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