drivers/staging/lustre/lustre/osc/osc_cache.c

   1 /*
   2  * GPL HEADER START
   3  *
   4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 only,
   8  * as published by the Free Software Foundation.
   9  *
  10  * This program is distributed in the hope that it will be useful, but
  11  * WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13  * General Public License version 2 for more details (a copy is included
  14  * in the LICENSE file that accompanied this code).
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * version 2 along with this program; If not, see
  18  * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
  19  *
  20  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
  21  * CA 95054 USA or visit www.sun.com if you need additional information or
  22  * have any questions.
  23  *
  24  * GPL HEADER END
  25  */
  26 /*
  27  * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
  28  * Use is subject to license terms.
  29  *
  30  * Copyright (c) 2012, 2015, Intel Corporation.
  31  *
  32  */
  33 /*
  34  * This file is part of Lustre, http://www.lustre.org/
  35  * Lustre is a trademark of Sun Microsystems, Inc.
  36  *
  37  * osc cache management.
  38  *
  39  * Author: Jinshan Xiong <jinshan.xiong@whamcloud.com>
  40  */
  41
  42 #define DEBUG_SUBSYSTEM S_OSC
  43
  44 #include "osc_cl_internal.h"
  45 #include "osc_internal.h"
  46
  47 static int extent_debug; /* set it to be true for more debug */
  48
  49 static void osc_update_pending(struct osc_object *obj, int cmd, int delta);
  50 static int osc_extent_wait(const struct lu_env *env, struct osc_extent *ext,
  51                            int state);
  52 static void osc_ap_completion(const struct lu_env *env, struct client_obd *cli,
  53                               struct osc_async_page *oap, int sent, int rc);
  54 static int osc_make_ready(const struct lu_env *env, struct osc_async_page *oap,
  55                           int cmd);
  56 static int osc_refresh_count(const struct lu_env *env,
  57                              struct osc_async_page *oap, int cmd);
  58 static int osc_io_unplug_async(const struct lu_env *env,
  59                                struct client_obd *cli, struct osc_object *osc);
  60 static void osc_free_grant(struct client_obd *cli, unsigned int nr_pages,
  61                            unsigned int lost_grant);
  62
  63 static void osc_extent_tree_dump0(int level, struct osc_object *obj,
  64                                   const char *func, int line);
  65 #define osc_extent_tree_dump(lvl, obj) \
  66         osc_extent_tree_dump0(lvl, obj, __func__, __LINE__)
  67
  68 /** \addtogroup osc
  69  *  @{
  70  */
  71
  72 /* ------------------ osc extent ------------------ */
  73 static inline char *ext_flags(struct osc_extent *ext, char *flags)
  74 {
  75         char *buf = flags;
  76         *buf++ = ext->oe_rw ? 'r' : 'w';
  77         if (ext->oe_intree)
  78                 *buf++ = 'i';
  79         if (ext->oe_srvlock)
  80                 *buf++ = 's';
  81         if (ext->oe_hp)
  82                 *buf++ = 'h';
  83         if (ext->oe_urgent)
  84                 *buf++ = 'u';
  85         if (ext->oe_memalloc)
  86                 *buf++ = 'm';
  87         if (ext->oe_trunc_pending)
  88                 *buf++ = 't';
  89         if (ext->oe_fsync_wait)
  90                 *buf++ = 'Y';
  91         *buf = 0;
  92         return flags;
  93 }
  94
  95 static inline char list_empty_marker(struct list_head *list)
  96 {
  97         return list_empty(list) ? '-' : '+';
  98 }
  99
 100 #define EXTSTR       "[%lu -> %lu/%lu]"
 101 #define EXTPARA(ext) (ext)->oe_start, (ext)->oe_end, (ext)->oe_max_end
 102 static const char *oes_strings[] = {
 103         "inv", "active", "cache", "locking", "lockdone", "rpc", "trunc", NULL };
 104
 105 #define OSC_EXTENT_DUMP(lvl, extent, fmt, ...) do {                           \
 106         struct osc_extent *__ext = (extent);                                  \
 107         char __buf[16];                                                       \
 108                                                                               \
 109         CDEBUG(lvl,                                                           \
 110                 "extent %p@{" EXTSTR ", "                                     \
 111                 "[%d|%d|%c|%s|%s|%p], [%d|%d|%c|%c|%p|%u|%p]} " fmt,          \
 112                 /* ----- extent part 0 ----- */                               \
 113                 __ext, EXTPARA(__ext),                                        \
 114                 /* ----- part 1 ----- */                                      \
 115                 atomic_read(&__ext->oe_refc),                                 \
 116                 atomic_read(&__ext->oe_users),                                \
 117                 list_empty_marker(&__ext->oe_link),                           \
 118                 oes_strings[__ext->oe_state], ext_flags(__ext, __buf),        \
 119                 __ext->oe_obj,                                                \
 120                 /* ----- part 2 ----- */                                      \
 121                 __ext->oe_grants, __ext->oe_nr_pages,                         \
 122                 list_empty_marker(&__ext->oe_pages),                          \
 123                 waitqueue_active(&__ext->oe_waitq) ? '+' : '-',               \
 124                 __ext->oe_osclock, __ext->oe_mppr, __ext->oe_owner,           \
 125                 /* ----- part 4 ----- */                                      \
 126                 ## __VA_ARGS__);                                              \
 127 } while (0)
 128
 129 #undef EASSERTF
 130 #define EASSERTF(expr, ext, fmt, args...) do {                          \
 131         if (!(expr)) {                                                  \
 132                 OSC_EXTENT_DUMP(D_ERROR, (ext), fmt, ##args);           \
 133                 osc_extent_tree_dump(D_ERROR, (ext)->oe_obj);           \
 134                 LASSERT(expr);                                          \
 135         }                                                               \
 136 } while (0)
 137
 138 #undef EASSERT
 139 #define EASSERT(expr, ext) EASSERTF(expr, ext, "\n")
 140
 141 static inline struct osc_extent *rb_extent(struct rb_node *n)
 142 {
 143         if (!n)
 144                 return NULL;
 145
 146         return container_of(n, struct osc_extent, oe_node);
 147 }
 148
 149 static inline struct osc_extent *next_extent(struct osc_extent *ext)
 150 {
 151         if (!ext)
 152                 return NULL;
 153
 154         LASSERT(ext->oe_intree);
 155         return rb_extent(rb_next(&ext->oe_node));
 156 }
 157
 158 static inline struct osc_extent *prev_extent(struct osc_extent *ext)
 159 {
 160         if (!ext)
 161                 return NULL;
 162
 163         LASSERT(ext->oe_intree);
 164         return rb_extent(rb_prev(&ext->oe_node));
 165 }
 166
 167 static inline struct osc_extent *first_extent(struct osc_object *obj)
 168 {
 169         return rb_extent(rb_first(&obj->oo_root));
 170 }
 171
 172 /* object must be locked by caller. */
 173 static int osc_extent_sanity_check0(struct osc_extent *ext,
 174                                     const char *func, const int line)
 175 {
 176         struct osc_object *obj = ext->oe_obj;
 177         struct osc_async_page *oap;
 178         int page_count;
 179         int rc = 0;
 180
 181         if (!osc_object_is_locked(obj)) {
 182                 rc = 9;
 183                 goto out;
 184         }
 185
 186         if (ext->oe_state >= OES_STATE_MAX) {
 187                 rc = 10;
 188                 goto out;
 189         }
 190
 191         if (atomic_read(&ext->oe_refc) <= 0) {
 192                 rc = 20;
 193                 goto out;
 194         }
 195
 196         if (atomic_read(&ext->oe_refc) < atomic_read(&ext->oe_users)) {
 197                 rc = 30;
 198                 goto out;
 199         }
 200
 201         switch (ext->oe_state) {
 202         case OES_INV:
 203                 if (ext->oe_nr_pages > 0 || !list_empty(&ext->oe_pages))
 204                         rc = 35;
 205                 else
 206                         rc = 0;
 207                 goto out;
 208         case OES_ACTIVE:
 209                 if (atomic_read(&ext->oe_users) == 0) {
 210                         rc = 40;
 211                         goto out;
 212                 }
 213                 if (ext->oe_hp) {
 214                         rc = 50;
 215                         goto out;
 216                 }
 217                 if (ext->oe_fsync_wait && !ext->oe_urgent) {
 218                         rc = 55;
 219                         goto out;
 220                 }
 221                 break;
 222         case OES_CACHE:
 223                 if (ext->oe_grants == 0) {
 224                         rc = 60;
 225                         goto out;
 226                 }
 227                 if (ext->oe_fsync_wait && !ext->oe_urgent && !ext->oe_hp) {
 228                         rc = 65;
 229                         goto out;
 230                 }
 231         default:
 232                 if (atomic_read(&ext->oe_users) > 0) {
 233                         rc = 70;
 234                         goto out;
 235                 }
 236         }
 237
 238         if (ext->oe_max_end < ext->oe_end || ext->oe_end < ext->oe_start) {
 239                 rc = 80;
 240                 goto out;
 241         }
 242
 243         if (!ext->oe_osclock && ext->oe_grants > 0) {
 244                 rc = 90;
 245                 goto out;
 246         }
 247
 248         if (ext->oe_osclock) {
 249                 struct cl_lock_descr *descr;
 250
 251                 descr = &ext->oe_osclock->cll_descr;
 252                 if (!(descr->cld_start <= ext->oe_start &&
 253                       descr->cld_end >= ext->oe_max_end)) {
 254                         rc = 100;
 255                         goto out;
 256                 }
 257         }
 258
 259         if (ext->oe_nr_pages > ext->oe_mppr) {
 260                 rc = 105;
 261                 goto out;
 262         }
 263
 264         /* Do not verify page list if extent is in RPC. This is because an
 265          * in-RPC extent is supposed to be exclusively accessible w/o lock.
 266          */
 267         if (ext->oe_state > OES_CACHE) {
 268                 rc = 0;
 269                 goto out;
 270         }
 271
 272         if (!extent_debug) {
 273                 rc = 0;
 274                 goto out;
 275         }
 276
 277         page_count = 0;
 278         list_for_each_entry(oap, &ext->oe_pages, oap_pending_item) {
 279                 pgoff_t index = oap2cl_page(oap)->cp_index;
 280                 ++page_count;
 281                 if (index > ext->oe_end || index < ext->oe_start) {
 282                         rc = 110;
 283                         goto out;
 284                 }
 285         }
 286         if (page_count != ext->oe_nr_pages) {
 287                 rc = 120;
 288                 goto out;
 289         }
 290
 291 out:
 292         if (rc != 0)
 293                 OSC_EXTENT_DUMP(D_ERROR, ext,
 294                                 "%s:%d sanity check %p failed with rc = %d\n",
 295                                 func, line, ext, rc);
 296         return rc;
 297 }
 298
 299 #define sanity_check_nolock(ext) \
 300         osc_extent_sanity_check0(ext, __func__, __LINE__)
 301
 302 #define sanity_check(ext) ({                                            \
 303         int __res;                                                      \
 304         osc_object_lock((ext)->oe_obj);                                 \
 305         __res = sanity_check_nolock(ext);                               \
 306         osc_object_unlock((ext)->oe_obj);                               \
 307         __res;                                                          \
 308 })
 309
 310 /**
 311  * sanity check - to make sure there is no overlapped extent in the tree.
 312  */
 313 static int osc_extent_is_overlapped(struct osc_object *obj,
 314                                     struct osc_extent *ext)
 315 {
 316         struct osc_extent *tmp;
 317
 318         LASSERT(osc_object_is_locked(obj));
 319
 320         if (!extent_debug)
 321                 return 0;
 322
 323         for (tmp = first_extent(obj); tmp; tmp = next_extent(tmp)) {
 324                 if (tmp == ext)
 325                         continue;
 326                 if (tmp->oe_end >= ext->oe_start &&
 327                     tmp->oe_start <= ext->oe_end)
 328                         return 1;
 329         }
 330         return 0;
 331 }
 332
 333 static void osc_extent_state_set(struct osc_extent *ext, int state)
 334 {
 335         LASSERT(osc_object_is_locked(ext->oe_obj));
 336         LASSERT(state >= OES_INV && state < OES_STATE_MAX);
 337
 338         /* Never try to sanity check a state changing extent :-) */
 339         /* LASSERT(sanity_check_nolock(ext) == 0); */
 340
 341         /* TODO: validate the state machine */
 342         ext->oe_state = state;
 343         wake_up_all(&ext->oe_waitq);
 344 }
 345
 346 static struct osc_extent *osc_extent_alloc(struct osc_object *obj)
 347 {
 348         struct osc_extent *ext;
 349
 350         ext = kmem_cache_zalloc(osc_extent_kmem, GFP_NOFS);
 351         if (!ext)
 352                 return NULL;
 353
 354         RB_CLEAR_NODE(&ext->oe_node);
 355         ext->oe_obj = obj;
 356         atomic_set(&ext->oe_refc, 1);
 357         atomic_set(&ext->oe_users, 0);
 358         INIT_LIST_HEAD(&ext->oe_link);
 359         ext->oe_state = OES_INV;
 360         INIT_LIST_HEAD(&ext->oe_pages);
 361         init_waitqueue_head(&ext->oe_waitq);
 362         ext->oe_osclock = NULL;
 363
 364         return ext;
 365 }
 366
 367 static void osc_extent_free(struct osc_extent *ext)
 368 {
 369         kmem_cache_free(osc_extent_kmem, ext);
 370 }
 371
 372 static struct osc_extent *osc_extent_get(struct osc_extent *ext)
 373 {
 374         LASSERT(atomic_read(&ext->oe_refc) >= 0);
 375         atomic_inc(&ext->oe_refc);
 376         return ext;
 377 }
 378
 379 static void osc_extent_put(const struct lu_env *env, struct osc_extent *ext)
 380 {
 381         LASSERT(atomic_read(&ext->oe_refc) > 0);
 382         if (atomic_dec_and_test(&ext->oe_refc)) {
 383                 LASSERT(list_empty(&ext->oe_link));
 384                 LASSERT(atomic_read(&ext->oe_users) == 0);
 385                 LASSERT(ext->oe_state == OES_INV);
 386                 LASSERT(!ext->oe_intree);
 387
 388                 if (ext->oe_osclock) {
 389                         cl_lock_put(env, ext->oe_osclock);
 390                         ext->oe_osclock = NULL;
 391                 }
 392                 osc_extent_free(ext);
 393         }
 394 }
 395
 396 /**
 397  * osc_extent_put_trust() is a special version of osc_extent_put() when
 398  * it's known that the caller is not the last user. This is to address the
 399  * problem of lacking of lu_env ;-).
 400  */
 401 static void osc_extent_put_trust(struct osc_extent *ext)
 402 {
 403         LASSERT(atomic_read(&ext->oe_refc) > 1);
 404         LASSERT(osc_object_is_locked(ext->oe_obj));
 405         atomic_dec(&ext->oe_refc);
 406 }
 407
 408 /**
 409  * Return the extent which includes pgoff @index, or return the greatest
 410  * previous extent in the tree.
 411  */
 412 static struct osc_extent *osc_extent_search(struct osc_object *obj,
 413                                             pgoff_t index)
 414 {
 415         struct rb_node *n = obj->oo_root.rb_node;
 416         struct osc_extent *tmp, *p = NULL;
 417
 418         LASSERT(osc_object_is_locked(obj));
 419         while (n) {
 420                 tmp = rb_extent(n);
 421                 if (index < tmp->oe_start) {
 422                         n = n->rb_left;
 423                 } else if (index > tmp->oe_end) {
 424                         p = rb_extent(n);
 425                         n = n->rb_right;
 426                 } else {
 427                         return tmp;
 428                 }
 429         }
 430         return p;
 431 }
 432
 433 /*
 434  * Return the extent covering @index, otherwise return NULL.
 435  * caller must have held object lock.
 436  */
 437 static struct osc_extent *osc_extent_lookup(struct osc_object *obj,
 438                                             pgoff_t index)
 439 {
 440         struct osc_extent *ext;
 441
 442         ext = osc_extent_search(obj, index);
 443         if (ext && ext->oe_start <= index && index <= ext->oe_end)
 444                 return osc_extent_get(ext);
 445         return NULL;
 446 }
 447
 448 /* caller must have held object lock. */
 449 static void osc_extent_insert(struct osc_object *obj, struct osc_extent *ext)
 450 {
 451         struct rb_node **n = &obj->oo_root.rb_node;
 452         struct rb_node *parent = NULL;
 453         struct osc_extent *tmp;
 454
 455         LASSERT(ext->oe_intree == 0);
 456         LASSERT(ext->oe_obj == obj);
 457         LASSERT(osc_object_is_locked(obj));
 458         while (*n) {
 459                 tmp = rb_extent(*n);
 460                 parent = *n;
 461
 462                 if (ext->oe_end < tmp->oe_start)
 463                         n = &(*n)->rb_left;
 464                 else if (ext->oe_start > tmp->oe_end)
 465                         n = &(*n)->rb_right;
 466                 else
 467                         EASSERTF(0, tmp, EXTSTR"\n", EXTPARA(ext));
 468         }
 469         rb_link_node(&ext->oe_node, parent, n);
 470         rb_insert_color(&ext->oe_node, &obj->oo_root);
 471         osc_extent_get(ext);
 472         ext->oe_intree = 1;
 473 }
 474
 475 /* caller must have held object lock. */
 476 static void osc_extent_erase(struct osc_extent *ext)
 477 {
 478         struct osc_object *obj = ext->oe_obj;
 479
 480         LASSERT(osc_object_is_locked(obj));
 481         if (ext->oe_intree) {
 482                 rb_erase(&ext->oe_node, &obj->oo_root);
 483                 ext->oe_intree = 0;
 484                 /* rbtree held a refcount */
 485                 osc_extent_put_trust(ext);
 486         }
 487 }
 488
 489 static struct osc_extent *osc_extent_hold(struct osc_extent *ext)
 490 {
 491         struct osc_object *obj = ext->oe_obj;
 492
 493         LASSERT(osc_object_is_locked(obj));
 494         LASSERT(ext->oe_state == OES_ACTIVE || ext->oe_state == OES_CACHE);
 495         if (ext->oe_state == OES_CACHE) {
 496                 osc_extent_state_set(ext, OES_ACTIVE);
 497                 osc_update_pending(obj, OBD_BRW_WRITE, -ext->oe_nr_pages);
 498         }
 499         atomic_inc(&ext->oe_users);
 500         list_del_init(&ext->oe_link);
 501         return osc_extent_get(ext);
 502 }
 503
 504 static void __osc_extent_remove(struct osc_extent *ext)
 505 {
 506         LASSERT(osc_object_is_locked(ext->oe_obj));
 507         LASSERT(list_empty(&ext->oe_pages));
 508         osc_extent_erase(ext);
 509         list_del_init(&ext->oe_link);
 510         osc_extent_state_set(ext, OES_INV);
 511         OSC_EXTENT_DUMP(D_CACHE, ext, "destroyed.\n");
 512 }
 513
 514 static void osc_extent_remove(struct osc_extent *ext)
 515 {
 516         struct osc_object *obj = ext->oe_obj;
 517
 518         osc_object_lock(obj);
 519         __osc_extent_remove(ext);
 520         osc_object_unlock(obj);
 521 }
 522
 523 /**
 524  * This function is used to merge extents to get better performance. It checks
 525  * if @cur and @victim are contiguous at chunk level.
 526  */
 527 static int osc_extent_merge(const struct lu_env *env, struct osc_extent *cur,
 528                             struct osc_extent *victim)
 529 {
 530         struct osc_object *obj = cur->oe_obj;
 531         pgoff_t chunk_start;
 532         pgoff_t chunk_end;
 533         int ppc_bits;
 534
 535         LASSERT(cur->oe_state == OES_CACHE);
 536         LASSERT(osc_object_is_locked(obj));
 537         if (!victim)
 538                 return -EINVAL;
 539
 540         if (victim->oe_state != OES_CACHE || victim->oe_fsync_wait)
 541                 return -EBUSY;
 542
 543         if (cur->oe_max_end != victim->oe_max_end)
 544                 return -ERANGE;
 545
 546         LASSERT(cur->oe_osclock == victim->oe_osclock);
 547         ppc_bits = osc_cli(obj)->cl_chunkbits - PAGE_SHIFT;
 548         chunk_start = cur->oe_start >> ppc_bits;
 549         chunk_end = cur->oe_end >> ppc_bits;
 550         if (chunk_start != (victim->oe_end >> ppc_bits) + 1 &&
 551             chunk_end + 1 != victim->oe_start >> ppc_bits)
 552                 return -ERANGE;
 553
 554         OSC_EXTENT_DUMP(D_CACHE, victim, "will be merged by %p.\n", cur);
 555
 556         cur->oe_start = min(cur->oe_start, victim->oe_start);
 557         cur->oe_end = max(cur->oe_end, victim->oe_end);
 558         cur->oe_grants += victim->oe_grants;
 559         cur->oe_nr_pages += victim->oe_nr_pages;
 560         /* only the following bits are needed to merge */
 561         cur->oe_urgent |= victim->oe_urgent;
 562         cur->oe_memalloc |= victim->oe_memalloc;
 563         list_splice_init(&victim->oe_pages, &cur->oe_pages);
 564         list_del_init(&victim->oe_link);
 565         victim->oe_nr_pages = 0;
 566
 567         osc_extent_get(victim);
 568         __osc_extent_remove(victim);
 569         osc_extent_put(env, victim);
 570
 571         OSC_EXTENT_DUMP(D_CACHE, cur, "after merging %p.\n", victim);
 572         return 0;
 573 }
 574
 575 /**
 576  * Drop user count of osc_extent, and unplug IO asynchronously.
 577  */
 578 void osc_extent_release(const struct lu_env *env, struct osc_extent *ext)
 579 {
 580         struct osc_object *obj = ext->oe_obj;
 581
 582         LASSERT(atomic_read(&ext->oe_users) > 0);
 583         LASSERT(sanity_check(ext) == 0);
 584         LASSERT(ext->oe_grants > 0);
 585
 586         if (atomic_dec_and_lock(&ext->oe_users, &obj->oo_lock)) {
 587                 LASSERT(ext->oe_state == OES_ACTIVE);
 588                 if (ext->oe_trunc_pending) {
 589                         /* a truncate process is waiting for this extent.
 590                          * This may happen due to a race, check
 591                          * osc_cache_truncate_start().
 592                          */
 593                         osc_extent_state_set(ext, OES_TRUNC);
 594                         ext->oe_trunc_pending = 0;
 595                 } else {
 596                         osc_extent_state_set(ext, OES_CACHE);
 597                         osc_update_pending(obj, OBD_BRW_WRITE,
 598                                            ext->oe_nr_pages);
 599
 600                         /* try to merge the previous and next extent. */
 601                         osc_extent_merge(env, ext, prev_extent(ext));
 602                         osc_extent_merge(env, ext, next_extent(ext));
 603
 604                         if (ext->oe_urgent)
 605                                 list_move_tail(&ext->oe_link,
 606                                                &obj->oo_urgent_exts);
 607                 }
 608                 osc_object_unlock(obj);
 609
 610                 osc_io_unplug_async(env, osc_cli(obj), obj);
 611         }
 612         osc_extent_put(env, ext);
 613 }
 614
 615 static inline int overlapped(struct osc_extent *ex1, struct osc_extent *ex2)
 616 {
 617         return !(ex1->oe_end < ex2->oe_start || ex2->oe_end < ex1->oe_start);
 618 }
 619
 620 /**
 621  * Find or create an extent which includes @index, core function to manage
 622  * extent tree.
 623  */
 624 static struct osc_extent *osc_extent_find(const struct lu_env *env,
 625                                           struct osc_object *obj, pgoff_t index,
 626                                           int *grants)
 627
 628 {
 629         struct client_obd *cli = osc_cli(obj);
 630         struct cl_lock *lock;
 631         struct osc_extent *cur;
 632         struct osc_extent *ext;
 633         struct osc_extent *conflict = NULL;
 634         struct osc_extent *found = NULL;
 635         pgoff_t chunk;
 636         pgoff_t max_end;
 637         int max_pages; /* max_pages_per_rpc */
 638         int chunksize;
 639         int ppc_bits; /* pages per chunk bits */
 640         int chunk_mask;
 641         int rc;
 642
 643         cur = osc_extent_alloc(obj);
 644         if (!cur)
 645                 return ERR_PTR(-ENOMEM);
 646
 647         lock = cl_lock_at_pgoff(env, osc2cl(obj), index, NULL, 1, 0);
 648         LASSERT(lock->cll_descr.cld_mode >= CLM_WRITE);
 649
 650         LASSERT(cli->cl_chunkbits >= PAGE_SHIFT);
 651         ppc_bits = cli->cl_chunkbits - PAGE_SHIFT;
 652         chunk_mask = ~((1 << ppc_bits) - 1);
 653         chunksize = 1 << cli->cl_chunkbits;
 654         chunk = index >> ppc_bits;
 655
 656         /* align end to rpc edge, rpc size may not be a power 2 integer. */
 657         max_pages = cli->cl_max_pages_per_rpc;
 658         LASSERT((max_pages & ~chunk_mask) == 0);
 659         max_end = index - (index % max_pages) + max_pages - 1;
 660         max_end = min_t(pgoff_t, max_end, lock->cll_descr.cld_end);
 661
 662         /* initialize new extent by parameters so far */
 663         cur->oe_max_end = max_end;
 664         cur->oe_start = index & chunk_mask;
 665         cur->oe_end = ((index + ~chunk_mask + 1) & chunk_mask) - 1;
 666         if (cur->oe_start < lock->cll_descr.cld_start)
 667                 cur->oe_start = lock->cll_descr.cld_start;
 668         if (cur->oe_end > max_end)
 669                 cur->oe_end = max_end;
 670         cur->oe_osclock = lock;
 671         cur->oe_grants = 0;
 672         cur->oe_mppr = max_pages;
 673
 674         /* grants has been allocated by caller */
 675         LASSERTF(*grants >= chunksize + cli->cl_extent_tax,
 676                  "%u/%u/%u.\n", *grants, chunksize, cli->cl_extent_tax);
 677         LASSERTF((max_end - cur->oe_start) < max_pages, EXTSTR"\n",
 678                  EXTPARA(cur));
 679
 680 restart:
 681         osc_object_lock(obj);
 682         ext = osc_extent_search(obj, cur->oe_start);
 683         if (!ext)
 684                 ext = first_extent(obj);
 685         while (ext) {
 686                 loff_t ext_chk_start = ext->oe_start >> ppc_bits;
 687                 loff_t ext_chk_end = ext->oe_end >> ppc_bits;
 688
 689                 LASSERT(sanity_check_nolock(ext) == 0);
 690                 if (chunk > ext_chk_end + 1)
 691                         break;
 692
 693                 /* if covering by different locks, no chance to match */
 694                 if (lock != ext->oe_osclock) {
 695                         EASSERTF(!overlapped(ext, cur), ext,
 696                                  EXTSTR"\n", EXTPARA(cur));
 697
 698                         ext = next_extent(ext);
 699                         continue;
 700                 }
 701
 702                 /* discontiguous chunks? */
 703                 if (chunk + 1 < ext_chk_start) {
 704                         ext = next_extent(ext);
 705                         continue;
 706                 }
 707
 708                 /* ok, from now on, ext and cur have these attrs:
 709                  * 1. covered by the same lock
 710                  * 2. contiguous at chunk level or overlapping.
 711                  */
 712
 713                 if (overlapped(ext, cur)) {
 714                         /* cur is the minimum unit, so overlapping means
 715                          * full contain.
 716                          */
 717                         EASSERTF((ext->oe_start <= cur->oe_start &&
 718                                   ext->oe_end >= cur->oe_end),
 719                                  ext, EXTSTR"\n", EXTPARA(cur));
 720
 721                         if (ext->oe_state > OES_CACHE || ext->oe_fsync_wait) {
 722                                 /* for simplicity, we wait for this extent to
 723                                  * finish before going forward.
 724                                  */
 725                                 conflict = osc_extent_get(ext);
 726                                 break;
 727                         }
 728
 729                         found = osc_extent_hold(ext);
 730                         break;
 731                 }
 732
 733                 /* non-overlapped extent */
 734                 if (ext->oe_state != OES_CACHE || ext->oe_fsync_wait) {
 735                         /* we can't do anything for a non OES_CACHE extent, or
 736                          * if there is someone waiting for this extent to be
 737                          * flushed, try next one.
 738                          */
 739                         ext = next_extent(ext);
 740                         continue;
 741                 }
 742
 743                 /* check if they belong to the same rpc slot before trying to
 744                  * merge. the extents are not overlapped and contiguous at
 745                  * chunk level to get here.
 746                  */
 747                 if (ext->oe_max_end != max_end) {
 748                         /* if they don't belong to the same RPC slot or
 749                          * max_pages_per_rpc has ever changed, do not merge.
 750                          */
 751                         ext = next_extent(ext);
 752                         continue;
 753                 }
 754
 755                 /* it's required that an extent must be contiguous at chunk
 756                  * level so that we know the whole extent is covered by grant
 757                  * (the pages in the extent are NOT required to be contiguous).
 758                  * Otherwise, it will be too much difficult to know which
 759                  * chunks have grants allocated.
 760                  */
 761
 762                 /* try to do front merge - extend ext's start */
 763                 if (chunk + 1 == ext_chk_start) {
 764                         /* ext must be chunk size aligned */
 765                         EASSERT((ext->oe_start & ~chunk_mask) == 0, ext);
 766
 767                         /* pull ext's start back to cover cur */
 768                         ext->oe_start = cur->oe_start;
 769                         ext->oe_grants += chunksize;
 770                         *grants -= chunksize;
 771
 772                         found = osc_extent_hold(ext);
 773                 } else if (chunk == ext_chk_end + 1) {
 774                         /* rear merge */
 775                         ext->oe_end = cur->oe_end;
 776                         ext->oe_grants += chunksize;
 777                         *grants -= chunksize;
 778
 779                         /* try to merge with the next one because we just fill
 780                          * in a gap
 781                          */
 782                         if (osc_extent_merge(env, ext, next_extent(ext)) == 0)
 783                                 /* we can save extent tax from next extent */
 784                                 *grants += cli->cl_extent_tax;
 785
 786                         found = osc_extent_hold(ext);
 787                 }
 788                 if (found)
 789                         break;
 790
 791                 ext = next_extent(ext);
 792         }
 793
 794         osc_extent_tree_dump(D_CACHE, obj);
 795         if (found) {
 796                 LASSERT(!conflict);
 797                 if (!IS_ERR(found)) {
 798                         LASSERT(found->oe_osclock == cur->oe_osclock);
 799                         OSC_EXTENT_DUMP(D_CACHE, found,
 800                                         "found caching ext for %lu.\n", index);
 801                 }
 802         } else if (!conflict) {
 803                 /* create a new extent */
 804                 EASSERT(osc_extent_is_overlapped(obj, cur) == 0, cur);
 805                 cur->oe_grants = chunksize + cli->cl_extent_tax;
 806                 *grants -= cur->oe_grants;
 807                 LASSERT(*grants >= 0);
 808
 809                 cur->oe_state = OES_CACHE;
 810                 found = osc_extent_hold(cur);
 811                 osc_extent_insert(obj, cur);
 812                 OSC_EXTENT_DUMP(D_CACHE, cur, "add into tree %lu/%lu.\n",
 813                                 index, lock->cll_descr.cld_end);
 814         }
 815         osc_object_unlock(obj);
 816
 817         if (conflict) {
 818                 LASSERT(!found);
 819
 820                 /* waiting for IO to finish. Please notice that it's impossible
 821                  * to be an OES_TRUNC extent.
 822                  */
 823                 rc = osc_extent_wait(env, conflict, OES_INV);
 824                 osc_extent_put(env, conflict);
 825                 conflict = NULL;
 826                 if (rc < 0) {
 827                         found = ERR_PTR(rc);
 828                         goto out;
 829                 }
 830
 831                 goto restart;
 832         }
 833
 834 out:
 835         osc_extent_put(env, cur);
 836         LASSERT(*grants >= 0);
 837         return found;
 838 }
 839
 840 /**
 841  * Called when IO is finished to an extent.
 842  */
 843 int osc_extent_finish(const struct lu_env *env, struct osc_extent *ext,
 844                       int sent, int rc)
 845 {
 846         struct client_obd *cli = osc_cli(ext->oe_obj);
 847         struct osc_async_page *oap;
 848         struct osc_async_page *tmp;
 849         int nr_pages = ext->oe_nr_pages;
 850         int lost_grant = 0;
 851         int blocksize = cli->cl_import->imp_obd->obd_osfs.os_bsize ? : 4096;
 852         __u64 last_off = 0;
 853         int last_count = -1;
 854
 855         OSC_EXTENT_DUMP(D_CACHE, ext, "extent finished.\n");
 856
 857         ext->oe_rc = rc ?: ext->oe_nr_pages;
 858         EASSERT(ergo(rc == 0, ext->oe_state == OES_RPC), ext);
 859         list_for_each_entry_safe(oap, tmp, &ext->oe_pages, oap_pending_item) {
 860                 list_del_init(&oap->oap_rpc_item);
 861                 list_del_init(&oap->oap_pending_item);
 862                 if (last_off <= oap->oap_obj_off) {
 863                         last_off = oap->oap_obj_off;
 864                         last_count = oap->oap_count;
 865                 }
 866
 867                 --ext->oe_nr_pages;
 868                 osc_ap_completion(env, cli, oap, sent, rc);
 869         }
 870         EASSERT(ext->oe_nr_pages == 0, ext);
 871
 872         if (!sent) {
 873                 lost_grant = ext->oe_grants;
 874         } else if (blocksize < PAGE_SIZE &&
 875                    last_count != PAGE_SIZE) {
 876                 /* For short writes we shouldn't count parts of pages that
 877                  * span a whole chunk on the OST side, or our accounting goes
 878                  * wrong.  Should match the code in filter_grant_check.
 879                  */
 880                 int offset = oap->oap_page_off & ~CFS_PAGE_MASK;
 881                 int count = oap->oap_count + (offset & (blocksize - 1));
 882                 int end = (offset + oap->oap_count) & (blocksize - 1);
 883
 884                 if (end)
 885                         count += blocksize - end;
 886
 887                 lost_grant = PAGE_SIZE - count;
 888         }
 889         if (ext->oe_grants > 0)
 890                 osc_free_grant(cli, nr_pages, lost_grant);
 891
 892         osc_extent_remove(ext);
 893         /* put the refcount for RPC */
 894         osc_extent_put(env, ext);
 895         return 0;
 896 }
 897
 898 static int extent_wait_cb(struct osc_extent *ext, int state)
 899 {
 900         int ret;
 901
 902         osc_object_lock(ext->oe_obj);
 903         ret = ext->oe_state == state;
 904         osc_object_unlock(ext->oe_obj);
 905
 906         return ret;
 907 }
 908
 909 /**
 910  * Wait for the extent's state to become @state.
 911  */
 912 static int osc_extent_wait(const struct lu_env *env, struct osc_extent *ext,
 913                            int state)
 914 {
 915         struct osc_object *obj = ext->oe_obj;
 916         struct l_wait_info lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(600), NULL,
 917                                                   LWI_ON_SIGNAL_NOOP, NULL);
 918         int rc = 0;
 919
 920         osc_object_lock(obj);
 921         LASSERT(sanity_check_nolock(ext) == 0);
 922         /* `Kick' this extent only if the caller is waiting for it to be
 923          * written out.
 924          */
 925         if (state == OES_INV && !ext->oe_urgent && !ext->oe_hp &&
 926             !ext->oe_trunc_pending) {
 927                 if (ext->oe_state == OES_ACTIVE) {
 928                         ext->oe_urgent = 1;
 929                 } else if (ext->oe_state == OES_CACHE) {
 930                         ext->oe_urgent = 1;
 931                         osc_extent_hold(ext);
 932                         rc = 1;
 933                 }
 934         }
 935         osc_object_unlock(obj);
 936         if (rc == 1)
 937                 osc_extent_release(env, ext);
 938
 939         /* wait for the extent until its state becomes @state */
 940         rc = l_wait_event(ext->oe_waitq, extent_wait_cb(ext, state), &lwi);
 941         if (rc == -ETIMEDOUT) {
 942                 OSC_EXTENT_DUMP(D_ERROR, ext,
 943                         "%s: wait ext to %d timedout, recovery in progress?\n",
 944                         osc_export(obj)->exp_obd->obd_name, state);
 945
 946                 lwi = LWI_INTR(LWI_ON_SIGNAL_NOOP, NULL);
 947                 rc = l_wait_event(ext->oe_waitq, extent_wait_cb(ext, state),
 948                                   &lwi);
 949         }
 950         if (rc == 0 && ext->oe_rc < 0)
 951                 rc = ext->oe_rc;
 952         return rc;
 953 }
 954
 955 /**
 956  * Discard pages with index greater than @size. If @ext is overlapped with
 957  * @size, then partial truncate happens.
 958  */
 959 static int osc_extent_truncate(struct osc_extent *ext, pgoff_t trunc_index,
 960                                bool partial)
 961 {
 962         struct cl_env_nest nest;
 963         struct lu_env *env;
 964         struct cl_io *io;
 965         struct osc_object *obj = ext->oe_obj;
 966         struct client_obd *cli = osc_cli(obj);
 967         struct osc_async_page *oap;
 968         struct osc_async_page *tmp;
 969         int pages_in_chunk = 0;
 970         int ppc_bits = cli->cl_chunkbits - PAGE_SHIFT;
 971         __u64 trunc_chunk = trunc_index >> ppc_bits;
 972         int grants = 0;
 973         int nr_pages = 0;
 974         int rc = 0;
 975
 976         LASSERT(sanity_check(ext) == 0);
 977         EASSERT(ext->oe_state == OES_TRUNC, ext);
 978         EASSERT(!ext->oe_urgent, ext);
 979
 980         /* Request new lu_env.
 981          * We can't use that env from osc_cache_truncate_start() because
 982          * it's from lov_io_sub and not fully initialized.
 983          */
 984         env = cl_env_nested_get(&nest);
 985         io  = &osc_env_info(env)->oti_io;
 986         io->ci_obj = cl_object_top(osc2cl(obj));
 987         rc = cl_io_init(env, io, CIT_MISC, io->ci_obj);
 988         if (rc < 0)
 989                 goto out;
 990
 991         /* discard all pages with index greater then trunc_index */
 992         list_for_each_entry_safe(oap, tmp, &ext->oe_pages, oap_pending_item) {
 993                 struct cl_page *sub = oap2cl_page(oap);
 994                 struct cl_page *page = cl_page_top(sub);
 995
 996                 LASSERT(list_empty(&oap->oap_rpc_item));
 997
 998                 /* only discard the pages with their index greater than
 999                  * trunc_index, and ...
1000                  */
1001                 if (sub->cp_index < trunc_index ||
1002                     (sub->cp_index == trunc_index && partial)) {
1003                         /* accounting how many pages remaining in the chunk
1004                          * so that we can calculate grants correctly. */
1005                         if (sub->cp_index >> ppc_bits == trunc_chunk)
1006                                 ++pages_in_chunk;
1007                         continue;
1008                 }
1009
1010                 list_del_init(&oap->oap_pending_item);
1011
1012                 cl_page_get(page);
1013                 lu_ref_add(&page->cp_reference, "truncate", current);
1014
1015                 if (cl_page_own(env, io, page) == 0) {
1016                         cl_page_unmap(env, io, page);
1017                         cl_page_discard(env, io, page);
1018                         cl_page_disown(env, io, page);
1019                 } else {
1020                         LASSERT(page->cp_state == CPS_FREEING);
1021                         LASSERT(0);
1022                 }
1023
1024                 lu_ref_del(&page->cp_reference, "truncate", current);
1025                 cl_page_put(env, page);
1026
1027                 --ext->oe_nr_pages;
1028                 ++nr_pages;
1029         }
1030         EASSERTF(ergo(ext->oe_start >= trunc_index + !!partial,
1031                       ext->oe_nr_pages == 0),
1032                 ext, "trunc_index %lu, partial %d\n", trunc_index, partial);
1033
1034         osc_object_lock(obj);
1035         if (ext->oe_nr_pages == 0) {
1036                 LASSERT(pages_in_chunk == 0);
1037                 grants = ext->oe_grants;
1038                 ext->oe_grants = 0;
1039         } else { /* calculate how many grants we can free */
1040                 int chunks = (ext->oe_end >> ppc_bits) - trunc_chunk;
1041                 pgoff_t last_index;
1042
1043                 /* if there is no pages in this chunk, we can also free grants
1044                  * for the last chunk
1045                  */
1046                 if (pages_in_chunk == 0) {
1047                         /* if this is the 1st chunk and no pages in this chunk,
1048                          * ext->oe_nr_pages must be zero, so we should be in
1049                          * the other if-clause.
1050                          */
1051                         LASSERT(trunc_chunk > 0);
1052                         --trunc_chunk;
1053                         ++chunks;
1054                 }
1055
1056                 /* this is what we can free from this extent */
1057                 grants = chunks << cli->cl_chunkbits;
1058                 ext->oe_grants -= grants;
1059                 last_index = ((trunc_chunk + 1) << ppc_bits) - 1;
1060                 ext->oe_end = min(last_index, ext->oe_max_end);
1061                 LASSERT(ext->oe_end >= ext->oe_start);
1062                 LASSERT(ext->oe_grants > 0);
1063         }
1064         osc_object_unlock(obj);
1065
1066         if (grants > 0 || nr_pages > 0)
1067                 osc_free_grant(cli, nr_pages, grants);
1068
1069 out:
1070         cl_io_fini(env, io);
1071         cl_env_nested_put(&nest, env);
1072         return rc;
1073 }
1074
1075 /**
1076  * This function is used to make the extent prepared for transfer.
1077  * A race with flushing page - ll_writepage() has to be handled cautiously.
1078  */
1079 static int osc_extent_make_ready(const struct lu_env *env,
1080                                  struct osc_extent *ext)
1081 {
1082         struct osc_async_page *oap;
1083         struct osc_async_page *last = NULL;
1084         struct osc_object *obj = ext->oe_obj;
1085         int page_count = 0;
1086         int rc;
1087
1088         /* we're going to grab page lock, so object lock must not be taken. */
1089         LASSERT(sanity_check(ext) == 0);
1090         /* in locking state, any process should not touch this extent. */
1091         EASSERT(ext->oe_state == OES_LOCKING, ext);
1092         EASSERT(ext->oe_owner, ext);
1093
1094         OSC_EXTENT_DUMP(D_CACHE, ext, "make ready\n");
1095
1096         list_for_each_entry(oap, &ext->oe_pages, oap_pending_item) {
1097                 ++page_count;
1098                 if (!last || last->oap_obj_off < oap->oap_obj_off)
1099                         last = oap;
1100
1101                 /* checking ASYNC_READY is race safe */
1102                 if ((oap->oap_async_flags & ASYNC_READY) != 0)
1103                         continue;
1104
1105                 rc = osc_make_ready(env, oap, OBD_BRW_WRITE);
1106                 switch (rc) {
1107                 case 0:
1108                         spin_lock(&oap->oap_lock);
1109                         oap->oap_async_flags |= ASYNC_READY;
1110                         spin_unlock(&oap->oap_lock);
1111                         break;
1112                 case -EALREADY:
1113                         LASSERT((oap->oap_async_flags & ASYNC_READY) != 0);
1114                         break;
1115                 default:
1116                         LASSERTF(0, "unknown return code: %d\n", rc);
1117                 }
1118         }
1119
1120         LASSERT(page_count == ext->oe_nr_pages);
1121         LASSERT(last);
1122         /* the last page is the only one we need to refresh its count by
1123          * the size of file.
1124          */
1125         if (!(last->oap_async_flags & ASYNC_COUNT_STABLE)) {
1126                 last->oap_count = osc_refresh_count(env, last, OBD_BRW_WRITE);
1127                 LASSERT(last->oap_count > 0);
1128                 LASSERT(last->oap_page_off + last->oap_count <= PAGE_SIZE);
1129                 last->oap_async_flags |= ASYNC_COUNT_STABLE;
1130         }
1131
1132         /* for the rest of pages, we don't need to call osf_refresh_count()
1133          * because it's known they are not the last page
1134          */
1135         list_for_each_entry(oap, &ext->oe_pages, oap_pending_item) {
1136                 if (!(oap->oap_async_flags & ASYNC_COUNT_STABLE)) {
1137                         oap->oap_count = PAGE_SIZE - oap->oap_page_off;
1138                         oap->oap_async_flags |= ASYNC_COUNT_STABLE;
1139                 }
1140         }
1141
1142         osc_object_lock(obj);
1143         osc_extent_state_set(ext, OES_RPC);
1144         osc_object_unlock(obj);
1145         /* get a refcount for RPC. */
1146         osc_extent_get(ext);
1147
1148         return 0;
1149 }
1150
1151 /**
1152  * Quick and simple version of osc_extent_find(). This function is frequently
1153  * called to expand the extent for the same IO. To expand the extent, the
1154  * page index must be in the same or next chunk of ext->oe_end.
1155  */
1156 static int osc_extent_expand(struct osc_extent *ext, pgoff_t index, int *grants)
1157 {
1158         struct osc_object *obj = ext->oe_obj;
1159         struct client_obd *cli = osc_cli(obj);
1160         struct osc_extent *next;
1161         int ppc_bits = cli->cl_chunkbits - PAGE_SHIFT;
1162         pgoff_t chunk = index >> ppc_bits;
1163         pgoff_t end_chunk;
1164         pgoff_t end_index;
1165         int chunksize = 1 << cli->cl_chunkbits;
1166         int rc = 0;
1167
1168         LASSERT(ext->oe_max_end >= index && ext->oe_start <= index);
1169         osc_object_lock(obj);
1170         LASSERT(sanity_check_nolock(ext) == 0);
1171         end_chunk = ext->oe_end >> ppc_bits;
1172         if (chunk > end_chunk + 1) {
1173                 rc = -ERANGE;
1174                 goto out;
1175         }
1176
1177         if (end_chunk >= chunk) {
1178                 rc = 0;
1179                 goto out;
1180         }
1181
1182         LASSERT(end_chunk + 1 == chunk);
1183         /* try to expand this extent to cover @index */
1184         end_index = min(ext->oe_max_end, ((chunk + 1) << ppc_bits) - 1);
1185
1186         next = next_extent(ext);
1187         if (next && next->oe_start <= end_index) {
1188                 /* complex mode - overlapped with the next extent,
1189                  * this case will be handled by osc_extent_find()
1190                  */
1191                 rc = -EAGAIN;
1192                 goto out;
1193         }
1194
1195         ext->oe_end = end_index;
1196         ext->oe_grants += chunksize;
1197         *grants -= chunksize;
1198         LASSERT(*grants >= 0);
1199         EASSERTF(osc_extent_is_overlapped(obj, ext) == 0, ext,
1200                  "overlapped after expanding for %lu.\n", index);
1201
1202 out:
1203         osc_object_unlock(obj);
1204         return rc;
1205 }
1206
1207 static void osc_extent_tree_dump0(int level, struct osc_object *obj,
1208                                   const char *func, int line)
1209 {
1210         struct osc_extent *ext;
1211         int cnt;
1212
1213         CDEBUG(level, "Dump object %p extents at %s:%d, mppr: %u.\n",
1214                obj, func, line, osc_cli(obj)->cl_max_pages_per_rpc);
1215
1216         /* osc_object_lock(obj); */
1217         cnt = 1;
1218         for (ext = first_extent(obj); ext; ext = next_extent(ext))
1219                 OSC_EXTENT_DUMP(level, ext, "in tree %d.\n", cnt++);
1220
1221         cnt = 1;
1222         list_for_each_entry(ext, &obj->oo_hp_exts, oe_link)
1223                 OSC_EXTENT_DUMP(level, ext, "hp %d.\n", cnt++);
1224
1225         cnt = 1;
1226         list_for_each_entry(ext, &obj->oo_urgent_exts, oe_link)
1227                 OSC_EXTENT_DUMP(level, ext, "urgent %d.\n", cnt++);
1228
1229         cnt = 1;
1230         list_for_each_entry(ext, &obj->oo_reading_exts, oe_link)
1231                 OSC_EXTENT_DUMP(level, ext, "reading %d.\n", cnt++);
1232         /* osc_object_unlock(obj); */
1233 }
1234
1235 /* ------------------ osc extent end ------------------ */
1236
1237 static inline int osc_is_ready(struct osc_object *osc)
1238 {
1239         return !list_empty(&osc->oo_ready_item) ||
1240                !list_empty(&osc->oo_hp_ready_item);
1241 }
1242
1243 #define OSC_IO_DEBUG(OSC, STR, args...)                                        \
1244         CDEBUG(D_CACHE, "obj %p ready %d|%c|%c wr %d|%c|%c rd %d|%c " STR,     \
1245                (OSC), osc_is_ready(OSC),                                       \
1246                list_empty_marker(&(OSC)->oo_hp_ready_item),                    \
1247                list_empty_marker(&(OSC)->oo_ready_item),                       \
1248                atomic_read(&(OSC)->oo_nr_writes),                              \
1249                list_empty_marker(&(OSC)->oo_hp_exts),                          \
1250                list_empty_marker(&(OSC)->oo_urgent_exts),                      \
1251                atomic_read(&(OSC)->oo_nr_reads),                               \
1252                list_empty_marker(&(OSC)->oo_reading_exts),                     \
1253                ##args)
1254
1255 static int osc_make_ready(const struct lu_env *env, struct osc_async_page *oap,
1256                           int cmd)
1257 {
1258         struct osc_page *opg = oap2osc_page(oap);
1259         struct cl_page *page = cl_page_top(oap2cl_page(oap));
1260         int result;
1261
1262         LASSERT(cmd == OBD_BRW_WRITE); /* no cached reads */
1263
1264         result = cl_page_make_ready(env, page, CRT_WRITE);
1265         if (result == 0)
1266                 opg->ops_submit_time = cfs_time_current();
1267         return result;
1268 }
1269
1270 static int osc_refresh_count(const struct lu_env *env,
1271                              struct osc_async_page *oap, int cmd)
1272 {
1273         struct osc_page *opg = oap2osc_page(oap);
1274         struct cl_page *page = oap2cl_page(oap);
1275         struct cl_object *obj;
1276         struct cl_attr *attr = &osc_env_info(env)->oti_attr;
1277
1278         int result;
1279         loff_t kms;
1280
1281         /* readpage queues with _COUNT_STABLE, shouldn't get here. */
1282         LASSERT(!(cmd & OBD_BRW_READ));
1283         obj = opg->ops_cl.cpl_obj;
1284
1285         cl_object_attr_lock(obj);
1286         result = cl_object_attr_get(env, obj, attr);
1287         cl_object_attr_unlock(obj);
1288         if (result < 0)
1289                 return result;
1290         kms = attr->cat_kms;
1291         if (cl_offset(obj, page->cp_index) >= kms)
1292                 /* catch race with truncate */
1293                 return 0;
1294         else if (cl_offset(obj, page->cp_index + 1) > kms)
1295                 /* catch sub-page write at end of file */
1296                 return kms % PAGE_SIZE;
1297         else
1298                 return PAGE_SIZE;
1299 }
1300
1301 static int osc_completion(const struct lu_env *env, struct osc_async_page *oap,
1302                           int cmd, int rc)
1303 {
1304         struct osc_page *opg = oap2osc_page(oap);
1305         struct cl_page *page = cl_page_top(oap2cl_page(oap));
1306         struct osc_object *obj = cl2osc(opg->ops_cl.cpl_obj);
1307         enum cl_req_type crt;
1308         int srvlock;
1309
1310         cmd &= ~OBD_BRW_NOQUOTA;
1311         LASSERT(equi(page->cp_state == CPS_PAGEIN,  cmd == OBD_BRW_READ));
1312         LASSERT(equi(page->cp_state == CPS_PAGEOUT, cmd == OBD_BRW_WRITE));
1313         LASSERT(opg->ops_transfer_pinned);
1314
1315         /*
1316          * page->cp_req can be NULL if io submission failed before
1317          * cl_req was allocated.
1318          */
1319         if (page->cp_req)
1320                 cl_req_page_done(env, page);
1321         LASSERT(!page->cp_req);
1322
1323         crt = cmd == OBD_BRW_READ ? CRT_READ : CRT_WRITE;
1324         /* Clear opg->ops_transfer_pinned before VM lock is released. */
1325         opg->ops_transfer_pinned = 0;
1326
1327         spin_lock(&obj->oo_seatbelt);
1328         LASSERT(opg->ops_submitter);
1329         LASSERT(!list_empty(&opg->ops_inflight));
1330         list_del_init(&opg->ops_inflight);
1331         opg->ops_submitter = NULL;
1332         spin_unlock(&obj->oo_seatbelt);
1333
1334         opg->ops_submit_time = 0;
1335         srvlock = oap->oap_brw_flags & OBD_BRW_SRVLOCK;
1336
1337         /* statistic */
1338         if (rc == 0 && srvlock) {
1339                 struct lu_device *ld = opg->ops_cl.cpl_obj->co_lu.lo_dev;
1340                 struct osc_stats *stats = &lu2osc_dev(ld)->od_stats;
1341                 int bytes = oap->oap_count;
1342
1343                 if (crt == CRT_READ)
1344                         stats->os_lockless_reads += bytes;
1345                 else
1346                         stats->os_lockless_writes += bytes;
1347         }
1348
1349         /*
1350          * This has to be the last operation with the page, as locks are
1351          * released in cl_page_completion() and nothing except for the
1352          * reference counter protects page from concurrent reclaim.
1353          */
1354         lu_ref_del(&page->cp_reference, "transfer", page);
1355
1356         cl_page_completion(env, page, crt, rc);
1357
1358         return 0;
1359 }
1360
1361 #define OSC_DUMP_GRANT(cli, fmt, args...) do {                                \
1362         struct client_obd *__tmp = (cli);                                     \
1363         CDEBUG(D_CACHE, "%s: { dirty: %ld/%ld dirty_pages: %d/%d "            \
1364                "dropped: %ld avail: %ld, reserved: %ld, flight: %d } " fmt,   \
1365                __tmp->cl_import->imp_obd->obd_name,                           \
1366                __tmp->cl_dirty, __tmp->cl_dirty_max,                          \
1367                atomic_read(&obd_dirty_pages), obd_max_dirty_pages,            \
1368                __tmp->cl_lost_grant, __tmp->cl_avail_grant,                   \
1369                __tmp->cl_reserved_grant, __tmp->cl_w_in_flight, ##args);      \
1370 } while (0)
1371
1372 /* caller must hold loi_list_lock */
1373 static void osc_consume_write_grant(struct client_obd *cli,
1374                                     struct brw_page *pga)
1375 {
1376         assert_spin_locked(&cli->cl_loi_list_lock.lock);
1377         LASSERT(!(pga->flag & OBD_BRW_FROM_GRANT));
1378         atomic_inc(&obd_dirty_pages);
1379         cli->cl_dirty += PAGE_SIZE;
1380         pga->flag |= OBD_BRW_FROM_GRANT;
1381         CDEBUG(D_CACHE, "using %lu grant credits for brw %p page %p\n",
1382                PAGE_SIZE, pga, pga->pg);
1383         osc_update_next_shrink(cli);
1384 }
1385
1386 /* the companion to osc_consume_write_grant, called when a brw has completed.
1387  * must be called with the loi lock held.
1388  */
1389 static void osc_release_write_grant(struct client_obd *cli,
1390                                     struct brw_page *pga)
1391 {
1392         assert_spin_locked(&cli->cl_loi_list_lock.lock);
1393         if (!(pga->flag & OBD_BRW_FROM_GRANT)) {
1394                 return;
1395         }
1396
1397         pga->flag &= ~OBD_BRW_FROM_GRANT;
1398         atomic_dec(&obd_dirty_pages);
1399         cli->cl_dirty -= PAGE_SIZE;
1400         if (pga->flag & OBD_BRW_NOCACHE) {
1401                 pga->flag &= ~OBD_BRW_NOCACHE;
1402                 atomic_dec(&obd_dirty_transit_pages);
1403                 cli->cl_dirty_transit -= PAGE_SIZE;
1404         }
1405 }
1406
1407 /**
1408  * To avoid sleeping with object lock held, it's good for us allocate enough
1409  * grants before entering into critical section.
1410  *
1411  * client_obd_list_lock held by caller
1412  */
1413 static int osc_reserve_grant(struct client_obd *cli, unsigned int bytes)
1414 {
1415         int rc = -EDQUOT;
1416
1417         if (cli->cl_avail_grant >= bytes) {
1418                 cli->cl_avail_grant -= bytes;
1419                 cli->cl_reserved_grant += bytes;
1420                 rc = 0;
1421         }
1422         return rc;
1423 }
1424
1425 static void __osc_unreserve_grant(struct client_obd *cli,
1426                                   unsigned int reserved, unsigned int unused)
1427 {
1428         /* it's quite normal for us to get more grant than reserved.
1429          * Thinking about a case that two extents merged by adding a new
1430          * chunk, we can save one extent tax. If extent tax is greater than
1431          * one chunk, we can save more grant by adding a new chunk
1432          */
1433         cli->cl_reserved_grant -= reserved;
1434         if (unused > reserved) {
1435                 cli->cl_avail_grant += reserved;
1436                 cli->cl_lost_grant  += unused - reserved;
1437         } else {
1438                 cli->cl_avail_grant += unused;
1439         }
1440 }
1441
1442 static void osc_unreserve_grant(struct client_obd *cli,
1443                                 unsigned int reserved, unsigned int unused)
1444 {
1445         client_obd_list_lock(&cli->cl_loi_list_lock);
1446         __osc_unreserve_grant(cli, reserved, unused);
1447         if (unused > 0)
1448                 osc_wake_cache_waiters(cli);
1449         client_obd_list_unlock(&cli->cl_loi_list_lock);
1450 }
1451
1452 /**
1453  * Free grant after IO is finished or canceled.
1454  *
1455  * @lost_grant is used to remember how many grants we have allocated but not
1456  * used, we should return these grants to OST. There're two cases where grants
1457  * can be lost:
1458  * 1. truncate;
1459  * 2. blocksize at OST is less than PAGE_SIZE and a partial page was
1460  *    written. In this case OST may use less chunks to serve this partial
1461  *    write. OSTs don't actually know the page size on the client side. so
1462  *    clients have to calculate lost grant by the blocksize on the OST.
1463  *    See filter_grant_check() for details.
1464  */
1465 static void osc_free_grant(struct client_obd *cli, unsigned int nr_pages,
1466                            unsigned int lost_grant)
1467 {
1468         int grant = (1 << cli->cl_chunkbits) + cli->cl_extent_tax;
1469
1470         client_obd_list_lock(&cli->cl_loi_list_lock);
1471         atomic_sub(nr_pages, &obd_dirty_pages);
1472         cli->cl_dirty -= nr_pages << PAGE_SHIFT;
1473         cli->cl_lost_grant += lost_grant;
1474         if (cli->cl_avail_grant < grant && cli->cl_lost_grant >= grant) {
1475                 /* borrow some grant from truncate to avoid the case that
1476                  * truncate uses up all avail grant
1477                  */
1478                 cli->cl_lost_grant -= grant;
1479                 cli->cl_avail_grant += grant;
1480         }
1481         osc_wake_cache_waiters(cli);
1482         client_obd_list_unlock(&cli->cl_loi_list_lock);
1483         CDEBUG(D_CACHE, "lost %u grant: %lu avail: %lu dirty: %lu\n",
1484                lost_grant, cli->cl_lost_grant,
1485                cli->cl_avail_grant, cli->cl_dirty);
1486 }
1487
1488 /**
1489  * The companion to osc_enter_cache(), called when @oap is no longer part of
1490  * the dirty accounting due to error.
1491  */
1492 static void osc_exit_cache(struct client_obd *cli, struct osc_async_page *oap)
1493 {
1494         client_obd_list_lock(&cli->cl_loi_list_lock);
1495         osc_release_write_grant(cli, &oap->oap_brw_page);
1496         client_obd_list_unlock(&cli->cl_loi_list_lock);
1497 }
1498
1499 /**
1500  * Non-blocking version of osc_enter_cache() that consumes grant only when it
1501  * is available.
1502  */
1503 static int osc_enter_cache_try(struct client_obd *cli,
1504                                struct osc_async_page *oap,
1505                                int bytes, int transient)
1506 {
1507         int rc;
1508
1509         OSC_DUMP_GRANT(cli, "need:%d.\n", bytes);
1510
1511         rc = osc_reserve_grant(cli, bytes);
1512         if (rc < 0)
1513                 return 0;
1514
1515         if (cli->cl_dirty + PAGE_SIZE <= cli->cl_dirty_max &&
1516             atomic_read(&obd_dirty_pages) + 1 <= obd_max_dirty_pages) {
1517                 osc_consume_write_grant(cli, &oap->oap_brw_page);
1518                 if (transient) {
1519                         cli->cl_dirty_transit += PAGE_SIZE;
1520                         atomic_inc(&obd_dirty_transit_pages);
1521                         oap->oap_brw_flags |= OBD_BRW_NOCACHE;
1522                 }
1523                 rc = 1;
1524         } else {
1525                 __osc_unreserve_grant(cli, bytes, bytes);
1526                 rc = 0;
1527         }
1528         return rc;
1529 }
1530
1531 static int ocw_granted(struct client_obd *cli, struct osc_cache_waiter *ocw)
1532 {
1533         int rc;
1534
1535         client_obd_list_lock(&cli->cl_loi_list_lock);
1536         rc = list_empty(&ocw->ocw_entry);
1537         client_obd_list_unlock(&cli->cl_loi_list_lock);
1538         return rc;
1539 }
1540
1541 /**
1542  * The main entry to reserve dirty page accounting. Usually the grant reserved
1543  * in this function will be freed in bulk in osc_free_grant() unless it fails
1544  * to add osc cache, in that case, it will be freed in osc_exit_cache().
1545  *
1546  * The process will be put into sleep if it's already run out of grant.
1547  */
1548 static int osc_enter_cache(const struct lu_env *env, struct client_obd *cli,
1549                            struct osc_async_page *oap, int bytes)
1550 {
1551         struct osc_object *osc = oap->oap_obj;
1552         struct lov_oinfo *loi = osc->oo_oinfo;
1553         struct osc_cache_waiter ocw;
1554         struct l_wait_info lwi = LWI_INTR(LWI_ON_SIGNAL_NOOP, NULL);
1555         int rc = -EDQUOT;
1556
1557         OSC_DUMP_GRANT(cli, "need:%d.\n", bytes);
1558
1559         client_obd_list_lock(&cli->cl_loi_list_lock);
1560
1561         /* force the caller to try sync io.  this can jump the list
1562          * of queued writes and create a discontiguous rpc stream
1563          */
1564         if (OBD_FAIL_CHECK(OBD_FAIL_OSC_NO_GRANT) ||
1565             cli->cl_dirty_max < PAGE_SIZE     ||
1566             cli->cl_ar.ar_force_sync || loi->loi_ar.ar_force_sync) {
1567                 rc = -EDQUOT;
1568                 goto out;
1569         }
1570
1571         /* Hopefully normal case - cache space and write credits available */
1572         if (osc_enter_cache_try(cli, oap, bytes, 0)) {
1573                 rc = 0;
1574                 goto out;
1575         }
1576
1577         /* We can get here for two reasons: too many dirty pages in cache, or
1578          * run out of grants. In both cases we should write dirty pages out.
1579          * Adding a cache waiter will trigger urgent write-out no matter what
1580          * RPC size will be.
1581          * The exiting condition is no avail grants and no dirty pages caching,
1582          * that really means there is no space on the OST.
1583          */
1584         init_waitqueue_head(&ocw.ocw_waitq);
1585         ocw.ocw_oap   = oap;
1586         ocw.ocw_grant = bytes;
1587         while (cli->cl_dirty > 0 || cli->cl_w_in_flight > 0) {
1588                 list_add_tail(&ocw.ocw_entry, &cli->cl_cache_waiters);
1589                 ocw.ocw_rc = 0;
1590                 client_obd_list_unlock(&cli->cl_loi_list_lock);
1591
1592                 osc_io_unplug_async(env, cli, NULL);
1593
1594                 CDEBUG(D_CACHE, "%s: sleeping for cache space @ %p for %p\n",
1595                        cli->cl_import->imp_obd->obd_name, &ocw, oap);
1596
1597                 rc = l_wait_event(ocw.ocw_waitq, ocw_granted(cli, &ocw), &lwi);
1598
1599                 client_obd_list_lock(&cli->cl_loi_list_lock);
1600
1601                 /* l_wait_event is interrupted by signal */
1602                 if (rc < 0) {
1603                         list_del_init(&ocw.ocw_entry);
1604                         goto out;
1605                 }
1606
1607                 LASSERT(list_empty(&ocw.ocw_entry));
1608                 rc = ocw.ocw_rc;
1609
1610                 if (rc != -EDQUOT)
1611                         goto out;
1612                 if (osc_enter_cache_try(cli, oap, bytes, 0)) {
1613                         rc = 0;
1614                         goto out;
1615                 }
1616         }
1617 out:
1618         client_obd_list_unlock(&cli->cl_loi_list_lock);
1619         OSC_DUMP_GRANT(cli, "returned %d.\n", rc);
1620         return rc;
1621 }
1622
1623 /* caller must hold loi_list_lock */
1624 void osc_wake_cache_waiters(struct client_obd *cli)
1625 {
1626         struct list_head *l, *tmp;
1627         struct osc_cache_waiter *ocw;
1628
1629         list_for_each_safe(l, tmp, &cli->cl_cache_waiters) {
1630                 ocw = list_entry(l, struct osc_cache_waiter, ocw_entry);
1631                 list_del_init(&ocw->ocw_entry);
1632
1633                 ocw->ocw_rc = -EDQUOT;
1634                 /* we can't dirty more */
1635                 if ((cli->cl_dirty + PAGE_SIZE > cli->cl_dirty_max) ||
1636                     (atomic_read(&obd_dirty_pages) + 1 >
1637                      obd_max_dirty_pages)) {
1638                         CDEBUG(D_CACHE, "no dirty room: dirty: %ld osc max %ld, sys max %d\n",
1639                                cli->cl_dirty,
1640                                cli->cl_dirty_max, obd_max_dirty_pages);
1641                         goto wakeup;
1642                 }
1643
1644                 ocw->ocw_rc = 0;
1645                 if (!osc_enter_cache_try(cli, ocw->ocw_oap, ocw->ocw_grant, 0))
1646                         ocw->ocw_rc = -EDQUOT;
1647
1648 wakeup:
1649                 CDEBUG(D_CACHE, "wake up %p for oap %p, avail grant %ld, %d\n",
1650                        ocw, ocw->ocw_oap, cli->cl_avail_grant, ocw->ocw_rc);
1651
1652                 wake_up(&ocw->ocw_waitq);
1653         }
1654 }
1655
1656 static int osc_max_rpc_in_flight(struct client_obd *cli, struct osc_object *osc)
1657 {
1658         int hprpc = !!list_empty(&osc->oo_hp_exts);
1659
1660         return rpcs_in_flight(cli) >= cli->cl_max_rpcs_in_flight + hprpc;
1661 }
1662
1663 /* This maintains the lists of pending pages to read/write for a given object
1664  * (lop).  This is used by osc_check_rpcs->osc_next_obj() and osc_list_maint()
1665  * to quickly find objects that are ready to send an RPC.
1666  */
1667 static int osc_makes_rpc(struct client_obd *cli, struct osc_object *osc,
1668                          int cmd)
1669 {
1670         int invalid_import = 0;
1671
1672         /* if we have an invalid import we want to drain the queued pages
1673          * by forcing them through rpcs that immediately fail and complete
1674          * the pages.  recovery relies on this to empty the queued pages
1675          * before canceling the locks and evicting down the llite pages
1676          */
1677         if (!cli->cl_import || cli->cl_import->imp_invalid)
1678                 invalid_import = 1;
1679
1680         if (cmd & OBD_BRW_WRITE) {
1681                 if (atomic_read(&osc->oo_nr_writes) == 0)
1682                         return 0;
1683                 if (invalid_import) {
1684                         CDEBUG(D_CACHE, "invalid import forcing RPC\n");
1685                         return 1;
1686                 }
1687                 if (!list_empty(&osc->oo_hp_exts)) {
1688                         CDEBUG(D_CACHE, "high prio request forcing RPC\n");
1689                         return 1;
1690                 }
1691                 if (!list_empty(&osc->oo_urgent_exts)) {
1692                         CDEBUG(D_CACHE, "urgent request forcing RPC\n");
1693                         return 1;
1694                 }
1695                 /* trigger a write rpc stream as long as there are dirtiers
1696                  * waiting for space.  as they're waiting, they're not going to
1697                  * create more pages to coalesce with what's waiting..
1698                  */
1699                 if (!list_empty(&cli->cl_cache_waiters)) {
1700                         CDEBUG(D_CACHE, "cache waiters forcing RPC\n");
1701                         return 1;
1702                 }
1703                 if (atomic_read(&osc->oo_nr_writes) >=
1704                     cli->cl_max_pages_per_rpc)
1705                         return 1;
1706         } else {
1707                 if (atomic_read(&osc->oo_nr_reads) == 0)
1708                         return 0;
1709                 if (invalid_import) {
1710                         CDEBUG(D_CACHE, "invalid import forcing RPC\n");
1711                         return 1;
1712                 }
1713                 /* all read are urgent. */
1714                 if (!list_empty(&osc->oo_reading_exts))
1715                         return 1;
1716         }
1717
1718         return 0;
1719 }
1720
1721 static void osc_update_pending(struct osc_object *obj, int cmd, int delta)
1722 {
1723         struct client_obd *cli = osc_cli(obj);
1724
1725         if (cmd & OBD_BRW_WRITE) {
1726                 atomic_add(delta, &obj->oo_nr_writes);
1727                 atomic_add(delta, &cli->cl_pending_w_pages);
1728                 LASSERT(atomic_read(&obj->oo_nr_writes) >= 0);
1729         } else {
1730                 atomic_add(delta, &obj->oo_nr_reads);
1731                 atomic_add(delta, &cli->cl_pending_r_pages);
1732                 LASSERT(atomic_read(&obj->oo_nr_reads) >= 0);
1733         }
1734         OSC_IO_DEBUG(obj, "update pending cmd %d delta %d.\n", cmd, delta);
1735 }
1736
1737 static int osc_makes_hprpc(struct osc_object *obj)
1738 {
1739         return !list_empty(&obj->oo_hp_exts);
1740 }
1741
1742 static void on_list(struct list_head *item, struct list_head *list, int should_be_on)
1743 {
1744         if (list_empty(item) && should_be_on)
1745                 list_add_tail(item, list);
1746         else if (!list_empty(item) && !should_be_on)
1747                 list_del_init(item);
1748 }
1749
1750 /* maintain the osc's cli list membership invariants so that osc_send_oap_rpc
1751  * can find pages to build into rpcs quickly
1752  */
1753 static int __osc_list_maint(struct client_obd *cli, struct osc_object *osc)
1754 {
1755         if (osc_makes_hprpc(osc)) {
1756                 /* HP rpc */
1757                 on_list(&osc->oo_ready_item, &cli->cl_loi_ready_list, 0);
1758                 on_list(&osc->oo_hp_ready_item, &cli->cl_loi_hp_ready_list, 1);
1759         } else {
1760                 on_list(&osc->oo_hp_ready_item, &cli->cl_loi_hp_ready_list, 0);
1761                 on_list(&osc->oo_ready_item, &cli->cl_loi_ready_list,
1762                         osc_makes_rpc(cli, osc, OBD_BRW_WRITE) ||
1763                         osc_makes_rpc(cli, osc, OBD_BRW_READ));
1764         }
1765
1766         on_list(&osc->oo_write_item, &cli->cl_loi_write_list,
1767                 atomic_read(&osc->oo_nr_writes) > 0);
1768
1769         on_list(&osc->oo_read_item, &cli->cl_loi_read_list,
1770                 atomic_read(&osc->oo_nr_reads) > 0);
1771
1772         return osc_is_ready(osc);
1773 }
1774
1775 static int osc_list_maint(struct client_obd *cli, struct osc_object *osc)
1776 {
1777         int is_ready;
1778
1779         client_obd_list_lock(&cli->cl_loi_list_lock);
1780         is_ready = __osc_list_maint(cli, osc);
1781         client_obd_list_unlock(&cli->cl_loi_list_lock);
1782
1783         return is_ready;
1784 }
1785
1786 /* this is trying to propagate async writeback errors back up to the
1787  * application.  As an async write fails we record the error code for later if
1788  * the app does an fsync.  As long as errors persist we force future rpcs to be
1789  * sync so that the app can get a sync error and break the cycle of queueing
1790  * pages for which writeback will fail.
1791  */
1792 static void osc_process_ar(struct osc_async_rc *ar, __u64 xid,
1793                            int rc)
1794 {
1795         if (rc) {
1796                 if (!ar->ar_rc)
1797                         ar->ar_rc = rc;
1798
1799                 ar->ar_force_sync = 1;
1800                 ar->ar_min_xid = ptlrpc_sample_next_xid();
1801                 return;
1802
1803         }
1804
1805         if (ar->ar_force_sync && (xid >= ar->ar_min_xid))
1806                 ar->ar_force_sync = 0;
1807 }
1808
1809 /* this must be called holding the loi list lock to give coverage to exit_cache,
1810  * async_flag maintenance, and oap_request
1811  */
1812 static void osc_ap_completion(const struct lu_env *env, struct client_obd *cli,
1813                               struct osc_async_page *oap, int sent, int rc)
1814 {
1815         struct osc_object *osc = oap->oap_obj;
1816         struct lov_oinfo *loi = osc->oo_oinfo;
1817         __u64 xid = 0;
1818
1819         if (oap->oap_request) {
1820                 xid = ptlrpc_req_xid(oap->oap_request);
1821                 ptlrpc_req_finished(oap->oap_request);
1822                 oap->oap_request = NULL;
1823         }
1824
1825         /* As the transfer for this page is being done, clear the flags */
1826         spin_lock(&oap->oap_lock);
1827         oap->oap_async_flags = 0;
1828         spin_unlock(&oap->oap_lock);
1829         oap->oap_interrupted = 0;
1830
1831         if (oap->oap_cmd & OBD_BRW_WRITE && xid > 0) {
1832                 client_obd_list_lock(&cli->cl_loi_list_lock);
1833                 osc_process_ar(&cli->cl_ar, xid, rc);
1834                 osc_process_ar(&loi->loi_ar, xid, rc);
1835                 client_obd_list_unlock(&cli->cl_loi_list_lock);
1836         }
1837
1838         rc = osc_completion(env, oap, oap->oap_cmd, rc);
1839         if (rc)
1840                 CERROR("completion on oap %p obj %p returns %d.\n",
1841                        oap, osc, rc);
1842 }
1843
1844 /**
1845  * Try to add extent to one RPC. We need to think about the following things:
1846  * - # of pages must not be over max_pages_per_rpc
1847  * - extent must be compatible with previous ones
1848  */
1849 static int try_to_add_extent_for_io(struct client_obd *cli,
1850                                     struct osc_extent *ext, struct list_head *rpclist,
1851                                     int *pc, unsigned int *max_pages)
1852 {
1853         struct osc_extent *tmp;
1854         struct osc_async_page *oap = list_first_entry(&ext->oe_pages,
1855                                                       struct osc_async_page,
1856                                                       oap_pending_item);
1857
1858         EASSERT((ext->oe_state == OES_CACHE || ext->oe_state == OES_LOCK_DONE),
1859                 ext);
1860
1861         *max_pages = max(ext->oe_mppr, *max_pages);
1862         if (*pc + ext->oe_nr_pages > *max_pages)
1863                 return 0;
1864
1865         list_for_each_entry(tmp, rpclist, oe_link) {
1866                 struct osc_async_page *oap2;
1867
1868                 oap2 = list_first_entry(&tmp->oe_pages, struct osc_async_page,
1869                                         oap_pending_item);
1870                 EASSERT(tmp->oe_owner == current, tmp);
1871                 if (oap2cl_page(oap)->cp_type != oap2cl_page(oap2)->cp_type) {
1872                         CDEBUG(D_CACHE, "Do not permit different type of IO"
1873                                         " for a same RPC\n");
1874                         return 0;
1875                 }
1876
1877                 if (tmp->oe_srvlock != ext->oe_srvlock ||
1878                     !tmp->oe_grants != !ext->oe_grants)
1879                         return 0;
1880
1881                 /* remove break for strict check */
1882                 break;
1883         }
1884
1885         *pc += ext->oe_nr_pages;
1886         list_move_tail(&ext->oe_link, rpclist);
1887         ext->oe_owner = current;
1888         return 1;
1889 }
1890
1891 /**
1892  * In order to prevent multiple ptlrpcd from breaking contiguous extents,
1893  * get_write_extent() takes all appropriate extents in atomic.
1894  *
1895  * The following policy is used to collect extents for IO:
1896  * 1. Add as many HP extents as possible;
1897  * 2. Add the first urgent extent in urgent extent list and take it out of
1898  *    urgent list;
1899  * 3. Add subsequent extents of this urgent extent;
1900  * 4. If urgent list is not empty, goto 2;
1901  * 5. Traverse the extent tree from the 1st extent;
1902  * 6. Above steps exit if there is no space in this RPC.
1903  */
1904 static int get_write_extents(struct osc_object *obj, struct list_head *rpclist)
1905 {
1906         struct client_obd *cli = osc_cli(obj);
1907         struct osc_extent *ext;
1908         struct osc_extent *temp;
1909         int page_count = 0;
1910         unsigned int max_pages = cli->cl_max_pages_per_rpc;
1911
1912         LASSERT(osc_object_is_locked(obj));
1913         list_for_each_entry_safe(ext, temp, &obj->oo_hp_exts, oe_link) {
1914                 LASSERT(ext->oe_state == OES_CACHE);
1915                 if (!try_to_add_extent_for_io(cli, ext, rpclist, &page_count,
1916                                               &max_pages))
1917                         return page_count;
1918                 EASSERT(ext->oe_nr_pages <= max_pages, ext);
1919         }
1920         if (page_count == max_pages)
1921                 return page_count;
1922
1923         while (!list_empty(&obj->oo_urgent_exts)) {
1924                 ext = list_entry(obj->oo_urgent_exts.next,
1925                                  struct osc_extent, oe_link);
1926                 if (!try_to_add_extent_for_io(cli, ext, rpclist, &page_count,
1927                                               &max_pages))
1928                         return page_count;
1929
1930                 if (!ext->oe_intree)
1931                         continue;
1932
1933                 while ((ext = next_extent(ext)) != NULL) {
1934                         if ((ext->oe_state != OES_CACHE) ||
1935                             (!list_empty(&ext->oe_link) &&
1936                              ext->oe_owner))
1937                                 continue;
1938
1939                         if (!try_to_add_extent_for_io(cli, ext, rpclist,
1940                                                       &page_count, &max_pages))
1941                                 return page_count;
1942                 }
1943         }
1944         if (page_count == max_pages)
1945                 return page_count;
1946
1947         ext = first_extent(obj);
1948         while (ext) {
1949                 if ((ext->oe_state != OES_CACHE) ||
1950                     /* this extent may be already in current rpclist */
1951                     (!list_empty(&ext->oe_link) && ext->oe_owner)) {
1952                         ext = next_extent(ext);
1953                         continue;
1954                 }
1955
1956                 if (!try_to_add_extent_for_io(cli, ext, rpclist, &page_count,
1957                                               &max_pages))
1958                         return page_count;
1959
1960                 ext = next_extent(ext);
1961         }
1962         return page_count;
1963 }
1964
1965 static int
1966 osc_send_write_rpc(const struct lu_env *env, struct client_obd *cli,
1967                    struct osc_object *osc)
1968         __must_hold(osc)
1969 {
1970         LIST_HEAD(rpclist);
1971         struct osc_extent *ext;
1972         struct osc_extent *tmp;
1973         struct osc_extent *first = NULL;
1974         u32 page_count = 0;
1975         int srvlock = 0;
1976         int rc = 0;
1977
1978         LASSERT(osc_object_is_locked(osc));
1979
1980         page_count = get_write_extents(osc, &rpclist);
1981         LASSERT(equi(page_count == 0, list_empty(&rpclist)));
1982
1983         if (list_empty(&rpclist))
1984                 return 0;
1985
1986         osc_update_pending(osc, OBD_BRW_WRITE, -page_count);
1987
1988         list_for_each_entry(ext, &rpclist, oe_link) {
1989                 LASSERT(ext->oe_state == OES_CACHE ||
1990                         ext->oe_state == OES_LOCK_DONE);
1991                 if (ext->oe_state == OES_CACHE)
1992                         osc_extent_state_set(ext, OES_LOCKING);
1993                 else
1994                         osc_extent_state_set(ext, OES_RPC);
1995         }
1996
1997         /* we're going to grab page lock, so release object lock because
1998          * lock order is page lock -> object lock.
1999          */
2000         osc_object_unlock(osc);
2001
2002         list_for_each_entry_safe(ext, tmp, &rpclist, oe_link) {
2003                 if (ext->oe_state == OES_LOCKING) {
2004                         rc = osc_extent_make_ready(env, ext);
2005                         if (unlikely(rc < 0)) {
2006                                 list_del_init(&ext->oe_link);
2007                                 osc_extent_finish(env, ext, 0, rc);
2008                                 continue;
2009                         }
2010                 }
2011                 if (!first) {
2012                         first = ext;
2013                         srvlock = ext->oe_srvlock;
2014                 } else {
2015                         LASSERT(srvlock == ext->oe_srvlock);
2016                 }
2017         }
2018
2019         if (!list_empty(&rpclist)) {
2020                 LASSERT(page_count > 0);
2021                 rc = osc_build_rpc(env, cli, &rpclist, OBD_BRW_WRITE);
2022                 LASSERT(list_empty(&rpclist));
2023         }
2024
2025         osc_object_lock(osc);
2026         return rc;
2027 }
2028
2029 /**
2030  * prepare pages for ASYNC io and put pages in send queue.
2031  *
2032  * \param cmd OBD_BRW_* macroses
2033  * \param lop pending pages
2034  *
2035  * \return zero if no page added to send queue.
2036  * \return 1 if pages successfully added to send queue.
2037  * \return negative on errors.
2038  */
2039 static int
2040 osc_send_read_rpc(const struct lu_env *env, struct client_obd *cli,
2041                   struct osc_object *osc)
2042         __must_hold(osc)
2043 {
2044         struct osc_extent *ext;
2045         struct osc_extent *next;
2046         LIST_HEAD(rpclist);
2047         int page_count = 0;
2048         unsigned int max_pages = cli->cl_max_pages_per_rpc;
2049         int rc = 0;
2050
2051         LASSERT(osc_object_is_locked(osc));
2052         list_for_each_entry_safe(ext, next, &osc->oo_reading_exts, oe_link) {
2053                 EASSERT(ext->oe_state == OES_LOCK_DONE, ext);
2054                 if (!try_to_add_extent_for_io(cli, ext, &rpclist, &page_count,
2055                                               &max_pages))
2056                         break;
2057                 osc_extent_state_set(ext, OES_RPC);
2058                 EASSERT(ext->oe_nr_pages <= max_pages, ext);
2059         }
2060         LASSERT(page_count <= max_pages);
2061
2062         osc_update_pending(osc, OBD_BRW_READ, -page_count);
2063
2064         if (!list_empty(&rpclist)) {
2065                 osc_object_unlock(osc);
2066
2067                 LASSERT(page_count > 0);
2068                 rc = osc_build_rpc(env, cli, &rpclist, OBD_BRW_READ);
2069                 LASSERT(list_empty(&rpclist));
2070
2071                 osc_object_lock(osc);
2072         }
2073         return rc;
2074 }
2075
2076 #define list_to_obj(list, item) ({                                            \
2077         struct list_head *__tmp = (list)->next;                               \
2078         list_del_init(__tmp);                                                 \
2079         list_entry(__tmp, struct osc_object, oo_##item);                      \
2080 })
2081
2082 /* This is called by osc_check_rpcs() to find which objects have pages that
2083  * we could be sending.  These lists are maintained by osc_makes_rpc().
2084  */
2085 static struct osc_object *osc_next_obj(struct client_obd *cli)
2086 {
2087         /* First return objects that have blocked locks so that they
2088          * will be flushed quickly and other clients can get the lock,
2089          * then objects which have pages ready to be stuffed into RPCs
2090          */
2091         if (!list_empty(&cli->cl_loi_hp_ready_list))
2092                 return list_to_obj(&cli->cl_loi_hp_ready_list, hp_ready_item);
2093         if (!list_empty(&cli->cl_loi_ready_list))
2094                 return list_to_obj(&cli->cl_loi_ready_list, ready_item);
2095
2096         /* then if we have cache waiters, return all objects with queued
2097          * writes.  This is especially important when many small files
2098          * have filled up the cache and not been fired into rpcs because
2099          * they don't pass the nr_pending/object threshold
2100          */
2101         if (!list_empty(&cli->cl_cache_waiters) &&
2102             !list_empty(&cli->cl_loi_write_list))
2103                 return list_to_obj(&cli->cl_loi_write_list, write_item);
2104
2105         /* then return all queued objects when we have an invalid import
2106          * so that they get flushed
2107          */
2108         if (!cli->cl_import || cli->cl_import->imp_invalid) {
2109                 if (!list_empty(&cli->cl_loi_write_list))
2110                         return list_to_obj(&cli->cl_loi_write_list, write_item);
2111                 if (!list_empty(&cli->cl_loi_read_list))
2112                         return list_to_obj(&cli->cl_loi_read_list, read_item);
2113         }
2114         return NULL;
2115 }
2116
2117 /* called with the loi list lock held */
2118 static void osc_check_rpcs(const struct lu_env *env, struct client_obd *cli)
2119         __must_hold(&cli->cl_loi_list_lock)
2120 {
2121         struct osc_object *osc;
2122         int rc = 0;
2123
2124         while ((osc = osc_next_obj(cli)) != NULL) {
2125                 struct cl_object *obj = osc2cl(osc);
2126                 struct lu_ref_link link;
2127
2128                 OSC_IO_DEBUG(osc, "%lu in flight\n", rpcs_in_flight(cli));
2129
2130                 if (osc_max_rpc_in_flight(cli, osc)) {
2131                         __osc_list_maint(cli, osc);
2132                         break;
2133                 }
2134
2135                 cl_object_get(obj);
2136                 client_obd_list_unlock(&cli->cl_loi_list_lock);
2137                 lu_object_ref_add_at(&obj->co_lu, &link, "check",
2138                                      current);
2139
2140                 /* attempt some read/write balancing by alternating between
2141                  * reads and writes in an object.  The makes_rpc checks here
2142                  * would be redundant if we were getting read/write work items
2143                  * instead of objects.  we don't want send_oap_rpc to drain a
2144                  * partial read pending queue when we're given this object to
2145                  * do io on writes while there are cache waiters
2146                  */
2147                 osc_object_lock(osc);
2148                 if (osc_makes_rpc(cli, osc, OBD_BRW_WRITE)) {
2149                         rc = osc_send_write_rpc(env, cli, osc);
2150                         if (rc < 0) {
2151                                 CERROR("Write request failed with %d\n", rc);
2152
2153                                 /* osc_send_write_rpc failed, mostly because of
2154                                  * memory pressure.
2155                                  *
2156                                  * It can't break here, because if:
2157                                  *  - a page was submitted by osc_io_submit, so
2158                                  *    page locked;
2159                                  *  - no request in flight
2160                                  *  - no subsequent request
2161                                  * The system will be in live-lock state,
2162                                  * because there is no chance to call
2163                                  * osc_io_unplug() and osc_check_rpcs() any
2164                                  * more. pdflush can't help in this case,
2165                                  * because it might be blocked at grabbing
2166                                  * the page lock as we mentioned.
2167                                  *
2168                                  * Anyway, continue to drain pages.
2169                                  */
2170                                 /* break; */
2171                         }
2172                 }
2173                 if (osc_makes_rpc(cli, osc, OBD_BRW_READ)) {
2174                         rc = osc_send_read_rpc(env, cli, osc);
2175                         if (rc < 0)
2176                                 CERROR("Read request failed with %d\n", rc);
2177                 }
2178                 osc_object_unlock(osc);
2179
2180                 osc_list_maint(cli, osc);
2181                 lu_object_ref_del_at(&obj->co_lu, &link, "check",
2182                                      current);
2183                 cl_object_put(env, obj);
2184
2185                 client_obd_list_lock(&cli->cl_loi_list_lock);
2186         }
2187 }
2188
2189 static int osc_io_unplug0(const struct lu_env *env, struct client_obd *cli,
2190                           struct osc_object *osc, int async)
2191 {
2192         int rc = 0;
2193
2194         if (osc && osc_list_maint(cli, osc) == 0)
2195                 return 0;
2196
2197         if (!async) {
2198                 /* disable osc_lru_shrink() temporarily to avoid
2199                  * potential stack overrun problem. LU-2859
2200                  */
2201                 atomic_inc(&cli->cl_lru_shrinkers);
2202                 client_obd_list_lock(&cli->cl_loi_list_lock);
2203                 osc_check_rpcs(env, cli);
2204                 client_obd_list_unlock(&cli->cl_loi_list_lock);
2205                 atomic_dec(&cli->cl_lru_shrinkers);
2206         } else {
2207                 CDEBUG(D_CACHE, "Queue writeback work for client %p.\n", cli);
2208                 LASSERT(cli->cl_writeback_work);
2209                 rc = ptlrpcd_queue_work(cli->cl_writeback_work);
2210         }
2211         return rc;
2212 }
2213
2214 static int osc_io_unplug_async(const struct lu_env *env,
2215                                struct client_obd *cli, struct osc_object *osc)
2216 {
2217         return osc_io_unplug0(env, cli, osc, 1);
2218 }
2219
2220 void osc_io_unplug(const struct lu_env *env, struct client_obd *cli,
2221                    struct osc_object *osc)
2222 {
2223         (void)osc_io_unplug0(env, cli, osc, 0);
2224 }
2225
2226 int osc_prep_async_page(struct osc_object *osc, struct osc_page *ops,
2227                         struct page *page, loff_t offset)
2228 {
2229         struct obd_export *exp = osc_export(osc);
2230         struct osc_async_page *oap = &ops->ops_oap;
2231
2232         if (!page)
2233                 return cfs_size_round(sizeof(*oap));
2234
2235         oap->oap_magic = OAP_MAGIC;
2236         oap->oap_cli = &exp->exp_obd->u.cli;
2237         oap->oap_obj = osc;
2238
2239         oap->oap_page = page;
2240         oap->oap_obj_off = offset;
2241         LASSERT(!(offset & ~CFS_PAGE_MASK));
2242
2243         if (!client_is_remote(exp) && capable(CFS_CAP_SYS_RESOURCE))
2244                 oap->oap_brw_flags = OBD_BRW_NOQUOTA;
2245
2246         INIT_LIST_HEAD(&oap->oap_pending_item);
2247         INIT_LIST_HEAD(&oap->oap_rpc_item);
2248
2249         spin_lock_init(&oap->oap_lock);
2250         CDEBUG(D_INFO, "oap %p page %p obj off %llu\n",
2251                oap, page, oap->oap_obj_off);
2252         return 0;
2253 }
2254
2255 int osc_queue_async_io(const struct lu_env *env, struct cl_io *io,
2256                        struct osc_page *ops)
2257 {
2258         struct osc_io *oio = osc_env_io(env);
2259         struct osc_extent *ext = NULL;
2260         struct osc_async_page *oap = &ops->ops_oap;
2261         struct client_obd *cli = oap->oap_cli;
2262         struct osc_object *osc = oap->oap_obj;
2263         pgoff_t index;
2264         int grants = 0;
2265         int brw_flags = OBD_BRW_ASYNC;
2266         int cmd = OBD_BRW_WRITE;
2267         int need_release = 0;
2268         int rc = 0;
2269
2270         if (oap->oap_magic != OAP_MAGIC)
2271                 return -EINVAL;
2272
2273         if (!cli->cl_import || cli->cl_import->imp_invalid)
2274                 return -EIO;
2275
2276         if (!list_empty(&oap->oap_pending_item) ||
2277             !list_empty(&oap->oap_rpc_item))
2278                 return -EBUSY;
2279
2280         /* Set the OBD_BRW_SRVLOCK before the page is queued. */
2281         brw_flags |= ops->ops_srvlock ? OBD_BRW_SRVLOCK : 0;
2282         if (!client_is_remote(osc_export(osc)) &&
2283             capable(CFS_CAP_SYS_RESOURCE)) {
2284                 brw_flags |= OBD_BRW_NOQUOTA;
2285                 cmd |= OBD_BRW_NOQUOTA;
2286         }
2287
2288         /* check if the file's owner/group is over quota */
2289         if (!(cmd & OBD_BRW_NOQUOTA)) {
2290                 struct cl_object *obj;
2291                 struct cl_attr *attr;
2292                 unsigned int qid[MAXQUOTAS];
2293
2294                 obj = cl_object_top(&osc->oo_cl);
2295                 attr = &osc_env_info(env)->oti_attr;
2296
2297                 cl_object_attr_lock(obj);
2298                 rc = cl_object_attr_get(env, obj, attr);
2299                 cl_object_attr_unlock(obj);
2300
2301                 qid[USRQUOTA] = attr->cat_uid;
2302                 qid[GRPQUOTA] = attr->cat_gid;
2303                 if (rc == 0 && osc_quota_chkdq(cli, qid) == NO_QUOTA)
2304                         rc = -EDQUOT;
2305                 if (rc)
2306                         return rc;
2307         }
2308
2309         oap->oap_cmd = cmd;
2310         oap->oap_page_off = ops->ops_from;
2311         oap->oap_count = ops->ops_to - ops->ops_from;
2312         oap->oap_async_flags = 0;
2313         oap->oap_brw_flags = brw_flags;
2314
2315         OSC_IO_DEBUG(osc, "oap %p page %p added for cmd %d\n",
2316                      oap, oap->oap_page, oap->oap_cmd & OBD_BRW_RWMASK);
2317
2318         index = oap2cl_page(oap)->cp_index;
2319
2320         /* Add this page into extent by the following steps:
2321          * 1. if there exists an active extent for this IO, mostly this page
2322          *    can be added to the active extent and sometimes we need to
2323          *    expand extent to accommodate this page;
2324          * 2. otherwise, a new extent will be allocated.
2325          */
2326
2327         ext = oio->oi_active;
2328         if (ext && ext->oe_start <= index && ext->oe_max_end >= index) {
2329                 /* one chunk plus extent overhead must be enough to write this
2330                  * page
2331                  */
2332                 grants = (1 << cli->cl_chunkbits) + cli->cl_extent_tax;
2333                 if (ext->oe_end >= index)
2334                         grants = 0;
2335
2336                 /* it doesn't need any grant to dirty this page */
2337                 client_obd_list_lock(&cli->cl_loi_list_lock);
2338                 rc = osc_enter_cache_try(cli, oap, grants, 0);
2339                 client_obd_list_unlock(&cli->cl_loi_list_lock);
2340                 if (rc == 0) { /* try failed */
2341                         grants = 0;
2342                         need_release = 1;
2343                 } else if (ext->oe_end < index) {
2344                         int tmp = grants;
2345                         /* try to expand this extent */
2346                         rc = osc_extent_expand(ext, index, &tmp);
2347                         if (rc < 0) {
2348                                 need_release = 1;
2349                                 /* don't free reserved grant */
2350                         } else {
2351                                 OSC_EXTENT_DUMP(D_CACHE, ext,
2352                                                 "expanded for %lu.\n", index);
2353                                 osc_unreserve_grant(cli, grants, tmp);
2354                                 grants = 0;
2355                         }
2356                 }
2357                 rc = 0;
2358         } else if (ext) {
2359                 /* index is located outside of active extent */
2360                 need_release = 1;
2361         }
2362         if (need_release) {
2363                 osc_extent_release(env, ext);
2364                 oio->oi_active = NULL;
2365                 ext = NULL;
2366         }
2367
2368         if (!ext) {
2369                 int tmp = (1 << cli->cl_chunkbits) + cli->cl_extent_tax;
2370
2371                 /* try to find new extent to cover this page */
2372                 LASSERT(!oio->oi_active);
2373                 /* we may have allocated grant for this page if we failed
2374                  * to expand the previous active extent.
2375                  */
2376                 LASSERT(ergo(grants > 0, grants >= tmp));
2377
2378                 rc = 0;
2379                 if (grants == 0) {
2380                         /* we haven't allocated grant for this page. */
2381                         rc = osc_enter_cache(env, cli, oap, tmp);
2382                         if (rc == 0)
2383                                 grants = tmp;
2384                 }
2385
2386                 tmp = grants;
2387                 if (rc == 0) {
2388                         ext = osc_extent_find(env, osc, index, &tmp);
2389                         if (IS_ERR(ext)) {
2390                                 LASSERT(tmp == grants);
2391                                 osc_exit_cache(cli, oap);
2392                                 rc = PTR_ERR(ext);
2393                                 ext = NULL;
2394                         } else {
2395                                 oio->oi_active = ext;
2396                         }
2397                 }
2398                 if (grants > 0)
2399                         osc_unreserve_grant(cli, grants, tmp);
2400         }
2401
2402         LASSERT(ergo(rc == 0, ext));
2403         if (ext) {
2404                 EASSERTF(ext->oe_end >= index && ext->oe_start <= index,
2405                          ext, "index = %lu.\n", index);
2406                 LASSERT((oap->oap_brw_flags & OBD_BRW_FROM_GRANT) != 0);
2407
2408                 osc_object_lock(osc);
2409                 if (ext->oe_nr_pages == 0)
2410                         ext->oe_srvlock = ops->ops_srvlock;
2411                 else
2412                         LASSERT(ext->oe_srvlock == ops->ops_srvlock);
2413                 ++ext->oe_nr_pages;
2414                 list_add_tail(&oap->oap_pending_item, &ext->oe_pages);
2415                 osc_object_unlock(osc);
2416         }
2417         return rc;
2418 }
2419
2420 int osc_teardown_async_page(const struct lu_env *env,
2421                             struct osc_object *obj, struct osc_page *ops)
2422 {
2423         struct osc_async_page *oap = &ops->ops_oap;
2424         struct osc_extent *ext = NULL;
2425         int rc = 0;
2426
2427         LASSERT(oap->oap_magic == OAP_MAGIC);
2428
2429         CDEBUG(D_INFO, "teardown oap %p page %p at index %lu.\n",
2430                oap, ops, oap2cl_page(oap)->cp_index);
2431
2432         osc_object_lock(obj);
2433         if (!list_empty(&oap->oap_rpc_item)) {
2434                 CDEBUG(D_CACHE, "oap %p is not in cache.\n", oap);
2435                 rc = -EBUSY;
2436         } else if (!list_empty(&oap->oap_pending_item)) {
2437                 ext = osc_extent_lookup(obj, oap2cl_page(oap)->cp_index);
2438                 /* only truncated pages are allowed to be taken out.
2439                  * See osc_extent_truncate() and osc_cache_truncate_start()
2440                  * for details.
2441                  */
2442                 if (ext && ext->oe_state != OES_TRUNC) {
2443                         OSC_EXTENT_DUMP(D_ERROR, ext, "trunc at %lu.\n",
2444                                         oap2cl_page(oap)->cp_index);
2445                         rc = -EBUSY;
2446                 }
2447         }
2448         osc_object_unlock(obj);
2449         if (ext)
2450                 osc_extent_put(env, ext);
2451         return rc;
2452 }
2453
2454 /**
2455  * This is called when a page is picked up by kernel to write out.
2456  *
2457  * We should find out the corresponding extent and add the whole extent
2458  * into urgent list. The extent may be being truncated or used, handle it
2459  * carefully.
2460  */
2461 int osc_flush_async_page(const struct lu_env *env, struct cl_io *io,
2462                          struct osc_page *ops)
2463 {
2464         struct osc_extent *ext = NULL;
2465         struct osc_object *obj = cl2osc(ops->ops_cl.cpl_obj);
2466         struct cl_page *cp = ops->ops_cl.cpl_page;
2467         pgoff_t index = cp->cp_index;
2468         struct osc_async_page *oap = &ops->ops_oap;
2469         bool unplug = false;
2470         int rc = 0;
2471
2472         osc_object_lock(obj);
2473         ext = osc_extent_lookup(obj, index);
2474         if (!ext) {
2475                 osc_extent_tree_dump(D_ERROR, obj);
2476                 LASSERTF(0, "page index %lu is NOT covered.\n", index);
2477         }
2478
2479         switch (ext->oe_state) {
2480         case OES_RPC:
2481         case OES_LOCK_DONE:
2482                 CL_PAGE_DEBUG(D_ERROR, env, cl_page_top(cp),
2483                               "flush an in-rpc page?\n");
2484                 LASSERT(0);
2485                 break;
2486         case OES_LOCKING:
2487                 /* If we know this extent is being written out, we should abort
2488                  * so that the writer can make this page ready. Otherwise, there
2489                  * exists a deadlock problem because other process can wait for
2490                  * page writeback bit holding page lock; and meanwhile in
2491                  * vvp_page_make_ready(), we need to grab page lock before
2492                  * really sending the RPC.
2493                  */
2494         case OES_TRUNC:
2495                 /* race with truncate, page will be redirtied */
2496         case OES_ACTIVE:
2497                 /* The extent is active so we need to abort and let the caller
2498                  * re-dirty the page. If we continued on here, and we were the
2499                  * one making the extent active, we could deadlock waiting for
2500                  * the page writeback to clear but it won't because the extent
2501                  * is active and won't be written out.
2502                  */
2503                 rc = -EAGAIN;
2504                 goto out;
2505         default:
2506                 break;
2507         }
2508
2509         rc = cl_page_prep(env, io, cl_page_top(cp), CRT_WRITE);
2510         if (rc)
2511                 goto out;
2512
2513         spin_lock(&oap->oap_lock);
2514         oap->oap_async_flags |= ASYNC_READY|ASYNC_URGENT;
2515         spin_unlock(&oap->oap_lock);
2516
2517         if (memory_pressure_get())
2518                 ext->oe_memalloc = 1;
2519
2520         ext->oe_urgent = 1;
2521         if (ext->oe_state == OES_CACHE) {
2522                 OSC_EXTENT_DUMP(D_CACHE, ext,
2523                                 "flush page %p make it urgent.\n", oap);
2524                 if (list_empty(&ext->oe_link))
2525                         list_add_tail(&ext->oe_link, &obj->oo_urgent_exts);
2526                 unplug = true;
2527         }
2528         rc = 0;
2529
2530 out:
2531         osc_object_unlock(obj);
2532         osc_extent_put(env, ext);
2533         if (unplug)
2534                 osc_io_unplug_async(env, osc_cli(obj), obj);
2535         return rc;
2536 }
2537
2538 /**
2539  * this is called when a sync waiter receives an interruption.  Its job is to
2540  * get the caller woken as soon as possible.  If its page hasn't been put in an
2541  * rpc yet it can dequeue immediately.  Otherwise it has to mark the rpc as
2542  * desiring interruption which will forcefully complete the rpc once the rpc
2543  * has timed out.
2544  */
2545 int osc_cancel_async_page(const struct lu_env *env, struct osc_page *ops)
2546 {
2547         struct osc_async_page *oap = &ops->ops_oap;
2548         struct osc_object *obj = oap->oap_obj;
2549         struct client_obd *cli = osc_cli(obj);
2550         struct osc_extent *ext;
2551         struct osc_extent *found = NULL;
2552         struct list_head *plist;
2553         pgoff_t index = oap2cl_page(oap)->cp_index;
2554         int rc = -EBUSY;
2555         int cmd;
2556
2557         LASSERT(!oap->oap_interrupted);
2558         oap->oap_interrupted = 1;
2559
2560         /* Find out the caching extent */
2561         osc_object_lock(obj);
2562         if (oap->oap_cmd & OBD_BRW_WRITE) {
2563                 plist = &obj->oo_urgent_exts;
2564                 cmd = OBD_BRW_WRITE;
2565         } else {
2566                 plist = &obj->oo_reading_exts;
2567                 cmd = OBD_BRW_READ;
2568         }
2569         list_for_each_entry(ext, plist, oe_link) {
2570                 if (ext->oe_start <= index && ext->oe_end >= index) {
2571                         LASSERT(ext->oe_state == OES_LOCK_DONE);
2572                         /* For OES_LOCK_DONE state extent, it has already held
2573                          * a refcount for RPC.
2574                          */
2575                         found = osc_extent_get(ext);
2576                         break;
2577                 }
2578         }
2579         if (found) {
2580                 list_del_init(&found->oe_link);
2581                 osc_update_pending(obj, cmd, -found->oe_nr_pages);
2582                 osc_object_unlock(obj);
2583
2584                 osc_extent_finish(env, found, 0, -EINTR);
2585                 osc_extent_put(env, found);
2586                 rc = 0;
2587         } else {
2588                 osc_object_unlock(obj);
2589                 /* ok, it's been put in an rpc. only one oap gets a request
2590                  * reference
2591                  */
2592                 if (oap->oap_request) {
2593                         ptlrpc_mark_interrupted(oap->oap_request);
2594                         ptlrpcd_wake(oap->oap_request);
2595                         ptlrpc_req_finished(oap->oap_request);
2596                         oap->oap_request = NULL;
2597                 }
2598         }
2599
2600         osc_list_maint(cli, obj);
2601         return rc;
2602 }
2603
2604 int osc_queue_sync_pages(const struct lu_env *env, struct osc_object *obj,
2605                          struct list_head *list, int cmd, int brw_flags)
2606 {
2607         struct client_obd *cli = osc_cli(obj);
2608         struct osc_extent *ext;
2609         struct osc_async_page *oap, *tmp;
2610         int page_count = 0;
2611         int mppr = cli->cl_max_pages_per_rpc;
2612         pgoff_t start = CL_PAGE_EOF;
2613         pgoff_t end = 0;
2614
2615         list_for_each_entry(oap, list, oap_pending_item) {
2616                 struct cl_page *cp = oap2cl_page(oap);
2617
2618                 if (cp->cp_index > end)
2619                         end = cp->cp_index;
2620                 if (cp->cp_index < start)
2621                         start = cp->cp_index;
2622                 ++page_count;
2623                 mppr <<= (page_count > mppr);
2624         }
2625
2626         ext = osc_extent_alloc(obj);
2627         if (!ext) {
2628                 list_for_each_entry_safe(oap, tmp, list, oap_pending_item) {
2629                         list_del_init(&oap->oap_pending_item);
2630                         osc_ap_completion(env, cli, oap, 0, -ENOMEM);
2631                 }
2632                 return -ENOMEM;
2633         }
2634
2635         ext->oe_rw = !!(cmd & OBD_BRW_READ);
2636         ext->oe_urgent = 1;
2637         ext->oe_start = start;
2638         ext->oe_end = ext->oe_max_end = end;
2639         ext->oe_obj = obj;
2640         ext->oe_srvlock = !!(brw_flags & OBD_BRW_SRVLOCK);
2641         ext->oe_nr_pages = page_count;
2642         ext->oe_mppr = mppr;
2643         list_splice_init(list, &ext->oe_pages);
2644
2645         osc_object_lock(obj);
2646         /* Reuse the initial refcount for RPC, don't drop it */
2647         osc_extent_state_set(ext, OES_LOCK_DONE);
2648         if (cmd & OBD_BRW_WRITE) {
2649                 list_add_tail(&ext->oe_link, &obj->oo_urgent_exts);
2650                 osc_update_pending(obj, OBD_BRW_WRITE, page_count);
2651         } else {
2652                 list_add_tail(&ext->oe_link, &obj->oo_reading_exts);
2653                 osc_update_pending(obj, OBD_BRW_READ, page_count);
2654         }
2655         osc_object_unlock(obj);
2656
2657         osc_io_unplug_async(env, cli, obj);
2658         return 0;
2659 }
2660
2661 /**
2662  * Called by osc_io_setattr_start() to freeze and destroy covering extents.
2663  */
2664 int osc_cache_truncate_start(const struct lu_env *env, struct osc_io *oio,
2665                              struct osc_object *obj, __u64 size)
2666 {
2667         struct client_obd *cli = osc_cli(obj);
2668         struct osc_extent *ext;
2669         struct osc_extent *temp;
2670         struct osc_extent *waiting = NULL;
2671         pgoff_t index;
2672         LIST_HEAD(list);
2673         int result = 0;
2674         bool partial;
2675
2676         /* pages with index greater or equal to index will be truncated. */
2677         index = cl_index(osc2cl(obj), size);
2678         partial = size > cl_offset(osc2cl(obj), index);
2679
2680 again:
2681         osc_object_lock(obj);
2682         ext = osc_extent_search(obj, index);
2683         if (!ext)
2684                 ext = first_extent(obj);
2685         else if (ext->oe_end < index)
2686                 ext = next_extent(ext);
2687         while (ext) {
2688                 EASSERT(ext->oe_state != OES_TRUNC, ext);
2689
2690                 if (ext->oe_state > OES_CACHE || ext->oe_urgent) {
2691                         /* if ext is in urgent state, it means there must exist
2692                          * a page already having been flushed by write_page().
2693                          * We have to wait for this extent because we can't
2694                          * truncate that page.
2695                          */
2696                         LASSERT(!ext->oe_hp);
2697                         OSC_EXTENT_DUMP(D_CACHE, ext,
2698                                         "waiting for busy extent\n");
2699                         waiting = osc_extent_get(ext);
2700                         break;
2701                 }
2702
2703                 OSC_EXTENT_DUMP(D_CACHE, ext, "try to trunc:%llu.\n", size);
2704
2705                 osc_extent_get(ext);
2706                 if (ext->oe_state == OES_ACTIVE) {
2707                         /* though we grab inode mutex for write path, but we
2708                          * release it before releasing extent(in osc_io_end()),
2709                          * so there is a race window that an extent is still
2710                          * in OES_ACTIVE when truncate starts.
2711                          */
2712                         LASSERT(!ext->oe_trunc_pending);
2713                         ext->oe_trunc_pending = 1;
2714                 } else {
2715                         EASSERT(ext->oe_state == OES_CACHE, ext);
2716                         osc_extent_state_set(ext, OES_TRUNC);
2717                         osc_update_pending(obj, OBD_BRW_WRITE,
2718                                            -ext->oe_nr_pages);
2719                 }
2720                 EASSERT(list_empty(&ext->oe_link), ext);
2721                 list_add_tail(&ext->oe_link, &list);
2722
2723                 ext = next_extent(ext);
2724         }
2725         osc_object_unlock(obj);
2726
2727         osc_list_maint(cli, obj);
2728
2729         list_for_each_entry_safe(ext, temp, &list, oe_link) {
2730                 int rc;
2731
2732                 list_del_init(&ext->oe_link);
2733
2734                 /* extent may be in OES_ACTIVE state because inode mutex
2735                  * is released before osc_io_end() in file write case
2736                  */
2737                 if (ext->oe_state != OES_TRUNC)
2738                         osc_extent_wait(env, ext, OES_TRUNC);
2739
2740                 rc = osc_extent_truncate(ext, index, partial);
2741                 if (rc < 0) {
2742                         if (result == 0)
2743                                 result = rc;
2744
2745                         OSC_EXTENT_DUMP(D_ERROR, ext,
2746                                         "truncate error %d\n", rc);
2747                 } else if (ext->oe_nr_pages == 0) {
2748                         osc_extent_remove(ext);
2749                 } else {
2750                         /* this must be an overlapped extent which means only
2751                          * part of pages in this extent have been truncated.
2752                          */
2753                         EASSERTF(ext->oe_start <= index, ext,
2754                                  "trunc index = %lu/%d.\n", index, partial);
2755                         /* fix index to skip this partially truncated extent */
2756                         index = ext->oe_end + 1;
2757                         partial = false;
2758
2759                         /* we need to hold this extent in OES_TRUNC state so
2760                          * that no writeback will happen. This is to avoid
2761                          * BUG 17397.
2762                          */
2763                         LASSERT(!oio->oi_trunc);
2764                         oio->oi_trunc = osc_extent_get(ext);
2765                         OSC_EXTENT_DUMP(D_CACHE, ext,
2766                                         "trunc at %llu\n", size);
2767                 }
2768                 osc_extent_put(env, ext);
2769         }
2770         if (waiting) {
2771                 int rc;
2772
2773                 /* ignore the result of osc_extent_wait the write initiator
2774                  * should take care of it.
2775                  */
2776                 rc = osc_extent_wait(env, waiting, OES_INV);
2777                 if (rc < 0)
2778                         OSC_EXTENT_DUMP(D_CACHE, waiting, "error: %d.\n", rc);
2779
2780                 osc_extent_put(env, waiting);
2781                 waiting = NULL;
2782                 goto again;
2783         }
2784         return result;
2785 }
2786
2787 /**
2788  * Called after osc_io_setattr_end to add oio->oi_trunc back to cache.
2789  */
2790 void osc_cache_truncate_end(const struct lu_env *env, struct osc_io *oio,
2791                             struct osc_object *obj)
2792 {
2793         struct osc_extent *ext = oio->oi_trunc;
2794
2795         oio->oi_trunc = NULL;
2796         if (ext) {
2797                 bool unplug = false;
2798
2799                 EASSERT(ext->oe_nr_pages > 0, ext);
2800                 EASSERT(ext->oe_state == OES_TRUNC, ext);
2801                 EASSERT(!ext->oe_urgent, ext);
2802
2803                 OSC_EXTENT_DUMP(D_CACHE, ext, "trunc -> cache.\n");
2804                 osc_object_lock(obj);
2805                 osc_extent_state_set(ext, OES_CACHE);
2806                 if (ext->oe_fsync_wait && !ext->oe_urgent) {
2807                         ext->oe_urgent = 1;
2808                         list_move_tail(&ext->oe_link, &obj->oo_urgent_exts);
2809                         unplug = true;
2810                 }
2811                 osc_update_pending(obj, OBD_BRW_WRITE, ext->oe_nr_pages);
2812                 osc_object_unlock(obj);
2813                 osc_extent_put(env, ext);
2814
2815                 if (unplug)
2816                         osc_io_unplug_async(env, osc_cli(obj), obj);
2817         }
2818 }
2819
2820 /**
2821  * Wait for extents in a specific range to be written out.
2822  * The caller must have called osc_cache_writeback_range() to issue IO
2823  * otherwise it will take a long time for this function to finish.
2824  *
2825  * Caller must hold inode_mutex , or cancel exclusive dlm lock so that
2826  * nobody else can dirty this range of file while we're waiting for
2827  * extents to be written.
2828  */
2829 int osc_cache_wait_range(const struct lu_env *env, struct osc_object *obj,
2830                          pgoff_t start, pgoff_t end)
2831 {
2832         struct osc_extent *ext;
2833         pgoff_t index = start;
2834         int result = 0;
2835
2836 again:
2837         osc_object_lock(obj);
2838         ext = osc_extent_search(obj, index);
2839         if (!ext)
2840                 ext = first_extent(obj);
2841         else if (ext->oe_end < index)
2842                 ext = next_extent(ext);
2843         while (ext) {
2844                 int rc;
2845
2846                 if (ext->oe_start > end)
2847                         break;
2848
2849                 if (!ext->oe_fsync_wait) {
2850                         ext = next_extent(ext);
2851                         continue;
2852                 }
2853
2854                 EASSERT(ergo(ext->oe_state == OES_CACHE,
2855                              ext->oe_hp || ext->oe_urgent), ext);
2856                 EASSERT(ergo(ext->oe_state == OES_ACTIVE,
2857                              !ext->oe_hp && ext->oe_urgent), ext);
2858
2859                 index = ext->oe_end + 1;
2860                 osc_extent_get(ext);
2861                 osc_object_unlock(obj);
2862
2863                 rc = osc_extent_wait(env, ext, OES_INV);
2864                 if (result == 0)
2865                         result = rc;
2866                 osc_extent_put(env, ext);
2867                 goto again;
2868         }
2869         osc_object_unlock(obj);
2870
2871         OSC_IO_DEBUG(obj, "sync file range.\n");
2872         return result;
2873 }
2874
2875 /**
2876  * Called to write out a range of osc object.
2877  *
2878  * @hp     : should be set this is caused by lock cancel;
2879  * @discard: is set if dirty pages should be dropped - file will be deleted or
2880  *         truncated, this implies there is no partially discarding extents.
2881  *
2882  * Return how many pages will be issued, or error code if error occurred.
2883  */
2884 int osc_cache_writeback_range(const struct lu_env *env, struct osc_object *obj,
2885                               pgoff_t start, pgoff_t end, int hp, int discard)
2886 {
2887         struct osc_extent *ext;
2888         LIST_HEAD(discard_list);
2889         bool unplug = false;
2890         int result = 0;
2891
2892         osc_object_lock(obj);
2893         ext = osc_extent_search(obj, start);
2894         if (!ext)
2895                 ext = first_extent(obj);
2896         else if (ext->oe_end < start)
2897                 ext = next_extent(ext);
2898         while (ext) {
2899                 if (ext->oe_start > end)
2900                         break;
2901
2902                 ext->oe_fsync_wait = 1;
2903                 switch (ext->oe_state) {
2904                 case OES_CACHE:
2905                         result += ext->oe_nr_pages;
2906                         if (!discard) {
2907                                 struct list_head *list = NULL;
2908
2909                                 if (hp) {
2910                                         EASSERT(!ext->oe_hp, ext);
2911                                         ext->oe_hp = 1;
2912                                         list = &obj->oo_hp_exts;
2913                                 } else if (!ext->oe_urgent) {
2914                                         ext->oe_urgent = 1;
2915                                         list = &obj->oo_urgent_exts;
2916                                 }
2917                                 if (list)
2918                                         list_move_tail(&ext->oe_link, list);
2919                                 unplug = true;
2920                         } else {
2921                                 /* the only discarder is lock cancelling, so
2922                                  * [start, end] must contain this extent
2923                                  */
2924                                 EASSERT(ext->oe_start >= start &&
2925                                         ext->oe_max_end <= end, ext);
2926                                 osc_extent_state_set(ext, OES_LOCKING);
2927                                 ext->oe_owner = current;
2928                                 list_move_tail(&ext->oe_link, &discard_list);
2929                                 osc_update_pending(obj, OBD_BRW_WRITE,
2930                                                    -ext->oe_nr_pages);
2931                         }
2932                         break;
2933                 case OES_ACTIVE:
2934                         /* It's pretty bad to wait for ACTIVE extents, because
2935                          * we don't know how long we will wait for it to be
2936                          * flushed since it may be blocked at awaiting more
2937                          * grants. We do this for the correctness of fsync.
2938                          */
2939                         LASSERT(hp == 0 && discard == 0);
2940                         ext->oe_urgent = 1;
2941                         break;
2942                 case OES_TRUNC:
2943                         /* this extent is being truncated, can't do anything
2944                          * for it now. it will be set to urgent after truncate
2945                          * is finished in osc_cache_truncate_end().
2946                          */
2947                 default:
2948                         break;
2949                 }
2950                 ext = next_extent(ext);
2951         }
2952         osc_object_unlock(obj);
2953
2954         LASSERT(ergo(!discard, list_empty(&discard_list)));
2955         if (!list_empty(&discard_list)) {
2956                 struct osc_extent *tmp;
2957                 int rc;
2958
2959                 osc_list_maint(osc_cli(obj), obj);
2960                 list_for_each_entry_safe(ext, tmp, &discard_list, oe_link) {
2961                         list_del_init(&ext->oe_link);
2962                         EASSERT(ext->oe_state == OES_LOCKING, ext);
2963
2964                         /* Discard caching pages. We don't actually write this
2965                          * extent out but we complete it as if we did.
2966                          */
2967                         rc = osc_extent_make_ready(env, ext);
2968                         if (unlikely(rc < 0)) {
2969                                 OSC_EXTENT_DUMP(D_ERROR, ext,
2970                                                 "make_ready returned %d\n", rc);
2971                                 if (result >= 0)
2972                                         result = rc;
2973                         }
2974
2975                         /* finish the extent as if the pages were sent */
2976                         osc_extent_finish(env, ext, 0, 0);
2977                 }
2978         }
2979
2980         if (unplug)
2981                 osc_io_unplug(env, osc_cli(obj), obj);
2982
2983         if (hp || discard) {
2984                 int rc;
2985
2986                 rc = osc_cache_wait_range(env, obj, start, end);
2987                 if (result >= 0 && rc < 0)
2988                         result = rc;
2989         }
2990
2991         OSC_IO_DEBUG(obj, "cache page out.\n");
2992         return result;
2993 }
2994
2995 /** @} osc */