);
 349 }
 350
 351 static int osc_setattr_interpret(const struct lu_env *env,
 352                                  struct ptlrpc_request *req,
 353                                  struct osc_async_args *aa, int rc)
 354 {
 355         struct ost_body *body;
 356         ENTRY;
 357
 358         if (rc != 0)
 359                 GOTO(out, rc);
 360
 361         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
 362         if (body == NULL)
 363                 GOTO(out, rc = -EPROTO);
 364
 365         *aa->aa_oi->oi_oa = body->oa;
 366 out:
 367         rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
 368         RETURN(rc);
 369 }
 370
 371 static int osc_setattr_async(struct obd_export *exp, struct obd_info *oinfo,
 372                              struct obd_trans_info *oti,
 373                              struct ptlrpc_request_set *rqset)
 374 {
 375         struct ptlrpc_request *req;
 376         struct osc_async_args *aa;
 377         int                    rc;
 378         ENTRY;
 379
 380         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
 381         if (req == NULL)
 382                 RETURN(-ENOMEM);
 383
 384         osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
 385         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
 386         if (rc) {
 387                 ptlrpc_request_free(req);
 388                 RETURN(rc);
 389         }
 390
 391         osc_pack_req_body(req, oinfo);
 392
 393         ptlrpc_request_set_replen(req);
 394
 395         if (oinfo->oi_oa->o_valid & OBD_MD_FLCOOKIE) {
 396                 LASSERT(oti);
 397                 oinfo->oi_oa->o_lcookie = *oti->oti_logcookies;
 398         }
 399
 400         /* do mds to ost setattr asynchronously */
 401         if (!rqset) {
 402                 /* Do not wait for response. */
 403                 ptlrpcd_add_req(req, PSCOPE_OTHER);
 404         } else {
 405                 req->rq_interpret_reply =
 406                         (ptlrpc_interpterer_t)osc_setattr_interpret;
 407
 408                 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
 409                 aa = ptlrpc_req_async_args(req);
 410                 aa->aa_oi = oinfo;
 411
 412                 ptlrpc_set_add_req(rqset, req);
 413         }
 414
 415         RETURN(0);
 416 }
 417
 418 int osc_real_create(struct obd_export *exp, struct obdo *oa,
 419                     struct lov_stripe_md **ea, struct obd_trans_info *oti)
 420 {
 421         struct ptlrpc_request *req;
 422         struct ost_body       *body;
 423         struct lov_stripe_md  *lsm;
 424         int                    rc;
 425         ENTRY;
 426
 427         LASSERT(oa);
 428         LASSERT(ea);
 429
 430         lsm = *ea;
 431         if (!lsm) {
 432                 rc = obd_alloc_memmd(exp, &lsm);
 433                 if (rc < 0)
 434                         RETURN(rc);
 435         }
 436
 437         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_CREATE);
 438         if (req == NULL)
 439                 GOTO(out, rc = -ENOMEM);
 440
 441         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_CREATE);
 442         if (rc) {
 443                 ptlrpc_request_free(req);
 444                 GOTO(out, rc);
 445         }
 446
 447         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
 448         LASSERT(body);
 449         body->oa = *oa;
 450
 451         ptlrpc_request_set_replen(req);
 452
 453         if ((oa->o_valid & OBD_MD_FLFLAGS) &&
 454             oa->o_flags == OBD_FL_DELORPHAN) {
 455                 DEBUG_REQ(D_HA, req,
 456                           "delorphan from OST integration");
 457                 /* Don't resend the delorphan req */
 458                 req->rq_no_resend = req->rq_no_delay = 1;
 459         }
 460
 461         rc = ptlrpc_queue_wait(req);
 462         if (rc)
 463                 GOTO(out_req, rc);
 464
 465         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
 466         if (body == NULL)
 467                 GOTO(out_req, rc = -EPROTO);
 468
 469         *oa = body->oa;
 470
 471         /* This should really be sent by the OST */
 472         oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
 473         oa->o_valid |= OBD_MD_FLBLKSZ;
 474
 475         /* XXX LOV STACKING: the lsm that is passed to us from LOV does not
 476          * have valid lsm_oinfo data structs, so don't go touching that.
 477          * This needs to be fixed in a big way.
 478          */
 479         lsm->lsm_object_id = oa->o_id;
 480         lsm->lsm_object_gr = oa->o_gr;
 481         *ea = lsm;
 482
 483         if (oti != NULL) {
 484                 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
 485
 486                 if (oa->o_valid & OBD_MD_FLCOOKIE) {
 487                         if (!oti->oti_logcookies)
 488                                 oti_alloc_cookies(oti, 1);
 489                         *oti->oti_logcookies = oa->o_lcookie;
 490                 }
 491         }
 492
 493         CDEBUG(D_HA, "transno: "LPD64"\n",
 494                lustre_msg_get_transno(req->rq_repmsg));
 495 out_req:
 496         ptlrpc_req_finished(req);
 497 out:
 498         if (rc && !*ea)
 499                 obd_free_memmd(exp, &lsm);
 500         RETURN(rc);
 501 }
 502
 503 static int osc_punch_interpret(const struct lu_env *env,
 504                                struct ptlrpc_request *req,
 505                                struct osc_punch_args *aa, int rc)
 506 {
 507         struct ost_body *body;
 508         ENTRY;
 509
 510         if (rc != 0)
 511                 GOTO(out, rc);
 512
 513         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
 514         if (body == NULL)
 515                 GOTO(out, rc = -EPROTO);
 516
 517         *aa->pa_oa = body->oa;
 518 out:
 519         rc = aa->pa_upcall(aa->pa_cookie, rc);
 520         RETURN(rc);
 521 }
 522
 523 int osc_punch_base(struct obd_export *exp, struct obdo *oa,
 524                    struct obd_capa *capa,
 525                    obd_enqueue_update_f upcall, void *cookie,
 526                    struct ptlrpc_request_set *rqset)
 527 {
 528         struct ptlrpc_request *req;
 529         struct osc_punch_args *aa;
 530         struct ost_body       *body;
 531         int                    rc;
 532         ENTRY;
 533
 534         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_PUNCH);
 535         if (req == NULL)
 536                 RETURN(-ENOMEM);
 537
 538         osc_set_capa_size(req, &RMF_CAPA1, capa);
 539         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_PUNCH);
 540         if (rc) {
 541                 ptlrpc_request_free(req);
 542                 RETURN(rc);
 543         }
 544         req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
 545         ptlrpc_at_set_req_timeout(req);
 546
 547         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
 548         LASSERT(body);
 549         body->oa = *oa;
 550         osc_pack_capa(req, body, capa);
 551
 552         ptlrpc_request_set_replen(req);
 553
 554
 555         req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_punch_interpret;
 556         CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
 557         aa = ptlrpc_req_async_args(req);
 558         aa->pa_oa     = oa;
 559         aa->pa_upcall = upcall;
 560         aa->pa_cookie = cookie;
 561         if (rqset == PTLRPCD_SET)
 562                 ptlrpcd_add_req(req, PSCOPE_OTHER);
 563         else
 564                 ptlrpc_set_add_req(rqset, req);
 565
 566         RETURN(0);
 567 }
 568
 569 static int osc_punch(struct obd_export *exp, struct obd_info *oinfo,
 570                      struct obd_trans_info *oti,
 571                      struct ptlrpc_request_set *rqset)
 572 {
 573         oinfo->oi_oa->o_size   = oinfo->oi_policy.l_extent.start;
 574         oinfo->oi_oa->o_blocks = oinfo->oi_policy.l_extent.end;
 575         oinfo->oi_oa->o_valid |= OBD_MD_FLSIZE | OBD_MD_FLBLOCKS;
 576         return osc_punch_base(exp, oinfo->oi_oa, oinfo->oi_capa,
 577                               oinfo->oi_cb_up, oinfo, rqset);
 578 }
 579
 580 static int osc_sync(struct obd_export *exp, struct obdo *oa,
 581                     struct lov_stripe_md *md, obd_size start, obd_size end,
 582                     void *capa)
 583 {
 584         struct ptlrpc_request *req;
 585         struct ost_body       *body;
 586         int                    rc;
 587         ENTRY;
 588
 589         if (!oa) {
 590                 CDEBUG(D_INFO, "oa NULL\n");
 591                 RETURN(-EINVAL);
 592         }
 593
 594         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SYNC);
 595         if (req == NULL)
 596                 RETURN(-ENOMEM);
 597
 598         osc_set_capa_size(req, &RMF_CAPA1, capa);
 599         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SYNC);
 600         if (rc) {
 601                 ptlrpc_request_free(req);
 602                 RETURN(rc);
 603         }
 604
 605         /* overload the size and blocks fields in the oa with start/end */
 606         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
 607         LASSERT(body);
 608         body->oa = *oa;
 609         body->oa.o_size = start;
 610         body->oa.o_blocks = end;
 611         body->oa.o_valid |= (OBD_MD_FLSIZE | OBD_MD_FLBLOCKS);
 612         osc_pack_capa(req, body, capa);
 613
 614         ptlrpc_request_set_replen(req);
 615
 616         rc = ptlrpc_queue_wait(req);
 617         if (rc)
 618                 GOTO(out, rc);
 619
 620         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
 621         if (body == NULL)
 622                 GOTO(out, rc = -EPROTO);
 623
 624         *oa = body->oa;
 625
 626         EXIT;
 627  out:
 628         ptlrpc_req_finished(req);
 629         return rc;
 630 }
 631
 632 /* Find and cancel locally locks matched by @mode in the resource found by
 633  * @objid. Found locks are added into @cancel list. Returns the amount of
 634  * locks added to @cancels list. */
 635 static int osc_resource_get_unused(struct obd_export *exp, struct obdo *oa,
 636                                    struct list_head *cancels, ldlm_mode_t mode,
 637                                    int lock_flags)
 638 {
 639         struct ldlm_namespace *ns = exp->exp_obd->obd_namespace;
 640         struct ldlm_res_id res_id;
 641         struct ldlm_resource *res;
 642         int count;
 643         ENTRY;
 644
 645         osc_build_res_name(oa->o_id, oa->o_gr, &res_id);
 646         res = ldlm_resource_get(ns, NULL, &res_id, 0, 0);
 647         if (res == NULL)
 648                 RETURN(0);
 649
 650         LDLM_RESOURCE_ADDREF(res);
 651         count = ldlm_cancel_resource_local(res, cancels, NULL, mode,
 652                                            lock_flags, 0, NULL);
 653         LDLM_RESOURCE_DELREF(res);
 654         ldlm_resource_putref(res);
 655         RETURN(count);
 656 }
 657
 658 static int osc_destroy_interpret(const struct lu_env *env,
 659                                  struct ptlrpc_request *req, void *data,
 660                                  int rc)
 661 {
 662         struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
 663
 664         atomic_dec(&cli->cl_destroy_in_flight);
 665         cfs_waitq_signal(&cli->cl_destroy_waitq);
 666         return 0;
 667 }
 668
 669 static int osc_can_send_destroy(struct client_obd *cli)
 670 {
 671         if (atomic_inc_return(&cli->cl_destroy_in_flight) <=
 672             cli->cl_max_rpcs_in_flight) {
 673                 /* The destroy request can be sent */
 674                 return 1;
 675         }
 676         if (atomic_dec_return(&cli->cl_destroy_in_flight) <
 677             cli->cl_max_rpcs_in_flight) {
 678                 /*
 679                  * The counter has been modified between the two atomic
 680                  * operations.
 681                  */
 682                 cfs_waitq_signal(&cli->cl_destroy_waitq);
 683         }
 684         return 0;
 685 }
 686
 687 /* Destroy requests can be async always on the client, and we don't even really
 688  * care about the return code since the client cannot do anything at all about
 689  * a destroy failure.
 690  * When the MDS is unlinking a filename, it saves the file objects into a
 691  * recovery llog, and these object records are cancelled when the OST reports
 692  * they were destroyed and sync'd to disk (i.e. transaction committed).
 693  * If the client dies, or the OST is down when the object should be destroyed,
 694  * the records are not cancelled, and when the OST reconnects to the MDS next,
 695  * it will retrieve the llog unlink logs and then sends the log cancellation
 696  * cookies to the MDS after committing destroy transactions. */
 697 static int osc_destroy(struct obd_export *exp, struct obdo *oa,
 698                        struct lov_stripe_md *ea, struct obd_trans_info *oti,
 699                        struct obd_export *md_export, void *capa)
 700 {
 701         struct client_obd     *cli = &exp->exp_obd->u.cli;
 702         struct ptlrpc_request *req;
 703         struct ost_body       *body;
 704         CFS_LIST_HEAD(cancels);
 705         int rc, count;
 706         ENTRY;
 707
 708         if (!oa) {
 709                 CDEBUG(D_INFO, "oa NULL\n");
 710                 RETURN(-EINVAL);
 711         }
 712
 713         count = osc_resource_get_unused(exp, oa, &cancels, LCK_PW,
 714                                         LDLM_FL_DISCARD_DATA);
 715
 716         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_DESTROY);
 717         if (req == NULL) {
 718                 ldlm_lock_list_put(&cancels, l_bl_ast, count);
 719                 RETURN(-ENOMEM);
 720         }
 721
 722         osc_set_capa_size(req, &RMF_CAPA1, (struct obd_capa *)capa);
 723         rc = ldlm_prep_elc_req(exp, req, LUSTRE_OST_VERSION, OST_DESTROY,
 724                                0, &cancels, count);
 725         if (rc) {
 726                 ptlrpc_request_free(req);
 727                 RETURN(rc);
 728         }
 729
 730         req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
 731         ptlrpc_at_set_req_timeout(req);
 732
 733         if (oti != NULL && oa->o_valid & OBD_MD_FLCOOKIE)
 734                 oa->o_lcookie = *oti->oti_logcookies;
 735         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
 736         LASSERT(body);
 737         body->oa = *oa;
 738
 739         osc_pack_capa(req, body, (struct obd_capa *)capa);
 740         ptlrpc_request_set_replen(req);
 741
 742         /* don't throttle destroy RPCs for the MDT */
 743         if (!(cli->cl_import->imp_connect_flags_orig & OBD_CONNECT_MDS)) {
 744                 req->rq_interpret_reply = osc_destroy_interpret;
 745                 if (!osc_can_send_destroy(cli)) {
 746                         struct l_wait_info lwi = { 0 };
 747
 748                         /*
 749                          * Wait until the number of on-going destroy RPCs drops
 750                          * under max_rpc_in_flight
 751                          */
 752                         l_wait_event_exclusive(cli->cl_destroy_waitq,
 753                                                osc_can_send_destroy(cli), &lwi);
 754                 }
 755         }
 756
 757         /* Do not wait for response */
 758         ptlrpcd_add_req(req, PSCOPE_OTHER);
 759         RETURN(0);
 760 }
 761
 762 static void osc_announce_cached(struct client_obd *cli, struct obdo *oa,
 763                                 long writing_bytes)
 764 {
 765         obd_flag bits = OBD_MD_FLBLOCKS|OBD_MD_FLGRANT;
 766
 767         LASSERT(!(oa->o_valid & bits));
 768
 769         oa->o_valid |= bits;
 770         client_obd_list_lock(&cli->cl_loi_list_lock);
 771         oa->o_dirty = cli->cl_dirty;
 772         if (cli->cl_dirty - cli->cl_dirty_transit > cli->cl_dirty_max) {
 773                 CERROR("dirty %lu - %lu > dirty_max %lu\n",
 774                        cli->cl_dirty, cli->cl_dirty_transit, cli->cl_dirty_max);
 775                 oa->o_undirty = 0;
 776         } else if (atomic_read(&obd_dirty_pages) -
 777                    atomic_read(&obd_dirty_transit_pages) > obd_max_dirty_pages){
 778                 CERROR("dirty %d - %d > system dirty_max %d\n",
 779                        atomic_read(&obd_dirty_pages),
 780                        atomic_read(&obd_dirty_transit_pages),
 781                        obd_max_dirty_pages);
 782                 oa->o_undirty = 0;
 783         } else if (cli->cl_dirty_max - cli->cl_dirty > 0x7fffffff) {
 784                 CERROR("dirty %lu - dirty_max %lu too big???\n",
 785                        cli->cl_dirty, cli->cl_dirty_max);
 786                 oa->o_undirty = 0;
 787         } else {
 788                 long max_in_flight = (cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT)*
 789                                 (cli->cl_max_rpcs_in_flight + 1);
 790                 oa->o_undirty = max(cli->cl_dirty_max, max_in_flight);
 791         }
 792         oa->o_grant = cli->cl_avail_grant;
 793         oa->o_dropped = cli->cl_lost_grant;
 794         cli->cl_lost_grant = 0;
 795         client_obd_list_unlock(&cli->cl_loi_list_lock);
 796         CDEBUG(D_CACHE,"dirty: "LPU64" undirty: %u dropped %u grant: "LPU64"\n",
 797                oa->o_dirty, oa->o_undirty, oa->o_dropped, oa->o_grant);
 798
 799 }
 800
 801 static void osc_update_next_shrink(struct client_obd *cli)
 802 {
 803         int time = GRANT_SHRINK_INTERVAL;
 804         cli->cl_next_shrink_grant = cfs_time_shift(time);
 805         CDEBUG(D_CACHE, "next time %ld to shrink grant \n",
 806                cli->cl_next_shrink_grant);
 807 }
 808
 809 /* caller must hold loi_list_lock */
 810 static void osc_consume_write_grant(struct client_obd *cli,
 811                                     struct brw_page *pga)
 812 {
 813         LASSERT_SPIN_LOCKED(&cli->cl_loi_list_lock);
 814         LASSERT(!(pga->flag & OBD_BRW_FROM_GRANT));
 815         atomic_inc(&obd_dirty_pages);
 816         cli->cl_dirty += CFS_PAGE_SIZE;
 817         cli->cl_avail_grant -= CFS_PAGE_SIZE;
 818         pga->flag |= OBD_BRW_FROM_GRANT;
 819         CDEBUG(D_CACHE, "using %lu grant credits for brw %p page %p\n",
 820                CFS_PAGE_SIZE, pga, pga->pg);
 821         LASSERT(cli->cl_avail_grant >= 0);
 822         osc_update_next_shrink(cli);
 823 }
 824
 825 /* the companion to osc_consume_write_grant, called when a brw has completed.
 826  * must be called with the loi lock held. */
 827 static void osc_release_write_grant(struct client_obd *cli,
 828                                     struct brw_page *pga, int sent)
 829 {
 830         int blocksize = cli->cl_import->imp_obd->obd_osfs.os_bsize ? : 4096;
 831         ENTRY;
 832
 833         LASSERT_SPIN_LOCKED(&cli->cl_loi_list_lock);
 834         if (!(pga->flag & OBD_BRW_FROM_GRANT)) {
 835                 EXIT;
 836                 return;
 837         }
 838
 839         pga->flag &= ~OBD_BRW_FROM_GRANT;
 840         atomic_dec(&obd_dirty_pages);
 841         cli->cl_dirty -= CFS_PAGE_SIZE;
 842         if (pga->flag & OBD_BRW_NOCACHE) {
 843                 pga->flag &= ~OBD_BRW_NOCACHE;
 844                 atomic_dec(&obd_dirty_transit_pages);
 845                 cli->cl_dirty_transit -= CFS_PAGE_SIZE;
 846         }
 847         if (!sent) {
 848                 cli->cl_lost_grant += CFS_PAGE_SIZE;
 849                 CDEBUG(D_CACHE, "lost grant: %lu avail grant: %lu dirty: %lu\n",
 850                        cli->cl_lost_grant, cli->cl_avail_grant, cli->cl_dirty);
 851         } else if (CFS_PAGE_SIZE != blocksize && pga->count != CFS_PAGE_SIZE) {
 852                 /* For short writes we shouldn't count parts of pages that
 853                  * span a whole block on the OST side, or our accounting goes
 854                  * wrong.  Should match the code in filter_grant_check. */
 855                 int offset = pga->off & ~CFS_PAGE_MASK;
 856                 int count = pga->count + (offset & (blocksize - 1));
 857                 int end = (offset + pga->count) & (blocksize - 1);
 858                 if (end)
 859                         count += blocksize - end;
 860
 861                 cli->cl_lost_grant += CFS_PAGE_SIZE - count;
 862                 CDEBUG(D_CACHE, "lost %lu grant: %lu avail: %lu dirty: %lu\n",
 863                        CFS_PAGE_SIZE - count, cli->cl_lost_grant,
 864                        cli->cl_avail_grant, cli->cl_dirty);
 865         }
 866
 867         EXIT;
 868 }
 869
 870 static unsigned long rpcs_in_flight(struct client_obd *cli)
 871 {
 872         return cli->cl_r_in_flight + cli->cl_w_in_flight;
 873 }
 874
 875 /* caller must hold loi_list_lock */
 876 void osc_wake_cache_waiters(struct client_obd *cli)
 877 {
 878         struct list_head *l, *tmp;
 879         struct osc_cache_waiter *ocw;
 880
 881         ENTRY;
 882         list_for_each_safe(l, tmp, &cli->cl_cache_waiters) {
 883                 /* if we can't dirty more, we must wait until some is written */
 884                 if ((cli->cl_dirty + CFS_PAGE_SIZE > cli->cl_dirty_max) ||
 885                    (atomic_read(&obd_dirty_pages) + 1 > obd_max_dirty_pages)) {
 886                         CDEBUG(D_CACHE, "no dirty room: dirty: %ld "
 887                                "osc max %ld, sys max %d\n", cli->cl_dirty,
 888                                cli->cl_dirty_max, obd_max_dirty_pages);
 889                         return;
 890                 }
 891
 892                 /* if still dirty cache but no grant wait for pending RPCs that
 893                  * may yet return us some grant before doing sync writes */
 894                 if (cli->cl_w_in_flight && cli->cl_avail_grant < CFS_PAGE_SIZE) {
 895                         CDEBUG(D_CACHE, "%u BRW writes in flight, no grant\n",
 896                                cli->cl_w_in_flight);
 897                         return;
 898                 }
 899
 900                 ocw = list_entry(l, struct osc_cache_waiter, ocw_entry);
 901                 list_del_init(&ocw->ocw_entry);
 902                 if (cli->cl_avail_grant < CFS_PAGE_SIZE) {
 903                         /* no more RPCs in flight to return grant, do sync IO */
 904                         ocw->ocw_rc = -EDQUOT;
 905                         CDEBUG(D_INODE, "wake oap %p for sync\n", ocw->ocw_oap);
 906                 } else {
 907                         osc_consume_write_grant(cli,
 908                                                 &ocw->ocw_oap->oap_brw_page);
 909                 }
 910
 911                 cfs_waitq_signal(&ocw->ocw_waitq);
 912         }
 913
 914         EXIT;
 915 }
 916
 917 static void __osc_update_grant(struct client_obd *cli, obd_size grant)
 918 {
 919         client_obd_list_lock(&cli->cl_loi_list_lock);
 920         cli->cl_avail_grant += grant;
 921         client_obd_list_unlock(&cli->cl_loi_list_lock);
 922 }
 923
 924 static void osc_update_grant(struct client_obd *cli, struct ost_body *body)
 925 {
 926         if (body->oa.o_valid & OBD_MD_FLGRANT) {
 927                 CDEBUG(D_CACHE, "got "LPU64" extra grant\n", body->oa.o_grant);
 928                 __osc_update_grant(cli, body->oa.o_grant);
 929         }
 930 }
 931
 932 static int osc_set_info_async(struct obd_export *exp, obd_count keylen,
 933                               void *key, obd_count vallen, void *val,
 934                               struct ptlrpc_request_set *set);
 935
 936 static int osc_shrink_grant_interpret(const struct lu_env *env,
 937                                       struct ptlrpc_request *req,
 938                                       void *aa, int rc)
 939 {
 940         struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
 941         struct obdo *oa = ((struct osc_grant_args *)aa)->aa_oa;
 942         struct ost_body *body;
 943
 944         if (rc != 0) {
 945                 __osc_update_grant(cli, oa->o_grant);
 946                 GOTO(out, rc);
 947         }
 948
 949         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
 950         LASSERT(body);
 951         osc_update_grant(cli, body);
 952 out:
 953         OBD_FREE_PTR(oa);
 954         return rc;
 955 }
 956
 957 static void osc_shrink_grant_local(struct client_obd *cli, struct obdo *oa)
 958 {
 959         client_obd_list_lock(&cli->cl_loi_list_lock);
 960         oa->o_grant = cli->cl_avail_grant / 4;
 961         cli->cl_avail_grant -= oa->o_grant;
 962         client_obd_list_unlock(&cli->cl_loi_list_lock);
 963         oa->o_flags |= OBD_FL_SHRINK_GRANT;
 964         osc_update_next_shrink(cli);
 965 }
 966
 967 static int osc_shrink_grant(struct client_obd *cli)
 968 {
 969         int    rc = 0;
 970         struct ost_body     *body;
 971         ENTRY;
 972
 973         OBD_ALLOC_PTR(body);
 974         if (!body)
 975                 RETURN(-ENOMEM);
 976
 977         osc_announce_cached(cli, &body->oa, 0);
 978         osc_shrink_grant_local(cli, &body->oa);
 979         rc = osc_set_info_async(cli->cl_import->imp_obd->obd_self_export,
 980                                 sizeof(KEY_GRANT_SHRINK), KEY_GRANT_SHRINK,
 981                                 sizeof(*body), body, NULL);
 982         if (rc != 0)
 983                 __osc_update_grant(cli, body->oa.o_grant);
 984         if (body)
 985                OBD_FREE_PTR(body);
 986         RETURN(rc);
 987 }
 988
 989 #define GRANT_SHRINK_LIMIT PTLRPC_MAX_BRW_SIZE
 990 static int osc_should_shrink_grant(struct client_obd *client)
 991 {
 992         cfs_time_t time = cfs_time_current();
 993         cfs_time_t next_shrink = client->cl_next_shrink_grant;
 994         if (cfs_time_aftereq(time, next_shrink - 5 * CFS_TICK)) {
 995                 if (client->cl_import->imp_state == LUSTRE_IMP_FULL &&
 996                     client->cl_avail_grant > GRANT_SHRINK_LIMIT)
 997                         return 1;
 998                 else
 999                         osc_update_next_shrink(client);
1000         }
1001         return 0;
1002 }
1003
1004 static int osc_grant_shrink_grant_cb(struct timeout_item *item, void *data)
1005 {
1006         struct client_obd *client;
1007
1008         list_for_each_entry(client, &item->ti_obd_list, cl_grant_shrink_list) {
1009                 if (osc_should_shrink_grant(client))
1010                         osc_shrink_grant(client);
1011         }
1012         return 0;
1013 }
1014
1015 static int osc_add_shrink_grant(struct client_obd *client)
1016 {
1017         int rc;
1018
1019         rc = ptlrpc_add_timeout_client(GRANT_SHRINK_INTERVAL,
1020                                          TIMEOUT_GRANT,
1021                                          osc_grant_shrink_grant_cb, NULL,
1022                                          &client->cl_grant_shrink_list);
1023         if (rc) {
1024                 CERROR("add grant client %s error %d\n",
1025                         client->cl_import->imp_obd->obd_name, rc);
1026                 return rc;
1027         }
1028         CDEBUG(D_CACHE, "add grant client %s \n",
1029                client->cl_import->imp_obd->obd_name);
1030         osc_update_next_shrink(client);
1031         return 0;
1032 }
1033
1034 static int osc_del_shrink_grant(struct client_obd *client)
1035 {
1036         return ptlrpc_del_timeout_client(&client->cl_grant_shrink_list,
1037                                          TIMEOUT_GRANT);
1038 }
1039
1040 static void osc_init_grant(struct client_obd *cli, struct obd_connect_data *ocd)
1041 {
1042         client_obd_list_lock(&cli->cl_loi_list_lock);
1043         cli->cl_avail_grant = ocd->ocd_grant;
1044         client_obd_list_unlock(&cli->cl_loi_list_lock);
1045
1046         if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT_SHRINK &&
1047             list_empty(&cli->cl_grant_shrink_list))
1048                 osc_add_shrink_grant(cli);
1049
1050         CDEBUG(D_CACHE, "setting cl_avail_grant: %ld cl_lost_grant: %ld \n",
1051                cli->cl_avail_grant, cli->cl_lost_grant);
1052         LASSERT(cli->cl_avail_grant >= 0);
1053 }
1054
1055 /* We assume that the reason this OSC got a short read is because it read
1056  * beyond the end of a stripe file; i.e. lustre is reading a sparse file
1057  * via the LOV, and it _knows_ it's reading inside the file, it's just that
1058  * this stripe never got written at or beyond this stripe offset yet. */
1059 static void handle_short_read(int nob_read, obd_count page_count,
1060                               struct brw_page **pga)
1061 {
1062         char *ptr;
1063         int i = 0;
1064
1065         /* skip bytes read OK */
1066         while (nob_read > 0) {
1067                 LASSERT (page_count > 0);
1068
1069                 if (pga[i]->count > nob_read) {
1070                         /* EOF inside this page */
1071                         ptr = cfs_kmap(pga[i]->pg) +
1072                                 (pga[i]->off & ~CFS_PAGE_MASK);
1073                         memset(ptr + nob_read, 0, pga[i]->count - nob_read);
1074                         cfs_kunmap(pga[i]->pg);
1075                         page_count--;
1076                         i++;
1077                         break;
1078                 }
1079
1080                 nob_read -= pga[i]->count;
1081                 page_count--;
1082                 i++;
1083         }
1084
1085         /* zero remaining pages */
1086         while (page_count-- > 0) {
1087                 ptr = cfs_kmap(pga[i]->pg) + (pga[i]->off & ~CFS_PAGE_MASK);
1088                 memset(ptr, 0, pga[i]->count);
1089                 cfs_kunmap(pga[i]->pg);
1090                 i++;
1091         }
1092 }
1093
1094 static int check_write_rcs(struct ptlrpc_request *req,
1095                            int requested_nob, int niocount,
1096                            obd_count page_count, struct brw_page **pga)
1097 {
1098         int    *remote_rcs, i;
1099
1100         /* return error if any niobuf was in error */
1101         remote_rcs = lustre_swab_repbuf(req, REQ_REC_OFF + 1,
1102                                         sizeof(*remote_rcs) * niocount, NULL);
1103         if (remote_rcs == NULL) {
1104                 CDEBUG(D_INFO, "Missing/short RC vector on BRW_WRITE reply\n");
1105                 return(-EPROTO);
1106         }
1107         if (lustre_msg_swabbed(req->rq_repmsg))
1108                 for (i = 0; i < niocount; i++)
1109                         __swab32s(&remote_rcs[i]);
1110
1111         for (i = 0; i < niocount; i++) {
1112                 if (remote_rcs[i] < 0)
1113                         return(remote_rcs[i]);
1114
1115                 if (remote_rcs[i] != 0) {
1116                         CDEBUG(D_INFO, "rc[%d] invalid (%d) req %p\n",
1117                                 i, remote_rcs[i], req);
1118                         return(-EPROTO);
1119                 }
1120         }
1121
1122         if (req->rq_bulk->bd_nob_transferred != requested_nob) {
1123                 CERROR("Unexpected # bytes transferred: %d (requested %d)\n",
1124                        req->rq_bulk->bd_nob_transferred, requested_nob);
1125                 return(-EPROTO);
1126         }
1127
1128         return (0);
1129 }
1130
1131 static inline int can_merge_pages(struct brw_page *p1, struct brw_page *p2)
1132 {
1133         if (p1->flag != p2->flag) {
1134                 unsigned mask = ~(OBD_BRW_FROM_GRANT|
1135                                   OBD_BRW_NOCACHE|OBD_BRW_SYNC);
1136
1137                 /* warn if we try to combine flags that we don't know to be
1138                  * safe to combine */
1139                 if ((p1->flag & mask) != (p2->flag & mask))
1140                         CERROR("is it ok to have flags 0x%x and 0x%x in the "
1141                                "same brw?\n", p1->flag, p2->flag);
1142                 return 0;
1143         }
1144
1145         return (p1->off + p1->count == p2->off);
1146 }
1147
1148 static obd_count osc_checksum_bulk(int nob, obd_count pg_count,
1149                                    struct brw_page **pga, int opc,
1150                                    cksum_type_t cksum_type)
1151 {
1152         __u32 cksum;
1153         int i = 0;
1154
1155         LASSERT (pg_count > 0);
1156         cksum = init_checksum(cksum_type);
1157         while (nob > 0 && pg_count > 0) {
1158                 unsigned char *ptr = cfs_kmap(pga[i]->pg);
1159                 int off = pga[i]->off & ~CFS_PAGE_MASK;
1160                 int count = pga[i]->count > nob ? nob : pga[i]->count;
1161
1162                 /* corrupt the data before we compute the checksum, to
1163                  * simulate an OST->client data error */
1164                 if (i == 0 && opc == OST_READ &&
1165                     OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_RECEIVE))
1166                         memcpy(ptr + off, "bad1", min(4, nob));
1167                 cksum = compute_checksum(cksum, ptr + off, count, cksum_type);
1168                 cfs_kunmap(pga[i]->pg);
1169                 LL_CDEBUG_PAGE(D_PAGE, pga[i]->pg, "off %d checksum %x\n",
1170                                off, cksum);
1171
1172                 nob -= pga[i]->count;
1173                 pg_count--;
1174                 i++;
1175         }
1176         /* For sending we only compute the wrong checksum instead
1177          * of corrupting the data so it is still correct on a redo */
1178         if (opc == OST_WRITE && OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_SEND))
1179                 cksum++;
1180
1181         return cksum;
1182 }
1183
1184 static int osc_brw_prep_request(int cmd, struct client_obd *cli,struct obdo *oa,
1185                                 struct lov_stripe_md *lsm, obd_count page_count,
1186                                 struct brw_page **pga,
1187                                 struct ptlrpc_request **reqp,
1188                                 struct obd_capa *ocapa, int reserve)
1189 {
1190         struct ptlrpc_request   *req;
1191         struct ptlrpc_bulk_desc *desc;
1192         struct ost_body         *body;
1193         struct obd_ioobj        *ioobj;
1194         struct niobuf_remote    *niobuf;
1195         int niocount, i, requested_nob, opc, rc;
1196         struct osc_brw_async_args *aa;
1197         struct req_capsule      *pill;
1198         struct brw_page *pg_prev;
1199
1200         ENTRY;
1201         if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ))
1202                 RETURN(-ENOMEM); /* Recoverable */
1203         if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ2))
1204                 RETURN(-EINVAL); /* Fatal */
1205
1206         if ((cmd & OBD_BRW_WRITE) != 0) {
1207                 opc = OST_WRITE;
1208                 req = ptlrpc_request_alloc_pool(cli->cl_import,
1209                                                 cli->cl_import->imp_rq_pool,
1210                                                 &RQF_OST_BRW);
1211         } else {
1212                 opc = OST_READ;
1213                 req = ptlrpc_request_alloc(cli->cl_import, &RQF_OST_BRW);
1214         }
1215         if (req == NULL)
1216                 RETURN(-ENOMEM);
1217
1218         for (niocount = i = 1; i < page_count; i++) {
1219                 if (!can_merge_pages(pga[i - 1], pga[i]))
1220                         niocount++;
1221         }
1222
1223         pill = &req->rq_pill;
1224         req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_CLIENT,
1225                              niocount * sizeof(*niobuf));
1226         osc_set_capa_size(req, &RMF_CAPA1, ocapa);
1227
1228         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, opc);
1229         if (rc) {
1230                 ptlrpc_request_free(req);
1231                 RETURN(rc);
1232         }
1233         req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
1234         ptlrpc_at_set_req_timeout(req);
1235
1236         if (opc == OST_WRITE)
1237                 desc = ptlrpc_prep_bulk_imp(req, page_count,
1238                                             BULK_GET_SOURCE, OST_BULK_PORTAL);
1239         else
1240                 desc = ptlrpc_prep_bulk_imp(req, page_count,
1241                                             BULK_PUT_SINK, OST_BULK_PORTAL);
1242
1243         if (desc == NULL)
1244                 GOTO(out, rc = -ENOMEM);
1245         /* NB request now owns desc and will free it when it gets freed */
1246
1247         body = req_capsule_client_get(pill, &RMF_OST_BODY);
1248         ioobj = req_capsule_client_get(pill, &RMF_OBD_IOOBJ);
1249         niobuf = req_capsule_client_get(pill, &RMF_NIOBUF_REMOTE);
1250         LASSERT(body && ioobj && niobuf);
1251
1252         body->oa = *oa;
1253
1254         obdo_to_ioobj(oa, ioobj);
1255         ioobj->ioo_bufcnt = niocount;
1256         osc_pack_capa(req, body, ocapa);
1257         LASSERT (page_count > 0);
1258         pg_prev = pga[0];
1259         for (requested_nob = i = 0; i < page_count; i++, niobuf++) {
1260                 struct brw_page *pg = pga[i];
1261
1262                 LASSERT(pg->count > 0);
1263                 LASSERTF((pg->off & ~CFS_PAGE_MASK) + pg->count <= CFS_PAGE_SIZE,
1264                          "i: %d pg: %p off: "LPU64", count: %u\n", i, pg,
1265                          pg->off, pg->count);
1266 #ifdef __linux__
1267                 LASSERTF(i == 0 || pg->off > pg_prev->off,
1268                          "i %d p_c %u pg %p [pri %lu ind %lu] off "LPU64
1269                          " prev_pg %p [pri %lu ind %lu] off "LPU64"\n",
1270                          i, page_count,
1271                          pg->pg, page_private(pg->pg), pg->pg->index, pg->off,
1272                          pg_prev->pg, page_private(pg_prev->pg),
1273                          pg_prev->pg->index, pg_prev->off);
1274 #else
1275                 LASSERTF(i == 0 || pg->off > pg_prev->off,
1276                          "i %d p_c %u\n", i, page_count);
1277 #endif
1278                 LASSERT((pga[0]->flag & OBD_BRW_SRVLOCK) ==
1279                         (pg->flag & OBD_BRW_SRVLOCK));
1280
1281                 ptlrpc_prep_bulk_page(desc, pg->pg, pg->off & ~CFS_PAGE_MASK,
1282                                       pg->count);
1283                 requested_nob += pg->count;
1284
1285                 if (i > 0 && can_merge_pages(pg_prev, pg)) {
1286                         niobuf--;
1287                         niobuf->len += pg->count;
1288                 } else {
1289                         niobuf->offset = pg->off;
1290                         niobuf->len    = pg->count;
1291                         niobuf->flags  = pg->flag;
1292                 }
1293                 pg_prev = pg;
1294         }
1295
1296         LASSERTF((void *)(niobuf - niocount) ==
1297                 lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF + 2,
1298                                niocount * sizeof(*niobuf)),
1299                 "want %p - real %p\n", lustre_msg_buf(req->rq_reqmsg,
1300                 REQ_REC_OFF + 2, niocount * sizeof(*niobuf)),
1301                 (void *)(niobuf - niocount));
1302
1303         osc_announce_cached(cli, &body->oa, opc == OST_WRITE ? requested_nob:0);
1304         if (osc_should_shrink_grant(cli))
1305                 osc_shrink_grant_local(cli, &body->oa);
1306
1307         /* size[REQ_REC_OFF] still sizeof (*body) */
1308         if (opc == OST_WRITE) {
1309                 if (unlikely(cli->cl_checksum) &&
1310                     !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1311                         /* store cl_cksum_type in a local variable since
1312                          * it can be changed via lprocfs */
1313                         cksum_type_t cksum_type = cli->cl_cksum_type;
1314
1315                         if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1316                                 oa->o_flags = body->oa.o_flags = 0;
1317                         body->oa.o_flags |= cksum_type_pack(cksum_type);
1318                         body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1319                         body->oa.o_cksum = osc_checksum_bulk(requested_nob,
1320                                                              page_count, pga,
1321                                                              OST_WRITE,
1322                                                              cksum_type);
1323                         CDEBUG(D_PAGE, "checksum at write origin: %x\n",
1324                                body->oa.o_cksum);
1325                         /* save this in 'oa', too, for later checking */
1326                         oa->o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1327                         oa->o_flags |= cksum_type_pack(cksum_type);
1328                 } else {
1329                         /* clear out the checksum flag, in case this is a
1330                          * resend but cl_checksum is no longer set. b=11238 */
1331                         oa->o_valid &= ~OBD_MD_FLCKSUM;
1332                 }
1333                 oa->o_cksum = body->oa.o_cksum;
1334                 /* 1 RC per niobuf */
1335                 req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_SERVER,
1336                                      sizeof(__u32) * niocount);
1337         } else {
1338                 if (unlikely(cli->cl_checksum) &&
1339                     !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1340                         if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1341                                 body->oa.o_flags = 0;
1342                         body->oa.o_flags |= cksum_type_pack(cli->cl_cksum_type);
1343                         body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1344                 }
1345                 req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_SERVER, 0);
1346                 /* 1 RC for the whole I/O */
1347         }
1348         ptlrpc_request_set_replen(req);
1349
1350         CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
1351         aa = ptlrpc_req_async_args(req);
1352         aa->aa_oa = oa;
1353         aa->aa_requested_nob = requested_nob;
1354         aa->aa_nio_count = niocount;
1355         aa->aa_page_count = page_count;
1356         aa->aa_resends = 0;
1357         aa->aa_ppga = pga;
1358         aa->aa_cli = cli;
1359         CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1360         if (ocapa && reserve)
1361                 aa->aa_ocapa = capa_get(ocapa);
1362
1363         *reqp = req;
1364         RETURN(0);
1365
1366  out:
1367         ptlrpc_req_finished(req);
1368         RETURN(rc);
1369 }
1370
1371 static int check_write_checksum(struct obdo *oa, const lnet_process_id_t *peer,
1372                                 __u32 client_cksum, __u32 server_cksum, int nob,
1373                                 obd_count page_count, struct brw_page **pga,
1374                                 cksum_type_t client_cksum_type)
1375 {
1376         __u32 new_cksum;
1377         char *msg;
1378         cksum_type_t cksum_type;
1379
1380         if (server_cksum == client_cksum) {
1381                 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1382                 return 0;
1383         }
1384
1385         if (oa->o_valid & OBD_MD_FLFLAGS)
1386                 cksum_type = cksum_type_unpack(oa->o_flags);
1387         else
1388                 cksum_type = OBD_CKSUM_CRC32;
1389
1390         new_cksum = osc_checksum_bulk(nob, page_count, pga, OST_WRITE,
1391                                       cksum_type);
1392
1393         if (cksum_type != client_cksum_type)
1394                 msg = "the server did not use the checksum type specified in "
1395                       "the original request - likely a protocol problem";
1396         else if (new_cksum == server_cksum)
1397                 msg = "changed on the client after we checksummed it - "
1398                       "likely false positive due to mmap IO (bug 11742)";
1399         else if (new_cksum == client_cksum)
1400                 msg = "changed in transit before arrival at OST";
1401         else
1402                 msg = "changed in transit AND doesn't match the original - "
1403                       "likely false positive due to mmap IO (bug 11742)";
1404
1405         LCONSOLE_ERROR_MSG(0x132, "BAD WRITE CHECKSUM: %s: from %s inum "
1406                            LPU64"/"LPU64" object "LPU64"/"LPU64" extent "
1407                            "["LPU64"-"LPU64"]\n",
1408                            msg, libcfs_nid2str(peer->nid),
1409                            oa->o_valid & OBD_MD_FLFID ? oa->o_fid : (__u64)0,
1410                            oa->o_valid & OBD_MD_FLFID ? oa->o_generation :
1411                                                         (__u64)0,
1412                            oa->o_id,
1413                            oa->o_valid & OBD_MD_FLGROUP ? oa->o_gr : (__u64)0,
1414                            pga[0]->off,
1415                            pga[page_count-1]->off + pga[page_count-1]->count - 1);
1416         CERROR("original client csum %x (type %x), server csum %x (type %x), "
1417                "client csum now %x\n", client_cksum, client_cksum_type,
1418                server_cksum, cksum_type, new_cksum);
1419         return 1;
1420 }
1421
1422 /* Note rc enters this function as number of bytes transferred */
1423 static int osc_brw_fini_request(struct ptlrpc_request *req, int rc)
1424 {
1425         struct osc_brw_async_args *aa = (void *)&req->rq_async_args;
1426         const lnet_process_id_t *peer =
1427                         &req->rq_import->imp_connection->c_peer;
1428         struct client_obd *cli = aa->aa_cli;
1429         struct ost_body *body;
1430         __u32 client_cksum = 0;
1431         ENTRY;
1432
1433         if (rc < 0 && rc != -EDQUOT)
1434                 RETURN(rc);
1435
1436         LASSERTF(req->rq_repmsg != NULL, "rc = %d\n", rc);
1437         body = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*body),
1438                                   lustre_swab_ost_body);
1439         if (body == NULL) {
1440                 CDEBUG(D_INFO, "Can't unpack body\n");
1441                 RETURN(-EPROTO);
1442         }
1443
1444         /* set/clear over quota flag for a uid/gid */
1445         if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE &&
1446             body->oa.o_valid & (OBD_MD_FLUSRQUOTA | OBD_MD_FLGRPQUOTA)) {
1447                 unsigned int qid[MAXQUOTAS] = { body->oa.o_uid, body->oa.o_gid };
1448
1449                 lquota_setdq(quota_interface, cli, qid, body->oa.o_valid,
1450                              body->oa.o_flags);
1451         }
1452
1453         if (rc < 0)
1454                 RETURN(rc);
1455
1456         if (aa->aa_oa->o_valid & OBD_MD_FLCKSUM)
1457                 client_cksum = aa->aa_oa->o_cksum; /* save for later */
1458
1459         osc_update_grant(cli, body);
1460
1461         if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
1462                 if (rc > 0) {
1463                         CERROR("Unexpected +ve rc %d\n", rc);
1464                         RETURN(-EPROTO);
1465                 }
1466                 LASSERT(req->rq_bulk->bd_nob == aa->aa_requested_nob);
1467
1468                 if (sptlrpc_cli_unwrap_bulk_write(req, req->rq_bulk))
1469                         RETURN(-EAGAIN);
1470
1471                 if ((aa->aa_oa->o_valid & OBD_MD_FLCKSUM) && client_cksum &&
1472                     check_write_checksum(&body->oa, peer, client_cksum,
1473                                          body->oa.o_cksum, aa->aa_requested_nob,
1474                                          aa->aa_page_count, aa->aa_ppga,
1475                                          cksum_type_unpack(aa->aa_oa->o_flags)))
1476                         RETURN(-EAGAIN);
1477
1478                 rc = check_write_rcs(req, aa->aa_requested_nob,aa->aa_nio_count,
1479                                      aa->aa_page_count, aa->aa_ppga);
1480                 GOTO(out, rc);
1481         }
1482
1483         /* The rest of this function executes only for OST_READs */
1484
1485         /* if unwrap_bulk failed, return -EAGAIN to retry */
1486         rc = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk, rc);
1487         if (rc < 0)
1488                 GOTO(out, rc = -EAGAIN);
1489
1490         if (rc > aa->aa_requested_nob) {
1491                 CERROR("Unexpected rc %d (%d requested)\n", rc,
1492                        aa->aa_requested_nob);
1493                 RETURN(-EPROTO);
1494         }
1495
1496         if (rc != req->rq_bulk->bd_nob_transferred) {
1497                 CERROR ("Unexpected rc %d (%d transferred)\n",
1498                         rc, req->rq_bulk->bd_nob_transferred);
1499                 return (-EPROTO);
1500         }
1501
1502         if (rc < aa->aa_requested_nob)
1503                 handle_short_read(rc, aa->aa_page_count, aa->aa_ppga);
1504
1505         if (body->oa.o_valid & OBD_MD_FLCKSUM) {
1506                 static int cksum_counter;
1507                 __u32      server_cksum = body->oa.o_cksum;
1508                 char      *via;
1509                 char      *router;
1510                 cksum_type_t cksum_type;
1511
1512                 if (body->oa.o_valid & OBD_MD_FLFLAGS)
1513                         cksum_type = cksum_type_unpack(body->oa.o_flags);
1514                 else
1515                         cksum_type = OBD_CKSUM_CRC32;
1516                 client_cksum = osc_checksum_bulk(rc, aa->aa_page_count,
1517                                                  aa->aa_ppga, OST_READ,
1518                                                  cksum_type);
1519
1520                 if (peer->nid == req->rq_bulk->bd_sender) {
1521                         via = router = "";
1522                 } else {
1523                         via = " via ";
1524                         router = libcfs_nid2str(req->rq_bulk->bd_sender);
1525                 }
1526
1527                 if (server_cksum == ~0 && rc > 0) {
1528                         CERROR("Protocol error: server %s set the 'checksum' "
1529                                "bit, but didn't send a checksum.  Not fatal, "
1530                                "but please notify on http://bugzilla.lustre.org/\n",
1531                                libcfs_nid2str(peer->nid));
1532                 } else if (server_cksum != client_cksum) {
1533                         LCONSOLE_ERROR_MSG(0x133, "%s: BAD READ CHECKSUM: from "
1534                                            "%s%s%s inum "LPU64"/"LPU64" object "
1535                                            LPU64"/"LPU64" extent "
1536                                            "["LPU64"-"LPU64"]\n",
1537                                            req->rq_import->imp_obd->obd_name,
1538                                            libcfs_nid2str(peer->nid),
1539                                            via, router,
1540                                            body->oa.o_valid & OBD_MD_FLFID ?
1541                                                 body->oa.o_fid : (__u64)0,
1542                                            body->oa.o_valid & OBD_MD_FLFID ?
1543                                                 body->oa.o_generation :(__u64)0,
1544                                            body->oa.o_id,
1545                                            body->oa.o_valid & OBD_MD_FLGROUP ?
1546                                                 body->oa.o_gr : (__u64)0,
1547                                            aa->aa_ppga[0]->off,
1548                                            aa->aa_ppga[aa->aa_page_count-1]->off +
1549                                            aa->aa_ppga[aa->aa_page_count-1]->count -
1550                                                                         1);
1551                         CERROR("client %x, server %x, cksum_type %x\n",
1552                                client_cksum, server_cksum, cksum_type);
1553                         cksum_counter = 0;
1554                         aa->aa_oa->o_cksum = client_cksum;
1555                         rc = -EAGAIN;
1556                 } else {
1557                         cksum_counter++;
1558                         CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1559                         rc = 0;
1560                 }
1561         } else if (unlikely(client_cksum)) {
1562                 static int cksum_missed;
1563
1564                 cksum_missed++;
1565                 if ((cksum_missed & (-cksum_missed)) == cksum_missed)
1566                         CERROR("Checksum %u requested from %s but not sent\n",
1567                                cksum_missed, libcfs_nid2str(peer->nid));
1568         } else {
1569                 rc = 0;
1570         }
1571 out:
1572         if (rc >= 0)
1573                 *aa->aa_oa = body->oa;
1574
1575         RETURN(rc);
1576 }
1577
1578 static int osc_brw_internal(int cmd, struct obd_export *exp, struct obdo *oa,
1579                             struct lov_stripe_md *lsm,
1580                             obd_count page_count, struct brw_page **pga,
1581                             struct obd_capa *ocapa)
1582 {
1583         struct ptlrpc_request *req;
1584         int                    rc;
1585         cfs_waitq_t            waitq;
1586         int                    resends = 0;
1587         struct l_wait_info     lwi;
1588
1589         ENTRY;
1590
1591         cfs_waitq_init(&waitq);
1592
1593 restart_bulk:
1594         rc = osc_brw_prep_request(cmd, &exp->exp_obd->u.cli, oa, lsm,
1595                                   page_count, pga, &req, ocapa, 0);
1596         if (rc != 0)
1597                 return (rc);
1598
1599         rc = ptlrpc_queue_wait(req);
1600
1601         if (rc == -ETIMEDOUT && req->rq_resend) {
1602                 DEBUG_REQ(D_HA, req,  "BULK TIMEOUT");
1603                 ptlrpc_req_finished(req);
1604                 goto restart_bulk;
1605         }
1606
1607         rc = osc_brw_fini_request(req, rc);
1608
1609         ptlrpc_req_finished(req);
1610         if (osc_recoverable_error(rc)) {
1611                 resends++;
1612                 if (!osc_should_resend(resends, &exp->exp_obd->u.cli)) {
1613                         CERROR("too many resend retries, returning error\n");
1614                         RETURN(-EIO);
1615                 }
1616
1617                 lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(resends), NULL, NULL, NULL);
1618                 l_wait_event(waitq, 0, &lwi);
1619
1620                 goto restart_bulk;
1621         }
1622
1623         RETURN (rc);
1624 }
1625
1626 int osc_brw_redo_request(struct ptlrpc_request *request,
1627                          struct osc_brw_async_args *aa)
1628 {
1629         struct ptlrpc_request *new_req;
1630         struct ptlrpc_request_set *set = request->rq_set;
1631         struct osc_brw_async_args *new_aa;
1632         struct osc_async_page *oap;
1633         int rc = 0;
1634         ENTRY;
1635
1636         if (!osc_should_resend(aa->aa_resends, aa->aa_cli)) {
1637                 CERROR("too many resend retries, returning error\n");
1638                 RETURN(-EIO);
1639         }
1640
1641         DEBUG_REQ(D_ERROR, request, "redo for recoverable error");
1642
1643         rc = osc_brw_prep_request(lustre_msg_get_opc(request->rq_reqmsg) ==
1644                                         OST_WRITE ? OBD_BRW_WRITE :OBD_BRW_READ,
1645                                   aa->aa_cli, aa->aa_oa,
1646                                   NULL /* lsm unused by osc currently */,
1647                                   aa->aa_page_count, aa->aa_ppga,
1648                                   &new_req, aa->aa_ocapa, 0);
1649         if (rc)
1650                 RETURN(rc);
1651
1652         client_obd_list_lock(&aa->aa_cli->cl_loi_list_lock);
1653
1654         list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
1655                 if (oap->oap_request != NULL) {
1656                         LASSERTF(request == oap->oap_request,
1657                                  "request %p != oap_request %p\n",
1658                                  request, oap->oap_request);
1659                         if (oap->oap_interrupted) {
1660                                 client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
1661                                 ptlrpc_req_finished(new_req);
1662                                 RETURN(-EINTR);
1663                         }
1664                 }
1665         }
1666         /* New request takes over pga and oaps from old request.
1667          * Note that copying a list_head doesn't work, need to move it... */
1668         aa->aa_resends++;
1669         new_req->rq_interpret_reply = request->rq_interpret_reply;
1670         new_req->rq_async_args = request->rq_async_args;
1671         new_req->rq_sent = cfs_time_current_sec() + aa->aa_resends;
1672
1673         new_aa = ptlrpc_req_async_args(new_req);
1674
1675         CFS_INIT_LIST_HEAD(&new_aa->aa_oaps);
1676         list_splice(&aa->aa_oaps, &new_aa->aa_oaps);
1677         CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1678
1679         list_for_each_entry(oap, &new_aa->aa_oaps, oap_rpc_item) {
1680                 if (oap->oap_request) {
1681                         ptlrpc_req_finished(oap->oap_request);
1682                         oap->oap_request = ptlrpc_request_addref(new_req);
1683                 }
1684         }
1685
1686         new_aa->aa_ocapa = aa->aa_ocapa;
1687         aa->aa_ocapa = NULL;
1688
1689         /* use ptlrpc_set_add_req is safe because interpret functions work
1690          * in check_set context. only one way exist with access to request
1691          * from different thread got -EINTR - this way protected with
1692          * cl_loi_list_lock */
1693         ptlrpc_set_add_req(set, new_req);
1694
1695         client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
1696
1697         DEBUG_REQ(D_INFO, new_req, "new request");
1698         RETURN(0);
1699 }
1700
1701 /*
1702  * ugh, we want disk allocation on the target to happen in offset order.  we'll
1703  * follow sedgewicks advice and stick to the dead simple shellsort -- it'll do
1704  * fine for our small page arrays and doesn't require allocation.  its an
1705  * insertion sort that swaps elements that are strides apart, shrinking the
1706  * stride down until its '1' and the array is sorted.
1707  */
1708 static void sort_brw_pages(struct brw_page **array, int num)
1709 {
1710         int stride, i, j;
1711         struct brw_page *tmp;
1712
1713         if (num == 1)
1714                 return;
1715         for (stride = 1; stride < num ; stride = (stride * 3) + 1)
1716                 ;
1717
1718         do {
1719                 stride /= 3;
1720                 for (i = stride ; i < num ; i++) {
1721                         tmp = array[i];
1722                         j = i;
1723                         while (j >= stride && array[j - stride]->off > tmp->off) {
1724                                 array[j] = array[j - stride];
1725                                 j -= stride;
1726                         }
1727                         array[j] = tmp;
1728                 }
1729         } while (stride > 1);
1730 }
1731
1732 static obd_count max_unfragmented_pages(struct brw_page **pg, obd_count pages)
1733 {
1734         int count = 1;
1735         int offset;
1736         int i = 0;
1737
1738         LASSERT (pages > 0);
1739         offset = pg[i]->off & ~CFS_PAGE_MASK;
1740
1741         for (;;) {
1742                 pages--;
1743                 if (pages == 0)         /* that's all */
1744                         return count;
1745
1746                 if (offset + pg[i]->count < CFS_PAGE_SIZE)
1747                         return count;   /* doesn't end on page boundary */
1748
1749                 i++;
1750                 offset = pg[i]->off & ~CFS_PAGE_MASK;
1751                 if (offset != 0)        /* doesn't start on page boundary */
1752                         return count;
1753
1754                 count++;
1755         }
1756 }
1757
1758 static struct brw_page **osc_build_ppga(struct brw_page *pga, obd_count count)
1759 {
1760         struct brw_page **ppga;
1761         int i;
1762
1763         OBD_ALLOC(ppga, sizeof(*ppga) * count);
1764         if (ppga == NULL)
1765                 return NULL;
1766
1767         for (i = 0; i < count; i++)
1768                 ppga[i] = pga + i;
1769         return ppga;
1770 }
1771
1772 static void osc_release_ppga(struct brw_page **ppga, obd_count count)
1773 {
1774         LASSERT(ppga != NULL);
1775         OBD_FREE(ppga, sizeof(*ppga) * count);
1776 }
1777
1778 static int osc_brw(int cmd, struct obd_export *exp, struct obd_info *oinfo,
1779                    obd_count page_count, struct brw_page *pga,
1780                    struct obd_trans_info *oti)
1781 {
1782         struct obdo *saved_oa = NULL;
1783         struct brw_page **ppga, **orig;
1784         struct obd_import *imp = class_exp2cliimp(exp);
1785         struct client_obd *cli;
1786         int rc, page_count_orig;
1787         ENTRY;
1788
1789         LASSERT((imp != NULL) && (imp->imp_obd != NULL));
1790         cli = &imp->imp_obd->u.cli;
1791
1792         if (cmd & OBD_BRW_CHECK) {
1793                 /* The caller just wants to know if there's a chance that this
1794                  * I/O can succeed */
1795
1796                 if (imp->imp_invalid)
1797                         RETURN(-EIO);
1798                 RETURN(0);
1799         }
1800
1801         /* test_brw with a failed create can trip this, maybe others. */
1802         LASSERT(cli->cl_max_pages_per_rpc);
1803
1804         rc = 0;
1805
1806         orig = ppga = osc_build_ppga(pga, page_count);
1807         if (ppga == NULL)
1808                 RETURN(-ENOMEM);
1809         page_count_orig = page_count;
1810
1811         sort_brw_pages(ppga, page_count);
1812         while (page_count) {
1813                 obd_count pages_per_brw;
1814
1815                 if (page_count > cli->cl_max_pages_per_rpc)
1816                         pages_per_brw = cli->cl_max_pages_per_rpc;
1817                 else
1818                         pages_per_brw = page_count;
1819
1820                 pages_per_brw = max_unfragmented_pages(ppga, pages_per_brw);
1821
1822                 if (saved_oa != NULL) {
1823                         /* restore previously saved oa */
1824                         *oinfo->oi_oa = *saved_oa;
1825                 } else if (page_count > pages_per_brw) {
1826                         /* save a copy of oa (brw will clobber it) */
1827                         OBDO_ALLOC(saved_oa);
1828                         if (saved_oa == NULL)
1829                                 GOTO(out, rc = -ENOMEM);
1830                         *saved_oa = *oinfo->oi_oa;
1831                 }
1832
1833                 rc = osc_brw_internal(cmd, exp, oinfo->oi_oa, oinfo->oi_md,
1834                                       pages_per_brw, ppga, oinfo->oi_capa);
1835
1836                 if (rc != 0)
1837                         break;
1838
1839                 page_count -= pages_per_brw;
1840                 ppga += pages_per_brw;
1841         }
1842
1843 out:
1844         osc_release_ppga(orig, page_count_orig);
1845
1846         if (saved_oa != NULL)
1847                 OBDO_FREE(saved_oa);
1848
1849         RETURN(rc);
1850 }
1851
1852 /* The companion to osc_enter_cache(), called when @oap is no longer part of
1853  * the dirty accounting.  Writeback completes or truncate happens before
1854  * writing starts.  Must be called with the loi lock held. */
1855 static void osc_exit_cache(struct client_obd *cli, struct osc_async_page *oap,
1856                            int sent)
1857 {
1858         osc_release_write_grant(cli, &oap->oap_brw_page, sent);
1859 }
1860
1861
1862 /* This maintains the lists of pending pages to read/write for a given object
1863  * (lop).  This is used by osc_check_rpcs->osc_next_loi() and loi_list_maint()
1864  * to quickly find objects that are ready to send an RPC. */
1865 static int lop_makes_rpc(struct client_obd *cli, struct loi_oap_pages *lop,
1866                          int cmd)
1867 {
1868         int optimal;
1869         ENTRY;
1870
1871         if (lop->lop_num_pending == 0)
1872                 RETURN(0);
1873
1874         /* if we have an invalid import we want to drain the queued pages
1875          * by forcing them through rpcs that immediately fail and complete
1876          * the pages.  recovery relies on this to empty the queued pages
1877          * before canceling the locks and evicting down the llite pages */
1878         if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
1879                 RETURN(1);
1880
1881         /* stream rpcs in queue order as long as as there is an urgent page
1882          * queued.  this is our cheap solution for good batching in the case
1883          * where writepage marks some random page in the middle of the file
1884          * as urgent because of, say, memory pressure */
1885         if (!list_empty(&lop->lop_urgent)) {
1886                 CDEBUG(D_CACHE, "urgent request forcing RPC\n");
1887                 RETURN(1);
1888         }
1889         /* fire off rpcs when we have 'optimal' rpcs as tuned for the wire. */
1890         optimal = cli->cl_max_pages_per_rpc;
1891         if (cmd & OBD_BRW_WRITE) {
1892                 /* trigger a write rpc stream as long as there are dirtiers
1893                  * waiting for space.  as they're waiting, they're not going to
1894                  * create more pages to coallesce with what's waiting.. */
1895                 if (!list_empty(&cli->cl_cache_waiters)) {
1896                         CDEBUG(D_CACHE, "cache waiters forcing RPC\n");
1897                         RETURN(1);
1898                 }
1899                 /* +16 to avoid triggering rpcs that would want to include pages
1900                  * that are being queued but which can't be made ready until
1901                  * the queuer finishes with the page. this is a wart for
1902                  * llite::commit_write() */
1903                 optimal += 16;
1904         }
1905         if (lop->lop_num_pending >= optimal)
1906                 RETURN(1);
1907
1908         RETURN(0);
1909 }
1910
1911 static int lop_makes_hprpc(struct loi_oap_pages *lop)
1912 {
1913         struct osc_async_page *oap;
1914         ENTRY;
1915
1916         if (list_empty(&lop->lop_urgent))
1917                 RETURN(0);
1918
1919         oap = list_entry(lop->lop_urgent.next,
1920                          struct osc_async_page, oap_urgent_item);
1921
1922         if (oap->oap_async_flags & ASYNC_HP) {
1923                 CDEBUG(D_CACHE, "hp request forcing RPC\n");
1924                 RETURN(1);
1925         }
1926
1927         RETURN(0);
1928 }
1929
1930 static void on_list(struct list_head *item, struct list_head *list,
1931                     int should_be_on)
1932 {
1933         if (list_empty(item) && should_be_on)
1934                 list_add_tail(item, list);
1935         else if (!list_empty(item) && !should_be_on)
1936                 list_del_init(item);
1937 }
1938
1939 /* maintain the loi's cli list membership invariants so that osc_send_oap_rpc
1940  * can find pages to build into rpcs quickly */
1941 void loi_list_maint(struct client_obd *cli, struct lov_oinfo *loi)
1942 {
1943         if (lop_makes_hprpc(&loi->loi_write_lop) ||
1944             lop_makes_hprpc(&loi->loi_read_lop)) {
1945                 /* HP rpc */
1946                 on_list(&loi->loi_ready_item, &cli->cl_loi_ready_list, 0);
1947                 on_list(&loi->loi_hp_ready_item, &cli->cl_loi_hp_ready_list, 1);
1948         } else {
1949                 on_list(&loi->loi_hp_ready_item, &cli->cl_loi_hp_ready_list, 0);
1950                 on_list(&loi->loi_ready_item, &cli->cl_loi_ready_list,
1951                         lop_makes_rpc(cli, &loi->loi_write_lop, OBD_BRW_WRITE)||
1952                         lop_makes_rpc(cli, &loi->loi_read_lop, OBD_BRW_READ));
1953         }
1954
1955         on_list(&loi->loi_write_item, &cli->cl_loi_write_list,
1956                 loi->loi_write_lop.lop_num_pending);
1957
1958         on_list(&loi->loi_read_item, &cli->cl_loi_read_list,
1959                 loi->loi_read_lop.lop_num_pending);
1960 }
1961
1962 static void lop_update_pending(struct client_obd *cli,
1963                                struct loi_oap_pages *lop, int cmd, int delta)
1964 {
1965         lop->lop_num_pending += delta;
1966         if (cmd & OBD_BRW_WRITE)
1967                 cli->cl_pending_w_pages += delta;
1968         else
1969                 cli->cl_pending_r_pages += delta;
1970 }
1971
1972 /**
1973  * this is called when a sync waiter receives an interruption.  Its job is to
1974  * get the caller woken as soon as possible.  If its page hasn't been put in an
1975  * rpc yet it can dequeue immediately.  Otherwise it has to mark the rpc as
1976  * desiring interruption which will forcefully complete the rpc once the rpc
1977  * has timed out.
1978  */
1979 int osc_oap_interrupted(const struct lu_env *env, struct osc_async_page *oap)
1980 {
1981         struct loi_oap_pages *lop;
1982         struct lov_oinfo *loi;
1983         int rc = -EBUSY;
1984         ENTRY;
1985
1986         LASSERT(!oap->oap_interrupted);
1987         oap->oap_interrupted = 1;
1988
1989         /* ok, it's been put in an rpc. only one oap gets a request reference */
1990         if (oap->oap_request != NULL) {
1991                 ptlrpc_mark_interrupted(oap->oap_request);
1992                 ptlrpcd_wake(oap->oap_request);
1993                 ptlrpc_req_finished(oap->oap_request);
1994                 oap->oap_request = NULL;
1995         }
1996
1997         /*
1998          * page completion may be called only if ->cpo_prep() method was
1999          * executed by osc_io_submit(), that also adds page the to pending list
2000          */
2001         if (!list_empty(&oap->oap_pending_item)) {
2002                 list_del_init(&oap->oap_pending_item);
2003                 list_del_init(&oap->oap_urgent_item);
2004
2005                 loi = oap->oap_loi;
2006                 lop = (oap->oap_cmd & OBD_BRW_WRITE) ?
2007                         &loi->loi_write_lop : &loi->loi_read_lop;
2008                 lop_update_pending(oap->oap_cli, lop, oap->oap_cmd, -1);
2009                 loi_list_maint(oap->oap_cli, oap->oap_loi);
2010                 rc = oap->oap_caller_ops->ap_completion(env,
2011                                           oap->oap_caller_data,
2012                                           oap->oap_cmd, NULL, -EINTR);
2013         }
2014
2015         RETURN(rc);
2016 }
2017
2018 /* this is trying to propogate async writeback errors back up to the
2019  * application.  As an async write fails we record the error code for later if
2020  * the app does an fsync.  As long as errors persist we force future rpcs to be
2021  * sync so that the app can get a sync error and break the cycle of queueing
2022  * pages for which writeback will fail. */
2023 static void osc_process_ar(struct osc_async_rc *ar, __u64 xid,
2024                            int rc)
2025 {
2026         if (rc) {
2027                 if (!ar->ar_rc)
2028                         ar->ar_rc = rc;
2029
2030                 ar->ar_force_sync = 1;
2031                 ar->ar_min_xid = ptlrpc_sample_next_xid();
2032                 return;
2033
2034         }
2035
2036         if (ar->ar_force_sync && (xid >= ar->ar_min_xid))
2037                 ar->ar_force_sync = 0;
2038 }
2039
2040 void osc_oap_to_pending(struct osc_async_page *oap)
2041 {
2042         struct loi_oap_pages *lop;
2043
2044         if (oap->oap_cmd & OBD_BRW_WRITE)
2045                 lop = &oap->oap_loi->loi_write_lop;
2046         else
2047                 lop = &oap->oap_loi->loi_read_lop;
2048
2049         if (oap->oap_async_flags & ASYNC_HP)
2050                 list_add(&oap->oap_urgent_item, &lop->lop_urgent);
2051         else if (oap->oap_async_flags & ASYNC_URGENT)
2052                 list_add_tail(&oap->oap_urgent_item, &lop->lop_urgent);
2053         list_add_tail(&oap->oap_pending_item, &lop->lop_pending);
2054         lop_update_pending(oap->oap_cli, lop, oap->oap_cmd, 1);
2055 }
2056
2057 /* this must be called holding the loi list lock to give coverage to exit_cache,
2058  * async_flag maintenance, and oap_request */
2059 static void osc_ap_completion(const struct lu_env *env,
2060                               struct client_obd *cli, struct obdo *oa,
2061                               struct osc_async_page *oap, int sent, int rc)
2062 {
2063         __u64 xid = 0;
2064
2065         ENTRY;
2066         if (oap->oap_request != NULL) {
2067                 xid = ptlrpc_req_xid(oap->oap_request);
2068                 ptlrpc_req_finished(oap->oap_request);
2069                 oap->oap_request = NULL;
2070         }
2071
2072         oap->oap_async_flags = 0;
2073         oap->oap_interrupted = 0;
2074
2075         if (oap->oap_cmd & OBD_BRW_WRITE) {
2076                 osc_process_ar(&cli->cl_ar, xid, rc);
2077                 osc_process_ar(&oap->oap_loi->loi_ar, xid, rc);
2078         }
2079
2080         if (rc == 0 && oa != NULL) {
2081                 if (oa->o_valid & OBD_MD_FLBLOCKS)
2082                         oap->oap_loi->loi_lvb.lvb_blocks = oa->o_blocks;
2083                 if (oa->o_valid & OBD_MD_FLMTIME)
2084                         oap->oap_loi->loi_lvb.lvb_mtime = oa->o_mtime;
2085                 if (oa->o_valid & OBD_MD_FLATIME)
2086                         oap->oap_loi->loi_lvb.lvb_atime = oa->o_atime;
2087                 if (oa->o_valid & OBD_MD_FLCTIME)
2088                         oap->oap_loi->loi_lvb.lvb_ctime = oa->o_ctime;
2089         }
2090
2091         rc = oap->oap_caller_ops->ap_completion(env, oap->oap_caller_data,
2092                                                 oap->oap_cmd, oa, rc);
2093
2094         /* ll_ap_completion (from llite) drops PG_locked. so, a new
2095          * I/O on the page could start, but OSC calls it under lock
2096          * and thus we can add oap back to pending safely */
2097         if (rc)
2098                 /* upper layer wants to leave the page on pending queue */
2099                 osc_oap_to_pending(oap);
2100         else
2101                 osc_exit_cache(cli, oap, sent);
2102         EXIT;
2103 }
2104
2105 static int brw_interpret(const struct lu_env *env,
2106                          struct ptlrpc_request *req, void *data, int rc)
2107 {
2108         struct osc_brw_async_args *aa = data;
2109         struct client_obd *cli;
2110         int async;
2111         ENTRY;
2112
2113         rc = osc_brw_fini_request(req, rc);
2114         CDEBUG(D_INODE, "request %p aa %p rc %d\n", req, aa, rc);
2115         if (osc_recoverable_error(rc)) {
2116                 rc = osc_brw_redo_request(req, aa);
2117                 if (rc == 0)
2118                         RETURN(0);
2119         }
2120
2121         if (aa->aa_ocapa) {
2122                 capa_put(aa->aa_ocapa);
2123                 aa->aa_ocapa = NULL;
2124         }
2125
2126         cli = aa->aa_cli;
2127
2128         client_obd_list_lock(&cli->cl_loi_list_lock);
2129
2130         /* We need to decrement before osc_ap_completion->osc_wake_cache_waiters
2131          * is called so we know whether to go to sync BRWs or wait for more
2132          * RPCs to complete */
2133         if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE)
2134                 cli->cl_w_in_flight--;
2135         else
2136                 cli->cl_r_in_flight--;
2137
2138         async = list_empty(&aa->aa_oaps);
2139         if (!async) { /* from osc_send_oap_rpc() */
2140                 struct osc_async_page *oap, *tmp;
2141                 /* the caller may re-use the oap after the completion call so
2142                  * we need to clean it up a little */
2143                 list_for_each_entry_safe(oap, tmp, &aa->aa_oaps, oap_rpc_item) {
2144                         list_del_init(&oap->oap_rpc_item);
2145                         osc_ap_completion(env, cli, aa->aa_oa, oap, 1, rc);
2146                 }
2147                 OBDO_FREE(aa->aa_oa);
2148         } else { /* from async_internal() */
2149                 int i;
2150                 for (i = 0; i < aa->aa_page_count; i++)
2151                         osc_release_write_grant(aa->aa_cli, aa->aa_ppga[i], 1);
2152         }
2153         osc_wake_cache_waiters(cli);
2154         osc_check_rpcs(env, cli);
2155         client_obd_list_unlock(&cli->cl_loi_list_lock);
2156         if (!async)
2157                 cl_req_completion(env, aa->aa_clerq, rc);
2158         osc_release_ppga(aa->aa_ppga, aa->aa_page_count);
2159         RETURN(rc);
2160 }
2161
2162 static struct ptlrpc_request *osc_build_req(const struct lu_env *env,
2163                                             struct client_obd *cli,
2164                                             struct list_head *rpc_list,
2165                                             int page_count, int cmd)
2166 {
2167         struct ptlrpc_request *req;
2168         struct brw_page **pga = NULL;
2169         struct osc_brw_async_args *aa;
2170         struct obdo *oa = NULL;
2171         const struct obd_async_page_ops *ops = NULL;
2172         void *caller_data = NULL;
2173         struct osc_async_page *oap;
2174         struct osc_async_page *tmp;
2175         struct ost_body *body;
2176         struct cl_req *clerq = NULL;
2177         enum cl_req_type crt = (cmd & OBD_BRW_WRITE) ? CRT_WRITE : CRT_READ;
2178         struct ldlm_lock *lock = NULL;
2179         struct cl_req_attr crattr;
2180         int i, rc;
2181
2182         ENTRY;
2183         LASSERT(!list_empty(rpc_list));
2184
2185         memset(&crattr, 0, sizeof crattr);
2186         OBD_ALLOC(pga, sizeof(*pga) * page_count);
2187         if (pga == NULL)
2188                 GOTO(out, req = ERR_PTR(-ENOMEM));
2189
2190         OBDO_ALLOC(oa);
2191         if (oa == NULL)
2192                 GOTO(out, req = ERR_PTR(-ENOMEM));
2193
2194         i = 0;
2195         list_for_each_entry(oap, rpc_list, oap_rpc_item) {
2196                 struct cl_page *page = osc_oap2cl_page(oap);
2197                 if (ops == NULL) {
2198                         ops = oap->oap_caller_ops;
2199                         caller_data = oap->oap_caller_data;
2200
2201                         clerq = cl_req_alloc(env, page, crt,
2202                                              1 /* only 1-object rpcs for
2203                                                 * now */);
2204                         if (IS_ERR(clerq))
2205                                 GOTO(out, req = (void *)clerq);
2206                         lock = oap->oap_ldlm_lock;
2207                 }
2208                 pga[i] = &oap->oap_brw_page;
2209                 pga[i]->off = oap->oap_obj_off + oap->oap_page_off;
2210                 CDEBUG(0, "put page %p index %lu oap %p flg %x to pga\n",
2211                        pga[i]->pg, cfs_page_index(oap->oap_page), oap, pga[i]->flag);
2212                 i++;
2213                 cl_req_page_add(env, clerq, page);
2214         }
2215
2216         /* always get the data for the obdo for the rpc */
2217         LASSERT(ops != NULL);
2218         crattr.cra_oa = oa;
2219         crattr.cra_capa = NULL;
2220         cl_req_attr_set(env, clerq, &crattr, ~0ULL);
2221         if (lock) {
2222                 oa->o_handle = lock->l_remote_handle;
2223                 oa->o_valid |= OBD_MD_FLHANDLE;
2224         }
2225
2226         rc = cl_req_prep(env, clerq);
2227         if (rc != 0) {
2228                 CERROR("cl_req_prep failed: %d\n", rc);
2229                 GOTO(out, req = ERR_PTR(rc));
2230         }
2231
2232         sort_brw_pages(pga, page_count);
2233         rc = osc_brw_prep_request(cmd, cli, oa, NULL, page_count,
2234                                   pga, &req, crattr.cra_capa, 1);
2235         if (rc != 0) {
2236                 CERROR("prep_req failed: %d\n", rc);
2237                 GOTO(out, req = ERR_PTR(rc));
2238         }
2239
2240         /* Need to update the timestamps after the request is built in case
2241          * we race with setattr (locally or in queue at OST).  If OST gets
2242          * later setattr before earlier BRW (as determined by the request xid),
2243          * the OST will not use BRW timestamps.  Sadly, there is no obvious
2244          * way to do this in a single call.  bug 10150 */
2245         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
2246         cl_req_attr_set(env, clerq, &crattr,
2247                         OBD_MD_FLMTIME|OBD_MD_FLCTIME|OBD_MD_FLATIME);
2248
2249         CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2250         aa = ptlrpc_req_async_args(req);
2251         CFS_INIT_LIST_HEAD(&aa->aa_oaps);
2252         list_splice(rpc_list, &aa->aa_oaps);
2253         CFS_INIT_LIST_HEAD(rpc_list);
2254         aa->aa_clerq = clerq;
2255 out:
2256         capa_put(crattr.cra_capa);
2257         if (IS_ERR(req)) {
2258                 if (oa)
2259                         OBDO_FREE(oa);
2260                 if (pga)
2261                         OBD_FREE(pga, sizeof(*pga) * page_count);
2262                 /* this should happen rarely and is pretty bad, it makes the
2263                  * pending list not follow the dirty order */
2264                 client_obd_list_lock(&cli->cl_loi_list_lock);
2265                 list_for_each_entry_safe(oap, tmp, rpc_list, oap_rpc_item) {
2266                         list_del_init(&oap->oap_rpc_item);
2267
2268                         /* queued sync pages can be torn down while the pages
2269                          * were between the pending list and the rpc */
2270                         if (oap->oap_interrupted) {
2271                                 CDEBUG(D_INODE, "oap %p interrupted\n", oap);
2272                                 osc_ap_completion(env, cli, NULL, oap, 0,
2273                                                   oap->oap_count);
2274                                 continue;
2275                         }
2276                         osc_ap_completion(env, cli, NULL, oap, 0, PTR_ERR(req));
2277                 }
2278                 if (clerq && !IS_ERR(clerq))
2279                         cl_req_completion(env, clerq, PTR_ERR(req));
2280         }
2281         RETURN(req);
2282 }
2283
2284 /**
2285  * prepare pages for ASYNC io and put pages in send queue.
2286  *
2287  * \param cli -
2288  * \param loi -
2289  * \param cmd - OBD_BRW_* macroses
2290  * \param lop - pending pages
2291  *
2292  * \return zero if pages successfully add to send queue.
2293  * \return not zere if error occurring.
2294  */
2295 static int
2296 osc_send_oap_rpc(const struct lu_env *env, struct client_obd *cli,
2297                  struct lov_oinfo *loi,
2298                  int cmd, struct loi_oap_pages *lop)
2299 {
2300         struct ptlrpc_request *req;
2301         obd_count page_count = 0;
2302         struct osc_async_page *oap = NULL, *tmp;
2303         struct osc_brw_async_args *aa;
2304         const struct obd_async_page_ops *ops;
2305         CFS_LIST_HEAD(rpc_list);
2306         unsigned int ending_offset;
2307         unsigned  starting_offset = 0;
2308         int srvlock = 0;
2309         struct cl_object *clob = NULL;
2310         ENTRY;
2311
2312         /* If there are HP OAPs we need to handle at least 1 of them,
2313          * move it the beginning of the pending list for that. */
2314         if (!list_empty(&lop->lop_urgent)) {
2315                 oap = list_entry(lop->lop_urgent.next,
2316                                  struct osc_async_page, oap_urgent_item);
2317                 if (oap->oap_async_flags & ASYNC_HP)
2318                         list_move(&oap->oap_pending_item, &lop->lop_pending);
2319         }
2320
2321         /* first we find the pages we're allowed to work with */
2322         list_for_each_entry_safe(oap, tmp, &lop->lop_pending,
2323                                  oap_pending_item) {
2324                 ops = oap->oap_caller_ops;
2325
2326                 LASSERTF(oap->oap_magic == OAP_MAGIC, "Bad oap magic: oap %p, "
2327                          "magic 0x%x\n", oap, oap->oap_magic);
2328
2329                 if (clob == NULL) {
2330                         /* pin object in memory, so that completion call-backs
2331                          * can be safely called under client_obd_list lock. */
2332                         clob = osc_oap2cl_page(oap)->cp_obj;
2333                         cl_object_get(clob);
2334                 }
2335
2336                 if (page_count != 0 &&
2337                     srvlock != !!(oap->oap_brw_flags & OBD_BRW_SRVLOCK)) {
2338                         CDEBUG(D_PAGE, "SRVLOCK flag mismatch,"
2339                                " oap %p, page %p, srvlock %u\n",
2340                                oap, oap->oap_brw_page.pg, (unsigned)!srvlock);
2341                         break;
2342                 }
2343                 /* in llite being 'ready' equates to the page being locked
2344                  * until completion unlocks it.  commit_write submits a page
2345                  * as not ready because its unlock will happen unconditionally
2346                  * as the call returns.  if we race with commit_write giving
2347                  * us that page we dont' want to create a hole in the page
2348                  * stream, so we stop and leave the rpc to be fired by
2349                  * another dirtier or kupdated interval (the not ready page
2350                  * will still be on the dirty list).  we could call in
2351                  * at the end of ll_file_write to process the queue again. */
2352                 if (!(oap->oap_async_flags & ASYNC_READY)) {
2353                         int rc = ops->ap_make_ready(env, oap->oap_caller_data,
2354                                                     cmd);
2355                         if (rc < 0)
2356                                 CDEBUG(D_INODE, "oap %p page %p returned %d "
2357                                                 "instead of ready\n", oap,
2358                                                 oap->oap_page, rc);
2359                         switch (rc) {
2360                         case -EAGAIN:
2361                                 /* llite is telling us that the page is still
2362                                  * in commit_write and that we should try
2363                                  * and put it in an rpc again later.  we
2364                                  * break out of the loop so we don't create
2365                                  * a hole in the sequence of pages in the rpc
2366                                  * stream.*/
2367                                 oap = NULL;
2368                                 break;
2369                         case -EINTR:
2370                                 /* the io isn't needed.. tell the checks
2371                                  * below to complete the rpc with EINTR */
2372                                 oap->oap_async_flags |= ASYNC_COUNT_STABLE;
2373                                 oap->oap_count = -EINTR;
2374                                 break;
2375                         case 0:
2376                                 oap->oap_async_flags |= ASYNC_READY;
2377                                 break;
2378                         default:
2379                                 LASSERTF(0, "oap %p page %p returned %d "
2380                                             "from make_ready\n", oap,
2381                                             oap->oap_page, rc);
2382                                 break;
2383                         }
2384                 }
2385                 if (oap == NULL)
2386                         break;
2387                 /*
2388                  * Page submitted for IO has to be locked. Either by
2389                  * ->ap_make_ready() or by higher layers.
2390                  */
2391 #if defined(__KERNEL__) && defined(__linux__)
2392                 {
2393                         struct cl_page *page;
2394
2395                         page = osc_oap2cl_page(oap);
2396
2397                         if (page->cp_type == CPT_CACHEABLE &&
2398                             !(PageLocked(oap->oap_page) &&
2399                               (CheckWriteback(oap->oap_page, cmd)))) {
2400                                 CDEBUG(D_PAGE, "page %p lost wb %lx/%x\n",
2401                                        oap->oap_page,
2402                                        (long)oap->oap_page->flags,
2403                                        oap->oap_async_flags);
2404                                 LBUG();
2405                         }
2406                 }
2407 #endif
2408                 /* If there is a gap at the start of this page, it can't merge
2409                  * with any previous page, so we'll hand the network a
2410                  * "fragmented" page array that it can't transfer in 1 RDMA */
2411                 if (page_count != 0 && oap->oap_page_off != 0)
2412                         break;
2413
2414                 /* take the page out of our book-keeping */
2415                 list_del_init(&oap->oap_pending_item);
2416                 lop_update_pending(cli, lop, cmd, -1);
2417                 list_del_init(&oap->oap_urgent_item);
2418
2419                 if (page_count == 0)
2420                         starting_offset = (oap->oap_obj_off+oap->oap_page_off) &
2421                                           (PTLRPC_MAX_BRW_SIZE - 1);
2422
2423                 /* ask the caller for the size of the io as the rpc leaves. */
2424                 if (!(oap->oap_async_flags & ASYNC_COUNT_STABLE)) {
2425                         oap->oap_count =
2426                                 ops->ap_refresh_count(env, oap->oap_caller_data,
2427                                                       cmd);
2428                         LASSERT(oap->oap_page_off + oap->oap_count <= CFS_PAGE_SIZE);
2429                 }
2430                 if (oap->oap_count <= 0) {
2431                         CDEBUG(D_CACHE, "oap %p count %d, completing\n", oap,
2432                                oap->oap_count);
2433                         osc_ap_completion(env, cli, NULL,
2434                                           oap, 0, oap->oap_count);
2435                         continue;
2436                 }
2437
2438                 /* now put the page back in our accounting */
2439                 list_add_tail(&oap->oap_rpc_item, &rpc_list);
2440                 if (page_count == 0)
2441                         srvlock = !!(oap->oap_brw_flags & OBD_BRW_SRVLOCK);
2442                 if (++page_count >= cli->cl_max_pages_per_rpc)
2443                         break;
2444
2445                 /* End on a PTLRPC_MAX_BRW_SIZE boundary.  We want full-sized
2446                  * RPCs aligned on PTLRPC_MAX_BRW_SIZE boundaries to help reads
2447                  * have the same alignment as the initial writes that allocated
2448                  * extents on the server. */
2449                 ending_offset = (oap->oap_obj_off + oap->oap_page_off +
2450                                  oap->oap_count) & (PTLRPC_MAX_BRW_SIZE - 1);
2451                 if (ending_offset == 0)
2452                         break;
2453
2454                 /* If there is a gap at the end of this page, it can't merge
2455                  * with any subsequent pages, so we'll hand the network a
2456                  * "fragmented" page array that it can't transfer in 1 RDMA */
2457                 if (oap->oap_page_off + oap->oap_count < CFS_PAGE_SIZE)
2458                         break;
2459         }
2460
2461         osc_wake_cache_waiters(cli);
2462
2463         loi_list_maint(cli, loi);
2464
2465         client_obd_list_unlock(&cli->cl_loi_list_lock);
2466
2467         if (clob != NULL)
2468                 cl_object_put(env, clob);
2469
2470         if (page_count == 0) {
2471                 client_obd_list_lock(&cli->cl_loi_list_lock);
2472                 RETURN(0);
2473         }
2474
2475         req = osc_build_req(env, cli, &rpc_list, page_count, cmd);
2476         if (IS_ERR(req)) {
2477                 LASSERT(list_empty(&rpc_list));
2478                 loi_list_maint(cli, loi);
2479                 RETURN(PTR_ERR(req));
2480         }
2481
2482         aa = ptlrpc_req_async_args(req);
2483
2484         if (cmd == OBD_BRW_READ) {
2485                 lprocfs_oh_tally_log2(&cli->cl_read_page_hist, page_count);
2486                 lprocfs_oh_tally(&cli->cl_read_rpc_hist, cli->cl_r_in_flight);
2487                 lprocfs_oh_tally_log2(&cli->cl_read_offset_hist,
2488                                       (starting_offset >> CFS_PAGE_SHIFT) + 1);
2489         } else {
2490                 lprocfs_oh_tally_log2(&cli->cl_write_page_hist, page_count);
2491                 lprocfs_oh_tally(&cli->cl_write_rpc_hist,
2492                                  cli->cl_w_in_flight);
2493                 lprocfs_oh_tally_log2(&cli->cl_write_offset_hist,
2494                                       (starting_offset >> CFS_PAGE_SHIFT) + 1);
2495         }
2496         ptlrpc_lprocfs_brw(req, aa->aa_requested_nob);
2497
2498         client_obd_list_lock(&cli->cl_loi_list_lock);
2499
2500         if (cmd == OBD_BRW_READ)
2501                 cli->cl_r_in_flight++;
2502         else
2503                 cli->cl_w_in_flight++;
2504
2505         /* queued sync pages can be torn down while the pages
2506          * were between the pending list and the rpc */
2507         tmp = NULL;
2508         list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
2509                 /* only one oap gets a request reference */
2510                 if (tmp == NULL)
2511                         tmp = oap;
2512                 if (oap->oap_interrupted && !req->rq_intr) {
2513                         CDEBUG(D_INODE, "oap %p in req %p interrupted\n",
2514                                oap, req);
2515                         ptlrpc_mark_interrupted(req);
2516                 }
2517         }
2518         if (tmp != NULL)
2519                 tmp->oap_request = ptlrpc_request_addref(req);
2520
2521         DEBUG_REQ(D_INODE, req, "%d pages, aa %p. now %dr/%dw in flight",
2522                   page_count, aa, cli->cl_r_in_flight, cli->cl_w_in_flight);
2523
2524         req->rq_interpret_reply = brw_interpret;
2525         ptlrpcd_add_req(req, PSCOPE_BRW);
2526         RETURN(1);
2527 }
2528
2529 #define LOI_DEBUG(LOI, STR, args...)                                     \
2530         CDEBUG(D_INODE, "loi ready %d wr %d:%d rd %d:%d " STR,           \
2531                !list_empty(&(LOI)->loi_ready_item) ||                    \
2532                !list_empty(&(LOI)->loi_hp_ready_item),                   \
2533                (LOI)->loi_write_lop.lop_num_pending,                     \
2534                !list_empty(&(LOI)->loi_write_lop.lop_urgent),            \
2535                (LOI)->loi_read_lop.lop_num_pending,                      \
2536                !list_empty(&(LOI)->loi_read_lop.lop_urgent),             \
2537                args)                                                     \
2538
2539 /* This is called by osc_check_rpcs() to find which objects have pages that
2540  * we could be sending.  These lists are maintained by lop_makes_rpc(). */
2541 struct lov_oinfo *osc_next_loi(struct client_obd *cli)
2542 {
2543         ENTRY;
2544
2545         /* First return objects that have blocked locks so that they
2546          * will be flushed quickly and other clients can get the lock,
2547          * then objects which have pages ready to be stuffed into RPCs */
2548         if (!list_empty(&cli->cl_loi_hp_ready_list))
2549                 RETURN(list_entry(cli->cl_loi_hp_ready_list.next,
2550                                   struct lov_oinfo, loi_hp_ready_item));
2551         if (!list_empty(&cli->cl_loi_ready_list))
2552                 RETURN(list_entry(cli->cl_loi_ready_list.next,
2553                                   struct lov_oinfo, loi_ready_item));
2554
2555         /* then if we have cache waiters, return all objects with queued
2556          * writes.  This is especially important when many small files
2557          * have filled up the cache and not been fired into rpcs because
2558          * they don't pass the nr_pending/object threshhold */
2559         if (!list_empty(&cli->cl_cache_waiters) &&
2560             !list_empty(&cli->cl_loi_write_list))
2561                 RETURN(list_entry(cli->cl_loi_write_list.next,
2562                                   struct lov_oinfo, loi_write_item));
2563
2564         /* then return all queued objects when we have an invalid import
2565          * so that they get flushed */
2566         if (cli->cl_import == NULL || cli->cl_import->imp_invalid) {
2567                 if (!list_empty(&cli->cl_loi_write_list))
2568                         RETURN(list_entry(cli->cl_loi_write_list.next,
2569                                           struct lov_oinfo, loi_write_item));
2570                 if (!list_empty(&cli->cl_loi_read_list))
2571                         RETURN(list_entry(cli->cl_loi_read_list.next,
2572                                           struct lov_oinfo, loi_read_item));
2573         }
2574         RETURN(NULL);
2575 }
2576
2577 static int osc_max_rpc_in_flight(struct client_obd *cli, struct lov_oinfo *loi)
2578 {
2579         struct osc_async_page *oap;
2580         int hprpc = 0;
2581
2582         if (!list_empty(&loi->loi_write_lop.lop_urgent)) {
2583                 oap = list_entry(loi->loi_write_lop.lop_urgent.next,
2584                                  struct osc_async_page, oap_urgent_item);
2585                 hprpc = !!(oap->oap_async_flags & ASYNC_HP);
2586         }
2587
2588         if (!hprpc && !list_empty(&loi->loi_read_lop.lop_urgent)) {
2589                 oap = list_entry(loi->loi_write_lop.lop_urgent.next,
2590                                  struct osc_async_page, oap_urgent_item);
2591                 hprpc = !!(oap->oap_async_flags & ASYNC_HP);
2592         }
2593
2594         return rpcs_in_flight(cli) >= cli->cl_max_rpcs_in_flight + hprpc;
2595 }
2596
2597 /* called with the loi list lock held */
2598 void osc_check_rpcs(const struct lu_env *env, struct client_obd *cli)
2599 {
2600         struct lov_oinfo *loi;
2601         int rc = 0, race_counter = 0;
2602         ENTRY;
2603
2604         while ((loi = osc_next_loi(cli)) != NULL) {
2605                 LOI_DEBUG(loi, "%lu in flight\n", rpcs_in_flight(cli));
2606
2607                 if (osc_max_rpc_in_flight(cli, loi))
2608                         break;
2609
2610                 /* attempt some read/write balancing by alternating between
2611                  * reads and writes in an object.  The makes_rpc checks here
2612                  * would be redundant if we were getting read/write work items
2613                  * instead of objects.  we don't want send_oap_rpc to drain a
2614                  * partial read pending queue when we're given this object to
2615                  * do io on writes while there are cache waiters */
2616                 if (lop_makes_rpc(cli, &loi->loi_write_lop, OBD_BRW_WRITE)) {
2617                         rc = osc_send_oap_rpc(env, cli, loi, OBD_BRW_WRITE,
2618                                               &loi->loi_write_lop);
2619                         if (rc < 0)
2620                                 break;
2621                         if (rc > 0)
2622                                 race_counter = 0;
2623                         else
2624                                 race_counter++;
2625                 }
2626                 if (lop_makes_rpc(cli, &loi->loi_read_lop, OBD_BRW_READ)) {
2627                         rc = osc_send_oap_rpc(env, cli, loi, OBD_BRW_READ,
2628                                               &loi->loi_read_lop);
2629                         if (rc < 0)
2630                                 break;
2631                         if (rc > 0)
2632                                 race_counter = 0;
2633                         else
2634                                 race_counter++;
2635                 }
2636
2637                 /* attempt some inter-object balancing by issueing rpcs
2638                  * for each object in turn */
2639                 if (!list_empty(&loi->loi_hp_ready_item))
2640                         list_del_init(&loi->loi_hp_ready_item);
2641                 if (!list_empty(&loi->loi_ready_item))
2642                         list_del_init(&loi->loi_ready_item);
2643                 if (!list_empty(&loi->loi_write_item))
2644                         list_del_init(&loi->loi_write_item);
2645                 if (!list_empty(&loi->loi_read_item))
2646                         list_del_init(&loi->loi_read_item);
2647
2648                 loi_list_maint(cli, loi);
2649
2650                 /* send_oap_rpc fails with 0 when make_ready tells it to
2651                  * back off.  llite's make_ready does this when it tries
2652                  * to lock a page queued for write that is already locked.
2653                  * we want to try sending rpcs from many objects, but we
2654                  * don't want to spin failing with 0.  */
2655                 if (race_counter == 10)
2656                         break;
2657         }
2658         EXIT;
2659 }
2660
2661 /* we're trying to queue a page in the osc so we're subject to the
2662  * 'cl_dirty_max' limit on the number of pages that can be queued in the osc.
2663  * If the osc's queued pages are already at that limit, then we want to sleep
2664  * until there is space in the osc's queue for us.  We also may be waiting for
2665  * write credits from the OST if there are RPCs in flight that may return some
2666  * before we fall back to sync writes.
2667  *
2668  * We need this know our allocation was granted in the presence of signals */
2669 static int ocw_granted(struct client_obd *cli, struct osc_cache_waiter *ocw)
2670 {
2671         int rc;
2672         ENTRY;
2673         client_obd_list_lock(&cli->cl_loi_list_lock);
2674         rc = list_empty(&ocw->ocw_entry) || rpcs_in_flight(cli) == 0;
2675         client_obd_list_unlock(&cli->cl_loi_list_lock);
2676         RETURN(rc);
2677 };
2678
2679 /**
2680  * Non-blocking version of osc_enter_cache() that consumes grant only when it
2681  * is available.
2682  */
2683 int osc_enter_cache_try(const struct lu_env *env,
2684                         struct client_obd *cli, struct lov_oinfo *loi,
2685                         struct osc_async_page *oap, int transient)
2686 {
2687         int has_grant;
2688
2689         has_grant = cli->cl_avail_grant >= CFS_PAGE_SIZE;
2690         if (has_grant) {
2691                 osc_consume_write_grant(cli, &oap->oap_brw_page);
2692                 if (transient) {
2693                         cli->cl_dirty_transit += CFS_PAGE_SIZE;
2694                         atomic_inc(&obd_dirty_transit_pages);
2695                         oap->oap_brw_flags |= OBD_BRW_NOCACHE;
2696                 }
2697         }
2698         return has_grant;
2699 }
2700
2701 /* Caller must hold loi_list_lock - we drop/regain it if we need to wait for
2702  * grant or cache space. */
2703 static int osc_enter_cache(const struct lu_env *env,
2704                            struct client_obd *cli, struct lov_oinfo *loi,
2705                            struct osc_async_page *oap)
2706 {
2707         struct osc_cache_waiter ocw;
2708         struct l_wait_info lwi = { 0 };
2709
2710         ENTRY;
2711
2712         CDEBUG(D_CACHE, "dirty: %ld/%d dirty_max: %ld/%d dropped: %lu "
2713                "grant: %lu\n", cli->cl_dirty, atomic_read(&obd_dirty_pages),
2714                cli->cl_dirty_max, obd_max_dirty_pages,
2715                cli->cl_lost_grant, cli->cl_avail_grant);
2716
2717         /* force the caller to try sync io.  this can jump the list
2718          * of queued writes and create a discontiguous rpc stream */
2719         if (cli->cl_dirty_max < CFS_PAGE_SIZE || cli->cl_ar.ar_force_sync ||
2720             loi->loi_ar.ar_force_sync)
2721                 RETURN(-EDQUOT);
2722
2723         /* Hopefully normal case - cache space and write credits available */
2724         if (cli->cl_dirty + CFS_PAGE_SIZE <= cli->cl_dirty_max &&
2725             atomic_read(&obd_dirty_pages) + 1 <= obd_max_dirty_pages &&
2726             osc_enter_cache_try(env, cli, loi, oap, 0))
2727                 RETURN(0);
2728
2729         /* Make sure that there are write rpcs in flight to wait for.  This
2730          * is a little silly as this object may not have any pending but
2731          * other objects sure might. */
2732         if (cli->cl_w_in_flight) {
2733                 list_add_tail(&ocw.ocw_entry, &cli->cl_cache_waiters);
2734                 cfs_waitq_init(&ocw.ocw_waitq);
2735                 ocw.ocw_oap = oap;
2736                 ocw.ocw_rc = 0;
2737
2738                 loi_list_maint(cli, loi);
2739                 osc_check_rpcs(env, cli);
2740                 client_obd_list_unlock(&cli->cl_loi_list_lock);
2741
2742                 CDEBUG(D_CACHE, "sleeping for cache space\n");
2743                 l_wait_event(ocw.ocw_waitq, ocw_granted(cli, &ocw), &lwi);
2744
2745                 client_obd_list_lock(&cli->cl_loi_list_lock);
2746                 if (!list_empty(&ocw.ocw_entry)) {
2747                         list_del(&ocw.ocw_entry);
2748                         RETURN(-EINTR);
2749                 }
2750                 RETURN(ocw.ocw_rc);
2751         }
2752
2753         RETURN(-EDQUOT);
2754 }
2755
2756
2757 int osc_prep_async_page(struct obd_export *exp, struct lov_stripe_md *lsm,
2758                         struct lov_oinfo *loi, cfs_page_t *page,
2759                         obd_off offset, const struct obd_async_page_ops *ops,
2760                         void *data, void **res, int nocache,
2761                         struct lustre_handle *lockh)
2762 {
2763         struct osc_async_page *oap;
2764
2765         ENTRY;
2766
2767         if (!page)
2768                 return size_round(sizeof(*oap));
2769
2770         oap = *res;
2771         oap->oap_magic = OAP_MAGIC;
2772         oap->oap_cli = &exp->exp_obd->u.cli;
2773         oap->oap_loi = loi;
2774
2775         oap->oap_caller_ops = ops;
2776         oap->oap_caller_data = data;
2777
2778         oap->oap_page = page;
2779         oap->oap_obj_off = offset;
2780         if (!client_is_remote(exp) &&
2781             cfs_capable(CFS_CAP_SYS_RESOURCE))
2782                 oap->oap_brw_flags = OBD_BRW_NOQUOTA;
2783
2784         LASSERT(!(offset & ~CFS_PAGE_MASK));
2785
2786         CFS_INIT_LIST_HEAD(&oap->oap_pending_item);
2787         CFS_INIT_LIST_HEAD(&oap->oap_urgent_item);
2788         CFS_INIT_LIST_HEAD(&oap->oap_rpc_item);
2789         CFS_INIT_LIST_HEAD(&oap->oap_page_list);
2790
2791         spin_lock_init(&oap->oap_lock);
2792         CDEBUG(D_CACHE, "oap %p page %p obj off "LPU64"\n", oap, page, offset);
2793         RETURN(0);
2794 }
2795
2796 struct osc_async_page *oap_from_cookie(void *cookie)
2797 {
2798         struct osc_async_page *oap = cookie;
2799         if (oap->oap_magic != OAP_MAGIC)
2800                 return ERR_PTR(-EINVAL);
2801         return oap;
2802 };
2803
2804 int osc_queue_async_io(const struct lu_env *env,
2805                        struct obd_export *exp, struct lov_stripe_md *lsm,
2806                        struct lov_oinfo *loi, void *cookie,
2807                        int cmd, obd_off off, int count,
2808                        obd_flag brw_flags, enum async_flags async_flags)
2809 {
2810         struct client_obd *cli = &exp->exp_obd->u.cli;
2811         struct osc_async_page *oap;
2812         int rc = 0;
2813         ENTRY;
2814
2815         oap = oap_from_cookie(cookie);
2816         if (IS_ERR(oap))
2817                 RETURN(PTR_ERR(oap));
2818
2819         if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
2820                 RETURN(-EIO);
2821
2822         if (!list_empty(&oap->oap_pending_item) ||
2823             !list_empty(&oap->oap_urgent_item) ||
2824             !list_empty(&oap->oap_rpc_item))
2825                 RETURN(-EBUSY);
2826
2827         /* check if the file's owner/group is over quota */
2828         if ((cmd & OBD_BRW_WRITE) && !(cmd & OBD_BRW_NOQUOTA)) {
2829                 struct cl_object *obj;
2830                 struct cl_attr    attr; /* XXX put attr into thread info */
2831                 unsigned int qid[MAXQUOTAS];
2832
2833                 obj = cl_object_top(osc_oap2cl_page(oap)->cp_obj);
2834
2835                 cl_object_attr_lock(obj);
2836                 rc = cl_object_attr_get(env, obj, &attr);
2837                 cl_object_attr_unlock(obj);
2838
2839                 qid[USRQUOTA] = attr.cat_uid;
2840                 qid[GRPQUOTA] = attr.cat_gid;
2841                 if (rc == 0 &&
2842                     lquota_chkdq(quota_interface, cli, qid) == NO_QUOTA)
2843                         rc = -EDQUOT;
2844                 if (rc)
2845                         RETURN(rc);
2846         }
2847
2848         if (loi == NULL)
2849                 loi = lsm->lsm_oinfo[0];
2850
2851         client_obd_list_lock(&cli->cl_loi_list_lock);
2852
2853         LASSERT(off + count <= CFS_PAGE_SIZE);
2854         oap->oap_cmd = cmd;
2855         oap->oap_page_off = off;
2856         oap->oap_count = count;
2857         oap->oap_brw_flags = brw_flags;
2858         oap->oap_async_flags = async_flags;
2859
2860         if (cmd & OBD_BRW_WRITE) {
2861                 rc = osc_enter_cache(env, cli, loi, oap);
2862                 if (rc) {
2863                         client_obd_list_unlock(&cli->cl_loi_list_lock);
2864                         RETURN(rc);
2865                 }
2866         }
2867
2868         osc_oap_to_pending(oap);
2869         loi_list_maint(cli, loi);
2870
2871         LOI_DEBUG(loi, "oap %p page %p added for cmd %d\n", oap, oap->oap_page,
2872                   cmd);
2873
2874         osc_check_rpcs(env, cli);
2875         client_obd_list_unlock(&cli->cl_loi_list_lock);
2876
2877         RETURN(0);
2878 }
2879
2880 /* aka (~was & now & flag), but this is more clear :) */
2881 #define SETTING(was, now, flag) (!(was & flag) && (now & flag))
2882
2883 int osc_set_async_flags_base(struct client_obd *cli,
2884                              struct lov_oinfo *loi, struct osc_async_page *oap,
2885                              obd_flag async_flags)
2886 {
2887         struct loi_oap_pages *lop;
2888         ENTRY;
2889
2890         LASSERT(!list_empty(&oap->oap_pending_item));
2891
2892         if (oap->oap_cmd & OBD_BRW_WRITE) {
2893                 lop = &loi->loi_write_lop;
2894         } else {
2895                 lop = &loi->loi_read_lop;
2896         }
2897
2898         if ((oap->oap_async_flags & async_flags) == async_flags)
2899                 RETURN(0);
2900
2901         if (SETTING(oap->oap_async_flags, async_flags, ASYNC_READY))
2902                 oap->oap_async_flags |= ASYNC_READY;
2903
2904         if (SETTING(oap->oap_async_flags, async_flags, ASYNC_URGENT) &&
2905             list_empty(&oap->oap_rpc_item)) {
2906                 if (oap->oap_async_flags & ASYNC_HP)
2907                         list_add(&oap->oap_urgent_item, &lop->lop_urgent);
2908                 else
2909                         list_add_tail(&oap->oap_urgent_item, &lop->lop_urgent);
2910                 oap->oap_async_flags |= ASYNC_URGENT;
2911                 loi_list_maint(cli, loi);
2912         }
2913
2914         LOI_DEBUG(loi, "oap %p page %p has flags %x\n", oap, oap->oap_page,
2915                         oap->oap_async_flags);
2916         RETURN(0);
2917 }
2918
2919 int osc_teardown_async_page(struct obd_export *exp,
2920                             struct lov_stripe_md *lsm,
2921                             struct lov_oinfo *loi, void *cookie)
2922 {
2923         struct client_obd *cli = &exp->exp_obd->u.cli;
2924         struct loi_oap_pages *lop;
2925         struct osc_async_page *oap;
2926         int rc = 0;
2927         ENTRY;
2928
2929         oap = oap_from_cookie(cookie);
2930         if (IS_ERR(oap))
2931                 RETURN(PTR_ERR(oap));
2932
2933         if (loi == NULL)
2934                 loi = lsm->lsm_oinfo[0];
2935
2936         if (oap->oap_cmd & OBD_BRW_WRITE) {
2937                 lop = &loi->loi_write_lop;
2938         } else {
2939                 lop = &loi->loi_read_lop;
2940         }
2941
2942         client_obd_list_lock(&cli->cl_loi_list_lock);
2943
2944         if (!list_empty(&oap->oap_rpc_item))
2945                 GOTO(out, rc = -EBUSY);
2946
2947         osc_exit_cache(cli, oap, 0);
2948         osc_wake_cache_waiters(cli);
2949
2950         if (!list_empty(&oap->oap_urgent_item)) {
2951                 list_del_init(&oap->oap_urgent_item);
2952                 oap->oap_async_flags &= ~(ASYNC_URGENT | ASYNC_HP);
2953         }
2954         if (!list_empty(&oap->oap_pending_item)) {
2955                 list_del_init(&oap->oap_pending_item);
2956                 lop_update_pending(cli, lop, oap->oap_cmd, -1);
2957         }
2958         loi_list_maint(cli, loi);
2959         LOI_DEBUG(loi, "oap %p page %p torn down\n", oap, oap->oap_page);
2960 out:
2961         client_obd_list_unlock(&cli->cl_loi_list_lock);
2962         RETURN(rc);
2963 }
2964
2965 static void osc_set_lock_data_with_check(struct ldlm_lock *lock,
2966                                          struct ldlm_enqueue_info *einfo,
2967                                          int flags)
2968 {
2969         void *data = einfo->ei_cbdata;
2970
2971         LASSERT(lock != NULL);
2972         LASSERT(lock->l_blocking_ast == einfo->ei_cb_bl);
2973         LASSERT(lock->l_resource->lr_type == einfo->ei_type);
2974         LASSERT(lock->l_completion_ast == einfo->ei_cb_cp);
2975         LASSERT(lock->l_glimpse_ast == einfo->ei_cb_gl);
2976
2977         lock_res_and_lock(lock);
2978         spin_lock(&osc_ast_guard);
2979         LASSERT(lock->l_ast_data == NULL || lock->l_ast_data == data);
2980         lock->l_ast_data = data;
2981         spin_unlock(&osc_ast_guard);
2982         unlock_res_and_lock(lock);
2983 }
2984
2985 static void osc_set_data_with_check(struct lustre_handle *lockh,
2986                                     struct ldlm_enqueue_info *einfo,
2987                                     int flags)
2988 {
2989         struct ldlm_lock *lock = ldlm_handle2lock(lockh);
2990
2991         if (lock != NULL) {
2992                 osc_set_lock_data_with_check(lock, einfo, flags);
2993                 LDLM_LOCK_PUT(lock);
2994         } else
2995                 CERROR("lockh %p, data %p - client evicted?\n",
2996                        lockh, einfo->ei_cbdata);
2997 }
2998
2999 static int osc_change_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
3000                              ldlm_iterator_t replace, void *data)
3001 {
3002         struct ldlm_res_id res_id;
3003         struct obd_device *obd = class_exp2obd(exp);
3004
3005         osc_build_res_name(lsm->lsm_object_id, lsm->lsm_object_gr, &res_id);
3006         ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
3007         return 0;
3008 }
3009
3010 static int osc_enqueue_fini(struct ptlrpc_request *req, struct ost_lvb *lvb,
3011                             obd_enqueue_update_f upcall, void *cookie,
3012                             int *flags, int rc)
3013 {
3014         int intent = *flags & LDLM_FL_HAS_INTENT;
3015         ENTRY;
3016
3017         if (intent) {
3018                 /* The request was created before ldlm_cli_enqueue call. */
3019                 if (rc == ELDLM_LOCK_ABORTED) {
3020                         struct ldlm_reply *rep;
3021                         rep = req_capsule_server_get(&req->rq_pill,
3022                                                      &RMF_DLM_REP);
3023
3024                         LASSERT(rep != NULL);
3025                         if (rep->lock_policy_res1)
3026                                 rc = rep->lock_policy_res1;
3027                 }
3028         }
3029
3030         if ((intent && rc == ELDLM_LOCK_ABORTED) || !rc) {
3031                 *flags |= LDLM_FL_LVB_READY;
3032                 CDEBUG(D_INODE,"got kms "LPU64" blocks "LPU64" mtime "LPU64"\n",
3033                        lvb->lvb_size, lvb->lvb_blocks, lvb->lvb_mtime);
3034         }
3035
3036         /* Call the update callback. */
3037         rc = (*upcall)(cookie, rc);
3038         RETURN(rc);
3039 }
3040
3041 static int osc_enqueue_interpret(const struct lu_env *env,
3042                                  struct ptlrpc_request *req,
3043                                  struct osc_enqueue_args *aa, int rc)
3044 {
3045         struct ldlm_lock *lock;
3046         struct lustre_handle handle;
3047         __u32 mode;
3048
3049         /* Make a local copy of a lock handle and a mode, because aa->oa_*
3050          * might be freed anytime after lock upcall has been called. */
3051         lustre_handle_copy(&handle, aa->oa_lockh);
3052         mode = aa->oa_ei->ei_mode;
3053
3054         /* ldlm_cli_enqueue is holding a reference on the lock, so it must
3055          * be valid. */
3056         lock = ldlm_handle2lock(&handle);
3057
3058         /* Take an additional reference so that a blocking AST that
3059          * ldlm_cli_enqueue_fini() might post for a failed lock, is guaranteed
3060          * to arrive after an upcall has been executed by
3061          * osc_enqueue_fini(). */
3062         ldlm_lock_addref(&handle, mode);
3063
3064         /* Complete obtaining the lock procedure. */
3065         rc = ldlm_cli_enqueue_fini(aa->oa_exp, req, aa->oa_ei->ei_type, 1,
3066                                    mode, aa->oa_flags, aa->oa_lvb,
3067                                    sizeof(*aa->oa_lvb), lustre_swab_ost_lvb,
3068                                    &handle, rc);
3069         /* Complete osc stuff. */
3070         rc = osc_enqueue_fini(req, aa->oa_lvb,
3071                               aa->oa_upcall, aa->oa_cookie, aa->oa_flags, rc);
3072
3073         OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_CANCEL_RACE, 10);
3074
3075         /* Release the lock for async request. */
3076         if (lustre_handle_is_used(&handle) && rc == ELDLM_OK)
3077                 /*
3078                  * Releases a reference taken by ldlm_cli_enqueue(), if it is
3079                  * not already released by
3080                  * ldlm_cli_enqueue_fini()->failed_lock_cleanup()
3081                  */
3082                 ldlm_lock_decref(&handle, mode);
3083
3084         LASSERTF(lock != NULL, "lockh %p, req %p, aa %p - client evicted?\n",
3085                  aa->oa_lockh, req, aa);
3086         ldlm_lock_decref(&handle, mode);
3087         LDLM_LOCK_PUT(lock);
3088         return rc;
3089 }
3090
3091 void osc_update_enqueue(struct lustre_handle *lov_lockhp,
3092                         struct lov_oinfo *loi, int flags,
3093                         struct ost_lvb *lvb, __u32 mode, int rc)
3094 {
3095         if (rc == ELDLM_OK) {
3096                 struct ldlm_lock *lock = ldlm_handle2lock(lov_lockhp);
3097                 __u64 tmp;
3098
3099                 LASSERT(lock != NULL);
3100                 loi->loi_lvb = *lvb;
3101                 tmp = loi->loi_lvb.lvb_size;
3102                 /* Extend KMS up to the end of this lock and no further
3103                  * A lock on [x,y] means a KMS of up to y + 1 bytes! */
3104                 if (tmp > lock->l_policy_data.l_extent.end)
3105                         tmp = lock->l_policy_data.l_extent.end + 1;
3106                 if (tmp >= loi->loi_kms) {
3107                         LDLM_DEBUG(lock, "lock acquired, setting rss="LPU64
3108                                    ", kms="LPU64, loi->loi_lvb.lvb_size, tmp);
3109                         loi_kms_set(loi, tmp);
3110                 } else {
3111                         LDLM_DEBUG(lock, "lock acquired, setting rss="
3112                                    LPU64"; leaving kms="LPU64", end="LPU64,
3113                                    loi->loi_lvb.lvb_size, loi->loi_kms,
3114                                    lock->l_policy_data.l_extent.end);
3115                 }
3116                 ldlm_lock_allow_match(lock);
3117                 LDLM_LOCK_PUT(lock);
3118         } else if (rc == ELDLM_LOCK_ABORTED && (flags & LDLM_FL_HAS_INTENT)) {
3119                 loi->loi_lvb = *lvb;
3120                 CDEBUG(D_INODE, "glimpsed, setting rss="LPU64"; leaving"
3121                        " kms="LPU64"\n", loi->loi_lvb.lvb_size, loi->loi_kms);
3122                 rc = ELDLM_OK;
3123         }
3124 }
3125 EXPORT_SYMBOL(osc_update_enqueue);
3126
3127 struct ptlrpc_request_set *PTLRPCD_SET = (void *)1;
3128
3129 /* When enqueuing asynchronously, locks are not ordered, we can obtain a lock
3130  * from the 2nd OSC before a lock from the 1st one. This does not deadlock with
3131  * other synchronous requests, however keeping some locks and trying to obtain
3132  * others may take a considerable amount of time in a case of ost failure; and
3133  * when other sync requests do not get released lock from a client, the client
3134  * is excluded from the cluster -- such scenarious make the life difficult, so
3135  * release locks just after they are obtained. */
3136 int osc_enqueue_base(struct obd_export *exp, struct ldlm_res_id *res_id,
3137                      int *flags, ldlm_policy_data_t *policy,
3138                      struct ost_lvb *lvb, int kms_valid,
3139                      obd_enqueue_update_f upcall, void *cookie,
3140                      struct ldlm_enqueue_info *einfo,
3141                      struct lustre_handle *lockh,
3142                      struct ptlrpc_request_set *rqset, int async)
3143 {
3144         struct obd_device *obd = exp->exp_obd;
3145         struct ptlrpc_request *req = NULL;
3146         int intent = *flags & LDLM_FL_HAS_INTENT;
3147         ldlm_mode_t mode;
3148         int rc;
3149         ENTRY;
3150
3151         /* Filesystem lock extents are extended to page boundaries so that
3152          * dealing with the page cache is a little smoother.  */
3153         policy->l_extent.start -= policy->l_extent.start & ~CFS_PAGE_MASK;
3154         policy->l_extent.end |= ~CFS_PAGE_MASK;
3155
3156         /*
3157          * kms is not valid when either object is completely fresh (so that no
3158          * locks are cached), or object was evicted. In the latter case cached
3159          * lock cannot be used, because it would prime inode state with
3160          * potentially stale LVB.
3161          */
3162         if (!kms_valid)
3163                 goto no_match;
3164
3165         /* Next, search for already existing extent locks that will cover us */
3166         /* If we're trying to read, we also search for an existing PW lock.  The
3167          * VFS and page cache already protect us locally, so lots of readers/
3168          * writers can share a single PW lock.
3169          *
3170          * There are problems with conversion deadlocks, so instead of
3171          * converting a read lock to a write lock, we'll just enqueue a new
3172          * one.
3173          *
3174          * At some point we should cancel the read lock instead of making them
3175          * send us a blocking callback, but there are problems with canceling
3176          * locks out from other users right now, too. */
3177         mode = einfo->ei_mode;
3178         if (einfo->ei_mode == LCK_PR)
3179                 mode |= LCK_PW;