lustre/quota/quota_context.c

   1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
   2  * vim:expandtab:shiftwidth=8:tabstop=8:
   3  *
   4  * GPL HEADER START
   5  *
   6  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2 only,
  10  * as published by the Free Software Foundation.
  11  *
  12  * This program is distributed in the hope that it will be useful, but
  13  * WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * General Public License version 2 for more details (a copy is included
  16  * in the LICENSE file that accompanied this code).
  17  *
  18  * You should have received a copy of the GNU General Public License
  19  * version 2 along with this program; If not, see
  20  * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
  21  *
  22  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
  23  * CA 95054 USA or visit www.sun.com if you need additional information or
  24  * have any questions.
  25  *
  26  * GPL HEADER END
  27  */
  28 /*
  29  * Copyright  2008 Sun Microsystems, Inc. All rights reserved
  30  * Use is subject to license terms.
  31  */
  32 /*
  33  * This file is part of Lustre, http://www.lustre.org/
  34  * Lustre is a trademark of Sun Microsystems, Inc.
  35  *
  36  * lustre/quota/quota_context.c
  37  *
  38  * Lustre Quota Context
  39  *
  40  * Author: Niu YaWei <niu@clusterfs.com>
  41  */
  42
  43 #ifndef EXPORT_SYMTAB
  44 # define EXPORT_SYMTAB
  45 #endif
  46
  47 #define DEBUG_SUBSYSTEM S_LQUOTA
  48
  49 #include <linux/version.h>
  50 #include <linux/fs.h>
  51 #include <asm/unistd.h>
  52 #include <linux/slab.h>
  53 #include <linux/quotaops.h>
  54 #include <linux/module.h>
  55 #include <linux/init.h>
  56
  57 #include <obd_class.h>
  58 #include <lustre_quota.h>
  59 #include <lustre_fsfilt.h>
  60 #include <class_hash.h>
  61 #include <lprocfs_status.h>
  62 #include "quota_internal.h"
  63
  64 #ifdef HAVE_QUOTA_SUPPORT
  65
  66 static lustre_hash_ops_t lqs_hash_ops;
  67
  68 unsigned long default_bunit_sz = 128 * 1024 * 1024; /* 128M bytes */
  69 unsigned long default_btune_ratio = 50;             /* 50 percentage */
  70 unsigned long default_iunit_sz = 5120;              /* 5120 inodes */
  71 unsigned long default_itune_ratio = 50;             /* 50 percentage */
  72
  73 cfs_mem_cache_t *qunit_cachep = NULL;
  74 struct list_head qunit_hash[NR_DQHASH];
  75 spinlock_t qunit_hash_lock = SPIN_LOCK_UNLOCKED;
  76
  77 /* please sync qunit_state with qunit_state_names */
  78 enum qunit_state {
  79         QUNIT_CREATED      = 0,   /* a qunit is created */
  80         QUNIT_IN_HASH      = 1,   /* a qunit is added into qunit hash, that means
  81                                    * a quota req will be sent or is flying */
  82         QUNIT_RM_FROM_HASH = 2,   /* a qunit is removed from qunit hash, that
  83                                    * means a quota req is handled and comes
  84                                    * back */
  85         QUNIT_FINISHED     = 3,   /* qunit can wake up all threads waiting
  86                                    * for it */
  87 };
  88
  89 static const char *qunit_state_names[] = {
  90         [QUNIT_CREATED]      = "CREATED",
  91         [QUNIT_IN_HASH]      = "IN_HASH",
  92         [QUNIT_RM_FROM_HASH] = "RM_FROM_HASH",
  93         [QUNIT_FINISHED]     = "FINISHED",
  94 };
  95
  96 struct lustre_qunit {
  97         struct list_head lq_hash;          /* Hash list in memory */
  98         atomic_t lq_refcnt;                /* Use count */
  99         struct lustre_quota_ctxt *lq_ctxt; /* Quota context this applies to */
 100         struct qunit_data lq_data;         /* See qunit_data */
 101         unsigned int lq_opc;               /* QUOTA_DQACQ, QUOTA_DQREL */
 102         cfs_waitq_t lq_waitq;              /* Threads waiting for this qunit */
 103         spinlock_t lq_lock;                /* Protect the whole structure */
 104         enum qunit_state lq_state;         /* Present the status of qunit */
 105         int lq_rc;                         /* The rc of lq_data */
 106 };
 107
 108 #define QUNIT_SET_STATE(qunit, state)                                   \
 109 do {                                                                    \
 110         spin_lock(&qunit->lq_lock);                                     \
 111         QDATA_DEBUG((&qunit->lq_data), "qunit(%p) lq_state(%s->%s), "   \
 112                     "lq_rc(%d)\n",                                      \
 113                     qunit, qunit_state_names[qunit->lq_state],          \
 114                     qunit_state_names[state], qunit->lq_rc);            \
 115         qunit->lq_state = state;                                        \
 116         spin_unlock(&qunit->lq_lock);                                   \
 117 } while(0)
 118
 119 #define QUNIT_SET_STATE_AND_RC(qunit, state, rc)                        \
 120 do {                                                                    \
 121         spin_lock(&qunit->lq_lock);                                     \
 122         qunit->lq_rc = rc;                                              \
 123         QDATA_DEBUG((&qunit->lq_data), "qunit(%p) lq_state(%s->%s), "   \
 124                     "lq_rc(%d)\n",                                      \
 125                     qunit, qunit_state_names[qunit->lq_state],          \
 126                     qunit_state_names[state], qunit->lq_rc);            \
 127         qunit->lq_state = state;                                        \
 128         spin_unlock(&qunit->lq_lock);                                   \
 129 } while(0)
 130
 131
 132 int should_translate_quota (struct obd_import *imp)
 133 {
 134         ENTRY;
 135
 136         LASSERT(imp);
 137         if (imp->imp_connect_data.ocd_connect_flags & OBD_CONNECT_QUOTA64)
 138                 RETURN(0);
 139         else
 140                 RETURN(1);
 141 }
 142
 143 void qunit_cache_cleanup(void)
 144 {
 145         int i;
 146         ENTRY;
 147
 148         spin_lock(&qunit_hash_lock);
 149         for (i = 0; i < NR_DQHASH; i++)
 150                 LASSERT(list_empty(qunit_hash + i));
 151         spin_unlock(&qunit_hash_lock);
 152
 153         if (qunit_cachep) {
 154                 int rc;
 155                 rc = cfs_mem_cache_destroy(qunit_cachep);
 156                 LASSERTF(rc == 0, "couldn't destory qunit_cache slab\n");
 157                 qunit_cachep = NULL;
 158         }
 159         EXIT;
 160 }
 161
 162 int qunit_cache_init(void)
 163 {
 164         int i;
 165         ENTRY;
 166
 167         LASSERT(qunit_cachep == NULL);
 168         qunit_cachep = cfs_mem_cache_create("ll_qunit_cache",
 169                                             sizeof(struct lustre_qunit),
 170                                             0, 0);
 171         if (!qunit_cachep)
 172                 RETURN(-ENOMEM);
 173
 174         spin_lock(&qunit_hash_lock);
 175         for (i = 0; i < NR_DQHASH; i++)
 176                 INIT_LIST_HEAD(qunit_hash + i);
 177         spin_unlock(&qunit_hash_lock);
 178         RETURN(0);
 179 }
 180
 181 static inline int
 182 qunit_hashfn(struct lustre_quota_ctxt *qctxt, struct qunit_data *qdata)
 183              __attribute__((__const__));
 184
 185 static inline int
 186 qunit_hashfn(struct lustre_quota_ctxt *qctxt, struct qunit_data *qdata)
 187 {
 188         unsigned int id = qdata->qd_id;
 189         unsigned int type = QDATA_IS_GRP(qdata);
 190
 191         unsigned long tmp = ((unsigned long)qctxt >> L1_CACHE_SHIFT) ^ id;
 192         tmp = (tmp * (MAXQUOTAS - type)) % NR_DQHASH;
 193         return tmp;
 194 }
 195
 196 /* caller must hold qunit_hash_lock */
 197 static inline struct lustre_qunit *find_qunit(unsigned int hashent,
 198                                               struct lustre_quota_ctxt *qctxt,
 199                                               struct qunit_data *qdata)
 200 {
 201         struct lustre_qunit *qunit = NULL;
 202         struct qunit_data *tmp;
 203
 204         LASSERT_SPIN_LOCKED(&qunit_hash_lock);
 205         list_for_each_entry(qunit, qunit_hash + hashent, lq_hash) {
 206                 tmp = &qunit->lq_data;
 207                 if (qunit->lq_ctxt == qctxt &&
 208                     qdata->qd_id == tmp->qd_id &&
 209                     (qdata->qd_flags & LQUOTA_QUNIT_FLAGS) ==
 210                     (tmp->qd_flags & LQUOTA_QUNIT_FLAGS))
 211                         return qunit;
 212         }
 213         return NULL;
 214 }
 215
 216 /* check_cur_qunit - check the current usage of qunit.
 217  * @qctxt: quota context
 218  * @qdata: the type of quota unit to be checked
 219  *
 220  * return: 1 - need acquire qunit;
 221  *         2 - need release qunit;
 222  *         0 - need do nothing.
 223  *       < 0 - error.
 224  */
 225 static int
 226 check_cur_qunit(struct obd_device *obd,
 227                 struct lustre_quota_ctxt *qctxt, struct qunit_data *qdata)
 228 {
 229         struct super_block *sb = qctxt->lqc_sb;
 230         unsigned long qunit_sz, tune_sz;
 231         __u64 usage, limit, limit_org, pending_write = 0;
 232         long long record = 0;
 233         struct obd_quotactl *qctl;
 234         struct lustre_qunit_size *lqs = NULL;
 235         int ret = 0;
 236         ENTRY;
 237
 238         if (!ll_sb_any_quota_active(sb))
 239                 RETURN(0);
 240
 241         spin_lock(&qctxt->lqc_lock);
 242         if (!qctxt->lqc_valid){
 243                 spin_unlock(&qctxt->lqc_lock);
 244                 RETURN(0);
 245         }
 246         spin_unlock(&qctxt->lqc_lock);
 247
 248         OBD_ALLOC_PTR(qctl);
 249         if (qctl == NULL)
 250                 RETURN(-ENOMEM);
 251
 252         /* get fs quota usage & limit */
 253         qctl->qc_cmd = Q_GETQUOTA;
 254         qctl->qc_id = qdata->qd_id;
 255         qctl->qc_type = QDATA_IS_GRP(qdata);
 256         ret = fsfilt_quotactl(obd, sb, qctl);
 257         if (ret) {
 258                 if (ret == -ESRCH)      /* no limit */
 259                         ret = 0;
 260                 else
 261                         CERROR("can't get fs quota usage! (rc:%d)\n", ret);
 262                 GOTO(out, ret);
 263         }
 264
 265         if (QDATA_IS_BLK(qdata)) {
 266                 usage = qctl->qc_dqblk.dqb_curspace;
 267                 limit = qctl->qc_dqblk.dqb_bhardlimit << QUOTABLOCK_BITS;
 268         } else {
 269                 usage = qctl->qc_dqblk.dqb_curinodes;
 270                 limit = qctl->qc_dqblk.dqb_ihardlimit;
 271         }
 272
 273         /* ignore the no quota limit case; and it can avoid creating
 274          * unnecessary lqs for uid/gid */
 275         if (!limit)
 276                 GOTO(out, ret = 0);
 277
 278         lqs = quota_search_lqs(LQS_KEY(QDATA_IS_GRP(qdata), qdata->qd_id),
 279                                qctxt, 0);
 280         if (IS_ERR(lqs) || lqs == NULL) {
 281                 CDEBUG(D_ERROR, "fail to find a lqs(%s id: %u)!\n",
 282                        QDATA_IS_GRP(qdata) ? "group" : "user", qdata->qd_id);
 283                 GOTO (out, ret = 0);
 284         }
 285         spin_lock(&lqs->lqs_lock);
 286
 287         if (QDATA_IS_BLK(qdata)) {
 288                 qunit_sz = lqs->lqs_bunit_sz;
 289                 tune_sz  = lqs->lqs_btune_sz;
 290                 pending_write = lqs->lqs_bwrite_pending;
 291                 record   = lqs->lqs_blk_rec;
 292                 LASSERT(!(qunit_sz % QUOTABLOCK_SIZE));
 293         } else {
 294                 /* we didn't need change inode qunit size now */
 295                 qunit_sz = lqs->lqs_iunit_sz;
 296                 tune_sz  = lqs->lqs_itune_sz;
 297                 pending_write = lqs->lqs_iwrite_pending;
 298                 record   = lqs->lqs_ino_rec;
 299         }
 300
 301         /* we don't count the MIN_QLIMIT */
 302         if ((limit == MIN_QLIMIT && !QDATA_IS_BLK(qdata)) ||
 303             (toqb(limit) == MIN_QLIMIT && QDATA_IS_BLK(qdata)))
 304                 limit = 0;
 305
 306         usage += pending_write;
 307         limit_org = limit;
 308         /* when a releasing quota req is sent, before it returned
 309            limit is assigned a small value. limit will overflow */
 310         if (limit + record < 0)
 311                 usage -= record;
 312         else
 313                 limit += record;
 314
 315         LASSERT(qdata->qd_count == 0);
 316         if (limit <= usage + tune_sz) {
 317                 while (qdata->qd_count + limit <=
 318                        usage + tune_sz)
 319                         qdata->qd_count += qunit_sz;
 320                 ret = 1;
 321         } else if (limit > usage + qunit_sz + tune_sz &&
 322                    limit_org > qdata->qd_count + qunit_sz) {
 323                 while (limit - qdata->qd_count > usage + qunit_sz + tune_sz &&
 324                        limit_org > qdata->qd_count + qunit_sz)
 325                         qdata->qd_count += qunit_sz;
 326                 ret = 2;
 327                 /* if there are other pending writes for this uid/gid, releasing
 328                  * quota is put off until the last pending write b=16645 */
 329                 if (ret == 2 && pending_write) {
 330                         CDEBUG(D_QUOTA, "delay quota release\n");
 331                         ret = 0;
 332                 }
 333         }
 334         CDEBUG(D_QUOTA, "type: %c, limit: "LPU64", usage: "LPU64
 335                ", pending_write: "LPU64", record: "LPD64
 336                ", qunit_sz: %lu, tune_sz: %lu, ret: %d.\n",
 337                QDATA_IS_BLK(qdata) ? 'b' : 'i', limit, usage, pending_write,
 338                (__s64)record, qunit_sz, tune_sz, ret);
 339         LASSERT(ret == 0 || qdata->qd_count);
 340
 341         spin_unlock(&lqs->lqs_lock);
 342         lqs_putref(lqs);
 343
 344         EXIT;
 345  out:
 346         OBD_FREE_PTR(qctl);
 347         return ret;
 348 }
 349
 350 /* compute the remaining quota for certain gid or uid b=11693 */
 351 int compute_remquota(struct obd_device *obd, struct lustre_quota_ctxt *qctxt,
 352                      struct qunit_data *qdata, int isblk)
 353 {
 354         struct super_block *sb = qctxt->lqc_sb;
 355         __u64 usage, limit;
 356         struct obd_quotactl *qctl;
 357         int ret = QUOTA_RET_OK;
 358         ENTRY;
 359
 360         if (!ll_sb_any_quota_active(sb))
 361                 RETURN(QUOTA_RET_NOQUOTA);
 362
 363         /* ignore root user */
 364         if (qdata->qd_id == 0 && QDATA_IS_GRP(qdata) == USRQUOTA)
 365                 RETURN(QUOTA_RET_NOLIMIT);
 366
 367         OBD_ALLOC_PTR(qctl);
 368         if (qctl == NULL)
 369                 RETURN(-ENOMEM);
 370
 371         /* get fs quota usage & limit */
 372         qctl->qc_cmd = Q_GETQUOTA;
 373         qctl->qc_id = qdata->qd_id;
 374         qctl->qc_type = QDATA_IS_GRP(qdata);
 375         ret = fsfilt_quotactl(obd, sb, qctl);
 376         if (ret) {
 377                 if (ret == -ESRCH)      /* no limit */
 378                         ret = QUOTA_RET_NOLIMIT;
 379                 else
 380                         CDEBUG(D_QUOTA, "can't get fs quota usage! (rc:%d)",
 381                                ret);
 382                 GOTO(out, ret);
 383         }
 384
 385         usage = isblk ? qctl->qc_dqblk.dqb_curspace :
 386                 qctl->qc_dqblk.dqb_curinodes;
 387         limit = isblk ? qctl->qc_dqblk.dqb_bhardlimit << QUOTABLOCK_BITS :
 388                 qctl->qc_dqblk.dqb_ihardlimit;
 389         if (!limit){            /* no limit */
 390                 ret = QUOTA_RET_NOLIMIT;
 391                 GOTO(out, ret);
 392         }
 393
 394         if (limit >= usage)
 395                 qdata->qd_count = limit - usage;
 396         else
 397                 qdata->qd_count = 0;
 398         EXIT;
 399 out:
 400         OBD_FREE_PTR(qctl);
 401         return ret;
 402 }
 403
 404 static struct lustre_qunit *alloc_qunit(struct lustre_quota_ctxt *qctxt,
 405                                         struct qunit_data *qdata, int opc)
 406 {
 407         struct lustre_qunit *qunit = NULL;
 408         ENTRY;
 409
 410         OBD_SLAB_ALLOC(qunit, qunit_cachep, CFS_ALLOC_IO, sizeof(*qunit));
 411         if (qunit == NULL)
 412                 RETURN(NULL);
 413
 414         INIT_LIST_HEAD(&qunit->lq_hash);
 415         init_waitqueue_head(&qunit->lq_waitq);
 416         atomic_set(&qunit->lq_refcnt, 1);
 417         qunit->lq_ctxt = qctxt;
 418         memcpy(&qunit->lq_data, qdata, sizeof(*qdata));
 419         qunit->lq_opc = opc;
 420         qunit->lq_lock = SPIN_LOCK_UNLOCKED;
 421         QUNIT_SET_STATE_AND_RC(qunit, QUNIT_CREATED, 0);
 422         RETURN(qunit);
 423 }
 424
 425 static inline void free_qunit(struct lustre_qunit *qunit)
 426 {
 427         OBD_SLAB_FREE(qunit, qunit_cachep, sizeof(*qunit));
 428 }
 429
 430 static inline void qunit_get(struct lustre_qunit *qunit)
 431 {
 432         atomic_inc(&qunit->lq_refcnt);
 433 }
 434
 435 static void qunit_put(struct lustre_qunit *qunit)
 436 {
 437         LASSERT(atomic_read(&qunit->lq_refcnt));
 438         if (atomic_dec_and_test(&qunit->lq_refcnt))
 439                 free_qunit(qunit);
 440 }
 441
 442 /* caller must hold qunit_hash_lock and release ref of qunit after using it */
 443 static struct lustre_qunit *dqacq_in_flight(struct lustre_quota_ctxt *qctxt,
 444                                             struct qunit_data *qdata)
 445 {
 446         unsigned int hashent = qunit_hashfn(qctxt, qdata);
 447         struct lustre_qunit *qunit;
 448         ENTRY;
 449
 450         LASSERT_SPIN_LOCKED(&qunit_hash_lock);
 451         qunit = find_qunit(hashent, qctxt, qdata);
 452         if (qunit)
 453                 qunit_get(qunit);
 454         RETURN(qunit);
 455 }
 456
 457 static void
 458 insert_qunit_nolock(struct lustre_quota_ctxt *qctxt, struct lustre_qunit *qunit)
 459 {
 460         struct list_head *head;
 461
 462         LASSERT(list_empty(&qunit->lq_hash));
 463         qunit_get(qunit);
 464         head = qunit_hash + qunit_hashfn(qctxt, &qunit->lq_data);
 465         list_add(&qunit->lq_hash, head);
 466         QUNIT_SET_STATE(qunit, QUNIT_IN_HASH);
 467 }
 468
 469 static void compute_lqs_after_removing_qunit(struct lustre_qunit *qunit)
 470 {
 471         struct lustre_qunit_size *lqs;
 472
 473         lqs = quota_search_lqs(LQS_KEY(QDATA_IS_GRP(&qunit->lq_data),
 474                                        qunit->lq_data.qd_id),
 475                                qunit->lq_ctxt, 0);
 476         if (lqs && !IS_ERR(lqs)) {
 477                 spin_lock(&lqs->lqs_lock);
 478                 if (qunit->lq_opc == QUOTA_DQACQ)
 479                         quota_compute_lqs(&qunit->lq_data, lqs, 0, 1);
 480                 if (qunit->lq_opc == QUOTA_DQREL)
 481                         quota_compute_lqs(&qunit->lq_data, lqs, 0, 0);
 482                 spin_unlock(&lqs->lqs_lock);
 483                 /* this is for quota_search_lqs */
 484                 lqs_putref(lqs);
 485                 /* this is for schedule_dqacq */
 486                 lqs_putref(lqs);
 487         }
 488 }
 489
 490 static void remove_qunit_nolock(struct lustre_qunit *qunit)
 491 {
 492         LASSERT(!list_empty(&qunit->lq_hash));
 493         LASSERT_SPIN_LOCKED(&qunit_hash_lock);
 494
 495         list_del_init(&qunit->lq_hash);
 496         QUNIT_SET_STATE(qunit, QUNIT_RM_FROM_HASH);
 497         qunit_put(qunit);
 498 }
 499
 500 #define INC_QLIMIT(limit, count) (limit == MIN_QLIMIT) ? \
 501                                  (limit = count) : (limit += count)
 502
 503
 504 /* FIXME check if this mds is the master of specified id */
 505 static int
 506 is_master(struct obd_device *obd, struct lustre_quota_ctxt *qctxt,
 507           unsigned int id, int type)
 508 {
 509         return qctxt->lqc_handler ? 1 : 0;
 510 }
 511
 512 static int
 513 schedule_dqacq(struct obd_device *obd, struct lustre_quota_ctxt *qctxt,
 514                struct qunit_data *qdata, int opc, int wait,
 515                struct obd_trans_info *oti);
 516
 517 static inline void qdata_to_oqaq(struct qunit_data *qdata,
 518                                  struct quota_adjust_qunit *oqaq)
 519 {
 520         LASSERT(qdata);
 521         LASSERT(oqaq);
 522
 523         oqaq->qaq_flags = qdata->qd_flags;
 524         oqaq->qaq_id    = qdata->qd_id;
 525         if (QDATA_IS_ADJBLK(qdata))
 526                 oqaq->qaq_bunit_sz = qdata->qd_qunit;
 527         if (QDATA_IS_ADJINO(qdata))
 528                 oqaq->qaq_iunit_sz = qdata->qd_qunit;
 529 }
 530
 531 static int
 532 dqacq_completion(struct obd_device *obd, struct lustre_quota_ctxt *qctxt,
 533                  struct qunit_data *qdata, int rc, int opc)
 534 {
 535         struct lustre_qunit *qunit = NULL;
 536         struct super_block *sb = qctxt->lqc_sb;
 537         int err = 0;
 538         struct quota_adjust_qunit *oqaq = NULL;
 539         int rc1 = 0;
 540         ENTRY;
 541
 542         LASSERT(qdata);
 543         QDATA_DEBUG(qdata, "obd(%s): complete %s quota req\n",
 544                     obd->obd_name, (opc == QUOTA_DQACQ) ? "acq" : "rel");
 545
 546         /* do it only when a releasing quota req more than 5MB b=18491 */
 547         if (opc == QUOTA_DQREL && qdata->qd_count >= 5242880)
 548                 OBD_FAIL_TIMEOUT(OBD_FAIL_QUOTA_DELAY_REL, 5);
 549
 550         /* update local operational quota file */
 551         if (rc == 0) {
 552                 __u64 count = QUSG(qdata->qd_count, QDATA_IS_BLK(qdata));
 553                 struct obd_quotactl *qctl;
 554                 __u64 *hardlimit;
 555
 556                 OBD_ALLOC_PTR(qctl);
 557                 if (qctl == NULL)
 558                         GOTO(out, err = -ENOMEM);
 559
 560                 /* acq/rel qunit for specified uid/gid is serialized,
 561                  * so there is no race between get fs quota limit and
 562                  * set fs quota limit */
 563                 qctl->qc_cmd = Q_GETQUOTA;
 564                 qctl->qc_id = qdata->qd_id;
 565                 qctl->qc_type = QDATA_IS_GRP(qdata);
 566                 err = fsfilt_quotactl(obd, sb, qctl);
 567                 if (err) {
 568                         CERROR("error get quota fs limit! (rc:%d)\n", err);
 569                         GOTO(out_mem, err);
 570                 }
 571
 572                 if (QDATA_IS_BLK(qdata)) {
 573                         qctl->qc_dqblk.dqb_valid = QIF_BLIMITS;
 574                         hardlimit = &qctl->qc_dqblk.dqb_bhardlimit;
 575                 } else {
 576                         qctl->qc_dqblk.dqb_valid = QIF_ILIMITS;
 577                         hardlimit = &qctl->qc_dqblk.dqb_ihardlimit;
 578                 }
 579
 580                 CDEBUG(D_QUOTA, "hardlimt: "LPU64"\n", *hardlimit);
 581
 582                 if (*hardlimit == 0)
 583                         goto out_mem;
 584
 585                 switch (opc) {
 586                 case QUOTA_DQACQ:
 587                         INC_QLIMIT(*hardlimit, count);
 588                         break;
 589                 case QUOTA_DQREL:
 590                         LASSERTF(count < *hardlimit,
 591                                  "id(%u) flag(%u) type(%c) isblk(%c) "
 592                                  "count("LPU64") qd_qunit("LPU64") "
 593                                  "hardlimit("LPU64").\n",
 594                                  qdata->qd_id, qdata->qd_flags,
 595                                  QDATA_IS_GRP(qdata) ? 'g' : 'u',
 596                                  QDATA_IS_BLK(qdata) ? 'b': 'i',
 597                                  qdata->qd_count, qdata->qd_qunit, *hardlimit);
 598                         *hardlimit -= count;
 599                         break;
 600                 default:
 601                         LBUG();
 602                 }
 603
 604                 /* clear quota limit */
 605                 if (count == 0)
 606                         *hardlimit = 0;
 607
 608                 qctl->qc_cmd = Q_SETQUOTA;
 609                 err = fsfilt_quotactl(obd, sb, qctl);
 610                 if (err)
 611                         CERROR("error set quota fs limit! (rc:%d)\n", err);
 612
 613                 QDATA_DEBUG(qdata, "%s completion\n",
 614                             opc == QUOTA_DQACQ ? "DQACQ" : "DQREL");
 615 out_mem:
 616                 OBD_FREE_PTR(qctl);
 617         } else if (rc == -EDQUOT) {
 618                 QDATA_DEBUG(qdata, "acquire qunit got EDQUOT.\n");
 619         } else if (rc == -EBUSY) {
 620                 QDATA_DEBUG(qdata, "it's is recovering, got EBUSY.\n");
 621         } else {
 622                 CERROR("acquire qunit got error! (rc:%d)\n", rc);
 623         }
 624 out:
 625         /* remove the qunit from hash */
 626         spin_lock(&qunit_hash_lock);
 627
 628         qunit = dqacq_in_flight(qctxt, qdata);
 629         /* this qunit has been removed by qctxt_cleanup() */
 630         if (!qunit) {
 631                 spin_unlock(&qunit_hash_lock);
 632                 QDATA_DEBUG(qdata, "%s is discarded because qunit isn't found\n",
 633                             opc == QUOTA_DQACQ ? "DQACQ" : "DQREL");
 634                 RETURN(err);
 635         }
 636
 637         LASSERT(opc == qunit->lq_opc);
 638         /* remove this qunit from lq_hash so that new processes cannot be added
 639          * to qunit->lq_waiters */
 640         remove_qunit_nolock(qunit);
 641         spin_unlock(&qunit_hash_lock);
 642
 643         compute_lqs_after_removing_qunit(qunit);
 644
 645
 646         if (rc == 0)
 647                 rc = QUOTA_REQ_RETURNED;
 648         QUNIT_SET_STATE_AND_RC(qunit, QUNIT_FINISHED, rc);
 649         /* wake up all waiters */
 650         wake_up(&qunit->lq_waitq);
 651
 652         /* this is for dqacq_in_flight() */
 653         qunit_put(qunit);
 654         /* this is for alloc_qunit() */
 655         qunit_put(qunit);
 656         if (rc < 0 && rc != -EDQUOT)
 657                  RETURN(err);
 658
 659         /* don't reschedule in such cases:
 660          *   - acq/rel failure and qunit isn't changed,
 661          *     but not for quota recovery.
 662          *   - local dqacq/dqrel.
 663          *   - local disk io failure.
 664          */
 665          OBD_ALLOC_PTR(oqaq);
 666          if (!oqaq)
 667                  RETURN(-ENOMEM);
 668          qdata_to_oqaq(qdata, oqaq);
 669          /* adjust the qunit size in slaves */
 670          rc1 = quota_adjust_slave_lqs(oqaq, qctxt);
 671          OBD_FREE_PTR(oqaq);
 672          if (rc1 < 0) {
 673                  CERROR("adjust slave's qunit size failed!(rc:%d)\n", rc1);
 674                  RETURN(rc1);
 675          }
 676          if (err || (rc < 0 && rc != -EBUSY && rc1 == 0) ||
 677              is_master(obd, qctxt, qdata->qd_id, QDATA_IS_GRP(qdata)))
 678                 RETURN(err);
 679
 680          if (opc == QUOTA_DQREL && qdata->qd_count >= 5242880)
 681                  OBD_FAIL_RETURN(OBD_FAIL_QUOTA_DELAY_REL, err);
 682
 683         /* reschedule another dqacq/dqrel if needed */
 684         qdata->qd_count = 0;
 685         qdata->qd_flags &= LQUOTA_QUNIT_FLAGS;
 686         rc1 = check_cur_qunit(obd, qctxt, qdata);
 687         if (rc1 > 0) {
 688                 int opc;
 689                 opc = rc1 == 1 ? QUOTA_DQACQ : QUOTA_DQREL;
 690                 rc1 = schedule_dqacq(obd, qctxt, qdata, opc, 0, NULL);
 691                 QDATA_DEBUG(qdata, "reschedudle opc(%d) rc(%d)\n", opc, rc1);
 692         }
 693         RETURN(err);
 694 }
 695
 696 struct dqacq_async_args {
 697         struct lustre_quota_ctxt *aa_ctxt;
 698         struct lustre_qunit *aa_qunit;
 699 };
 700
 701 static int dqacq_interpret(struct ptlrpc_request *req, void *data, int rc)
 702 {
 703         struct dqacq_async_args *aa = (struct dqacq_async_args *)data;
 704         struct lustre_quota_ctxt *qctxt = aa->aa_ctxt;
 705         struct lustre_qunit *qunit = aa->aa_qunit;
 706         struct obd_device *obd = req->rq_import->imp_obd;
 707         struct qunit_data *qdata = NULL;
 708         int rc1 = 0;
 709         ENTRY;
 710
 711         LASSERT(req);
 712         LASSERT(req->rq_import);
 713
 714         /* there are several forms of qunit(historic causes), so we need to
 715          * adjust qunit from slaves to the same form here */
 716         OBD_ALLOC(qdata, sizeof(struct qunit_data));
 717         if (!qdata)
 718                 RETURN(-ENOMEM);
 719
 720         /* if a quota req timeouts or is dropped, we should update quota
 721          * statistics which will be handled in dqacq_completion. And in
 722          * this situation we should get qdata from request instead of
 723          * reply */
 724         rc1 = quota_get_qdata(req, qdata,
 725                               (rc != 0) ? QUOTA_REQUEST : QUOTA_REPLY,
 726                               QUOTA_IMPORT);
 727         if (rc1 < 0) {
 728                 DEBUG_REQ(D_ERROR, req,
 729                           "error unpacking qunit_data(rc: %d)\n", rc1);
 730                 GOTO(exit, rc = rc1);
 731         }
 732
 733         QDATA_DEBUG(qdata, "qdata: interpret rc(%d).\n", rc);
 734         QDATA_DEBUG((&qunit->lq_data), "lq_data: \n");
 735
 736         if (qdata->qd_id != qunit->lq_data.qd_id ||
 737             OBD_FAIL_CHECK_ONCE(OBD_FAIL_QUOTA_RET_QDATA)) {
 738                 CDEBUG(D_ERROR, "the returned qd_id isn't expected!"
 739                        "(qdata: %u, lq_data: %u)\n", qdata->qd_id,
 740                        qunit->lq_data.qd_id);
 741                 qdata->qd_id = qunit->lq_data.qd_id;
 742                 rc = -EPROTO;
 743         }
 744         if (QDATA_IS_GRP(qdata) != QDATA_IS_GRP(&qunit->lq_data)) {
 745                 CDEBUG(D_ERROR, "the returned grp/usr isn't expected!"
 746                        "(qdata: %u, lq_data: %u)\n", qdata->qd_flags,
 747                        qunit->lq_data.qd_flags);
 748                 if (QDATA_IS_GRP(&qunit->lq_data))
 749                         QDATA_SET_GRP(qdata);
 750                 else
 751                         QDATA_CLR_GRP(qdata);
 752                 rc = -EPROTO;
 753         }
 754         if (qdata->qd_count > qunit->lq_data.qd_count) {
 755                 CDEBUG(D_ERROR, "the returned qd_count isn't expected!"
 756                        "(qdata: "LPU64", lq_data: "LPU64")\n", qdata->qd_count,
 757                        qunit->lq_data.qd_count);
 758                 rc = -EPROTO;
 759         }
 760
 761         rc = dqacq_completion(obd, qctxt, qdata, rc,
 762                               lustre_msg_get_opc(req->rq_reqmsg));
 763
 764 exit:
 765         OBD_FREE(qdata, sizeof(struct qunit_data));
 766
 767         RETURN(rc);
 768 }
 769
 770 /* check if quota master is online */
 771 int check_qm(struct lustre_quota_ctxt *qctxt)
 772 {
 773         int rc;
 774         ENTRY;
 775
 776         spin_lock(&qctxt->lqc_lock);
 777         /* quit waiting when mds is back or qctxt is cleaned up */
 778         rc = qctxt->lqc_import || !qctxt->lqc_valid;
 779         spin_unlock(&qctxt->lqc_lock);
 780
 781         RETURN(rc);
 782 }
 783
 784 /* wake up all waiting threads when lqc_import is NULL */
 785 void dqacq_interrupt(struct lustre_quota_ctxt *qctxt)
 786 {
 787         struct lustre_qunit *qunit, *tmp;
 788         int i;
 789         ENTRY;
 790
 791         spin_lock(&qunit_hash_lock);
 792         for (i = 0; i < NR_DQHASH; i++) {
 793                 list_for_each_entry_safe(qunit, tmp, &qunit_hash[i], lq_hash) {
 794                         if (qunit->lq_ctxt != qctxt)
 795                                 continue;
 796
 797                         /* Wake up all waiters. Do not change lq_state.
 798                          * The waiters will check lq_rc which is kept as 0
 799                          * if no others change it, then the waiters will return
 800                          * -EAGAIN to caller who can perform related quota
 801                          * acq/rel if necessary. */
 802                         wake_up_all(&qunit->lq_waitq);
 803                 }
 804         }
 805         spin_unlock(&qunit_hash_lock);
 806         EXIT;
 807 }
 808
 809 static int got_qunit(struct lustre_qunit *qunit)
 810 {
 811         struct lustre_quota_ctxt *qctxt = qunit->lq_ctxt;
 812         int rc = 0;
 813         ENTRY;
 814
 815         spin_lock(&qunit->lq_lock);
 816         switch (qunit->lq_state) {
 817         case QUNIT_IN_HASH:
 818         case QUNIT_RM_FROM_HASH:
 819                 break;
 820         case QUNIT_FINISHED:
 821                 rc = 1;
 822                 break;
 823         default:
 824                 CERROR("invalid qunit state %d\n", qunit->lq_state);
 825         }
 826         spin_unlock(&qunit->lq_lock);
 827
 828         if (!rc) {
 829                 spin_lock(&qctxt->lqc_lock);
 830                 rc = !qctxt->lqc_import || !qctxt->lqc_valid;
 831                 spin_unlock(&qctxt->lqc_lock);
 832         }
 833
 834         RETURN(rc);
 835 }
 836
 837 static int
 838 schedule_dqacq(struct obd_device *obd, struct lustre_quota_ctxt *qctxt,
 839                struct qunit_data *qdata, int opc, int wait,
 840                struct obd_trans_info *oti)
 841 {
 842         struct lustre_qunit *qunit, *empty;
 843         struct l_wait_info lwi = { 0 };
 844         struct ptlrpc_request *req;
 845         struct dqacq_async_args *aa;
 846         int size[2] = { sizeof(struct ptlrpc_body), 0 };
 847         struct obd_import *imp = NULL;
 848         struct lustre_qunit_size *lqs = NULL;
 849         struct timeval work_start;
 850         struct timeval work_end;
 851         long timediff;
 852         int rc = 0;
 853         ENTRY;
 854
 855         LASSERT(opc == QUOTA_DQACQ || opc == QUOTA_DQREL);
 856         do_gettimeofday(&work_start);
 857         if ((empty = alloc_qunit(qctxt, qdata, opc)) == NULL)
 858                 RETURN(-ENOMEM);
 859
 860         spin_lock(&qunit_hash_lock);
 861         qunit = dqacq_in_flight(qctxt, qdata);
 862         if (qunit) {
 863                 spin_unlock(&qunit_hash_lock);
 864                 qunit_put(empty);
 865
 866                 goto wait_completion;
 867         }
 868         qunit = empty;
 869         qunit_get(qunit);
 870         insert_qunit_nolock(qctxt, qunit);
 871         spin_unlock(&qunit_hash_lock);
 872
 873         lqs = quota_search_lqs(LQS_KEY(QDATA_IS_GRP(qdata), qdata->qd_id),
 874                                qctxt, 0);
 875         if (lqs && !IS_ERR(lqs)) {
 876                 spin_lock(&lqs->lqs_lock);
 877                 quota_compute_lqs(qdata, lqs, 1, (opc == QUOTA_DQACQ) ? 1 : 0);
 878                 /* when this qdata returned from mds, it will call lqs_putref */
 879                 lqs_getref(lqs);
 880                 spin_unlock(&lqs->lqs_lock);
 881                 /* this is for quota_search_lqs */
 882                 lqs_putref(lqs);
 883         } else {
 884                 CDEBUG(D_ERROR, "Can't find the lustre qunit size!\n");
 885         }
 886
 887         QDATA_DEBUG(qdata, "obd(%s): send %s quota req\n",
 888                     obd->obd_name, (opc == QUOTA_DQACQ) ? "acq" : "rel");
 889         /* master is going to dqacq/dqrel from itself */
 890         if (is_master(obd, qctxt, qdata->qd_id, QDATA_IS_GRP(qdata))) {
 891                 int rc2;
 892                 QDATA_DEBUG(qdata, "local %s.\n",
 893                             opc == QUOTA_DQACQ ? "DQACQ" : "DQREL");
 894                 QDATA_SET_CHANGE_QS(qdata);
 895                 rc = qctxt->lqc_handler(obd, qdata, opc);
 896                 rc2 = dqacq_completion(obd, qctxt, qdata, rc, opc);
 897                 /* this is for qunit_get() */
 898                 qunit_put(qunit);
 899
 900                 do_gettimeofday(&work_end);
 901                 timediff = cfs_timeval_sub(&work_end, &work_start, NULL);
 902                 if (opc == QUOTA_DQACQ)
 903                         lprocfs_counter_add(qctxt->lqc_stats,
 904                                             wait ? LQUOTA_SYNC_ACQ : LQUOTA_ASYNC_ACQ,
 905                                             timediff);
 906                 else
 907                         lprocfs_counter_add(qctxt->lqc_stats,
 908                                             wait ? LQUOTA_SYNC_REL : LQUOTA_ASYNC_REL,
 909                                             timediff);
 910                 RETURN(rc ? rc : rc2);
 911         }
 912
 913         spin_lock(&qctxt->lqc_lock);
 914         if (!qctxt->lqc_import) {
 915                 spin_unlock(&qctxt->lqc_lock);
 916                 QDATA_DEBUG(qdata, "lqc_import is invalid.\n");
 917
 918                 spin_lock(&qunit_hash_lock);
 919                 remove_qunit_nolock(qunit);
 920                 spin_unlock(&qunit_hash_lock);
 921
 922                 compute_lqs_after_removing_qunit(qunit);
 923
 924                 QUNIT_SET_STATE_AND_RC(qunit, QUNIT_FINISHED, -EAGAIN);
 925                 wake_up(&qunit->lq_waitq);
 926
 927                 /* this is for qunit_get() */
 928                 qunit_put(qunit);
 929                 /* this for alloc_qunit() */
 930                 qunit_put(qunit);
 931                 spin_lock(&qctxt->lqc_lock);
 932                 if (wait && !qctxt->lqc_import) {
 933                         spin_unlock(&qctxt->lqc_lock);
 934
 935                         LASSERT(oti && oti->oti_thread &&
 936                                 oti->oti_thread->t_watchdog);
 937
 938                         lc_watchdog_disable(oti->oti_thread->t_watchdog);
 939                         CDEBUG(D_QUOTA, "sleep for quota master\n");
 940                         l_wait_event(qctxt->lqc_wait_for_qmaster,
 941                                      check_qm(qctxt), &lwi);
 942                         CDEBUG(D_QUOTA, "wake up when quota master is back\n");
 943                         lc_watchdog_touch(oti->oti_thread->t_watchdog,
 944                                  GET_TIMEOUT(oti->oti_thread->t_svc));
 945                 } else {
 946                         spin_unlock(&qctxt->lqc_lock);
 947                 }
 948
 949                 RETURN(-EAGAIN);
 950         }
 951         imp = class_import_get(qctxt->lqc_import);
 952         spin_unlock(&qctxt->lqc_lock);
 953
 954         /* build dqacq/dqrel request */
 955         LASSERT(imp);
 956         size[1] = quota_get_qunit_data_size(imp->
 957                                             imp_connect_data.ocd_connect_flags);
 958
 959         req = ptlrpc_prep_req(imp, LUSTRE_MDS_VERSION, opc, 2,
 960                               size, NULL);
 961         if (!req) {
 962                 dqacq_completion(obd, qctxt, qdata, -ENOMEM, opc);
 963                 class_import_put(imp);
 964                 /* this is for qunit_get() */
 965                 qunit_put(qunit);
 966                 RETURN(-ENOMEM);
 967         }
 968
 969         rc = quota_copy_qdata(req, qdata, QUOTA_REQUEST, QUOTA_IMPORT);
 970         if (rc < 0) {
 971                 CDEBUG(D_ERROR, "Can't pack qunit_data(rc: %d)\n", rc);
 972                 dqacq_completion(obd, qctxt, qdata, rc, opc);
 973                 class_import_put(imp);
 974                 /* this is for qunit_get() */
 975                 qunit_put(qunit);
 976                 RETURN(rc);
 977         }
 978         ptlrpc_req_set_repsize(req, 2, size);
 979         req->rq_no_resend = req->rq_no_delay = 1;
 980         class_import_put(imp);
 981
 982         CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
 983         aa = ptlrpc_req_async_args(req);
 984         aa->aa_ctxt = qctxt;
 985         aa->aa_qunit = qunit;
 986
 987         req->rq_interpret_reply = dqacq_interpret;
 988         ptlrpcd_add_req(req);
 989
 990         QDATA_DEBUG(qdata, "%s scheduled.\n",
 991                     opc == QUOTA_DQACQ ? "DQACQ" : "DQREL");
 992 wait_completion:
 993         if (wait && qunit) {
 994                 struct qunit_data *p = &qunit->lq_data;
 995
 996                 QDATA_DEBUG(p, "qunit(%p) is waiting for dqacq.\n", qunit);
 997                 l_wait_event(qunit->lq_waitq, got_qunit(qunit), &lwi);
 998                 /* rc = -EAGAIN, it means the quota master isn't ready yet
 999                  * rc = QUOTA_REQ_RETURNED, it means a quota req is finished;
1000                  * rc = -EDQUOT, it means out of quota
1001                  * rc = -EBUSY, it means recovery is happening
1002                  * other rc < 0, it means real errors, functions who call
1003                  * schedule_dqacq should take care of this */
1004                 spin_lock(&qunit->lq_lock);
1005                 rc = qunit->lq_rc;
1006                 spin_unlock(&qunit->lq_lock);
1007                 CDEBUG(D_QUOTA, "qunit(%p) finishes waiting. (rc:%d)\n",
1008                        qunit, rc);
1009         }
1010
1011         qunit_put(qunit);
1012         do_gettimeofday(&work_end);
1013         timediff = cfs_timeval_sub(&work_end, &work_start, NULL);
1014         if (opc == QUOTA_DQACQ)
1015                 lprocfs_counter_add(qctxt->lqc_stats,
1016                                     wait ? LQUOTA_SYNC_ACQ : LQUOTA_ASYNC_ACQ,
1017                                     timediff);
1018         else
1019                 lprocfs_counter_add(qctxt->lqc_stats,
1020                                     wait ? LQUOTA_SYNC_REL : LQUOTA_ASYNC_REL,
1021                                     timediff);
1022
1023         RETURN(rc);
1024 }
1025
1026 int
1027 qctxt_adjust_qunit(struct obd_device *obd, struct lustre_quota_ctxt *qctxt,
1028                    uid_t uid, gid_t gid, __u32 isblk, int wait,
1029                    struct obd_trans_info *oti)
1030 {
1031         int rc = 0, i = USRQUOTA;
1032         __u32 id[MAXQUOTAS] = { uid, gid };
1033         struct qunit_data qdata[MAXQUOTAS];
1034         ENTRY;
1035
1036         if (quota_is_set(obd, uid, gid, isblk ? QB_SET : QI_SET) == 0)
1037                 RETURN(0);
1038
1039         for (i = 0; i < MAXQUOTAS; i++) {
1040                 qdata[i].qd_id = id[i];
1041                 qdata[i].qd_flags = i;
1042                 if (isblk)
1043                         QDATA_SET_BLK(&qdata[i]);
1044                 qdata[i].qd_count = 0;
1045
1046                 rc = check_cur_qunit(obd, qctxt, &qdata[i]);
1047                 if (rc > 0) {
1048                         int opc;
1049                         /* need acquire or release */
1050                         opc = rc == 1 ? QUOTA_DQACQ : QUOTA_DQREL;
1051                         rc = schedule_dqacq(obd, qctxt, &qdata[i], opc,
1052                                             wait,oti);
1053                         if (rc < 0)
1054                                 RETURN(rc);
1055                 } else if (wait == 1) {
1056                         /* when wait equates 1, that means mds_quota_acquire
1057                          * or filter_quota_acquire is calling it. */
1058                         rc = qctxt_wait_pending_dqacq(qctxt, id[i], i, isblk);
1059                         if (rc < 0)
1060                                 RETURN(rc);
1061                 }
1062         }
1063
1064         RETURN(rc);
1065 }
1066
1067 int
1068 qctxt_wait_pending_dqacq(struct lustre_quota_ctxt *qctxt, unsigned int id,
1069                          unsigned short type, int isblk)
1070 {
1071         struct lustre_qunit *qunit = NULL;
1072         struct qunit_data qdata;
1073         struct timeval work_start;
1074         struct timeval work_end;
1075         long timediff;
1076         struct l_wait_info lwi = { 0 };
1077         int rc = 0;
1078         ENTRY;
1079
1080         do_gettimeofday(&work_start);
1081         qdata.qd_id = id;
1082         qdata.qd_flags = type;
1083         if (isblk)
1084                 QDATA_SET_BLK(&qdata);
1085         qdata.qd_count = 0;
1086
1087         spin_lock(&qunit_hash_lock);
1088         qunit = dqacq_in_flight(qctxt, &qdata);
1089         spin_unlock(&qunit_hash_lock);
1090
1091         if (qunit) {
1092                 struct qunit_data *p = &qunit->lq_data;
1093
1094                 QDATA_DEBUG(p, "qunit(%p) is waiting for dqacq.\n", qunit);
1095                 l_wait_event(qunit->lq_waitq, got_qunit(qunit), &lwi);
1096                 CDEBUG(D_QUOTA, "qunit(%p) finishes waiting. (rc:%d)\n",
1097                        qunit, qunit->lq_rc);
1098                 /* keep same as schedule_dqacq() b=17030 */
1099                 spin_lock(&qunit->lq_lock);
1100                 rc = qunit->lq_rc;
1101                 spin_unlock(&qunit->lq_lock);
1102                 /* this is for dqacq_in_flight() */
1103                 qunit_put(qunit);
1104                 do_gettimeofday(&work_end);
1105                 timediff = cfs_timeval_sub(&work_end, &work_start, NULL);
1106                 lprocfs_counter_add(qctxt->lqc_stats,
1107                                     isblk ? LQUOTA_WAIT_PENDING_BLK_QUOTA :
1108                                             LQUOTA_WAIT_PENDING_INO_QUOTA,
1109                                     timediff);
1110         } else {
1111                 do_gettimeofday(&work_end);
1112                 timediff = cfs_timeval_sub(&work_end, &work_start, NULL);
1113                 lprocfs_counter_add(qctxt->lqc_stats,
1114                                     isblk ? LQUOTA_NOWAIT_PENDING_BLK_QUOTA :
1115                                             LQUOTA_NOWAIT_PENDING_INO_QUOTA,
1116                                     timediff);
1117         }
1118
1119         RETURN(rc);
1120 }
1121
1122 int
1123 qctxt_init(struct obd_device *obd, dqacq_handler_t handler)
1124 {
1125         struct lustre_quota_ctxt *qctxt = &obd->u.obt.obt_qctxt;
1126         struct super_block *sb = obd->u.obt.obt_sb;
1127         int rc = 0;
1128         ENTRY;
1129
1130         LASSERT(qctxt);
1131
1132         rc = ptlrpcd_addref();
1133         if (rc)
1134                 RETURN(rc);
1135
1136         cfs_waitq_init(&qctxt->lqc_wait_for_qmaster);
1137         cfs_waitq_init(&qctxt->lqc_lqs_waitq);
1138         atomic_set(&qctxt->lqc_lqs, 0);
1139         spin_lock_init(&qctxt->lqc_lock);
1140         spin_lock(&qctxt->lqc_lock);
1141         qctxt->lqc_handler = handler;
1142         qctxt->lqc_sb = sb;
1143         qctxt->lqc_import = NULL;
1144         qctxt->lqc_recovery = 0;
1145         qctxt->lqc_switch_qs = 1; /* Change qunit size in default setting */
1146         qctxt->lqc_valid = 1;
1147         qctxt->lqc_cqs_boundary_factor = 4;
1148         qctxt->lqc_cqs_least_bunit = PTLRPC_MAX_BRW_SIZE;
1149         qctxt->lqc_cqs_least_iunit = 2;
1150         qctxt->lqc_cqs_qs_factor = 2;
1151         qctxt->lqc_flags = 0;
1152         QUOTA_MASTER_UNREADY(qctxt);
1153         qctxt->lqc_bunit_sz = default_bunit_sz;
1154         qctxt->lqc_btune_sz = default_bunit_sz / 100 * default_btune_ratio;
1155         qctxt->lqc_iunit_sz = default_iunit_sz;
1156         qctxt->lqc_itune_sz = default_iunit_sz * default_itune_ratio / 100;
1157         qctxt->lqc_switch_seconds = 300; /* enlarging will wait 5 minutes
1158                                           * after the last shrinking */
1159         qctxt->lqc_sync_blk = 0;
1160         spin_unlock(&qctxt->lqc_lock);
1161
1162         qctxt->lqc_lqs_hash = lustre_hash_init("LQS_HASH",
1163                                                HASH_LQS_CUR_BITS,
1164                                                HASH_LQS_MAX_BITS,
1165                                                &lqs_hash_ops, 0);
1166         if (!qctxt->lqc_lqs_hash)
1167                 CERROR("initialize hash lqs for %s error!\n", obd->obd_name);
1168
1169 #ifdef LPROCFS
1170         if (lquota_proc_setup(obd, is_master(obd, qctxt, 0, 0)))
1171                 CERROR("initialize proc for %s error!\n", obd->obd_name);
1172 #endif
1173
1174         RETURN(rc);
1175 }
1176
1177 static int check_lqs(struct lustre_quota_ctxt *qctxt)
1178 {
1179         int rc;
1180         ENTRY;
1181
1182         rc = !atomic_read(&qctxt->lqc_lqs);
1183
1184         RETURN(rc);
1185 }
1186
1187
1188 void hash_put_lqs(void *obj, void *data)
1189 {
1190         lqs_putref((struct lustre_qunit_size *)obj);
1191 }
1192
1193 void qctxt_cleanup(struct lustre_quota_ctxt *qctxt, int force)
1194 {
1195         struct lustre_qunit *qunit, *tmp;
1196         struct list_head tmp_list;
1197         struct l_wait_info lwi = { 0 };
1198         int i;
1199         ENTRY;
1200
1201         INIT_LIST_HEAD(&tmp_list);
1202
1203         spin_lock(&qctxt->lqc_lock);
1204         qctxt->lqc_valid = 0;
1205         spin_unlock(&qctxt->lqc_lock);
1206
1207         spin_lock(&qunit_hash_lock);
1208         for (i = 0; i < NR_DQHASH; i++) {
1209                 list_for_each_entry_safe(qunit, tmp, &qunit_hash[i], lq_hash) {
1210                         if (qunit->lq_ctxt != qctxt)
1211                                 continue;
1212                         remove_qunit_nolock(qunit);
1213                         list_add(&qunit->lq_hash, &tmp_list);
1214                 }
1215         }
1216         spin_unlock(&qunit_hash_lock);
1217
1218         list_for_each_entry_safe(qunit, tmp, &tmp_list, lq_hash) {
1219                 list_del_init(&qunit->lq_hash);
1220                 compute_lqs_after_removing_qunit(qunit);
1221
1222                 /* wake up all waiters */
1223                 QUNIT_SET_STATE_AND_RC(qunit, QUNIT_FINISHED, 0);
1224                 wake_up(&qunit->lq_waitq);
1225                 qunit_put(qunit);
1226         }
1227
1228         /* after qctxt_cleanup, qctxt might be freed, then check_qm() is
1229          * unpredicted. So we must wait until lqc_wait_for_qmaster is empty */
1230         while (cfs_waitq_active(&qctxt->lqc_wait_for_qmaster)) {
1231                 cfs_waitq_signal(&qctxt->lqc_wait_for_qmaster);
1232                 cfs_schedule_timeout(CFS_TASK_INTERRUPTIBLE,
1233                                      cfs_time_seconds(1));
1234         }
1235
1236         lustre_hash_for_each_safe(qctxt->lqc_lqs_hash, hash_put_lqs, NULL);
1237         l_wait_event(qctxt->lqc_lqs_waitq, check_lqs(qctxt), &lwi);
1238         lustre_hash_exit(qctxt->lqc_lqs_hash);
1239
1240         ptlrpcd_decref();
1241
1242 #ifdef LPROCFS
1243         if (lquota_proc_cleanup(qctxt))
1244                 CERROR("cleanup proc error!\n");
1245 #endif
1246
1247         EXIT;
1248 }
1249
1250 struct qslave_recov_thread_data {
1251         struct obd_device *obd;
1252         struct lustre_quota_ctxt *qctxt;
1253         struct completion comp;
1254 };
1255
1256 /* FIXME only recovery block quota by now */
1257 static int qslave_recovery_main(void *arg)
1258 {
1259         struct qslave_recov_thread_data *data = arg;
1260         struct obd_device *obd = data->obd;
1261         struct lustre_quota_ctxt *qctxt = data->qctxt;
1262         unsigned int type;
1263         int rc = 0;
1264         ENTRY;
1265
1266         ptlrpc_daemonize("qslave_recovd");
1267
1268         complete(&data->comp);
1269
1270         if (qctxt->lqc_recovery)
1271                 RETURN(0);
1272         qctxt->lqc_recovery = 1;
1273
1274         for (type = USRQUOTA; type < MAXQUOTAS; type++) {
1275                 struct qunit_data qdata;
1276                 struct quota_info *dqopt = sb_dqopt(qctxt->lqc_sb);
1277                 struct list_head id_list;
1278                 struct dquot_id *dqid, *tmp;
1279                 int ret;
1280
1281                 LOCK_DQONOFF_MUTEX(dqopt);
1282                 if (!ll_sb_has_quota_active(qctxt->lqc_sb, type)) {
1283                         UNLOCK_DQONOFF_MUTEX(dqopt);
1284                         break;
1285                 }
1286
1287                 LASSERT(dqopt->files[type] != NULL);
1288                 INIT_LIST_HEAD(&id_list);
1289 #ifndef KERNEL_SUPPORTS_QUOTA_READ
1290                 rc = fsfilt_qids(obd, dqopt->files[type], NULL, type, &id_list);
1291 #else
1292                 rc = fsfilt_qids(obd, NULL, dqopt->files[type], type, &id_list);
1293 #endif
1294                 UNLOCK_DQONOFF_MUTEX(dqopt);
1295                 if (rc)
1296                         CERROR("Get ids from quota file failed. (rc:%d)\n", rc);
1297
1298                 list_for_each_entry_safe(dqid, tmp, &id_list, di_link) {
1299                         list_del_init(&dqid->di_link);
1300                         /* skip slave recovery on itself */
1301                         if (is_master(obd, qctxt, dqid->di_id, type))
1302                                 goto free;
1303                         if (rc && rc != -EBUSY)
1304                                 goto free;
1305
1306                         qdata.qd_id = dqid->di_id;
1307                         qdata.qd_flags = type;
1308                         QDATA_SET_BLK(&qdata);
1309                         qdata.qd_count = 0;
1310
1311                         ret = check_cur_qunit(obd, qctxt, &qdata);
1312                         if (ret > 0) {
1313                                 int opc;
1314                                 opc = ret == 1 ? QUOTA_DQACQ : QUOTA_DQREL;
1315                                 rc = schedule_dqacq(obd, qctxt, &qdata, opc,
1316                                                     0, NULL);
1317                                 if (rc == -EDQUOT)
1318                                         rc = 0;
1319                         } else {
1320                                 rc = 0;
1321                         }
1322
1323                         if (rc)
1324                                 CDEBUG(rc == -EBUSY ? D_QUOTA : D_ERROR,
1325                                        "qslave recovery failed! (id:%d type:%d "
1326                                        " rc:%d)\n", dqid->di_id, type, rc);
1327 free:
1328                         OBD_FREE_PTR(dqid);
1329                 }
1330         }
1331
1332         qctxt->lqc_recovery = 0;
1333         RETURN(rc);
1334 }
1335
1336 void
1337 qslave_start_recovery(struct obd_device *obd, struct lustre_quota_ctxt *qctxt)
1338 {
1339         struct qslave_recov_thread_data data;
1340         int rc;
1341         ENTRY;
1342
1343         if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1344                 goto exit;
1345
1346         data.obd = obd;
1347         data.qctxt = qctxt;
1348         init_completion(&data.comp);
1349
1350         rc = kernel_thread(qslave_recovery_main, &data, CLONE_VM|CLONE_FILES);
1351         if (rc < 0) {
1352                 CERROR("Cannot start quota recovery thread: rc %d\n", rc);
1353                 goto exit;
1354         }
1355         wait_for_completion(&data.comp);
1356 exit:
1357         EXIT;
1358 }
1359
1360
1361 /*
1362  * lqs<->qctxt hash operations
1363  */
1364
1365 /* string hashing using djb2 hash algorithm */
1366 static unsigned
1367 lqs_hash(lustre_hash_t *lh, void *key, unsigned mask)
1368 {
1369         struct quota_adjust_qunit *lqs_key;
1370         unsigned hash;
1371         ENTRY;
1372
1373         LASSERT(key);
1374         lqs_key = (struct quota_adjust_qunit *)key;
1375         hash = (QAQ_IS_GRP(lqs_key) ? 5381 : 5387) * lqs_key->qaq_id;
1376
1377         RETURN(hash & mask);
1378 }
1379
1380 static int
1381 lqs_compare(void *key, struct hlist_node *hnode)
1382 {
1383         struct lustre_qunit_size *q;
1384         int rc;
1385         ENTRY;
1386
1387         LASSERT(key);
1388         q = hlist_entry(hnode, struct lustre_qunit_size, lqs_hash);
1389
1390         spin_lock(&q->lqs_lock);
1391         rc = (q->lqs_key == *((unsigned long long *)key));
1392         spin_unlock(&q->lqs_lock);
1393
1394         RETURN(rc);
1395 }
1396
1397 static void *
1398 lqs_get(struct hlist_node *hnode)
1399 {
1400         struct lustre_qunit_size *q =
1401             hlist_entry(hnode, struct lustre_qunit_size, lqs_hash);
1402         ENTRY;
1403
1404         if (atomic_inc_return(&q->lqs_refcount) == 2) /* quota_search_lqs */
1405                 atomic_inc(&q->lqs_ctxt->lqc_lqs);
1406         CDEBUG(D_QUOTA, "lqs=%p refcount %d\n",
1407                q, atomic_read(&q->lqs_refcount));
1408
1409         RETURN(q);
1410 }
1411
1412 static void *
1413 lqs_put(struct hlist_node *hnode)
1414 {
1415         struct lustre_qunit_size *q =
1416             hlist_entry(hnode, struct lustre_qunit_size, lqs_hash);
1417         ENTRY;
1418
1419         LASSERT(atomic_read(&q->lqs_refcount) > 0);
1420
1421         if (atomic_dec_return(&q->lqs_refcount) == 1)
1422                 if (atomic_dec_and_test(&q->lqs_ctxt->lqc_lqs))
1423                         cfs_waitq_signal(&q->lqs_ctxt->lqc_lqs_waitq);
1424
1425         CDEBUG(D_QUOTA, "lqs=%p refcount %d\n",
1426                q, atomic_read(&q->lqs_refcount));
1427
1428         RETURN(q);
1429 }
1430
1431 static void
1432 lqs_exit(struct hlist_node *hnode)
1433 {
1434         struct lustre_qunit_size *q;
1435         ENTRY;
1436
1437         q = hlist_entry(hnode, struct lustre_qunit_size, lqs_hash);
1438         /*
1439          * Nothing should be left. User of lqs put it and
1440          * lqs also was deleted from table by this time
1441          * so we should have 0 refs.
1442          */
1443         LASSERTF(atomic_read(&q->lqs_refcount) == 0,
1444                  "Busy lqs %p with %d refs\n", q,
1445                  atomic_read(&q->lqs_refcount));
1446         OBD_FREE_PTR(q);
1447         EXIT;
1448 }
1449
1450 static lustre_hash_ops_t lqs_hash_ops = {
1451         .lh_hash    = lqs_hash,
1452         .lh_compare = lqs_compare,
1453         .lh_get     = lqs_get,
1454         .lh_put     = lqs_put,
1455         .lh_exit    = lqs_exit
1456 };
1457 #endif /* HAVE_QUOTA_SUPPORT */