drivers/staging/lustre/lustre/llite/statahead.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) 2011, 2015, Intel Corporation.
  31  */
  32 /*
  33  * This file is part of Lustre, http://www.lustre.org/
  34  * Lustre is a trademark of Sun Microsystems, Inc.
  35  */
  36
  37 #include <linux/fs.h>
  38 #include <linux/sched.h>
  39 #include <linux/mm.h>
  40 #include <linux/highmem.h>
  41 #include <linux/pagemap.h>
  42
  43 #define DEBUG_SUBSYSTEM S_LLITE
  44
  45 #include "../include/obd_support.h"
  46 #include "../include/lustre_lite.h"
  47 #include "../include/lustre_dlm.h"
  48 #include "llite_internal.h"
  49
  50 #define SA_OMITTED_ENTRY_MAX 8ULL
  51
  52 enum se_stat {
  53         /** negative values are for error cases */
  54         SA_ENTRY_INIT = 0,      /** init entry */
  55         SA_ENTRY_SUCC = 1,      /** stat succeed */
  56         SA_ENTRY_INVA = 2,      /** invalid entry */
  57         SA_ENTRY_DEST = 3,      /** entry to be destroyed */
  58 };
  59
  60 struct ll_sa_entry {
  61         /* link into sai->sai_entries */
  62         struct list_head              se_link;
  63         /* link into sai->sai_entries_{received,stated} */
  64         struct list_head              se_list;
  65         /* link into sai hash table locally */
  66         struct list_head              se_hash;
  67         /* entry reference count */
  68         atomic_t            se_refcount;
  69         /* entry index in the sai */
  70         __u64              se_index;
  71         /* low layer ldlm lock handle */
  72         __u64              se_handle;
  73         /* entry status */
  74         enum se_stat       se_stat;
  75         /* entry size, contains name */
  76         int                  se_size;
  77         /* pointer to async getattr enqueue info */
  78         struct md_enqueue_info *se_minfo;
  79         /* pointer to the async getattr request */
  80         struct ptlrpc_request  *se_req;
  81         /* pointer to the target inode */
  82         struct inode       *se_inode;
  83         /* entry name */
  84         struct qstr          se_qstr;
  85 };
  86
  87 static unsigned int sai_generation;
  88 static DEFINE_SPINLOCK(sai_generation_lock);
  89
  90 /*
  91  * The entry only can be released by the caller, it is necessary to hold lock.
  92  */
  93 static inline int ll_sa_entry_stated(struct ll_sa_entry *entry)
  94 {
  95         smp_rmb();
  96         return (entry->se_stat != SA_ENTRY_INIT);
  97 }
  98
  99 static inline int ll_sa_entry_hash(int val)
 100 {
 101         return val & LL_SA_CACHE_MASK;
 102 }
 103
 104 /*
 105  * Insert entry to hash SA table.
 106  */
 107 static inline void
 108 ll_sa_entry_enhash(struct ll_statahead_info *sai, struct ll_sa_entry *entry)
 109 {
 110         int i = ll_sa_entry_hash(entry->se_qstr.hash);
 111
 112         spin_lock(&sai->sai_cache_lock[i]);
 113         list_add_tail(&entry->se_hash, &sai->sai_cache[i]);
 114         spin_unlock(&sai->sai_cache_lock[i]);
 115 }
 116
 117 /*
 118  * Remove entry from SA table.
 119  */
 120 static inline void
 121 ll_sa_entry_unhash(struct ll_statahead_info *sai, struct ll_sa_entry *entry)
 122 {
 123         int i = ll_sa_entry_hash(entry->se_qstr.hash);
 124
 125         spin_lock(&sai->sai_cache_lock[i]);
 126         list_del_init(&entry->se_hash);
 127         spin_unlock(&sai->sai_cache_lock[i]);
 128 }
 129
 130 static inline int agl_should_run(struct ll_statahead_info *sai,
 131                                  struct inode *inode)
 132 {
 133         return (inode && S_ISREG(inode->i_mode) && sai->sai_agl_valid);
 134 }
 135
 136 static inline int sa_sent_full(struct ll_statahead_info *sai)
 137 {
 138         return atomic_read(&sai->sai_cache_count) >= sai->sai_max;
 139 }
 140
 141 static inline int sa_received_empty(struct ll_statahead_info *sai)
 142 {
 143         return list_empty(&sai->sai_entries_received);
 144 }
 145
 146 static inline int agl_list_empty(struct ll_statahead_info *sai)
 147 {
 148         return list_empty(&sai->sai_entries_agl);
 149 }
 150
 151 /**
 152  * (1) hit ratio less than 80%
 153  * or
 154  * (2) consecutive miss more than 8
 155  * then means low hit.
 156  */
 157 static inline int sa_low_hit(struct ll_statahead_info *sai)
 158 {
 159         return ((sai->sai_hit > 7 && sai->sai_hit < 4 * sai->sai_miss) ||
 160                 (sai->sai_consecutive_miss > 8));
 161 }
 162
 163 /*
 164  * If the given index is behind of statahead window more than
 165  * SA_OMITTED_ENTRY_MAX, then it is old.
 166  */
 167 static inline int is_omitted_entry(struct ll_statahead_info *sai, __u64 index)
 168 {
 169         return ((__u64)sai->sai_max + index + SA_OMITTED_ENTRY_MAX <
 170                  sai->sai_index);
 171 }
 172
 173 /*
 174  * Insert it into sai_entries tail when init.
 175  */
 176 static struct ll_sa_entry *
 177 ll_sa_entry_alloc(struct ll_statahead_info *sai, __u64 index,
 178                   const char *name, int len)
 179 {
 180         struct ll_inode_info *lli;
 181         struct ll_sa_entry   *entry;
 182         int                entry_size;
 183         char             *dname;
 184
 185         entry_size = sizeof(struct ll_sa_entry) + (len & ~3) + 4;
 186         entry = kzalloc(entry_size, GFP_NOFS);
 187         if (unlikely(!entry))
 188                 return ERR_PTR(-ENOMEM);
 189
 190         CDEBUG(D_READA, "alloc sa entry %.*s(%p) index %llu\n",
 191                len, name, entry, index);
 192
 193         entry->se_index = index;
 194
 195         /*
 196          * Statahead entry reference rules:
 197          *
 198          * 1) When statahead entry is initialized, its reference is set as 2.
 199          *    One reference is used by the directory scanner. When the scanner
 200          *    searches the statahead cache for the given name, it can perform
 201          *    lockless hash lookup (only the scanner can remove entry from hash
 202          *    list), and once found, it needn't to call "atomic_inc()" for the
 203          *    entry reference. So the performance is improved. After using the
 204          *    statahead entry, the scanner will call "atomic_dec()" to drop the
 205          *    reference held when initialization. If it is the last reference,
 206          *    the statahead entry will be freed.
 207          *
 208          * 2) All other threads, including statahead thread and ptlrpcd thread,
 209          *    when they process the statahead entry, the reference for target
 210          *    should be held to guarantee the entry will not be released by the
 211          *    directory scanner. After processing the entry, these threads will
 212          *    drop the entry reference. If it is the last reference, the entry
 213          *    will be freed.
 214          *
 215          *    The second reference when initializes the statahead entry is used
 216          *    by the statahead thread, following the rule 2).
 217          */
 218         atomic_set(&entry->se_refcount, 2);
 219         entry->se_stat = SA_ENTRY_INIT;
 220         entry->se_size = entry_size;
 221         dname = (char *)entry + sizeof(struct ll_sa_entry);
 222         memcpy(dname, name, len);
 223         dname[len] = 0;
 224         entry->se_qstr.hash = full_name_hash(name, len);
 225         entry->se_qstr.len = len;
 226         entry->se_qstr.name = dname;
 227
 228         lli = ll_i2info(sai->sai_inode);
 229         spin_lock(&lli->lli_sa_lock);
 230         list_add_tail(&entry->se_link, &sai->sai_entries);
 231         INIT_LIST_HEAD(&entry->se_list);
 232         ll_sa_entry_enhash(sai, entry);
 233         spin_unlock(&lli->lli_sa_lock);
 234
 235         atomic_inc(&sai->sai_cache_count);
 236
 237         return entry;
 238 }
 239
 240 /*
 241  * Used by the directory scanner to search entry with name.
 242  *
 243  * Only the caller can remove the entry from hash, so it is unnecessary to hold
 244  * hash lock. It is caller's duty to release the init refcount on the entry, so
 245  * it is also unnecessary to increase refcount on the entry.
 246  */
 247 static struct ll_sa_entry *
 248 ll_sa_entry_get_byname(struct ll_statahead_info *sai, const struct qstr *qstr)
 249 {
 250         struct ll_sa_entry *entry;
 251         int i = ll_sa_entry_hash(qstr->hash);
 252
 253         list_for_each_entry(entry, &sai->sai_cache[i], se_hash) {
 254                 if (entry->se_qstr.hash == qstr->hash &&
 255                     entry->se_qstr.len == qstr->len &&
 256                     memcmp(entry->se_qstr.name, qstr->name, qstr->len) == 0)
 257                         return entry;
 258         }
 259         return NULL;
 260 }
 261
 262 /*
 263  * Used by the async getattr request callback to find entry with index.
 264  *
 265  * Inside lli_sa_lock to prevent others to change the list during the search.
 266  * It needs to increase entry refcount before returning to guarantee that the
 267  * entry cannot be freed by others.
 268  */
 269 static struct ll_sa_entry *
 270 ll_sa_entry_get_byindex(struct ll_statahead_info *sai, __u64 index)
 271 {
 272         struct ll_sa_entry *entry;
 273
 274         list_for_each_entry(entry, &sai->sai_entries, se_link) {
 275                 if (entry->se_index == index) {
 276                         LASSERT(atomic_read(&entry->se_refcount) > 0);
 277                         atomic_inc(&entry->se_refcount);
 278                         return entry;
 279                 }
 280                 if (entry->se_index > index)
 281                         break;
 282         }
 283         return NULL;
 284 }
 285
 286 static void ll_sa_entry_cleanup(struct ll_statahead_info *sai,
 287                                 struct ll_sa_entry *entry)
 288 {
 289         struct md_enqueue_info *minfo = entry->se_minfo;
 290         struct ptlrpc_request  *req   = entry->se_req;
 291
 292         if (minfo) {
 293                 entry->se_minfo = NULL;
 294                 ll_intent_release(&minfo->mi_it);
 295                 iput(minfo->mi_dir);
 296                 kfree(minfo);
 297         }
 298
 299         if (req) {
 300                 entry->se_req = NULL;
 301                 ptlrpc_req_finished(req);
 302         }
 303 }
 304
 305 static void ll_sa_entry_put(struct ll_statahead_info *sai,
 306                             struct ll_sa_entry *entry)
 307 {
 308         if (atomic_dec_and_test(&entry->se_refcount)) {
 309                 CDEBUG(D_READA, "free sa entry %.*s(%p) index %llu\n",
 310                        entry->se_qstr.len, entry->se_qstr.name, entry,
 311                        entry->se_index);
 312
 313                 LASSERT(list_empty(&entry->se_link));
 314                 LASSERT(list_empty(&entry->se_list));
 315                 LASSERT(list_empty(&entry->se_hash));
 316
 317                 ll_sa_entry_cleanup(sai, entry);
 318                 iput(entry->se_inode);
 319
 320                 kfree(entry);
 321                 atomic_dec(&sai->sai_cache_count);
 322         }
 323 }
 324
 325 static inline void
 326 do_sa_entry_fini(struct ll_statahead_info *sai, struct ll_sa_entry *entry)
 327 {
 328         struct ll_inode_info *lli = ll_i2info(sai->sai_inode);
 329
 330         LASSERT(!list_empty(&entry->se_hash));
 331         LASSERT(!list_empty(&entry->se_link));
 332
 333         ll_sa_entry_unhash(sai, entry);
 334
 335         spin_lock(&lli->lli_sa_lock);
 336         entry->se_stat = SA_ENTRY_DEST;
 337         list_del_init(&entry->se_link);
 338         if (likely(!list_empty(&entry->se_list)))
 339                 list_del_init(&entry->se_list);
 340         spin_unlock(&lli->lli_sa_lock);
 341
 342         ll_sa_entry_put(sai, entry);
 343 }
 344
 345 /*
 346  * Delete it from sai_entries_stated list when fini.
 347  */
 348 static void
 349 ll_sa_entry_fini(struct ll_statahead_info *sai, struct ll_sa_entry *entry)
 350 {
 351         struct ll_sa_entry *pos, *next;
 352
 353         if (entry)
 354                 do_sa_entry_fini(sai, entry);
 355
 356         /* drop old entry, only 'scanner' process does this, no need to lock */
 357         list_for_each_entry_safe(pos, next, &sai->sai_entries, se_link) {
 358                 if (!is_omitted_entry(sai, pos->se_index))
 359                         break;
 360                 do_sa_entry_fini(sai, pos);
 361         }
 362 }
 363
 364 /*
 365  * Inside lli_sa_lock.
 366  */
 367 static void
 368 do_sa_entry_to_stated(struct ll_statahead_info *sai,
 369                       struct ll_sa_entry *entry, enum se_stat stat)
 370 {
 371         struct ll_sa_entry *se;
 372         struct list_head         *pos = &sai->sai_entries_stated;
 373
 374         if (!list_empty(&entry->se_list))
 375                 list_del_init(&entry->se_list);
 376
 377         list_for_each_entry_reverse(se, &sai->sai_entries_stated, se_list) {
 378                 if (se->se_index < entry->se_index) {
 379                         pos = &se->se_list;
 380                         break;
 381                 }
 382         }
 383
 384         list_add(&entry->se_list, pos);
 385         entry->se_stat = stat;
 386 }
 387
 388 /*
 389  * Move entry to sai_entries_stated and sort with the index.
 390  * \retval 1    -- entry to be destroyed.
 391  * \retval 0    -- entry is inserted into stated list.
 392  */
 393 static int
 394 ll_sa_entry_to_stated(struct ll_statahead_info *sai,
 395                       struct ll_sa_entry *entry, enum se_stat stat)
 396 {
 397         struct ll_inode_info *lli = ll_i2info(sai->sai_inode);
 398         int                ret = 1;
 399
 400         ll_sa_entry_cleanup(sai, entry);
 401
 402         spin_lock(&lli->lli_sa_lock);
 403         if (likely(entry->se_stat != SA_ENTRY_DEST)) {
 404                 do_sa_entry_to_stated(sai, entry, stat);
 405                 ret = 0;
 406         }
 407         spin_unlock(&lli->lli_sa_lock);
 408
 409         return ret;
 410 }
 411
 412 /*
 413  * Insert inode into the list of sai_entries_agl.
 414  */
 415 static void ll_agl_add(struct ll_statahead_info *sai,
 416                        struct inode *inode, int index)
 417 {
 418         struct ll_inode_info *child  = ll_i2info(inode);
 419         struct ll_inode_info *parent = ll_i2info(sai->sai_inode);
 420         int                added  = 0;
 421
 422         spin_lock(&child->lli_agl_lock);
 423         if (child->lli_agl_index == 0) {
 424                 child->lli_agl_index = index;
 425                 spin_unlock(&child->lli_agl_lock);
 426
 427                 LASSERT(list_empty(&child->lli_agl_list));
 428
 429                 igrab(inode);
 430                 spin_lock(&parent->lli_agl_lock);
 431                 if (list_empty(&sai->sai_entries_agl))
 432                         added = 1;
 433                 list_add_tail(&child->lli_agl_list, &sai->sai_entries_agl);
 434                 spin_unlock(&parent->lli_agl_lock);
 435         } else {
 436                 spin_unlock(&child->lli_agl_lock);
 437         }
 438
 439         if (added > 0)
 440                 wake_up(&sai->sai_agl_thread.t_ctl_waitq);
 441 }
 442
 443 static struct ll_statahead_info *ll_sai_alloc(void)
 444 {
 445         struct ll_statahead_info *sai;
 446         int                    i;
 447
 448         sai = kzalloc(sizeof(*sai), GFP_NOFS);
 449         if (!sai)
 450                 return NULL;
 451
 452         atomic_set(&sai->sai_refcount, 1);
 453
 454         spin_lock(&sai_generation_lock);
 455         sai->sai_generation = ++sai_generation;
 456         if (unlikely(sai_generation == 0))
 457                 sai->sai_generation = ++sai_generation;
 458         spin_unlock(&sai_generation_lock);
 459
 460         sai->sai_max = LL_SA_RPC_MIN;
 461         sai->sai_index = 1;
 462         init_waitqueue_head(&sai->sai_waitq);
 463         init_waitqueue_head(&sai->sai_thread.t_ctl_waitq);
 464         init_waitqueue_head(&sai->sai_agl_thread.t_ctl_waitq);
 465
 466         INIT_LIST_HEAD(&sai->sai_entries);
 467         INIT_LIST_HEAD(&sai->sai_entries_received);
 468         INIT_LIST_HEAD(&sai->sai_entries_stated);
 469         INIT_LIST_HEAD(&sai->sai_entries_agl);
 470
 471         for (i = 0; i < LL_SA_CACHE_SIZE; i++) {
 472                 INIT_LIST_HEAD(&sai->sai_cache[i]);
 473                 spin_lock_init(&sai->sai_cache_lock[i]);
 474         }
 475         atomic_set(&sai->sai_cache_count, 0);
 476
 477         return sai;
 478 }
 479
 480 static inline struct ll_statahead_info *
 481 ll_sai_get(struct ll_statahead_info *sai)
 482 {
 483         atomic_inc(&sai->sai_refcount);
 484         return sai;
 485 }
 486
 487 static void ll_sai_put(struct ll_statahead_info *sai)
 488 {
 489         struct inode     *inode = sai->sai_inode;
 490         struct ll_inode_info *lli   = ll_i2info(inode);
 491
 492         if (atomic_dec_and_lock(&sai->sai_refcount, &lli->lli_sa_lock)) {
 493                 struct ll_sa_entry *entry, *next;
 494
 495                 if (unlikely(atomic_read(&sai->sai_refcount) > 0)) {
 496                         /* It is race case, the interpret callback just hold
 497                          * a reference count
 498                          */
 499                         spin_unlock(&lli->lli_sa_lock);
 500                         return;
 501                 }
 502
 503                 LASSERT(!lli->lli_opendir_key);
 504                 LASSERT(thread_is_stopped(&sai->sai_thread));
 505                 LASSERT(thread_is_stopped(&sai->sai_agl_thread));
 506
 507                 lli->lli_sai = NULL;
 508                 lli->lli_opendir_pid = 0;
 509                 spin_unlock(&lli->lli_sa_lock);
 510
 511                 if (sai->sai_sent > sai->sai_replied)
 512                         CDEBUG(D_READA, "statahead for dir "DFID
 513                               " does not finish: [sent:%llu] [replied:%llu]\n",
 514                               PFID(&lli->lli_fid),
 515                               sai->sai_sent, sai->sai_replied);
 516
 517                 list_for_each_entry_safe(entry, next, &sai->sai_entries,
 518                                          se_link)
 519                         do_sa_entry_fini(sai, entry);
 520
 521                 LASSERT(list_empty(&sai->sai_entries));
 522                 LASSERT(list_empty(&sai->sai_entries_received));
 523                 LASSERT(list_empty(&sai->sai_entries_stated));
 524
 525                 LASSERT(atomic_read(&sai->sai_cache_count) == 0);
 526                 LASSERT(list_empty(&sai->sai_entries_agl));
 527
 528                 iput(inode);
 529                 kfree(sai);
 530         }
 531 }
 532
 533 /* Do NOT forget to drop inode refcount when into sai_entries_agl. */
 534 static void ll_agl_trigger(struct inode *inode, struct ll_statahead_info *sai)
 535 {
 536         struct ll_inode_info *lli   = ll_i2info(inode);
 537         __u64            index = lli->lli_agl_index;
 538         int                rc;
 539
 540         LASSERT(list_empty(&lli->lli_agl_list));
 541
 542         /* AGL maybe fall behind statahead with one entry */
 543         if (is_omitted_entry(sai, index + 1)) {
 544                 lli->lli_agl_index = 0;
 545                 iput(inode);
 546                 return;
 547         }
 548
 549         /* Someone is in glimpse (sync or async), do nothing. */
 550         rc = down_write_trylock(&lli->lli_glimpse_sem);
 551         if (rc == 0) {
 552                 lli->lli_agl_index = 0;
 553                 iput(inode);
 554                 return;
 555         }
 556
 557         /*
 558          * Someone triggered glimpse within 1 sec before.
 559          * 1) The former glimpse succeeded with glimpse lock granted by OST, and
 560          *    if the lock is still cached on client, AGL needs to do nothing. If
 561          *    it is cancelled by other client, AGL maybe cannot obtain new lock
 562          *    for no glimpse callback triggered by AGL.
 563          * 2) The former glimpse succeeded, but OST did not grant glimpse lock.
 564          *    Under such case, it is quite possible that the OST will not grant
 565          *    glimpse lock for AGL also.
 566          * 3) The former glimpse failed, compared with other two cases, it is
 567          *    relative rare. AGL can ignore such case, and it will not muchly
 568          *    affect the performance.
 569          */
 570         if (lli->lli_glimpse_time != 0 &&
 571             time_before(cfs_time_shift(-1), lli->lli_glimpse_time)) {
 572                 up_write(&lli->lli_glimpse_sem);
 573                 lli->lli_agl_index = 0;
 574                 iput(inode);
 575                 return;
 576         }
 577
 578         CDEBUG(D_READA, "Handling (init) async glimpse: inode = "
 579                DFID", idx = %llu\n", PFID(&lli->lli_fid), index);
 580
 581         cl_agl(inode);
 582         lli->lli_agl_index = 0;
 583         lli->lli_glimpse_time = cfs_time_current();
 584         up_write(&lli->lli_glimpse_sem);
 585
 586         CDEBUG(D_READA, "Handled (init) async glimpse: inode= "
 587                DFID", idx = %llu, rc = %d\n",
 588                PFID(&lli->lli_fid), index, rc);
 589
 590         iput(inode);
 591 }
 592
 593 static void ll_post_statahead(struct ll_statahead_info *sai)
 594 {
 595         struct inode       *dir   = sai->sai_inode;
 596         struct inode       *child;
 597         struct ll_inode_info   *lli   = ll_i2info(dir);
 598         struct ll_sa_entry     *entry;
 599         struct md_enqueue_info *minfo;
 600         struct lookup_intent   *it;
 601         struct ptlrpc_request  *req;
 602         struct mdt_body *body;
 603         int                  rc    = 0;
 604
 605         spin_lock(&lli->lli_sa_lock);
 606         if (unlikely(list_empty(&sai->sai_entries_received))) {
 607                 spin_unlock(&lli->lli_sa_lock);
 608                 return;
 609         }
 610         entry = list_entry(sai->sai_entries_received.next,
 611                            struct ll_sa_entry, se_list);
 612         atomic_inc(&entry->se_refcount);
 613         list_del_init(&entry->se_list);
 614         spin_unlock(&lli->lli_sa_lock);
 615
 616         LASSERT(entry->se_handle != 0);
 617
 618         minfo = entry->se_minfo;
 619         it = &minfo->mi_it;
 620         req = entry->se_req;
 621         body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
 622         if (!body) {
 623                 rc = -EFAULT;
 624                 goto out;
 625         }
 626
 627         child = entry->se_inode;
 628         if (!child) {
 629                 /*
 630                  * lookup.
 631                  */
 632                 LASSERT(fid_is_zero(&minfo->mi_data.op_fid2));
 633
 634                 /* XXX: No fid in reply, this is probably cross-ref case.
 635                  * SA can't handle it yet.
 636                  */
 637                 if (body->valid & OBD_MD_MDS) {
 638                         rc = -EAGAIN;
 639                         goto out;
 640                 }
 641         } else {
 642                 /*
 643                  * revalidate.
 644                  */
 645                 /* unlinked and re-created with the same name */
 646                 if (unlikely(!lu_fid_eq(&minfo->mi_data.op_fid2, &body->fid1))) {
 647                         entry->se_inode = NULL;
 648                         iput(child);
 649                         child = NULL;
 650                 }
 651         }
 652
 653         it->d.lustre.it_lock_handle = entry->se_handle;
 654         rc = md_revalidate_lock(ll_i2mdexp(dir), it, ll_inode2fid(dir), NULL);
 655         if (rc != 1) {
 656                 rc = -EAGAIN;
 657                 goto out;
 658         }
 659
 660         rc = ll_prep_inode(&child, req, dir->i_sb, it);
 661         if (rc)
 662                 goto out;
 663
 664         CDEBUG(D_DLMTRACE, "%s: setting l_data to inode "DFID"%p\n",
 665                ll_get_fsname(child->i_sb, NULL, 0),
 666                PFID(ll_inode2fid(child)), child);
 667         ll_set_lock_data(ll_i2sbi(dir)->ll_md_exp, child, it, NULL);
 668
 669         entry->se_inode = child;
 670
 671         if (agl_should_run(sai, child))
 672                 ll_agl_add(sai, child, entry->se_index);
 673
 674 out:
 675         /* The "ll_sa_entry_to_stated()" will drop related ldlm ibits lock
 676          * reference count by calling "ll_intent_drop_lock()" in spite of the
 677          * above operations failed or not. Do not worry about calling
 678          * "ll_intent_drop_lock()" more than once.
 679          */
 680         rc = ll_sa_entry_to_stated(sai, entry,
 681                                    rc < 0 ? SA_ENTRY_INVA : SA_ENTRY_SUCC);
 682         if (rc == 0 && entry->se_index == sai->sai_index_wait)
 683                 wake_up(&sai->sai_waitq);
 684         ll_sa_entry_put(sai, entry);
 685 }
 686
 687 static int ll_statahead_interpret(struct ptlrpc_request *req,
 688                                   struct md_enqueue_info *minfo, int rc)
 689 {
 690         struct lookup_intent     *it  = &minfo->mi_it;
 691         struct inode         *dir = minfo->mi_dir;
 692         struct ll_inode_info     *lli = ll_i2info(dir);
 693         struct ll_statahead_info *sai = NULL;
 694         struct ll_sa_entry       *entry;
 695         __u64                     handle = 0;
 696         int                    wakeup;
 697
 698         if (it_disposition(it, DISP_LOOKUP_NEG))
 699                 rc = -ENOENT;
 700
 701         if (rc == 0) {
 702                 /* release ibits lock ASAP to avoid deadlock when statahead
 703                  * thread enqueues lock on parent in readdir and another
 704                  * process enqueues lock on child with parent lock held, eg.
 705                  * unlink.
 706                  */
 707                 handle = it->d.lustre.it_lock_handle;
 708                 ll_intent_drop_lock(it);
 709         }
 710
 711         spin_lock(&lli->lli_sa_lock);
 712         /* stale entry */
 713         if (unlikely(!lli->lli_sai ||
 714                      lli->lli_sai->sai_generation != minfo->mi_generation)) {
 715                 spin_unlock(&lli->lli_sa_lock);
 716                 rc = -ESTALE;
 717                 goto out;
 718         } else {
 719                 sai = ll_sai_get(lli->lli_sai);
 720                 if (unlikely(!thread_is_running(&sai->sai_thread))) {
 721                         sai->sai_replied++;
 722                         spin_unlock(&lli->lli_sa_lock);
 723                         rc = -EBADFD;
 724                         goto out;
 725                 }
 726
 727                 entry = ll_sa_entry_get_byindex(sai, minfo->mi_cbdata);
 728                 if (!entry) {
 729                         sai->sai_replied++;
 730                         spin_unlock(&lli->lli_sa_lock);
 731                         rc = -EIDRM;
 732                         goto out;
 733                 }
 734
 735                 if (rc != 0) {
 736                         do_sa_entry_to_stated(sai, entry, SA_ENTRY_INVA);
 737                         wakeup = (entry->se_index == sai->sai_index_wait);
 738                 } else {
 739                         entry->se_minfo = minfo;
 740                         entry->se_req = ptlrpc_request_addref(req);
 741                         /* Release the async ibits lock ASAP to avoid deadlock
 742                          * when statahead thread tries to enqueue lock on parent
 743                          * for readpage and other tries to enqueue lock on child
 744                          * with parent's lock held, for example: unlink.
 745                          */
 746                         entry->se_handle = handle;
 747                         wakeup = list_empty(&sai->sai_entries_received);
 748                         list_add_tail(&entry->se_list,
 749                                       &sai->sai_entries_received);
 750                 }
 751                 sai->sai_replied++;
 752                 spin_unlock(&lli->lli_sa_lock);
 753
 754                 ll_sa_entry_put(sai, entry);
 755                 if (wakeup)
 756                         wake_up(&sai->sai_thread.t_ctl_waitq);
 757         }
 758
 759 out:
 760         if (rc != 0) {
 761                 ll_intent_release(it);
 762                 iput(dir);
 763                 kfree(minfo);
 764         }
 765         if (sai)
 766                 ll_sai_put(sai);
 767         return rc;
 768 }
 769
 770 static void sa_args_fini(struct md_enqueue_info *minfo,
 771                          struct ldlm_enqueue_info *einfo)
 772 {
 773         LASSERT(minfo && einfo);
 774         iput(minfo->mi_dir);
 775         kfree(minfo);
 776         kfree(einfo);
 777 }
 778
 779 /**
 780  * prepare arguments for async stat RPC.
 781  */
 782 static int sa_args_init(struct inode *dir, struct inode *child,
 783                         struct ll_sa_entry *entry, struct md_enqueue_info **pmi,
 784                         struct ldlm_enqueue_info **pei)
 785 {
 786         struct qstr           *qstr = &entry->se_qstr;
 787         struct ll_inode_info     *lli  = ll_i2info(dir);
 788         struct md_enqueue_info   *minfo;
 789         struct ldlm_enqueue_info *einfo;
 790         struct md_op_data       *op_data;
 791
 792         einfo = kzalloc(sizeof(*einfo), GFP_NOFS);
 793         if (!einfo)
 794                 return -ENOMEM;
 795
 796         minfo = kzalloc(sizeof(*minfo), GFP_NOFS);
 797         if (!minfo) {
 798                 kfree(einfo);
 799                 return -ENOMEM;
 800         }
 801
 802         op_data = ll_prep_md_op_data(&minfo->mi_data, dir, child, qstr->name,
 803                                      qstr->len, 0, LUSTRE_OPC_ANY, NULL);
 804         if (IS_ERR(op_data)) {
 805                 kfree(einfo);
 806                 kfree(minfo);
 807                 return PTR_ERR(op_data);
 808         }
 809
 810         minfo->mi_it.it_op = IT_GETATTR;
 811         minfo->mi_dir = igrab(dir);
 812         minfo->mi_cb = ll_statahead_interpret;
 813         minfo->mi_generation = lli->lli_sai->sai_generation;
 814         minfo->mi_cbdata = entry->se_index;
 815
 816         einfo->ei_type   = LDLM_IBITS;
 817         einfo->ei_mode   = it_to_lock_mode(&minfo->mi_it);
 818         einfo->ei_cb_bl  = ll_md_blocking_ast;
 819         einfo->ei_cb_cp  = ldlm_completion_ast;
 820         einfo->ei_cb_gl  = NULL;
 821         einfo->ei_cbdata = NULL;
 822
 823         *pmi = minfo;
 824         *pei = einfo;
 825
 826         return 0;
 827 }
 828
 829 static int do_sa_lookup(struct inode *dir, struct ll_sa_entry *entry)
 830 {
 831         struct md_enqueue_info   *minfo;
 832         struct ldlm_enqueue_info *einfo;
 833         int                    rc;
 834
 835         rc = sa_args_init(dir, NULL, entry, &minfo, &einfo);
 836         if (rc)
 837                 return rc;
 838
 839         rc = md_intent_getattr_async(ll_i2mdexp(dir), minfo, einfo);
 840         if (rc < 0)
 841                 sa_args_fini(minfo, einfo);
 842
 843         return rc;
 844 }
 845
 846 /**
 847  * similar to ll_revalidate_it().
 848  * \retval      1 -- dentry valid
 849  * \retval      0 -- will send stat-ahead request
 850  * \retval others -- prepare stat-ahead request failed
 851  */
 852 static int do_sa_revalidate(struct inode *dir, struct ll_sa_entry *entry,
 853                             struct dentry *dentry)
 854 {
 855         struct inode         *inode = d_inode(dentry);
 856         struct lookup_intent      it = { .it_op = IT_GETATTR,
 857                                          .d.lustre.it_lock_handle = 0 };
 858         struct md_enqueue_info   *minfo;
 859         struct ldlm_enqueue_info *einfo;
 860         int rc;
 861
 862         if (unlikely(!inode))
 863                 return 1;
 864
 865         if (d_mountpoint(dentry))
 866                 return 1;
 867
 868         entry->se_inode = igrab(inode);
 869         rc = md_revalidate_lock(ll_i2mdexp(dir), &it, ll_inode2fid(inode),
 870                                 NULL);
 871         if (rc == 1) {
 872                 entry->se_handle = it.d.lustre.it_lock_handle;
 873                 ll_intent_release(&it);
 874                 return 1;
 875         }
 876
 877         rc = sa_args_init(dir, inode, entry, &minfo, &einfo);
 878         if (rc) {
 879                 entry->se_inode = NULL;
 880                 iput(inode);
 881                 return rc;
 882         }
 883
 884         rc = md_intent_getattr_async(ll_i2mdexp(dir), minfo, einfo);
 885         if (rc < 0) {
 886                 entry->se_inode = NULL;
 887                 iput(inode);
 888                 sa_args_fini(minfo, einfo);
 889         }
 890
 891         return rc;
 892 }
 893
 894 static void ll_statahead_one(struct dentry *parent, const char *entry_name,
 895                              int entry_name_len)
 896 {
 897         struct inode         *dir    = d_inode(parent);
 898         struct ll_inode_info     *lli    = ll_i2info(dir);
 899         struct ll_statahead_info *sai    = lli->lli_sai;
 900         struct dentry       *dentry = NULL;
 901         struct ll_sa_entry       *entry;
 902         int                    rc;
 903         int                    rc1;
 904
 905         entry = ll_sa_entry_alloc(sai, sai->sai_index, entry_name,
 906                                   entry_name_len);
 907         if (IS_ERR(entry))
 908                 return;
 909
 910         dentry = d_lookup(parent, &entry->se_qstr);
 911         if (!dentry) {
 912                 rc = do_sa_lookup(dir, entry);
 913         } else {
 914                 rc = do_sa_revalidate(dir, entry, dentry);
 915                 if (rc == 1 && agl_should_run(sai, d_inode(dentry)))
 916                         ll_agl_add(sai, d_inode(dentry), entry->se_index);
 917
 918                 dput(dentry);
 919         }
 920
 921         if (rc) {
 922                 rc1 = ll_sa_entry_to_stated(sai, entry,
 923                                         rc < 0 ? SA_ENTRY_INVA : SA_ENTRY_SUCC);
 924                 if (rc1 == 0 && entry->se_index == sai->sai_index_wait)
 925                         wake_up(&sai->sai_waitq);
 926         } else {
 927                 sai->sai_sent++;
 928         }
 929
 930         sai->sai_index++;
 931         /* drop one refcount on entry by ll_sa_entry_alloc */
 932         ll_sa_entry_put(sai, entry);
 933 }
 934
 935 static int ll_agl_thread(void *arg)
 936 {
 937         struct dentry       *parent = arg;
 938         struct inode         *dir    = d_inode(parent);
 939         struct ll_inode_info     *plli   = ll_i2info(dir);
 940         struct ll_inode_info     *clli;
 941         struct ll_sb_info       *sbi    = ll_i2sbi(dir);
 942         struct ll_statahead_info *sai    = ll_sai_get(plli->lli_sai);
 943         struct ptlrpc_thread     *thread = &sai->sai_agl_thread;
 944         struct l_wait_info      lwi    = { 0 };
 945
 946         thread->t_pid = current_pid();
 947         CDEBUG(D_READA, "agl thread started: sai %p, parent %pd\n",
 948                sai, parent);
 949
 950         atomic_inc(&sbi->ll_agl_total);
 951         spin_lock(&plli->lli_agl_lock);
 952         sai->sai_agl_valid = 1;
 953         if (thread_is_init(thread))
 954                 /* If someone else has changed the thread state
 955                  * (e.g. already changed to SVC_STOPPING), we can't just
 956                  * blindly overwrite that setting.
 957                  */
 958                 thread_set_flags(thread, SVC_RUNNING);
 959         spin_unlock(&plli->lli_agl_lock);
 960         wake_up(&thread->t_ctl_waitq);
 961
 962         while (1) {
 963                 l_wait_event(thread->t_ctl_waitq,
 964                              !list_empty(&sai->sai_entries_agl) ||
 965                              !thread_is_running(thread),
 966                              &lwi);
 967
 968                 if (!thread_is_running(thread))
 969                         break;
 970
 971                 spin_lock(&plli->lli_agl_lock);
 972                 /* The statahead thread maybe help to process AGL entries,
 973                  * so check whether list empty again.
 974                  */
 975                 if (!list_empty(&sai->sai_entries_agl)) {
 976                         clli = list_entry(sai->sai_entries_agl.next,
 977                                           struct ll_inode_info, lli_agl_list);
 978                         list_del_init(&clli->lli_agl_list);
 979                         spin_unlock(&plli->lli_agl_lock);
 980                         ll_agl_trigger(&clli->lli_vfs_inode, sai);
 981                 } else {
 982                         spin_unlock(&plli->lli_agl_lock);
 983                 }
 984         }
 985
 986         spin_lock(&plli->lli_agl_lock);
 987         sai->sai_agl_valid = 0;
 988         while (!list_empty(&sai->sai_entries_agl)) {
 989                 clli = list_entry(sai->sai_entries_agl.next,
 990                                   struct ll_inode_info, lli_agl_list);
 991                 list_del_init(&clli->lli_agl_list);
 992                 spin_unlock(&plli->lli_agl_lock);
 993                 clli->lli_agl_index = 0;
 994                 iput(&clli->lli_vfs_inode);
 995                 spin_lock(&plli->lli_agl_lock);
 996         }
 997         thread_set_flags(thread, SVC_STOPPED);
 998         spin_unlock(&plli->lli_agl_lock);
 999         wake_up(&thread->t_ctl_waitq);
1000         ll_sai_put(sai);
1001         CDEBUG(D_READA, "agl thread stopped: sai %p, parent %pd\n",
1002                sai, parent);
1003         return 0;
1004 }
1005
1006 static void ll_start_agl(struct dentry *parent, struct ll_statahead_info *sai)
1007 {
1008         struct ptlrpc_thread *thread = &sai->sai_agl_thread;
1009         struct l_wait_info    lwi    = { 0 };
1010         struct ll_inode_info  *plli;
1011         struct task_struct *task;
1012
1013         CDEBUG(D_READA, "start agl thread: sai %p, parent %pd\n",
1014                sai, parent);
1015
1016         plli = ll_i2info(d_inode(parent));
1017         task = kthread_run(ll_agl_thread, parent, "ll_agl_%u",
1018                            plli->lli_opendir_pid);
1019         if (IS_ERR(task)) {
1020                 CERROR("can't start ll_agl thread, rc: %ld\n", PTR_ERR(task));
1021                 thread_set_flags(thread, SVC_STOPPED);
1022                 return;
1023         }
1024
1025         l_wait_event(thread->t_ctl_waitq,
1026                      thread_is_running(thread) || thread_is_stopped(thread),
1027                      &lwi);
1028 }
1029
1030 static int ll_statahead_thread(void *arg)
1031 {
1032         struct dentry       *parent = arg;
1033         struct inode         *dir    = d_inode(parent);
1034         struct ll_inode_info     *plli   = ll_i2info(dir);
1035         struct ll_inode_info     *clli;
1036         struct ll_sb_info       *sbi    = ll_i2sbi(dir);
1037         struct ll_statahead_info *sai    = ll_sai_get(plli->lli_sai);
1038         struct ptlrpc_thread     *thread = &sai->sai_thread;
1039         struct ptlrpc_thread *agl_thread = &sai->sai_agl_thread;
1040         struct page           *page;
1041         __u64                pos    = 0;
1042         int                    first  = 0;
1043         int                    rc     = 0;
1044         struct ll_dir_chain       chain;
1045         struct l_wait_info      lwi    = { 0 };
1046
1047         thread->t_pid = current_pid();
1048         CDEBUG(D_READA, "statahead thread starting: sai %p, parent %pd\n",
1049                sai, parent);
1050
1051         if (sbi->ll_flags & LL_SBI_AGL_ENABLED)
1052                 ll_start_agl(parent, sai);
1053
1054         atomic_inc(&sbi->ll_sa_total);
1055         spin_lock(&plli->lli_sa_lock);
1056         if (thread_is_init(thread))
1057                 /* If someone else has changed the thread state
1058                  * (e.g. already changed to SVC_STOPPING), we can't just
1059                  * blindly overwrite that setting.
1060                  */
1061                 thread_set_flags(thread, SVC_RUNNING);
1062         spin_unlock(&plli->lli_sa_lock);
1063         wake_up(&thread->t_ctl_waitq);
1064
1065         ll_dir_chain_init(&chain);
1066         page = ll_get_dir_page(dir, pos, &chain);
1067
1068         while (1) {
1069                 struct lu_dirpage *dp;
1070                 struct lu_dirent  *ent;
1071
1072                 if (IS_ERR(page)) {
1073                         rc = PTR_ERR(page);
1074                         CDEBUG(D_READA, "error reading dir "DFID" at %llu/%llu: [rc %d] [parent %u]\n",
1075                                PFID(ll_inode2fid(dir)), pos, sai->sai_index,
1076                                rc, plli->lli_opendir_pid);
1077                         goto out;
1078                 }
1079
1080                 dp = page_address(page);
1081                 for (ent = lu_dirent_start(dp); ent;
1082                      ent = lu_dirent_next(ent)) {
1083                         __u64 hash;
1084                         int namelen;
1085                         char *name;
1086
1087                         hash = le64_to_cpu(ent->lde_hash);
1088                         if (unlikely(hash < pos))
1089                                 /*
1090                                  * Skip until we find target hash value.
1091                                  */
1092                                 continue;
1093
1094                         namelen = le16_to_cpu(ent->lde_namelen);
1095                         if (unlikely(namelen == 0))
1096                                 /*
1097                                  * Skip dummy record.
1098                                  */
1099                                 continue;
1100
1101                         name = ent->lde_name;
1102                         if (name[0] == '.') {
1103                                 if (namelen == 1) {
1104                                         /*
1105                                          * skip "."
1106                                          */
1107                                         continue;
1108                                 } else if (name[1] == '.' && namelen == 2) {
1109                                         /*
1110                                          * skip ".."
1111                                          */
1112                                         continue;
1113                                 } else if (!sai->sai_ls_all) {
1114                                         /*
1115                                          * skip hidden files.
1116                                          */
1117                                         sai->sai_skip_hidden++;
1118                                         continue;
1119                                 }
1120                         }
1121
1122                         /*
1123                          * don't stat-ahead first entry.
1124                          */
1125                         if (unlikely(++first == 1))
1126                                 continue;
1127
1128 keep_it:
1129                         l_wait_event(thread->t_ctl_waitq,
1130                                      !sa_sent_full(sai) ||
1131                                      !list_empty(&sai->sai_entries_received) ||
1132                                      !list_empty(&sai->sai_entries_agl) ||
1133                                      !thread_is_running(thread),
1134                                      &lwi);
1135
1136 interpret_it:
1137                         while (!list_empty(&sai->sai_entries_received))
1138                                 ll_post_statahead(sai);
1139
1140                         if (unlikely(!thread_is_running(thread))) {
1141                                 ll_release_page(page, 0);
1142                                 rc = 0;
1143                                 goto out;
1144                         }
1145
1146                         /* If no window for metadata statahead, but there are
1147                          * some AGL entries to be triggered, then try to help
1148                          * to process the AGL entries.
1149                          */
1150                         if (sa_sent_full(sai)) {
1151                                 spin_lock(&plli->lli_agl_lock);
1152                                 while (!list_empty(&sai->sai_entries_agl)) {
1153                                         clli = list_entry(sai->sai_entries_agl.next,
1154                                                           struct ll_inode_info, lli_agl_list);
1155                                         list_del_init(&clli->lli_agl_list);
1156                                         spin_unlock(&plli->lli_agl_lock);
1157                                         ll_agl_trigger(&clli->lli_vfs_inode,
1158                                                        sai);
1159
1160                                         if (!list_empty(&sai->sai_entries_received))
1161                                                 goto interpret_it;
1162
1163                                         if (unlikely(
1164                                                 !thread_is_running(thread))) {
1165                                                 ll_release_page(page, 0);
1166                                                 rc = 0;
1167                                                 goto out;
1168                                         }
1169
1170                                         if (!sa_sent_full(sai))
1171                                                 goto do_it;
1172
1173                                         spin_lock(&plli->lli_agl_lock);
1174                                 }
1175                                 spin_unlock(&plli->lli_agl_lock);
1176
1177                                 goto keep_it;
1178                         }
1179
1180 do_it:
1181                         ll_statahead_one(parent, name, namelen);
1182                 }
1183                 pos = le64_to_cpu(dp->ldp_hash_end);
1184                 if (pos == MDS_DIR_END_OFF) {
1185                         /*
1186                          * End of directory reached.
1187                          */
1188                         ll_release_page(page, 0);
1189                         while (1) {
1190                                 l_wait_event(thread->t_ctl_waitq,
1191                                              !list_empty(&sai->sai_entries_received) ||
1192                                              sai->sai_sent == sai->sai_replied ||
1193                                              !thread_is_running(thread),
1194                                              &lwi);
1195
1196                                 while (!list_empty(&sai->sai_entries_received))
1197                                         ll_post_statahead(sai);
1198
1199                                 if (unlikely(!thread_is_running(thread))) {
1200                                         rc = 0;
1201                                         goto out;
1202                                 }
1203
1204                                 if (sai->sai_sent == sai->sai_replied &&
1205                                     list_empty(&sai->sai_entries_received))
1206                                         break;
1207                         }
1208
1209                         spin_lock(&plli->lli_agl_lock);
1210                         while (!list_empty(&sai->sai_entries_agl) &&
1211                                thread_is_running(thread)) {
1212                                 clli = list_entry(sai->sai_entries_agl.next,
1213                                                   struct ll_inode_info, lli_agl_list);
1214                                 list_del_init(&clli->lli_agl_list);
1215                                 spin_unlock(&plli->lli_agl_lock);
1216                                 ll_agl_trigger(&clli->lli_vfs_inode, sai);
1217                                 spin_lock(&plli->lli_agl_lock);
1218                         }
1219                         spin_unlock(&plli->lli_agl_lock);
1220
1221                         rc = 0;
1222                         goto out;
1223                 } else if (1) {
1224                         /*
1225                          * chain is exhausted.
1226                          * Normal case: continue to the next page.
1227                          */
1228                         ll_release_page(page, le32_to_cpu(dp->ldp_flags) &
1229                                               LDF_COLLIDE);
1230                         page = ll_get_dir_page(dir, pos, &chain);
1231                 } else {
1232                         LASSERT(le32_to_cpu(dp->ldp_flags) & LDF_COLLIDE);
1233                         ll_release_page(page, 1);
1234                         /*
1235                          * go into overflow page.
1236                          */
1237                 }
1238         }
1239
1240 out:
1241         if (sai->sai_agl_valid) {
1242                 spin_lock(&plli->lli_agl_lock);
1243                 thread_set_flags(agl_thread, SVC_STOPPING);
1244                 spin_unlock(&plli->lli_agl_lock);
1245                 wake_up(&agl_thread->t_ctl_waitq);
1246
1247                 CDEBUG(D_READA, "stop agl thread: sai %p pid %u\n",
1248                        sai, (unsigned int)agl_thread->t_pid);
1249                 l_wait_event(agl_thread->t_ctl_waitq,
1250                              thread_is_stopped(agl_thread),
1251                              &lwi);
1252         } else {
1253                 /* Set agl_thread flags anyway. */
1254                 thread_set_flags(&sai->sai_agl_thread, SVC_STOPPED);
1255         }
1256         ll_dir_chain_fini(&chain);
1257         spin_lock(&plli->lli_sa_lock);
1258         if (!list_empty(&sai->sai_entries_received)) {
1259                 thread_set_flags(thread, SVC_STOPPING);
1260                 spin_unlock(&plli->lli_sa_lock);
1261
1262                 /* To release the resources held by received entries. */
1263                 while (!list_empty(&sai->sai_entries_received))
1264                         ll_post_statahead(sai);
1265
1266                 spin_lock(&plli->lli_sa_lock);
1267         }
1268         thread_set_flags(thread, SVC_STOPPED);
1269         spin_unlock(&plli->lli_sa_lock);
1270         wake_up(&sai->sai_waitq);
1271         wake_up(&thread->t_ctl_waitq);
1272         ll_sai_put(sai);
1273         dput(parent);
1274         CDEBUG(D_READA, "statahead thread stopped: sai %p, parent %pd\n",
1275                sai, parent);
1276         return rc;
1277 }
1278
1279 /**
1280  * called in ll_file_release().
1281  */
1282 void ll_stop_statahead(struct inode *dir, void *key)
1283 {
1284         struct ll_inode_info *lli = ll_i2info(dir);
1285
1286         if (unlikely(!key))
1287                 return;
1288
1289         spin_lock(&lli->lli_sa_lock);
1290         if (lli->lli_opendir_key != key || lli->lli_opendir_pid == 0) {
1291                 spin_unlock(&lli->lli_sa_lock);
1292                 return;
1293         }
1294
1295         lli->lli_opendir_key = NULL;
1296
1297         if (lli->lli_sai) {
1298                 struct l_wait_info lwi = { 0 };
1299                 struct ptlrpc_thread *thread = &lli->lli_sai->sai_thread;
1300
1301                 if (!thread_is_stopped(thread)) {
1302                         thread_set_flags(thread, SVC_STOPPING);
1303                         spin_unlock(&lli->lli_sa_lock);
1304                         wake_up(&thread->t_ctl_waitq);
1305
1306                         CDEBUG(D_READA, "stop statahead thread: sai %p pid %u\n",
1307                                lli->lli_sai, (unsigned int)thread->t_pid);
1308                         l_wait_event(thread->t_ctl_waitq,
1309                                      thread_is_stopped(thread),
1310                                      &lwi);
1311                 } else {
1312                         spin_unlock(&lli->lli_sa_lock);
1313                 }
1314
1315                 /*
1316                  * Put the ref which was held when first statahead_enter.
1317                  * It maybe not the last ref for some statahead requests
1318                  * maybe inflight.
1319                  */
1320                 ll_sai_put(lli->lli_sai);
1321         } else {
1322                 lli->lli_opendir_pid = 0;
1323                 spin_unlock(&lli->lli_sa_lock);
1324         }
1325 }
1326
1327 enum {
1328         /**
1329          * not first dirent, or is "."
1330          */
1331         LS_NONE_FIRST_DE = 0,
1332         /**
1333          * the first non-hidden dirent
1334          */
1335         LS_FIRST_DE,
1336         /**
1337          * the first hidden dirent, that is "."
1338          */
1339         LS_FIRST_DOT_DE
1340 };
1341
1342 static int is_first_dirent(struct inode *dir, struct dentry *dentry)
1343 {
1344         struct ll_dir_chain   chain;
1345         struct qstr       *target = &dentry->d_name;
1346         struct page       *page;
1347         __u64            pos    = 0;
1348         int                dot_de;
1349         int                rc     = LS_NONE_FIRST_DE;
1350
1351         ll_dir_chain_init(&chain);
1352         page = ll_get_dir_page(dir, pos, &chain);
1353
1354         while (1) {
1355                 struct lu_dirpage *dp;
1356                 struct lu_dirent  *ent;
1357
1358                 if (IS_ERR(page)) {
1359                         struct ll_inode_info *lli = ll_i2info(dir);
1360
1361                         rc = PTR_ERR(page);
1362                         CERROR("error reading dir "DFID" at %llu: [rc %d] [parent %u]\n",
1363                                PFID(ll_inode2fid(dir)), pos,
1364                                rc, lli->lli_opendir_pid);
1365                         break;
1366                 }
1367
1368                 dp = page_address(page);
1369                 for (ent = lu_dirent_start(dp); ent;
1370                      ent = lu_dirent_next(ent)) {
1371                         __u64 hash;
1372                         int namelen;
1373                         char *name;
1374
1375                         hash = le64_to_cpu(ent->lde_hash);
1376                         /* The ll_get_dir_page() can return any page containing
1377                          * the given hash which may be not the start hash.
1378                          */
1379                         if (unlikely(hash < pos))
1380                                 continue;
1381
1382                         namelen = le16_to_cpu(ent->lde_namelen);
1383                         if (unlikely(namelen == 0))
1384                                 /*
1385                                  * skip dummy record.
1386                                  */
1387                                 continue;
1388
1389                         name = ent->lde_name;
1390                         if (name[0] == '.') {
1391                                 if (namelen == 1)
1392                                         /*
1393                                          * skip "."
1394                                          */
1395                                         continue;
1396                                 else if (name[1] == '.' && namelen == 2)
1397                                         /*
1398                                          * skip ".."
1399                                          */
1400                                         continue;
1401                                 else
1402                                         dot_de = 1;
1403                         } else {
1404                                 dot_de = 0;
1405                         }
1406
1407                         if (dot_de && target->name[0] != '.') {
1408                                 CDEBUG(D_READA, "%.*s skip hidden file %.*s\n",
1409                                        target->len, target->name,
1410                                        namelen, name);
1411                                 continue;
1412                         }
1413
1414                         if (target->len != namelen ||
1415                             memcmp(target->name, name, namelen) != 0)
1416                                 rc = LS_NONE_FIRST_DE;
1417                         else if (!dot_de)
1418                                 rc = LS_FIRST_DE;
1419                         else
1420                                 rc = LS_FIRST_DOT_DE;
1421
1422                         ll_release_page(page, 0);
1423                         goto out;
1424                 }
1425                 pos = le64_to_cpu(dp->ldp_hash_end);
1426                 if (pos == MDS_DIR_END_OFF) {
1427                         /*
1428                          * End of directory reached.
1429                          */
1430                         ll_release_page(page, 0);
1431                         break;
1432                 } else if (1) {
1433                         /*
1434                          * chain is exhausted
1435                          * Normal case: continue to the next page.
1436                          */
1437                         ll_release_page(page, le32_to_cpu(dp->ldp_flags) &
1438                                               LDF_COLLIDE);
1439                         page = ll_get_dir_page(dir, pos, &chain);
1440                 } else {
1441                         /*
1442                          * go into overflow page.
1443                          */
1444                         LASSERT(le32_to_cpu(dp->ldp_flags) & LDF_COLLIDE);
1445                         ll_release_page(page, 1);
1446                 }
1447         }
1448
1449 out:
1450         ll_dir_chain_fini(&chain);
1451         return rc;
1452 }
1453
1454 static void
1455 ll_sai_unplug(struct ll_statahead_info *sai, struct ll_sa_entry *entry)
1456 {
1457         struct ptlrpc_thread *thread = &sai->sai_thread;
1458         struct ll_sb_info    *sbi    = ll_i2sbi(sai->sai_inode);
1459         int                hit;
1460
1461         if (entry && entry->se_stat == SA_ENTRY_SUCC)
1462                 hit = 1;
1463         else
1464                 hit = 0;
1465
1466         ll_sa_entry_fini(sai, entry);
1467         if (hit) {
1468                 sai->sai_hit++;
1469                 sai->sai_consecutive_miss = 0;
1470                 sai->sai_max = min(2 * sai->sai_max, sbi->ll_sa_max);
1471         } else {
1472                 struct ll_inode_info *lli = ll_i2info(sai->sai_inode);
1473
1474                 sai->sai_miss++;
1475                 sai->sai_consecutive_miss++;
1476                 if (sa_low_hit(sai) && thread_is_running(thread)) {
1477                         atomic_inc(&sbi->ll_sa_wrong);
1478                         CDEBUG(D_READA, "Statahead for dir " DFID " hit ratio too low: hit/miss %llu/%llu, sent/replied %llu/%llu, stopping statahead thread\n",
1479                                PFID(&lli->lli_fid), sai->sai_hit,
1480                                sai->sai_miss, sai->sai_sent,
1481                                sai->sai_replied);
1482                         spin_lock(&lli->lli_sa_lock);
1483                         if (!thread_is_stopped(thread))
1484                                 thread_set_flags(thread, SVC_STOPPING);
1485                         spin_unlock(&lli->lli_sa_lock);
1486                 }
1487         }
1488
1489         if (!thread_is_stopped(thread))
1490                 wake_up(&thread->t_ctl_waitq);
1491 }
1492
1493 /**
1494  * Start statahead thread if this is the first dir entry.
1495  * Otherwise if a thread is started already, wait it until it is ahead of me.
1496  * \retval 1       -- find entry with lock in cache, the caller needs to do
1497  *                  nothing.
1498  * \retval 0       -- find entry in cache, but without lock, the caller needs
1499  *                  refresh from MDS.
1500  * \retval others  -- the caller need to process as non-statahead.
1501  */
1502 int do_statahead_enter(struct inode *dir, struct dentry **dentryp,
1503                        int only_unplug)
1504 {
1505         struct ll_inode_info     *lli   = ll_i2info(dir);
1506         struct ll_statahead_info *sai   = lli->lli_sai;
1507         struct dentry       *parent;
1508         struct ll_sa_entry       *entry;
1509         struct ptlrpc_thread     *thread;
1510         struct l_wait_info      lwi   = { 0 };
1511         struct task_struct *task;
1512         int                    rc    = 0;
1513         struct ll_inode_info     *plli;
1514
1515         LASSERT(lli->lli_opendir_pid == current_pid());
1516
1517         if (sai) {
1518                 thread = &sai->sai_thread;
1519                 if (unlikely(thread_is_stopped(thread) &&
1520                              list_empty(&sai->sai_entries_stated))) {
1521                         /* to release resource */
1522                         ll_stop_statahead(dir, lli->lli_opendir_key);
1523                         return -EAGAIN;
1524                 }
1525
1526                 if ((*dentryp)->d_name.name[0] == '.') {
1527                         if (sai->sai_ls_all ||
1528                             sai->sai_miss_hidden >= sai->sai_skip_hidden) {
1529                                 /*
1530                                  * Hidden dentry is the first one, or statahead
1531                                  * thread does not skip so many hidden dentries
1532                                  * before "sai_ls_all" enabled as below.
1533                                  */
1534                         } else {
1535                                 if (!sai->sai_ls_all)
1536                                         /*
1537                                          * It maybe because hidden dentry is not
1538                                          * the first one, "sai_ls_all" was not
1539                                          * set, then "ls -al" missed. Enable
1540                                          * "sai_ls_all" for such case.
1541                                          */
1542                                         sai->sai_ls_all = 1;
1543
1544                                 /*
1545                                  * Such "getattr" has been skipped before
1546                                  * "sai_ls_all" enabled as above.
1547                                  */
1548                                 sai->sai_miss_hidden++;
1549                                 return -EAGAIN;
1550                         }
1551                 }
1552
1553                 entry = ll_sa_entry_get_byname(sai, &(*dentryp)->d_name);
1554                 if (!entry || only_unplug) {
1555                         ll_sai_unplug(sai, entry);
1556                         return entry ? 1 : -EAGAIN;
1557                 }
1558
1559                 if (!ll_sa_entry_stated(entry)) {
1560                         sai->sai_index_wait = entry->se_index;
1561                         lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(30), NULL,
1562                                                LWI_ON_SIGNAL_NOOP, NULL);
1563                         rc = l_wait_event(sai->sai_waitq,
1564                                           ll_sa_entry_stated(entry) ||
1565                                           thread_is_stopped(thread),
1566                                           &lwi);
1567                         if (rc < 0) {
1568                                 ll_sai_unplug(sai, entry);
1569                                 return -EAGAIN;
1570                         }
1571                 }
1572
1573                 if (entry->se_stat == SA_ENTRY_SUCC && entry->se_inode) {
1574                         struct inode *inode = entry->se_inode;
1575                         struct lookup_intent it = { .it_op = IT_GETATTR,
1576                                                     .d.lustre.it_lock_handle =
1577                                                      entry->se_handle };
1578                         __u64 bits;
1579
1580                         rc = md_revalidate_lock(ll_i2mdexp(dir), &it,
1581                                                 ll_inode2fid(inode), &bits);
1582                         if (rc == 1) {
1583                                 if (!d_inode(*dentryp)) {
1584                                         struct dentry *alias;
1585
1586                                         alias = ll_splice_alias(inode,
1587                                                                 *dentryp);
1588                                         if (IS_ERR(alias)) {
1589                                                 ll_sai_unplug(sai, entry);
1590                                                 return PTR_ERR(alias);
1591                                         }
1592                                         *dentryp = alias;
1593                                 } else if (d_inode(*dentryp) != inode) {
1594                                         /* revalidate, but inode is recreated */
1595                                         CDEBUG(D_READA, "%s: stale dentry %pd inode "DFID", statahead inode "DFID"\n",
1596                                                ll_get_fsname(d_inode(*dentryp)->i_sb, NULL, 0),
1597                                                *dentryp,
1598                                                PFID(ll_inode2fid(d_inode(*dentryp))),
1599                                                PFID(ll_inode2fid(inode)));
1600                                         ll_sai_unplug(sai, entry);
1601                                         return -ESTALE;
1602                                 } else {
1603                                         iput(inode);
1604                                 }
1605                                 entry->se_inode = NULL;
1606
1607                                 if ((bits & MDS_INODELOCK_LOOKUP) &&
1608                                     d_lustre_invalid(*dentryp))
1609                                         d_lustre_revalidate(*dentryp);
1610                                 ll_intent_release(&it);
1611                         }
1612                 }
1613
1614                 ll_sai_unplug(sai, entry);
1615                 return rc;
1616         }
1617
1618         /* I am the "lli_opendir_pid" owner, only me can set "lli_sai". */
1619         rc = is_first_dirent(dir, *dentryp);
1620         if (rc == LS_NONE_FIRST_DE) {
1621                 /* It is not "ls -{a}l" operation, no need statahead for it. */
1622                 rc = -EAGAIN;
1623                 goto out;
1624         }
1625
1626         sai = ll_sai_alloc();
1627         if (!sai) {
1628                 rc = -ENOMEM;
1629                 goto out;
1630         }
1631
1632         sai->sai_ls_all = (rc == LS_FIRST_DOT_DE);
1633         sai->sai_inode = igrab(dir);
1634         if (unlikely(!sai->sai_inode)) {
1635                 CWARN("Do not start stat ahead on dying inode "DFID"\n",
1636                       PFID(&lli->lli_fid));
1637                 rc = -ESTALE;
1638                 goto out;
1639         }
1640
1641         /* get parent reference count here, and put it in ll_statahead_thread */
1642         parent = dget((*dentryp)->d_parent);
1643         if (unlikely(sai->sai_inode != d_inode(parent))) {
1644                 struct ll_inode_info *nlli = ll_i2info(d_inode(parent));
1645
1646                 CWARN("Race condition, someone changed %pd just now: old parent "DFID", new parent "DFID"\n",
1647                       *dentryp,
1648                       PFID(&lli->lli_fid), PFID(&nlli->lli_fid));
1649                 dput(parent);
1650                 iput(sai->sai_inode);
1651                 rc = -EAGAIN;
1652                 goto out;
1653         }
1654
1655         CDEBUG(D_READA, "start statahead thread: sai %p, parent %pd\n",
1656                sai, parent);
1657
1658         /* The sai buffer already has one reference taken at allocation time,
1659          * but as soon as we expose the sai by attaching it to the lli that
1660          * default reference can be dropped by another thread calling
1661          * ll_stop_statahead. We need to take a local reference to protect
1662          * the sai buffer while we intend to access it.
1663          */
1664         ll_sai_get(sai);
1665         lli->lli_sai = sai;
1666
1667         plli = ll_i2info(d_inode(parent));
1668         task = kthread_run(ll_statahead_thread, parent, "ll_sa_%u",
1669                            plli->lli_opendir_pid);
1670         thread = &sai->sai_thread;
1671         if (IS_ERR(task)) {
1672                 rc = PTR_ERR(task);
1673                 CERROR("can't start ll_sa thread, rc: %d\n", rc);
1674                 dput(parent);
1675                 lli->lli_opendir_key = NULL;
1676                 thread_set_flags(thread, SVC_STOPPED);
1677                 thread_set_flags(&sai->sai_agl_thread, SVC_STOPPED);
1678                 /* Drop both our own local reference and the default
1679                  * reference from allocation time.
1680                  */
1681                 ll_sai_put(sai);
1682                 ll_sai_put(sai);
1683                 LASSERT(!lli->lli_sai);
1684                 return -EAGAIN;
1685         }
1686
1687         l_wait_event(thread->t_ctl_waitq,
1688                      thread_is_running(thread) || thread_is_stopped(thread),
1689                      &lwi);
1690         ll_sai_put(sai);
1691
1692         /*
1693          * We don't stat-ahead for the first dirent since we are already in
1694          * lookup.
1695          */
1696         return -EAGAIN;
1697
1698 out:
1699         kfree(sai);
1700         spin_lock(&lli->lli_sa_lock);
1701         lli->lli_opendir_key = NULL;
1702         lli->lli_opendir_pid = 0;
1703         spin_unlock(&lli->lli_sa_lock);
1704         return rc;
1705 }