sys/vfs/hammer/hammer_object.c

   1 /*
   2  * Copyright (c) 2007-2008 The DragonFly Project.  All rights reserved.
   3  *
   4  * This code is derived from software contributed to The DragonFly Project
   5  * by Matthew Dillon <dillon@backplane.com>
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  *
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in
  15  *    the documentation and/or other materials provided with the
  16  *    distribution.
  17  * 3. Neither the name of The DragonFly Project nor the names of its
  18  *    contributors may be used to endorse or promote products derived
  19  *    from this software without specific, prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
  25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
  27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
  29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
  31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  32  * SUCH DAMAGE.
  33  *
  34  * $DragonFly: src/sys/vfs/hammer/hammer_object.c,v 1.97 2008/09/23 22:28:56 dillon Exp $
  35  */
  36
  37 #include "hammer.h"
  38
  39 static int hammer_mem_lookup(hammer_cursor_t cursor);
  40 static int hammer_mem_first(hammer_cursor_t cursor);
  41 static int hammer_frontend_trunc_callback(hammer_record_t record,
  42                                 void *data __unused);
  43 static int hammer_bulk_scan_callback(hammer_record_t record, void *data);
  44 static int hammer_record_needs_overwrite_delete(hammer_record_t record);
  45 static int hammer_delete_general(hammer_cursor_t cursor, hammer_inode_t ip,
  46                       hammer_btree_leaf_elm_t leaf);
  47
  48 struct rec_trunc_info {
  49         u_int16_t       rec_type;
  50         int64_t         trunc_off;
  51 };
  52
  53 struct hammer_bulk_info {
  54         hammer_record_t record;
  55         struct hammer_btree_leaf_elm leaf;
  56 };
  57
  58 /*
  59  * Red-black tree support.  Comparison code for insertion.
  60  */
  61 static int
  62 hammer_rec_rb_compare(hammer_record_t rec1, hammer_record_t rec2)
  63 {
  64         if (rec1->leaf.base.rec_type < rec2->leaf.base.rec_type)
  65                 return(-1);
  66         if (rec1->leaf.base.rec_type > rec2->leaf.base.rec_type)
  67                 return(1);
  68
  69         if (rec1->leaf.base.key < rec2->leaf.base.key)
  70                 return(-1);
  71         if (rec1->leaf.base.key > rec2->leaf.base.key)
  72                 return(1);
  73
  74         /*
  75          * Never match against an item deleted by the front-end.
  76          *
  77          * rec1 is greater then rec2 if rec1 is marked deleted.
  78          * rec1 is less then rec2 if rec2 is marked deleted.
  79          *
  80          * Multiple deleted records may be present, do not return 0
  81          * if both are marked deleted.
  82          */
  83         if (rec1->flags & HAMMER_RECF_DELETED_FE)
  84                 return(1);
  85         if (rec2->flags & HAMMER_RECF_DELETED_FE)
  86                 return(-1);
  87
  88         return(0);
  89 }
  90
  91 /*
  92  * Basic record comparison code similar to hammer_btree_cmp().
  93  */
  94 static int
  95 hammer_rec_cmp(hammer_base_elm_t elm, hammer_record_t rec)
  96 {
  97         if (elm->rec_type < rec->leaf.base.rec_type)
  98                 return(-3);
  99         if (elm->rec_type > rec->leaf.base.rec_type)
 100                 return(3);
 101
 102         if (elm->key < rec->leaf.base.key)
 103                 return(-2);
 104         if (elm->key > rec->leaf.base.key)
 105                 return(2);
 106
 107         /*
 108          * Never match against an item deleted by the front-end.
 109          * elm is less then rec if rec is marked deleted.
 110          */
 111         if (rec->flags & HAMMER_RECF_DELETED_FE)
 112                 return(-1);
 113         return(0);
 114 }
 115
 116 /*
 117  * Ranged scan to locate overlapping record(s).  This is used by
 118  * hammer_ip_get_bulk() to locate an overlapping record.  We have
 119  * to use a ranged scan because the keys for data records with the
 120  * same file base offset can be different due to differing data_len's.
 121  *
 122  * NOTE: The base file offset of a data record is (key - data_len), not (key).
 123  */
 124 static int
 125 hammer_rec_overlap_cmp(hammer_record_t rec, void *data)
 126 {
 127         struct hammer_bulk_info *info = data;
 128         hammer_btree_leaf_elm_t leaf = &info->leaf;
 129
 130         if (rec->leaf.base.rec_type < leaf->base.rec_type)
 131                 return(-3);
 132         if (rec->leaf.base.rec_type > leaf->base.rec_type)
 133                 return(3);
 134
 135         /*
 136          * Overlap compare
 137          */
 138         if (leaf->base.rec_type == HAMMER_RECTYPE_DATA) {
 139                 /* rec_beg >= leaf_end */
 140                 if (rec->leaf.base.key - rec->leaf.data_len >= leaf->base.key)
 141                         return(2);
 142                 /* rec_end <= leaf_beg */
 143                 if (rec->leaf.base.key <= leaf->base.key - leaf->data_len)
 144                         return(-2);
 145         } else {
 146                 if (rec->leaf.base.key < leaf->base.key)
 147                         return(-2);
 148                 if (rec->leaf.base.key > leaf->base.key)
 149                         return(2);
 150         }
 151
 152         /*
 153          * We have to return 0 at this point, even if DELETED_FE is set,
 154          * because returning anything else will cause the scan to ignore
 155          * one of the branches when we really want it to check both.
 156          */
 157         return(0);
 158 }
 159
 160 /*
 161  * RB_SCAN comparison code for hammer_mem_first().  The argument order
 162  * is reversed so the comparison result has to be negated.  key_beg and
 163  * key_end are both range-inclusive.
 164  *
 165  * Localized deletions are not cached in-memory.
 166  */
 167 static
 168 int
 169 hammer_rec_scan_cmp(hammer_record_t rec, void *data)
 170 {
 171         hammer_cursor_t cursor = data;
 172         int r;
 173
 174         r = hammer_rec_cmp(&cursor->key_beg, rec);
 175         if (r > 1)
 176                 return(-1);
 177         r = hammer_rec_cmp(&cursor->key_end, rec);
 178         if (r < -1)
 179                 return(1);
 180         return(0);
 181 }
 182
 183 /*
 184  * This compare function is used when simply looking up key_beg.
 185  */
 186 static
 187 int
 188 hammer_rec_find_cmp(hammer_record_t rec, void *data)
 189 {
 190         hammer_cursor_t cursor = data;
 191         int r;
 192
 193         r = hammer_rec_cmp(&cursor->key_beg, rec);
 194         if (r > 1)
 195                 return(-1);
 196         if (r < -1)
 197                 return(1);
 198         return(0);
 199 }
 200
 201 /*
 202  * Locate blocks within the truncation range.  Partial blocks do not count.
 203  */
 204 static
 205 int
 206 hammer_rec_trunc_cmp(hammer_record_t rec, void *data)
 207 {
 208         struct rec_trunc_info *info = data;
 209
 210         if (rec->leaf.base.rec_type < info->rec_type)
 211                 return(-1);
 212         if (rec->leaf.base.rec_type > info->rec_type)
 213                 return(1);
 214
 215         switch(rec->leaf.base.rec_type) {
 216         case HAMMER_RECTYPE_DB:
 217                 /*
 218                  * DB record key is not beyond the truncation point, retain.
 219                  */
 220                 if (rec->leaf.base.key < info->trunc_off)
 221                         return(-1);
 222                 break;
 223         case HAMMER_RECTYPE_DATA:
 224                 /*
 225                  * DATA record offset start is not beyond the truncation point,
 226                  * retain.
 227                  */
 228                 if (rec->leaf.base.key - rec->leaf.data_len < info->trunc_off)
 229                         return(-1);
 230                 break;
 231         default:
 232                 panic("hammer_rec_trunc_cmp: unexpected record type");
 233         }
 234
 235         /*
 236          * The record start is >= the truncation point, return match,
 237          * the record should be destroyed.
 238          */
 239         return(0);
 240 }
 241
 242 RB_GENERATE(hammer_rec_rb_tree, hammer_record, rb_node, hammer_rec_rb_compare);
 243
 244 /*
 245  * Allocate a record for the caller to finish filling in.  The record is
 246  * returned referenced.
 247  */
 248 hammer_record_t
 249 hammer_alloc_mem_record(hammer_inode_t ip, int data_len)
 250 {
 251         hammer_record_t record;
 252         hammer_mount_t hmp;
 253
 254         hmp = ip->hmp;
 255         ++hammer_count_records;
 256         record = kmalloc(sizeof(*record), hmp->m_misc,
 257                          M_WAITOK | M_ZERO | M_USE_RESERVE);
 258         record->flush_state = HAMMER_FST_IDLE;
 259         record->ip = ip;
 260         record->leaf.base.btype = HAMMER_BTREE_TYPE_RECORD;
 261         record->leaf.data_len = data_len;
 262         hammer_ref(&record->lock);
 263
 264         if (data_len) {
 265                 record->data = kmalloc(data_len, hmp->m_misc, M_WAITOK | M_ZERO);
 266                 record->flags |= HAMMER_RECF_ALLOCDATA;
 267                 ++hammer_count_record_datas;
 268         }
 269
 270         return (record);
 271 }
 272
 273 void
 274 hammer_wait_mem_record_ident(hammer_record_t record, const char *ident)
 275 {
 276         while (record->flush_state == HAMMER_FST_FLUSH) {
 277                 record->flags |= HAMMER_RECF_WANTED;
 278                 tsleep(record, 0, ident, 0);
 279         }
 280 }
 281
 282 /*
 283  * Called from the backend, hammer_inode.c, after a record has been
 284  * flushed to disk.  The record has been exclusively locked by the
 285  * caller and interlocked with BE.
 286  *
 287  * We clean up the state, unlock, and release the record (the record
 288  * was referenced by the fact that it was in the HAMMER_FST_FLUSH state).
 289  */
 290 void
 291 hammer_flush_record_done(hammer_record_t record, int error)
 292 {
 293         hammer_inode_t target_ip;
 294
 295         KKASSERT(record->flush_state == HAMMER_FST_FLUSH);
 296         KKASSERT(record->flags & HAMMER_RECF_INTERLOCK_BE);
 297
 298         if (error) {
 299                 /*
 300                  * An error occured, the backend was unable to sync the
 301                  * record to its media.  Leave the record intact.
 302                  */
 303                 hammer_critical_error(record->ip->hmp, record->ip, error,
 304                                       "while flushing record");
 305         }
 306
 307         --record->flush_group->refs;
 308         record->flush_group = NULL;
 309
 310         if (record->flags & HAMMER_RECF_DELETED_BE) {
 311                 if ((target_ip = record->target_ip) != NULL) {
 312                         TAILQ_REMOVE(&target_ip->target_list, record,
 313                                      target_entry);
 314                         record->target_ip = NULL;
 315                         hammer_test_inode(target_ip);
 316                 }
 317                 record->flush_state = HAMMER_FST_IDLE;
 318         } else {
 319                 if (record->target_ip) {
 320                         record->flush_state = HAMMER_FST_SETUP;
 321                         hammer_test_inode(record->ip);
 322                         hammer_test_inode(record->target_ip);
 323                 } else {
 324                         record->flush_state = HAMMER_FST_IDLE;
 325                 }
 326         }
 327         record->flags &= ~HAMMER_RECF_INTERLOCK_BE;
 328         if (record->flags & HAMMER_RECF_WANTED) {
 329                 record->flags &= ~HAMMER_RECF_WANTED;
 330                 wakeup(record);
 331         }
 332         hammer_rel_mem_record(record);
 333 }
 334
 335 /*
 336  * Release a memory record.  Records marked for deletion are immediately
 337  * removed from the RB-Tree but otherwise left intact until the last ref
 338  * goes away.
 339  */
 340 void
 341 hammer_rel_mem_record(struct hammer_record *record)
 342 {
 343         hammer_mount_t hmp;
 344         hammer_reserve_t resv;
 345         hammer_inode_t ip;
 346         hammer_inode_t target_ip;
 347
 348         hammer_unref(&record->lock);
 349
 350         if (record->lock.refs == 0) {
 351                 /*
 352                  * Upon release of the last reference wakeup any waiters.
 353                  * The record structure may get destroyed so callers will
 354                  * loop up and do a relookup.
 355                  *
 356                  * WARNING!  Record must be removed from RB-TREE before we
 357                  * might possibly block.  hammer_test_inode() can block!
 358                  */
 359                 ip = record->ip;
 360                 hmp = ip->hmp;
 361
 362                 /*
 363                  * Upon release of the last reference a record marked deleted
 364                  * is destroyed.
 365                  */
 366                 if (record->flags & HAMMER_RECF_DELETED_FE) {
 367                         KKASSERT(ip->lock.refs > 0);
 368                         KKASSERT(record->flush_state != HAMMER_FST_FLUSH);
 369
 370                         /*
 371                          * target_ip may have zero refs, we have to ref it
 372                          * to prevent it from being ripped out from under
 373                          * us.
 374                          */
 375                         if ((target_ip = record->target_ip) != NULL) {
 376                                 TAILQ_REMOVE(&target_ip->target_list,
 377                                              record, target_entry);
 378                                 record->target_ip = NULL;
 379                                 hammer_ref(&target_ip->lock);
 380                         }
 381
 382                         if (record->flags & HAMMER_RECF_ONRBTREE) {
 383                                 RB_REMOVE(hammer_rec_rb_tree,
 384                                           &record->ip->rec_tree,
 385                                           record);
 386                                 KKASSERT(ip->rsv_recs > 0);
 387                                 --hmp->rsv_recs;
 388                                 --ip->rsv_recs;
 389                                 hmp->rsv_databytes -= record->leaf.data_len;
 390                                 record->flags &= ~HAMMER_RECF_ONRBTREE;
 391
 392                                 if (RB_EMPTY(&record->ip->rec_tree)) {
 393                                         record->ip->flags &= ~HAMMER_INODE_XDIRTY;
 394                                         record->ip->sync_flags &= ~HAMMER_INODE_XDIRTY;
 395                                         hammer_test_inode(record->ip);
 396                                 }
 397                         }
 398
 399                         /*
 400                          * We must wait for any direct-IO to complete before
 401                          * we can destroy the record because the bio may
 402                          * have a reference to it.
 403                          */
 404                         if (record->flags &
 405                            (HAMMER_RECF_DIRECT_IO | HAMMER_RECF_DIRECT_INVAL)) {
 406                                 hammer_io_direct_wait(record);
 407                         }
 408
 409
 410                         /*
 411                          * Do this test after removing record from the B-Tree.
 412                          */
 413                         if (target_ip) {
 414                                 hammer_test_inode(target_ip);
 415                                 hammer_rel_inode(target_ip, 0);
 416                         }
 417
 418                         if (record->flags & HAMMER_RECF_ALLOCDATA) {
 419                                 --hammer_count_record_datas;
 420                                 kfree(record->data, hmp->m_misc);
 421                                 record->flags &= ~HAMMER_RECF_ALLOCDATA;
 422                         }
 423
 424                         /*
 425                          * Release the reservation.  If the record was not
 426                          * committed return the reservation before
 427                          * releasing it.
 428                          */
 429                         if ((resv = record->resv) != NULL) {
 430 #if 0
 431                                 if ((record->flags & HAMMER_RECF_COMMITTED) == 0) {
 432                                         hammer_blockmap_reserve_undo(
 433                                                 hmp, resv,
 434                                                 record->leaf.data_offset,
 435                                                 record->leaf.data_len);
 436                                 }
 437 #endif
 438                                 hammer_blockmap_reserve_complete(hmp, resv);
 439                                 record->resv = NULL;
 440                         }
 441                         record->data = NULL;
 442                         --hammer_count_records;
 443                         kfree(record, hmp->m_misc);
 444                 }
 445         }
 446 }
 447
 448 /*
 449  * Record visibility depends on whether the record is being accessed by
 450  * the backend or the frontend.
 451  *
 452  * Return non-zero if the record is visible, zero if it isn't or if it is
 453  * deleted.
 454  *
 455  * If HAMMER_CURSOR_DELETE_VISIBILITY is set we allow deleted memory
 456  * records to be returned.  This is so pending deletions are detected
 457  * when using an iterator to locate an unused hash key, or when we need
 458  * to locate historical records on-disk to destroy.
 459  */
 460 static __inline
 461 int
 462 hammer_ip_iterate_mem_good(hammer_cursor_t cursor, hammer_record_t record)
 463 {
 464         if (cursor->flags & HAMMER_CURSOR_DELETE_VISIBILITY)
 465                 return(1);
 466         if (cursor->flags & HAMMER_CURSOR_BACKEND) {
 467                 if (record->flags & HAMMER_RECF_DELETED_BE)
 468                         return(0);
 469         } else {
 470                 if (record->flags & HAMMER_RECF_DELETED_FE)
 471                         return(0);
 472         }
 473         return(1);
 474 }
 475
 476 /*
 477  * This callback is used as part of the RB_SCAN function for in-memory
 478  * records.  We terminate it (return -1) as soon as we get a match.
 479  *
 480  * This routine is used by frontend code.
 481  *
 482  * The primary compare code does not account for ASOF lookups.  This
 483  * code handles that case as well as a few others.
 484  */
 485 static
 486 int
 487 hammer_rec_scan_callback(hammer_record_t rec, void *data)
 488 {
 489         hammer_cursor_t cursor = data;
 490
 491         /*
 492          * We terminate on success, so this should be NULL on entry.
 493          */
 494         KKASSERT(cursor->iprec == NULL);
 495
 496         /*
 497          * Skip if the record was marked deleted.
 498          */
 499         if (hammer_ip_iterate_mem_good(cursor, rec) == 0)
 500                 return(0);
 501
 502         /*
 503          * Skip if not visible due to our as-of TID
 504          */
 505         if (cursor->flags & HAMMER_CURSOR_ASOF) {
 506                 if (cursor->asof < rec->leaf.base.create_tid)
 507                         return(0);
 508                 if (rec->leaf.base.delete_tid &&
 509                     cursor->asof >= rec->leaf.base.delete_tid) {
 510                         return(0);
 511                 }
 512         }
 513
 514         /*
 515          * ref the record.  The record is protected from backend B-Tree
 516          * interactions by virtue of the cursor's IP lock.
 517          */
 518         hammer_ref(&rec->lock);
 519
 520         /*
 521          * The record may have been deleted while we were blocked.
 522          */
 523         if (hammer_ip_iterate_mem_good(cursor, rec) == 0) {
 524                 hammer_rel_mem_record(rec);
 525                 return(0);
 526         }
 527
 528         /*
 529          * Set the matching record and stop the scan.
 530          */
 531         cursor->iprec = rec;
 532         return(-1);
 533 }
 534
 535
 536 /*
 537  * Lookup an in-memory record given the key specified in the cursor.  Works
 538  * just like hammer_btree_lookup() but operates on an inode's in-memory
 539  * record list.
 540  *
 541  * The lookup must fail if the record is marked for deferred deletion.
 542  */
 543 static
 544 int
 545 hammer_mem_lookup(hammer_cursor_t cursor)
 546 {
 547         int error;
 548
 549         KKASSERT(cursor->ip);
 550         if (cursor->iprec) {
 551                 hammer_rel_mem_record(cursor->iprec);
 552                 cursor->iprec = NULL;
 553         }
 554         hammer_rec_rb_tree_RB_SCAN(&cursor->ip->rec_tree, hammer_rec_find_cmp,
 555                                    hammer_rec_scan_callback, cursor);
 556
 557         if (cursor->iprec == NULL)
 558                 error = ENOENT;
 559         else
 560                 error = 0;
 561         return(error);
 562 }
 563
 564 /*
 565  * hammer_mem_first() - locate the first in-memory record matching the
 566  * cursor within the bounds of the key range.
 567  */
 568 static
 569 int
 570 hammer_mem_first(hammer_cursor_t cursor)
 571 {
 572         hammer_inode_t ip;
 573
 574         ip = cursor->ip;
 575         KKASSERT(ip != NULL);
 576
 577         if (cursor->iprec) {
 578                 hammer_rel_mem_record(cursor->iprec);
 579                 cursor->iprec = NULL;
 580         }
 581
 582         hammer_rec_rb_tree_RB_SCAN(&ip->rec_tree, hammer_rec_scan_cmp,
 583                                    hammer_rec_scan_callback, cursor);
 584
 585         /*
 586          * Adjust scan.node and keep it linked into the RB-tree so we can
 587          * hold the cursor through third party modifications of the RB-tree.
 588          */
 589         if (cursor->iprec)
 590                 return(0);
 591         return(ENOENT);
 592 }
 593
 594 /************************************************************************
 595  *                   HAMMER IN-MEMORY RECORD FUNCTIONS                  *
 596  ************************************************************************
 597  *
 598  * These functions manipulate in-memory records.  Such records typically
 599  * exist prior to being committed to disk or indexed via the on-disk B-Tree.
 600  */
 601
 602 /*
 603  * Add a directory entry (dip,ncp) which references inode (ip).
 604  *
 605  * Note that the low 32 bits of the namekey are set temporarily to create
 606  * a unique in-memory record, and may be modified a second time when the
 607  * record is synchronized to disk.  In particular, the low 32 bits cannot be
 608  * all 0's when synching to disk, which is not handled here.
 609  *
 610  * NOTE: bytes does not include any terminating \0 on name, and name might
 611  * not be terminated.
 612  */
 613 int
 614 hammer_ip_add_directory(struct hammer_transaction *trans,
 615                      struct hammer_inode *dip, const char *name, int bytes,
 616                      struct hammer_inode *ip)
 617 {
 618         struct hammer_cursor cursor;
 619         hammer_record_t record;
 620         int error;
 621         u_int32_t max_iterations;
 622
 623         record = hammer_alloc_mem_record(dip, HAMMER_ENTRY_SIZE(bytes));
 624
 625         record->type = HAMMER_MEM_RECORD_ADD;
 626         record->leaf.base.localization = dip->obj_localization +
 627                                          HAMMER_LOCALIZE_MISC;
 628         record->leaf.base.obj_id = dip->obj_id;
 629         record->leaf.base.key = hammer_directory_namekey(dip, name, bytes,
 630                                                          &max_iterations);
 631         record->leaf.base.rec_type = HAMMER_RECTYPE_DIRENTRY;
 632         record->leaf.base.obj_type = ip->ino_leaf.base.obj_type;
 633         record->data->entry.obj_id = ip->obj_id;
 634         record->data->entry.localization = ip->obj_localization;
 635         bcopy(name, record->data->entry.name, bytes);
 636
 637         ++ip->ino_data.nlinks;
 638         hammer_modify_inode(ip, HAMMER_INODE_DDIRTY);
 639
 640         /*
 641          * Find an unused namekey.  Both the in-memory record tree and
 642          * the B-Tree are checked.  We do not want historically deleted
 643          * names to create a collision as our iteration space may be limited,
 644          * and since create_tid wouldn't match anyway an ASOF search
 645          * must be used to locate collisions.
 646          *
 647          * delete-visibility is set so pending deletions do not give us
 648          * a false-negative on our ability to use an iterator.
 649          *
 650          * The iterator must not rollover the key.  Directory keys only
 651          * use the positive key space.
 652          */
 653         hammer_init_cursor(trans, &cursor, &dip->cache[1], dip);
 654         cursor.key_beg = record->leaf.base;
 655         cursor.flags |= HAMMER_CURSOR_ASOF;
 656         cursor.flags |= HAMMER_CURSOR_DELETE_VISIBILITY;
 657         cursor.asof = ip->obj_asof;
 658
 659         while (hammer_ip_lookup(&cursor) == 0) {
 660                 ++record->leaf.base.key;
 661                 KKASSERT(record->leaf.base.key > 0);
 662                 cursor.key_beg.key = record->leaf.base.key;
 663                 if (--max_iterations == 0) {
 664                         hammer_rel_mem_record(record);
 665                         error = ENOSPC;
 666                         goto failed;
 667                 }
 668         }
 669
 670         /*
 671          * The target inode and the directory entry are bound together.
 672          */
 673         record->target_ip = ip;
 674         record->flush_state = HAMMER_FST_SETUP;
 675         TAILQ_INSERT_TAIL(&ip->target_list, record, target_entry);
 676
 677         /*
 678          * The inode now has a dependancy and must be taken out of the idle
 679          * state.  An inode not in an idle state is given an extra reference.
 680          *
 681          * When transitioning to a SETUP state flag for an automatic reflush
 682          * when the dependancies are disposed of if someone is waiting on
 683          * the inode.
 684          */
 685         if (ip->flush_state == HAMMER_FST_IDLE) {
 686                 hammer_ref(&ip->lock);
 687                 ip->flush_state = HAMMER_FST_SETUP;
 688                 if (ip->flags & HAMMER_INODE_FLUSHW)
 689                         ip->flags |= HAMMER_INODE_REFLUSH;
 690         }
 691         error = hammer_mem_add(record);
 692         if (error == 0) {
 693                 dip->ino_data.mtime = trans->time;
 694                 hammer_modify_inode(dip, HAMMER_INODE_MTIME);
 695         }
 696 failed:
 697         hammer_done_cursor(&cursor);
 698         return(error);
 699 }
 700
 701 /*
 702  * Delete the directory entry and update the inode link count.  The
 703  * cursor must be seeked to the directory entry record being deleted.
 704  *
 705  * The related inode should be share-locked by the caller.  The caller is
 706  * on the frontend.  It could also be NULL indicating that the directory
 707  * entry being removed has no related inode.
 708  *
 709  * This function can return EDEADLK requiring the caller to terminate
 710  * the cursor, any locks, wait on the returned record, and retry.
 711  */
 712 int
 713 hammer_ip_del_directory(struct hammer_transaction *trans,
 714                      hammer_cursor_t cursor, struct hammer_inode *dip,
 715                      struct hammer_inode *ip)
 716 {
 717         hammer_record_t record;
 718         int error;
 719
 720         if (hammer_cursor_inmem(cursor)) {
 721                 /*
 722                  * In-memory (unsynchronized) records can simply be freed.
 723                  * Even though the HAMMER_RECF_DELETED_FE flag is ignored
 724                  * by the backend, we must still avoid races against the
 725                  * backend potentially syncing the record to the media.
 726                  *
 727                  * We cannot call hammer_ip_delete_record(), that routine may
 728                  * only be called from the backend.
 729                  */
 730                 record = cursor->iprec;
 731                 if (record->flags & HAMMER_RECF_INTERLOCK_BE) {
 732                         KKASSERT(cursor->deadlk_rec == NULL);
 733                         hammer_ref(&record->lock);
 734                         cursor->deadlk_rec = record;
 735                         error = EDEADLK;
 736                 } else {
 737                         KKASSERT(record->type == HAMMER_MEM_RECORD_ADD);
 738                         record->flags |= HAMMER_RECF_DELETED_FE;
 739                         error = 0;
 740                 }
 741         } else {
 742                 /*
 743                  * If the record is on-disk we have to queue the deletion by
 744                  * the record's key.  This also causes lookups to skip the
 745                  * record.
 746                  */
 747                 KKASSERT(dip->flags &
 748                          (HAMMER_INODE_ONDISK | HAMMER_INODE_DONDISK));
 749                 record = hammer_alloc_mem_record(dip, 0);
 750                 record->type = HAMMER_MEM_RECORD_DEL;
 751                 record->leaf.base = cursor->leaf->base;
 752
 753                 /*
 754                  * ip may be NULL, indicating the deletion of a directory
 755                  * entry which has no related inode.
 756                  */
 757                 record->target_ip = ip;
 758                 if (ip) {
 759                         record->flush_state = HAMMER_FST_SETUP;
 760                         TAILQ_INSERT_TAIL(&ip->target_list, record,
 761                                           target_entry);
 762                 } else {
 763                         record->flush_state = HAMMER_FST_IDLE;
 764                 }
 765
 766                 /*
 767                  * The inode now has a dependancy and must be taken out of
 768                  * the idle state.  An inode not in an idle state is given
 769                  * an extra reference.
 770                  *
 771                  * When transitioning to a SETUP state flag for an automatic
 772                  * reflush when the dependancies are disposed of if someone
 773                  * is waiting on the inode.
 774                  */
 775                 if (ip && ip->flush_state == HAMMER_FST_IDLE) {
 776                         hammer_ref(&ip->lock);
 777                         ip->flush_state = HAMMER_FST_SETUP;
 778                         if (ip->flags & HAMMER_INODE_FLUSHW)
 779                                 ip->flags |= HAMMER_INODE_REFLUSH;
 780                 }
 781
 782                 error = hammer_mem_add(record);
 783         }
 784
 785         /*
 786          * One less link.  The file may still be open in the OS even after
 787          * all links have gone away.
 788          *
 789          * We have to terminate the cursor before syncing the inode to
 790          * avoid deadlocking against ourselves.  XXX this may no longer
 791          * be true.
 792          *
 793          * If nlinks drops to zero and the vnode is inactive (or there is
 794          * no vnode), call hammer_inode_unloadable_check() to zonk the
 795          * inode.  If we don't do this here the inode will not be destroyed
 796          * on-media until we unmount.
 797          */
 798         if (error == 0) {
 799                 if (ip)
 800                         --ip->ino_data.nlinks;  /* do before we might block */
 801                 dip->ino_data.mtime = trans->time;
 802                 hammer_modify_inode(dip, HAMMER_INODE_MTIME);
 803                 if (ip) {
 804                         hammer_modify_inode(ip, HAMMER_INODE_DDIRTY);
 805                         if (ip->ino_data.nlinks == 0 &&
 806                             (ip->vp == NULL || (ip->vp->v_flag & VINACTIVE))) {
 807                                 hammer_done_cursor(cursor);
 808                                 hammer_inode_unloadable_check(ip, 1);
 809                                 hammer_flush_inode(ip, 0);
 810                         }
 811                 }
 812
 813         }
 814         return(error);
 815 }
 816
 817 /*
 818  * Add a record to an inode.
 819  *
 820  * The caller must allocate the record with hammer_alloc_mem_record(ip) and
 821  * initialize the following additional fields:
 822  *
 823  * The related inode should be share-locked by the caller.  The caller is
 824  * on the frontend.
 825  *
 826  * record->rec.entry.base.base.key
 827  * record->rec.entry.base.base.rec_type
 828  * record->rec.entry.base.base.data_len
 829  * record->data         (a copy will be kmalloc'd if it cannot be embedded)
 830  */
 831 int
 832 hammer_ip_add_record(struct hammer_transaction *trans, hammer_record_t record)
 833 {
 834         hammer_inode_t ip = record->ip;
 835         int error;
 836
 837         KKASSERT(record->leaf.base.localization != 0);
 838         record->leaf.base.obj_id = ip->obj_id;
 839         record->leaf.base.obj_type = ip->ino_leaf.base.obj_type;
 840         error = hammer_mem_add(record);
 841         return(error);
 842 }
 843
 844 /*
 845  * Locate a bulk record in-memory.  Bulk records allow disk space to be
 846  * reserved so the front-end can flush large data writes without having
 847  * to queue the BIO to the flusher.  Only the related record gets queued
 848  * to the flusher.
 849  */
 850
 851 static hammer_record_t
 852 hammer_ip_get_bulk(hammer_inode_t ip, off_t file_offset, int bytes)
 853 {
 854         struct hammer_bulk_info info;
 855
 856         bzero(&info, sizeof(info));
 857         info.leaf.base.obj_id = ip->obj_id;
 858         info.leaf.base.key = file_offset + bytes;
 859         info.leaf.base.create_tid = 0;
 860         info.leaf.base.delete_tid = 0;
 861         info.leaf.base.rec_type = HAMMER_RECTYPE_DATA;
 862         info.leaf.base.obj_type = 0;                            /* unused */
 863         info.leaf.base.btype = HAMMER_BTREE_TYPE_RECORD;        /* unused */
 864         info.leaf.base.localization = ip->obj_localization +    /* unused */
 865                                       HAMMER_LOCALIZE_MISC;
 866         info.leaf.data_len = bytes;
 867
 868         hammer_rec_rb_tree_RB_SCAN(&ip->rec_tree, hammer_rec_overlap_cmp,
 869                                    hammer_bulk_scan_callback, &info);
 870
 871         return(info.record);    /* may be NULL */
 872 }
 873
 874 /*
 875  * Take records vetted by overlap_cmp.  The first non-deleted record
 876  * (if any) stops the scan.
 877  */
 878 static int
 879 hammer_bulk_scan_callback(hammer_record_t record, void *data)
 880 {
 881         struct hammer_bulk_info *info = data;
 882
 883         if (record->flags & HAMMER_RECF_DELETED_FE)
 884                 return(0);
 885         hammer_ref(&record->lock);
 886         info->record = record;
 887         return(-1);                     /* stop scan */
 888 }
 889
 890 /*
 891  * Reserve blockmap space placemarked with an in-memory record.
 892  *
 893  * This routine is called by the frontend in order to be able to directly
 894  * flush a buffer cache buffer.  The frontend has locked the related buffer
 895  * cache buffers and we should be able to manipulate any overlapping
 896  * in-memory records.
 897  *
 898  * The caller is responsible for adding the returned record.
 899  */
 900 hammer_record_t
 901 hammer_ip_add_bulk(hammer_inode_t ip, off_t file_offset, void *data, int bytes,
 902                    int *errorp)
 903 {
 904         hammer_record_t record;
 905         hammer_record_t conflict;
 906         int zone;
 907
 908         /*
 909          * Deal with conflicting in-memory records.  We cannot have multiple
 910          * in-memory records for the same base offset without seriously
 911          * confusing the backend, including but not limited to the backend
 912          * issuing delete-create-delete or create-delete-create sequences
 913          * and asserting on the delete_tid being the same as the create_tid.
 914          *
 915          * If we encounter a record with the backend interlock set we cannot
 916          * immediately delete it without confusing the backend.
 917          */
 918         while ((conflict = hammer_ip_get_bulk(ip, file_offset, bytes)) !=NULL) {
 919                 if (conflict->flags & HAMMER_RECF_INTERLOCK_BE) {
 920                         conflict->flags |= HAMMER_RECF_WANTED;
 921                         tsleep(conflict, 0, "hmrrc3", 0);
 922                 } else {
 923                         conflict->flags |= HAMMER_RECF_DELETED_FE;
 924                 }
 925                 hammer_rel_mem_record(conflict);
 926         }
 927
 928         /*
 929          * Create a record to cover the direct write.  This is called with
 930          * the related BIO locked so there should be no possible conflict.
 931          *
 932          * The backend is responsible for finalizing the space reserved in
 933          * this record.
 934          *
 935          * XXX bytes not aligned, depend on the reservation code to
 936          * align the reservation.
 937          */
 938         record = hammer_alloc_mem_record(ip, 0);
 939         zone = (bytes >= HAMMER_BUFSIZE) ? HAMMER_ZONE_LARGE_DATA_INDEX :
 940                                            HAMMER_ZONE_SMALL_DATA_INDEX;
 941         record->resv = hammer_blockmap_reserve(ip->hmp, zone, bytes,
 942                                                &record->leaf.data_offset,
 943                                                errorp);
 944         if (record->resv == NULL) {
 945                 kprintf("hammer_ip_add_bulk: reservation failed\n");
 946                 hammer_rel_mem_record(record);
 947                 return(NULL);
 948         }
 949         record->type = HAMMER_MEM_RECORD_DATA;
 950         record->leaf.base.rec_type = HAMMER_RECTYPE_DATA;
 951         record->leaf.base.obj_type = ip->ino_leaf.base.obj_type;
 952         record->leaf.base.obj_id = ip->obj_id;
 953         record->leaf.base.key = file_offset + bytes;
 954         record->leaf.base.localization = ip->obj_localization +
 955                                          HAMMER_LOCALIZE_MISC;
 956         record->leaf.data_len = bytes;
 957         hammer_crc_set_leaf(data, &record->leaf);
 958         KKASSERT(*errorp == 0);
 959         return(record);
 960 }
 961
 962 /*
 963  * Frontend truncation code.  Scan in-memory records only.  On-disk records
 964  * and records in a flushing state are handled by the backend.  The vnops
 965  * setattr code will handle the block containing the truncation point.
 966  *
 967  * Partial blocks are not deleted.
 968  */
 969 int
 970 hammer_ip_frontend_trunc(struct hammer_inode *ip, off_t file_size)
 971 {
 972         struct rec_trunc_info info;
 973
 974         switch(ip->ino_data.obj_type) {
 975         case HAMMER_OBJTYPE_REGFILE:
 976                 info.rec_type = HAMMER_RECTYPE_DATA;
 977                 break;
 978         case HAMMER_OBJTYPE_DBFILE:
 979                 info.rec_type = HAMMER_RECTYPE_DB;
 980                 break;
 981         default:
 982                 return(EINVAL);
 983         }
 984         info.trunc_off = file_size;
 985         hammer_rec_rb_tree_RB_SCAN(&ip->rec_tree, hammer_rec_trunc_cmp,
 986                                    hammer_frontend_trunc_callback, &info);
 987         return(0);
 988 }
 989
 990 static int
 991 hammer_frontend_trunc_callback(hammer_record_t record, void *data __unused)
 992 {
 993         if (record->flags & HAMMER_RECF_DELETED_FE)
 994                 return(0);
 995         if (record->flush_state == HAMMER_FST_FLUSH)
 996                 return(0);
 997         KKASSERT((record->flags & HAMMER_RECF_INTERLOCK_BE) == 0);
 998         hammer_ref(&record->lock);
 999         record->flags |= HAMMER_RECF_DELETED_FE;
1000         hammer_rel_mem_record(record);
1001         return(0);
1002 }
1003
1004 /*
1005  * Return 1 if the caller must check for and delete existing records
1006  * before writing out a new data record.
1007  *
1008  * Return 0 if the caller can just insert the record into the B-Tree without
1009  * checking.
1010  */
1011 static int
1012 hammer_record_needs_overwrite_delete(hammer_record_t record)
1013 {
1014         hammer_inode_t ip = record->ip;
1015         int64_t file_offset;
1016         int r;
1017
1018         if (ip->ino_data.obj_type == HAMMER_OBJTYPE_DBFILE)
1019                 file_offset = record->leaf.base.key;
1020         else
1021                 file_offset = record->leaf.base.key - record->leaf.data_len;
1022         r = (file_offset < ip->save_trunc_off);
1023         if (ip->ino_data.obj_type == HAMMER_OBJTYPE_DBFILE) {
1024                 if (ip->save_trunc_off <= record->leaf.base.key)
1025                         ip->save_trunc_off = record->leaf.base.key + 1;
1026         } else {
1027                 if (ip->save_trunc_off < record->leaf.base.key)
1028                         ip->save_trunc_off = record->leaf.base.key;
1029         }
1030         return(r);
1031 }
1032
1033 /*
1034  * Backend code.  Sync a record to the media.
1035  */
1036 int
1037 hammer_ip_sync_record_cursor(hammer_cursor_t cursor, hammer_record_t record)
1038 {
1039         hammer_transaction_t trans = cursor->trans;
1040         int64_t file_offset;
1041         int bytes;
1042         void *bdata;
1043         int error;
1044         int doprop;
1045
1046         KKASSERT(record->flush_state == HAMMER_FST_FLUSH);
1047         KKASSERT(record->flags & HAMMER_RECF_INTERLOCK_BE);
1048         KKASSERT(record->leaf.base.localization != 0);
1049
1050         /*
1051          * Any direct-write related to the record must complete before we
1052          * can sync the record to the on-disk media.
1053          */
1054         if (record->flags & (HAMMER_RECF_DIRECT_IO | HAMMER_RECF_DIRECT_INVAL))
1055                 hammer_io_direct_wait(record);
1056
1057         /*
1058          * If this is a bulk-data record placemarker there may be an existing
1059          * record on-disk, indicating a data overwrite.  If there is the
1060          * on-disk record must be deleted before we can insert our new record.
1061          *
1062          * We've synthesized this record and do not know what the create_tid
1063          * on-disk is, nor how much data it represents.
1064          *
1065          * Keep in mind that (key) for data records is (base_offset + len),
1066          * not (base_offset).  Also, we only want to get rid of on-disk
1067          * records since we are trying to sync our in-memory record, call
1068          * hammer_ip_delete_range() with truncating set to 1 to make sure
1069          * it skips in-memory records.
1070          *
1071          * It is ok for the lookup to return ENOENT.
1072          *
1073          * NOTE OPTIMIZATION: sync_trunc_off is used to determine if we have
1074          * to call hammer_ip_delete_range() or not.  This also means we must
1075          * update sync_trunc_off() as we write.
1076          */
1077         if (record->type == HAMMER_MEM_RECORD_DATA &&
1078             hammer_record_needs_overwrite_delete(record)) {
1079                 file_offset = record->leaf.base.key - record->leaf.data_len;
1080                 bytes = (record->leaf.data_len + HAMMER_BUFMASK) &
1081                         ~HAMMER_BUFMASK;
1082                 KKASSERT((file_offset & HAMMER_BUFMASK) == 0);
1083                 error = hammer_ip_delete_range(
1084                                 cursor, record->ip,
1085                                 file_offset, file_offset + bytes - 1,
1086                                 1);
1087                 if (error && error != ENOENT)
1088                         goto done;
1089         }
1090
1091         /*
1092          * If this is a general record there may be an on-disk version
1093          * that must be deleted before we can insert the new record.
1094          */
1095         if (record->type == HAMMER_MEM_RECORD_GENERAL) {
1096                 error = hammer_delete_general(cursor, record->ip,
1097                                               &record->leaf);
1098                 if (error && error != ENOENT)
1099                         goto done;
1100         }
1101
1102         /*
1103          * Setup the cursor.
1104          */
1105         hammer_normalize_cursor(cursor);
1106         cursor->key_beg = record->leaf.base;
1107         cursor->flags &= ~HAMMER_CURSOR_INITMASK;
1108         cursor->flags |= HAMMER_CURSOR_BACKEND;
1109         cursor->flags &= ~HAMMER_CURSOR_INSERT;
1110
1111         /*
1112          * Records can wind up on-media before the inode itself is on-media.
1113          * Flag the case.
1114          */
1115         record->ip->flags |= HAMMER_INODE_DONDISK;
1116
1117         /*
1118          * If we are deleting a directory entry an exact match must be
1119          * found on-disk.
1120          */
1121         if (record->type == HAMMER_MEM_RECORD_DEL) {
1122                 error = hammer_btree_lookup(cursor);
1123                 if (error == 0) {
1124                         KKASSERT(cursor->iprec == NULL);
1125                         error = hammer_ip_delete_record(cursor, record->ip,
1126                                                         trans->tid);
1127                         if (error == 0) {
1128                                 record->flags |= HAMMER_RECF_DELETED_FE;
1129                                 record->flags |= HAMMER_RECF_DELETED_BE;
1130                                 record->flags |= HAMMER_RECF_COMMITTED;
1131                         }
1132                 }
1133                 goto done;
1134         }
1135
1136         /*
1137          * We are inserting.
1138          *
1139          * Issue a lookup to position the cursor and locate the cluster.  The
1140          * target key should not exist.  If we are creating a directory entry
1141          * we may have to iterate the low 32 bits of the key to find an unused
1142          * key.
1143          */
1144         hammer_sync_lock_sh(trans);
1145         cursor->flags |= HAMMER_CURSOR_INSERT;
1146         error = hammer_btree_lookup(cursor);
1147         if (hammer_debug_inode)
1148                 kprintf("DOINSERT LOOKUP %d\n", error);
1149         if (error == 0) {
1150                 kprintf("hammer_ip_sync_record: duplicate rec "
1151                         "at (%016llx)\n", record->leaf.base.key);
1152                 Debugger("duplicate record1");
1153                 error = EIO;
1154         }
1155 #if 0
1156         if (record->type == HAMMER_MEM_RECORD_DATA)
1157                 kprintf("sync_record  %016llx ---------------- %016llx %d\n",
1158                         record->leaf.base.key - record->leaf.data_len,
1159                         record->leaf.data_offset, error);
1160 #endif
1161
1162         if (error != ENOENT)
1163                 goto done_unlock;
1164
1165         /*
1166          * Allocate the record and data.  The result buffers will be
1167          * marked as being modified and further calls to
1168          * hammer_modify_buffer() will result in unneeded UNDO records.
1169          *
1170          * Support zero-fill records (data == NULL and data_len != 0)
1171          */
1172         if (record->type == HAMMER_MEM_RECORD_DATA) {
1173                 /*
1174                  * The data portion of a bulk-data record has already been
1175                  * committed to disk, we need only adjust the layer2
1176                  * statistics in the same transaction as our B-Tree insert.
1177                  */
1178                 KKASSERT(record->leaf.data_offset != 0);
1179                 error = hammer_blockmap_finalize(trans,
1180                                                  record->resv,
1181                                                  record->leaf.data_offset,
1182                                                  record->leaf.data_len);
1183         } else if (record->data && record->leaf.data_len) {
1184                 /*
1185                  * Wholely cached record, with data.  Allocate the data.
1186                  */
1187                 bdata = hammer_alloc_data(trans, record->leaf.data_len,
1188                                           record->leaf.base.rec_type,
1189                                           &record->leaf.data_offset,
1190                                           &cursor->data_buffer, &error);
1191                 if (bdata == NULL)
1192                         goto done_unlock;
1193                 hammer_crc_set_leaf(record->data, &record->leaf);
1194                 hammer_modify_buffer(trans, cursor->data_buffer, NULL, 0);
1195                 bcopy(record->data, bdata, record->leaf.data_len);
1196                 hammer_modify_buffer_done(cursor->data_buffer);
1197         } else {
1198                 /*
1199                  * Wholely cached record, without data.
1200                  */
1201                 record->leaf.data_offset = 0;
1202                 record->leaf.data_crc = 0;
1203         }
1204
1205         error = hammer_btree_insert(cursor, &record->leaf, &doprop);
1206         if (hammer_debug_inode && error)
1207                 kprintf("BTREE INSERT error %d @ %016llx:%d key %016llx\n", error, cursor->node->node_offset, cursor->index, record->leaf.base.key);
1208
1209         /*
1210          * Our record is on-disk, normally mark the in-memory version as
1211          * deleted.  If the record represented a directory deletion but
1212          * we had to sync a valid directory entry to disk we must convert
1213          * the record to a covering delete so the frontend does not have
1214          * visibility on the synced entry.
1215          */
1216         if (error == 0) {
1217                 if (doprop) {
1218                         hammer_btree_do_propagation(cursor,
1219                                                     record->ip->pfsm,
1220                                                     &record->leaf);
1221                 }
1222                 if (record->flags & HAMMER_RECF_CONVERT_DELETE) {
1223                         KKASSERT(record->type == HAMMER_MEM_RECORD_ADD);
1224                         record->flags &= ~HAMMER_RECF_DELETED_FE;
1225                         record->type = HAMMER_MEM_RECORD_DEL;
1226                         KKASSERT(record->flush_state == HAMMER_FST_FLUSH);
1227                         record->flags &= ~HAMMER_RECF_CONVERT_DELETE;
1228                         /* hammer_flush_record_done takes care of the rest */
1229                 } else {
1230                         record->flags |= HAMMER_RECF_DELETED_FE;
1231                         record->flags |= HAMMER_RECF_DELETED_BE;
1232                 }
1233                 record->flags |= HAMMER_RECF_COMMITTED;
1234         } else {
1235                 if (record->leaf.data_offset) {
1236                         hammer_blockmap_free(trans, record->leaf.data_offset,
1237                                              record->leaf.data_len);
1238                 }
1239         }
1240 done_unlock:
1241         hammer_sync_unlock(trans);
1242 done:
1243         return(error);
1244 }
1245
1246 /*
1247  * Add the record to the inode's rec_tree.  The low 32 bits of a directory
1248  * entry's key is used to deal with hash collisions in the upper 32 bits.
1249  * A unique 64 bit key is generated in-memory and may be regenerated a
1250  * second time when the directory record is flushed to the on-disk B-Tree.
1251  *
1252  * A referenced record is passed to this function.  This function
1253  * eats the reference.  If an error occurs the record will be deleted.
1254  *
1255  * A copy of the temporary record->data pointer provided by the caller
1256  * will be made.
1257  */
1258 int
1259 hammer_mem_add(hammer_record_t record)
1260 {
1261         hammer_mount_t hmp = record->ip->hmp;
1262
1263         /*
1264          * Make a private copy of record->data
1265          */
1266         if (record->data)
1267                 KKASSERT(record->flags & HAMMER_RECF_ALLOCDATA);
1268
1269         /*
1270          * Insert into the RB tree.  A unique key should have already
1271          * been selected if this is a directory entry.
1272          */
1273         if (RB_INSERT(hammer_rec_rb_tree, &record->ip->rec_tree, record)) {
1274                 record->flags |= HAMMER_RECF_DELETED_FE;
1275                 hammer_rel_mem_record(record);
1276                 return (EEXIST);
1277         }
1278         ++hmp->count_newrecords;
1279         ++hmp->rsv_recs;
1280         ++record->ip->rsv_recs;
1281         record->ip->hmp->rsv_databytes += record->leaf.data_len;
1282         record->flags |= HAMMER_RECF_ONRBTREE;
1283         hammer_modify_inode(record->ip, HAMMER_INODE_XDIRTY);
1284         hammer_rel_mem_record(record);
1285         return(0);
1286 }
1287
1288 /************************************************************************
1289  *                   HAMMER INODE MERGED-RECORD FUNCTIONS               *
1290  ************************************************************************
1291  *
1292  * These functions augment the B-Tree scanning functions in hammer_btree.c
1293  * by merging in-memory records with on-disk records.
1294  */
1295
1296 /*
1297  * Locate a particular record either in-memory or on-disk.
1298  *
1299  * NOTE: This is basically a standalone routine, hammer_ip_next() may
1300  * NOT be called to iterate results.
1301  */
1302 int
1303 hammer_ip_lookup(hammer_cursor_t cursor)
1304 {
1305         int error;
1306
1307         /*
1308          * If the element is in-memory return it without searching the
1309          * on-disk B-Tree
1310          */
1311         KKASSERT(cursor->ip);
1312         error = hammer_mem_lookup(cursor);
1313         if (error == 0) {
1314                 cursor->leaf = &cursor->iprec->leaf;
1315                 return(error);
1316         }
1317         if (error != ENOENT)
1318                 return(error);
1319
1320         /*
1321          * If the inode has on-disk components search the on-disk B-Tree.
1322          */
1323         if ((cursor->ip->flags & (HAMMER_INODE_ONDISK|HAMMER_INODE_DONDISK)) == 0)
1324                 return(error);
1325         error = hammer_btree_lookup(cursor);
1326         if (error == 0)
1327                 error = hammer_btree_extract(cursor, HAMMER_CURSOR_GET_LEAF);
1328         return(error);
1329 }
1330
1331 /*
1332  * Locate the first record within the cursor's key_beg/key_end range,
1333  * restricted to a particular inode.  0 is returned on success, ENOENT
1334  * if no records matched the requested range, or some other error.
1335  *
1336  * When 0 is returned hammer_ip_next() may be used to iterate additional
1337  * records within the requested range.
1338  *
1339  * This function can return EDEADLK, requiring the caller to terminate
1340  * the cursor and try again.
1341  */
1342 int
1343 hammer_ip_first(hammer_cursor_t cursor)
1344 {
1345         hammer_inode_t ip = cursor->ip;
1346         int error;
1347
1348         KKASSERT(ip != NULL);
1349
1350         /*
1351          * Clean up fields and setup for merged scan
1352          */
1353         cursor->flags &= ~HAMMER_CURSOR_RETEST;
1354         cursor->flags |= HAMMER_CURSOR_ATEDISK | HAMMER_CURSOR_ATEMEM;
1355         cursor->flags |= HAMMER_CURSOR_DISKEOF | HAMMER_CURSOR_MEMEOF;
1356         if (cursor->iprec) {
1357                 hammer_rel_mem_record(cursor->iprec);
1358                 cursor->iprec = NULL;
1359         }
1360
1361         /*
1362          * Search the on-disk B-Tree.  hammer_btree_lookup() only does an
1363          * exact lookup so if we get ENOENT we have to call the iterate
1364          * function to validate the first record after the begin key.
1365          *
1366          * The ATEDISK flag is used by hammer_btree_iterate to determine
1367          * whether it must index forwards or not.  It is also used here
1368          * to select the next record from in-memory or on-disk.
1369          *
1370          * EDEADLK can only occur if the lookup hit an empty internal
1371          * element and couldn't delete it.  Since this could only occur
1372          * in-range, we can just iterate from the failure point.
1373          */
1374         if (ip->flags & (HAMMER_INODE_ONDISK|HAMMER_INODE_DONDISK)) {
1375                 error = hammer_btree_lookup(cursor);
1376                 if (error == ENOENT || error == EDEADLK) {
1377                         cursor->flags &= ~HAMMER_CURSOR_ATEDISK;
1378                         if (hammer_debug_general & 0x2000)
1379                                 kprintf("error %d node %p %016llx index %d\n", error, cursor->node, cursor->node->node_offset, cursor->index);
1380                         error = hammer_btree_iterate(cursor);
1381                 }
1382                 if (error && error != ENOENT)
1383                         return(error);
1384                 if (error == 0) {
1385                         cursor->flags &= ~HAMMER_CURSOR_DISKEOF;
1386                         cursor->flags &= ~HAMMER_CURSOR_ATEDISK;
1387                 } else {
1388                         cursor->flags |= HAMMER_CURSOR_ATEDISK;
1389                 }
1390         }
1391
1392         /*
1393          * Search the in-memory record list (Red-Black tree).  Unlike the
1394          * B-Tree search, mem_first checks for records in the range.
1395          */
1396         error = hammer_mem_first(cursor);
1397         if (error && error != ENOENT)
1398                 return(error);
1399         if (error == 0) {
1400                 cursor->flags &= ~HAMMER_CURSOR_MEMEOF;
1401                 cursor->flags &= ~HAMMER_CURSOR_ATEMEM;
1402                 if (hammer_ip_iterate_mem_good(cursor, cursor->iprec) == 0)
1403                         cursor->flags |= HAMMER_CURSOR_ATEMEM;
1404         }
1405
1406         /*
1407          * This will return the first matching record.
1408          */
1409         return(hammer_ip_next(cursor));
1410 }
1411
1412 /*
1413  * Retrieve the next record in a merged iteration within the bounds of the
1414  * cursor.  This call may be made multiple times after the cursor has been
1415  * initially searched with hammer_ip_first().
1416  *
1417  * 0 is returned on success, ENOENT if no further records match the
1418  * requested range, or some other error code is returned.
1419  */
1420 int
1421 hammer_ip_next(hammer_cursor_t cursor)
1422 {
1423         hammer_btree_elm_t elm;
1424         hammer_record_t rec, save;
1425         int error;
1426         int r;
1427
1428 next_btree:
1429         /*
1430          * Load the current on-disk and in-memory record.  If we ate any
1431          * records we have to get the next one.
1432          *
1433          * If we deleted the last on-disk record we had scanned ATEDISK will
1434          * be clear and RETEST will be set, forcing a call to iterate.  The
1435          * fact that ATEDISK is clear causes iterate to re-test the 'current'
1436          * element.  If ATEDISK is set, iterate will skip the 'current'
1437          * element.
1438          *
1439          * Get the next on-disk record
1440          */
1441         if (cursor->flags & (HAMMER_CURSOR_ATEDISK|HAMMER_CURSOR_RETEST)) {
1442                 if ((cursor->flags & HAMMER_CURSOR_DISKEOF) == 0) {
1443                         error = hammer_btree_iterate(cursor);
1444                         cursor->flags &= ~HAMMER_CURSOR_RETEST;
1445                         if (error == 0) {
1446                                 cursor->flags &= ~HAMMER_CURSOR_ATEDISK;
1447                                 hammer_cache_node(&cursor->ip->cache[1],
1448                                                   cursor->node);
1449                         } else {
1450                                 cursor->flags |= HAMMER_CURSOR_DISKEOF |
1451                                                  HAMMER_CURSOR_ATEDISK;
1452                         }
1453                 }
1454         }
1455
1456 next_memory:
1457         /*
1458          * Get the next in-memory record.
1459          *
1460          * hammer_rec_scan_cmp:  Is the record still in our general range,
1461          *                       (non-inclusive of snapshot exclusions)?
1462          * hammer_rec_scan_callback: Is the record in our snapshot?
1463          */
1464         if (cursor->flags & HAMMER_CURSOR_ATEMEM) {
1465                 if ((cursor->flags & HAMMER_CURSOR_MEMEOF) == 0) {
1466                         save = cursor->iprec;
1467                         cursor->iprec = NULL;
1468                         rec = save ? hammer_rec_rb_tree_RB_NEXT(save) : NULL;
1469                         while (rec) {
1470                                 if (hammer_rec_scan_cmp(rec, cursor) != 0)
1471                                         break;
1472                                 if (hammer_rec_scan_callback(rec, cursor) != 0)
1473                                         break;
1474                                 rec = hammer_rec_rb_tree_RB_NEXT(rec);
1475                         }
1476                         if (save)
1477                                 hammer_rel_mem_record(save);
1478                         if (cursor->iprec) {
1479                                 KKASSERT(cursor->iprec == rec);
1480                                 cursor->flags &= ~HAMMER_CURSOR_ATEMEM;
1481                         } else {
1482                                 cursor->flags |= HAMMER_CURSOR_MEMEOF;
1483                         }
1484                 }
1485         }
1486
1487         /*
1488          * The memory record may have become stale while being held in
1489          * cursor->iprec.  We are interlocked against the backend on
1490          * with regards to B-Tree entries.
1491          */
1492         if ((cursor->flags & HAMMER_CURSOR_ATEMEM) == 0) {
1493                 if (hammer_ip_iterate_mem_good(cursor, cursor->iprec) == 0) {
1494                         cursor->flags |= HAMMER_CURSOR_ATEMEM;
1495                         goto next_memory;
1496                 }
1497         }
1498
1499         /*
1500          * Extract either the disk or memory record depending on their
1501          * relative position.
1502          */
1503         error = 0;
1504         switch(cursor->flags & (HAMMER_CURSOR_ATEDISK | HAMMER_CURSOR_ATEMEM)) {
1505         case 0:
1506                 /*
1507                  * Both entries valid.   Compare the entries and nominally
1508                  * return the first one in the sort order.  Numerous cases
1509                  * require special attention, however.
1510                  */
1511                 elm = &cursor->node->ondisk->elms[cursor->index];
1512                 r = hammer_btree_cmp(&elm->base, &cursor->iprec->leaf.base);
1513
1514                 /*
1515                  * If the two entries differ only by their key (-2/2) or
1516                  * create_tid (-1/1), and are DATA records, we may have a
1517                  * nominal match.  We have to calculate the base file
1518                  * offset of the data.
1519                  */
1520                 if (r <= 2 && r >= -2 && r != 0 &&
1521                     cursor->ip->ino_data.obj_type == HAMMER_OBJTYPE_REGFILE &&
1522                     cursor->iprec->type == HAMMER_MEM_RECORD_DATA) {
1523                         int64_t base1 = elm->leaf.base.key - elm->leaf.data_len;
1524                         int64_t base2 = cursor->iprec->leaf.base.key -
1525                                         cursor->iprec->leaf.data_len;
1526                         if (base1 == base2)
1527                                 r = 0;
1528                 }
1529
1530                 if (r < 0) {
1531                         error = hammer_btree_extract(cursor,
1532                                                      HAMMER_CURSOR_GET_LEAF);
1533                         cursor->flags |= HAMMER_CURSOR_ATEDISK;
1534                         break;
1535                 }
1536
1537                 /*
1538                  * If the entries match exactly the memory entry is either
1539                  * an on-disk directory entry deletion or a bulk data
1540                  * overwrite.  If it is a directory entry deletion we eat
1541                  * both entries.
1542                  *
1543                  * For the bulk-data overwrite case it is possible to have
1544                  * visibility into both, which simply means the syncer
1545                  * hasn't gotten around to doing the delete+insert sequence
1546                  * on the B-Tree.  Use the memory entry and throw away the
1547                  * on-disk entry.
1548                  *
1549                  * If the in-memory record is not either of these we
1550                  * probably caught the syncer while it was syncing it to
1551                  * the media.  Since we hold a shared lock on the cursor,
1552                  * the in-memory record had better be marked deleted at
1553                  * this point.
1554                  */
1555                 if (r == 0) {
1556                         if (cursor->iprec->type == HAMMER_MEM_RECORD_DEL) {
1557                                 if ((cursor->flags & HAMMER_CURSOR_DELETE_VISIBILITY) == 0) {
1558                                         cursor->flags |= HAMMER_CURSOR_ATEDISK;
1559                                         cursor->flags |= HAMMER_CURSOR_ATEMEM;
1560                                         goto next_btree;
1561                                 }
1562                         } else if (cursor->iprec->type == HAMMER_MEM_RECORD_DATA) {
1563                                 if ((cursor->flags & HAMMER_CURSOR_DELETE_VISIBILITY) == 0) {
1564                                         cursor->flags |= HAMMER_CURSOR_ATEDISK;
1565                                 }
1566                                 /* fall through to memory entry */
1567                         } else {
1568                                 panic("hammer_ip_next: duplicate mem/b-tree entry %p %d %08x", cursor->iprec, cursor->iprec->type, cursor->iprec->flags);
1569                                 cursor->flags |= HAMMER_CURSOR_ATEMEM;
1570                                 goto next_memory;
1571                         }
1572                 }
1573                 /* fall through to the memory entry */
1574         case HAMMER_CURSOR_ATEDISK:
1575                 /*
1576                  * Only the memory entry is valid.
1577                  */
1578                 cursor->leaf = &cursor->iprec->leaf;
1579                 cursor->flags |= HAMMER_CURSOR_ATEMEM;
1580
1581                 /*
1582                  * If the memory entry is an on-disk deletion we should have
1583                  * also had found a B-Tree record.  If the backend beat us
1584                  * to it it would have interlocked the cursor and we should
1585                  * have seen the in-memory record marked DELETED_FE.
1586                  */
1587                 if (cursor->iprec->type == HAMMER_MEM_RECORD_DEL &&
1588                     (cursor->flags & HAMMER_CURSOR_DELETE_VISIBILITY) == 0) {
1589                         panic("hammer_ip_next: del-on-disk with no b-tree entry iprec %p flags %08x", cursor->iprec, cursor->iprec->flags);
1590                 }
1591                 break;
1592         case HAMMER_CURSOR_ATEMEM:
1593                 /*
1594                  * Only the disk entry is valid
1595                  */
1596                 error = hammer_btree_extract(cursor, HAMMER_CURSOR_GET_LEAF);
1597                 cursor->flags |= HAMMER_CURSOR_ATEDISK;
1598                 break;
1599         default:
1600                 /*
1601                  * Neither entry is valid
1602                  *
1603                  * XXX error not set properly
1604                  */
1605                 cursor->leaf = NULL;
1606                 error = ENOENT;
1607                 break;
1608         }
1609         return(error);
1610 }
1611
1612 /*
1613  * Resolve the cursor->data pointer for the current cursor position in
1614  * a merged iteration.
1615  */
1616 int
1617 hammer_ip_resolve_data(hammer_cursor_t cursor)
1618 {
1619         hammer_record_t record;
1620         int error;
1621
1622         if (hammer_cursor_inmem(cursor)) {
1623                 /*
1624                  * The data associated with an in-memory record is usually
1625                  * kmalloced, but reserve-ahead data records will have an
1626                  * on-disk reference.
1627                  *
1628                  * NOTE: Reserve-ahead data records must be handled in the
1629                  * context of the related high level buffer cache buffer
1630                  * to interlock against async writes.
1631                  */
1632                 record = cursor->iprec;
1633                 cursor->data = record->data;
1634                 error = 0;
1635                 if (cursor->data == NULL) {
1636                         KKASSERT(record->leaf.base.rec_type ==
1637                                  HAMMER_RECTYPE_DATA);
1638                         cursor->data = hammer_bread_ext(cursor->trans->hmp,
1639                                                     record->leaf.data_offset,
1640                                                     record->leaf.data_len,
1641                                                     &error,
1642                                                     &cursor->data_buffer);
1643                 }
1644         } else {
1645                 cursor->leaf = &cursor->node->ondisk->elms[cursor->index].leaf;
1646                 error = hammer_btree_extract(cursor, HAMMER_CURSOR_GET_DATA);
1647         }
1648         return(error);
1649 }
1650
1651 /*
1652  * Backend truncation / record replacement - delete records in range.
1653  *
1654  * Delete all records within the specified range for inode ip.  In-memory
1655  * records still associated with the frontend are ignored.
1656  *
1657  * If truncating is non-zero in-memory records associated with the back-end
1658  * are ignored.  If truncating is > 1 we can return EWOULDBLOCK.
1659  *
1660  * NOTES:
1661  *
1662  *      * An unaligned range will cause new records to be added to cover
1663  *        the edge cases. (XXX not implemented yet).
1664  *
1665  *      * Replacement via reservations (see hammer_ip_sync_record_cursor())
1666  *        also do not deal with unaligned ranges.
1667  *
1668  *      * ran_end is inclusive (e.g. 0,1023 instead of 0,1024).
1669  *
1670  *      * Record keys for regular file data have to be special-cased since
1671  *        they indicate the end of the range (key = base + bytes).
1672  *
1673  *      * This function may be asked to delete ridiculously huge ranges, for
1674  *        example if someone truncates or removes a 1TB regular file.  We
1675  *        must be very careful on restarts and we may have to stop w/
1676  *        EWOULDBLOCK to avoid blowing out the buffer cache.
1677  */
1678 int
1679 hammer_ip_delete_range(hammer_cursor_t cursor, hammer_inode_t ip,
1680                        int64_t ran_beg, int64_t ran_end, int truncating)
1681 {
1682         hammer_transaction_t trans = cursor->trans;
1683         hammer_btree_leaf_elm_t leaf;
1684         int error;
1685         int64_t off;
1686         int64_t tmp64;
1687
1688 #if 0
1689         kprintf("delete_range %p %016llx-%016llx\n", ip, ran_beg, ran_end);
1690 #endif
1691
1692         KKASSERT(trans->type == HAMMER_TRANS_FLS);
1693 retry:
1694         hammer_normalize_cursor(cursor);
1695         cursor->key_beg.localization = ip->obj_localization +
1696                                        HAMMER_LOCALIZE_MISC;
1697         cursor->key_beg.obj_id = ip->obj_id;
1698         cursor->key_beg.create_tid = 0;
1699         cursor->key_beg.delete_tid = 0;
1700         cursor->key_beg.obj_type = 0;
1701
1702         if (ip->ino_data.obj_type == HAMMER_OBJTYPE_DBFILE) {
1703                 cursor->key_beg.key = ran_beg;
1704                 cursor->key_beg.rec_type = HAMMER_RECTYPE_DB;
1705         } else {
1706                 /*
1707                  * The key in the B-Tree is (base+bytes), so the first possible
1708                  * matching key is ran_beg + 1.
1709                  */
1710                 cursor->key_beg.key = ran_beg + 1;
1711                 cursor->key_beg.rec_type = HAMMER_RECTYPE_DATA;
1712         }
1713
1714         cursor->key_end = cursor->key_beg;
1715         if (ip->ino_data.obj_type == HAMMER_OBJTYPE_DBFILE) {
1716                 cursor->key_end.key = ran_end;
1717         } else {
1718                 tmp64 = ran_end + MAXPHYS + 1;  /* work around GCC-4 bug */
1719                 if (tmp64 < ran_end)
1720                         cursor->key_end.key = 0x7FFFFFFFFFFFFFFFLL;
1721                 else
1722                         cursor->key_end.key = ran_end + MAXPHYS + 1;
1723         }
1724
1725         cursor->asof = ip->obj_asof;
1726         cursor->flags &= ~HAMMER_CURSOR_INITMASK;
1727         cursor->flags |= HAMMER_CURSOR_ASOF;
1728         cursor->flags |= HAMMER_CURSOR_DELETE_VISIBILITY;
1729         cursor->flags |= HAMMER_CURSOR_BACKEND;
1730         cursor->flags |= HAMMER_CURSOR_END_INCLUSIVE;
1731
1732         error = hammer_ip_first(cursor);
1733
1734         /*
1735          * Iterate through matching records and mark them as deleted.
1736          */
1737         while (error == 0) {
1738                 leaf = cursor->leaf;
1739
1740                 KKASSERT(leaf->base.delete_tid == 0);
1741                 KKASSERT(leaf->base.obj_id == ip->obj_id);
1742
1743                 /*
1744                  * There may be overlap cases for regular file data.  Also
1745                  * remember the key for a regular file record is (base + len),
1746                  * NOT (base).
1747                  *
1748                  * Note that do to duplicates (mem & media) allowed by
1749                  * DELETE_VISIBILITY, off can wind up less then ran_beg.
1750                  */
1751                 if (leaf->base.rec_type == HAMMER_RECTYPE_DATA) {
1752                         off = leaf->base.key - leaf->data_len;
1753                         /*
1754                          * Check the left edge case.  We currently do not
1755                          * split existing records.
1756                          */
1757                         if (off < ran_beg && leaf->base.key > ran_beg) {
1758                                 panic("hammer left edge case %016llx %d\n",
1759                                         leaf->base.key, leaf->data_len);
1760                         }
1761
1762                         /*
1763                          * Check the right edge case.  Note that the
1764                          * record can be completely out of bounds, which
1765                          * terminates the search.
1766                          *
1767                          * base->key is exclusive of the right edge while
1768                          * ran_end is inclusive of the right edge.  The
1769                          * (key - data_len) left boundary is inclusive.
1770                          *
1771                          * XXX theory-check this test at some point, are
1772                          * we missing a + 1 somewhere?  Note that ran_end
1773                          * could overflow.
1774                          */
1775                         if (leaf->base.key - 1 > ran_end) {
1776                                 if (leaf->base.key - leaf->data_len > ran_end)
1777                                         break;
1778                                 panic("hammer right edge case\n");
1779                         }
1780                 } else {
1781                         off = leaf->base.key;
1782                 }
1783
1784                 /*
1785                  * Delete the record.  When truncating we do not delete
1786                  * in-memory (data) records because they represent data
1787                  * written after the truncation.
1788                  *
1789                  * This will also physically destroy the B-Tree entry and
1790                  * data if the retention policy dictates.  The function
1791                  * will set HAMMER_CURSOR_RETEST to cause hammer_ip_next()
1792                  * to retest the new 'current' element.
1793                  */
1794                 if (truncating == 0 || hammer_cursor_ondisk(cursor)) {
1795                         error = hammer_ip_delete_record(cursor, ip, trans->tid);
1796                         /*
1797                          * If we have built up too many meta-buffers we risk
1798                          * deadlocking the kernel and must stop.  This can
1799                          * occur when deleting ridiculously huge files.
1800                          * sync_trunc_off is updated so the next cycle does
1801                          * not re-iterate records we have already deleted.
1802                          *
1803                          * This is only done with formal truncations.
1804                          */
1805                         if (truncating > 1 && error == 0 &&
1806                             hammer_flusher_meta_limit(ip->hmp)) {
1807                                 ip->sync_trunc_off = off;
1808                                 error = EWOULDBLOCK;
1809                         }
1810                 }
1811                 if (error)
1812                         break;
1813                 ran_beg = off;  /* for restart */
1814                 error = hammer_ip_next(cursor);
1815         }
1816         if (cursor->node)
1817                 hammer_cache_node(&ip->cache[1], cursor->node);
1818
1819         if (error == EDEADLK) {
1820                 hammer_done_cursor(cursor);
1821                 error = hammer_init_cursor(trans, cursor, &ip->cache[1], ip);
1822                 if (error == 0)
1823                         goto retry;
1824         }
1825         if (error == ENOENT)
1826                 error = 0;
1827         return(error);
1828 }
1829
1830 /*
1831  * This backend function deletes the specified record on-disk, similar to
1832  * delete_range but for a specific record.  Unlike the exact deletions
1833  * used when deleting a directory entry this function uses an ASOF search
1834  * like delete_range.
1835  *
1836  * This function may be called with ip->obj_asof set for a slave snapshot,
1837  * so don't use it.  We always delete non-historical records only.
1838  */
1839 static int
1840 hammer_delete_general(hammer_cursor_t cursor, hammer_inode_t ip,
1841                       hammer_btree_leaf_elm_t leaf)
1842 {
1843         hammer_transaction_t trans = cursor->trans;
1844         int error;
1845
1846         KKASSERT(trans->type == HAMMER_TRANS_FLS);
1847 retry:
1848         hammer_normalize_cursor(cursor);
1849         cursor->key_beg = leaf->base;
1850         cursor->asof = HAMMER_MAX_TID;
1851         cursor->flags &= ~HAMMER_CURSOR_INITMASK;
1852         cursor->flags |= HAMMER_CURSOR_ASOF;
1853         cursor->flags |= HAMMER_CURSOR_BACKEND;
1854         cursor->flags &= ~HAMMER_CURSOR_INSERT;
1855
1856         error = hammer_btree_lookup(cursor);
1857         if (error == 0) {
1858                 error = hammer_ip_delete_record(cursor, ip, trans->tid);
1859         }
1860         if (error == EDEADLK) {
1861                 hammer_done_cursor(cursor);
1862                 error = hammer_init_cursor(trans, cursor, &ip->cache[1], ip);
1863                 if (error == 0)
1864                         goto retry;
1865         }
1866         return(error);
1867 }
1868
1869 /*
1870  * This function deletes remaining auxillary records when an inode is
1871  * being deleted.  This function explicitly does not delete the
1872  * inode record, directory entry, data, or db records.  Those must be
1873  * properly disposed of prior to this call.
1874  */
1875 int
1876 hammer_ip_delete_clean(hammer_cursor_t cursor, hammer_inode_t ip, int *countp)
1877 {
1878         hammer_transaction_t trans = cursor->trans;
1879         hammer_btree_leaf_elm_t leaf;
1880         int error;
1881
1882         KKASSERT(trans->type == HAMMER_TRANS_FLS);
1883 retry:
1884         hammer_normalize_cursor(cursor);
1885         cursor->key_beg.localization = ip->obj_localization +
1886                                        HAMMER_LOCALIZE_MISC;
1887         cursor->key_beg.obj_id = ip->obj_id;
1888         cursor->key_beg.create_tid = 0;
1889         cursor->key_beg.delete_tid = 0;
1890         cursor->key_beg.obj_type = 0;
1891         cursor->key_beg.rec_type = HAMMER_RECTYPE_CLEAN_START;
1892         cursor->key_beg.key = HAMMER_MIN_KEY;
1893
1894         cursor->key_end = cursor->key_beg;
1895         cursor->key_end.rec_type = HAMMER_RECTYPE_MAX;
1896         cursor->key_end.key = HAMMER_MAX_KEY;
1897
1898         cursor->asof = ip->obj_asof;
1899         cursor->flags &= ~HAMMER_CURSOR_INITMASK;
1900         cursor->flags |= HAMMER_CURSOR_END_INCLUSIVE | HAMMER_CURSOR_ASOF;
1901         cursor->flags |= HAMMER_CURSOR_DELETE_VISIBILITY;
1902         cursor->flags |= HAMMER_CURSOR_BACKEND;
1903
1904         error = hammer_ip_first(cursor);
1905
1906         /*
1907          * Iterate through matching records and mark them as deleted.
1908          */
1909         while (error == 0) {
1910                 leaf = cursor->leaf;
1911
1912                 KKASSERT(leaf->base.delete_tid == 0);
1913
1914                 /*
1915                  * Mark the record and B-Tree entry as deleted.  This will
1916                  * also physically delete the B-Tree entry, record, and
1917                  * data if the retention policy dictates.  The function
1918                  * will set HAMMER_CURSOR_RETEST to cause hammer_ip_next()
1919                  * to retest the new 'current' element.
1920                  *
1921                  * Directory entries (and delete-on-disk directory entries)
1922                  * must be synced and cannot be deleted.
1923                  */
1924                 error = hammer_ip_delete_record(cursor, ip, trans->tid);
1925                 ++*countp;
1926                 if (error)
1927                         break;
1928                 error = hammer_ip_next(cursor);
1929         }
1930         if (cursor->node)
1931                 hammer_cache_node(&ip->cache[1], cursor->node);
1932         if (error == EDEADLK) {
1933                 hammer_done_cursor(cursor);
1934                 error = hammer_init_cursor(trans, cursor, &ip->cache[1], ip);
1935                 if (error == 0)
1936                         goto retry;
1937         }
1938         if (error == ENOENT)
1939                 error = 0;
1940         return(error);
1941 }
1942
1943 /*
1944  * Delete the record at the current cursor.  On success the cursor will
1945  * be positioned appropriately for an iteration but may no longer be at
1946  * a leaf node.
1947  *
1948  * This routine is only called from the backend.
1949  *
1950  * NOTE: This can return EDEADLK, requiring the caller to terminate the
1951  * cursor and retry.
1952  */
1953 int
1954 hammer_ip_delete_record(hammer_cursor_t cursor, hammer_inode_t ip,
1955                         hammer_tid_t tid)
1956 {
1957         hammer_record_t iprec;
1958         hammer_mount_t hmp;
1959         int error;
1960
1961         KKASSERT(cursor->flags & HAMMER_CURSOR_BACKEND);
1962         KKASSERT(tid != 0);
1963         hmp = cursor->node->hmp;
1964
1965         /*
1966          * In-memory (unsynchronized) records can simply be freed.  This
1967          * only occurs in range iterations since all other records are
1968          * individually synchronized.  Thus there should be no confusion with
1969          * the interlock.
1970          *
1971          * An in-memory record may be deleted before being committed to disk,
1972          * but could have been accessed in the mean time.  The reservation
1973          * code will deal with the case.
1974          */
1975         if (hammer_cursor_inmem(cursor)) {
1976                 iprec = cursor->iprec;
1977                 KKASSERT((iprec->flags & HAMMER_RECF_INTERLOCK_BE) ==0);
1978                 iprec->flags |= HAMMER_RECF_DELETED_FE;
1979                 iprec->flags |= HAMMER_RECF_DELETED_BE;
1980                 return(0);
1981         }
1982
1983         /*
1984          * On-disk records are marked as deleted by updating their delete_tid.
1985          * This does not effect their position in the B-Tree (which is based
1986          * on their create_tid).
1987          *
1988          * Frontend B-Tree operations track inodes so we tell
1989          * hammer_delete_at_cursor() not to.
1990          */
1991         error = hammer_btree_extract(cursor, HAMMER_CURSOR_GET_LEAF);
1992
1993         if (error == 0) {
1994                 error = hammer_delete_at_cursor(
1995                                 cursor,
1996                                 HAMMER_DELETE_ADJUST | hammer_nohistory(ip),
1997                                 cursor->trans->tid,
1998                                 cursor->trans->time32,
1999                                 0, NULL);
2000         }
2001         return(error);
2002 }
2003
2004 /*
2005  * Delete the B-Tree element at the current cursor and do any necessary
2006  * mirror propagation.
2007  *
2008  * The cursor must be properly positioned for an iteration on return but
2009  * may be pointing at an internal element.
2010  *
2011  * An element can be un-deleted by passing a delete_tid of 0 with
2012  * HAMMER_DELETE_ADJUST.
2013  */
2014 int
2015 hammer_delete_at_cursor(hammer_cursor_t cursor, int delete_flags,
2016                         hammer_tid_t delete_tid, u_int32_t delete_ts,
2017                         int track, int64_t *stat_bytes)
2018 {
2019         struct hammer_btree_leaf_elm save_leaf;
2020         hammer_transaction_t trans;
2021         hammer_btree_leaf_elm_t leaf;
2022         hammer_node_t node;
2023         hammer_btree_elm_t elm;
2024         hammer_off_t data_offset;
2025         int32_t data_len;
2026         u_int16_t rec_type;
2027         int error;
2028         int icount;
2029         int doprop;
2030
2031         error = hammer_cursor_upgrade(cursor);
2032         if (error)
2033                 return(error);
2034
2035         trans = cursor->trans;
2036         node = cursor->node;
2037         elm = &node->ondisk->elms[cursor->index];
2038         leaf = &elm->leaf;
2039         KKASSERT(elm->base.btype == HAMMER_BTREE_TYPE_RECORD);
2040
2041         hammer_sync_lock_sh(trans);
2042         doprop = 0;
2043         icount = 0;
2044
2045         /*
2046          * Adjust the delete_tid.  Update the mirror_tid propagation field
2047          * as well.  delete_tid can be 0 (undelete -- used by mirroring).
2048          */
2049         if (delete_flags & HAMMER_DELETE_ADJUST) {
2050                 if (elm->base.rec_type == HAMMER_RECTYPE_INODE) {
2051                         if (elm->leaf.base.delete_tid == 0 && delete_tid)
2052                                 icount = -1;
2053                         if (elm->leaf.base.delete_tid && delete_tid == 0)
2054                                 icount = 1;
2055                 }
2056
2057                 hammer_modify_node(trans, node, elm, sizeof(*elm));
2058                 elm->leaf.base.delete_tid = delete_tid;
2059                 elm->leaf.delete_ts = delete_ts;
2060                 hammer_modify_node_done(node);
2061
2062                 if (elm->leaf.base.delete_tid > node->ondisk->mirror_tid) {
2063                         hammer_modify_node_field(trans, node, mirror_tid);
2064                         node->ondisk->mirror_tid = elm->leaf.base.delete_tid;
2065                         hammer_modify_node_done(node);
2066                         doprop = 1;
2067                         if (hammer_debug_general & 0x0002) {
2068                                 kprintf("delete_at_cursor: propagate %016llx"
2069                                         " @%016llx\n",
2070                                         elm->leaf.base.delete_tid,
2071                                         node->node_offset);
2072                         }
2073                 }
2074
2075                 /*
2076                  * Adjust for the iteration.  We have deleted the current
2077                  * element and want to clear ATEDISK so the iteration does
2078                  * not skip the element after, which now becomes the current
2079                  * element.  This element must be re-tested if doing an
2080                  * iteration, which is handled by the RETEST flag.
2081                  */
2082                 if ((cursor->flags & HAMMER_CURSOR_DISKEOF) == 0) {
2083                         cursor->flags |= HAMMER_CURSOR_RETEST;
2084                         cursor->flags &= ~HAMMER_CURSOR_ATEDISK;
2085                 }
2086
2087                 /*
2088                  * An on-disk record cannot have the same delete_tid
2089                  * as its create_tid.  In a chain of record updates
2090                  * this could result in a duplicate record.
2091                  */
2092                 KKASSERT(elm->leaf.base.delete_tid !=
2093                          elm->leaf.base.create_tid);
2094         }
2095
2096         /*
2097          * Destroy the B-Tree element if asked (typically if a nohistory
2098          * file or mount, or when called by the pruning code).
2099          *
2100          * Adjust the ATEDISK flag to properly support iterations.
2101          */
2102         if (delete_flags & HAMMER_DELETE_DESTROY) {
2103                 data_offset = elm->leaf.data_offset;
2104                 data_len = elm->leaf.data_len;
2105                 rec_type = elm->leaf.base.rec_type;
2106                 if (doprop) {
2107                         save_leaf = elm->leaf;
2108                         leaf = &save_leaf;
2109                 }
2110                 if (elm->base.rec_type == HAMMER_RECTYPE_INODE &&
2111                     elm->leaf.base.delete_tid == 0) {
2112                         icount = -1;
2113                 }
2114
2115                 error = hammer_btree_delete(cursor);
2116                 if (error == 0) {
2117                         /*
2118                          * The deletion moves the next element (if any) to
2119                          * the current element position.  We must clear
2120                          * ATEDISK so this element is not skipped and we
2121                          * must set RETEST to force any iteration to re-test
2122                          * the element.
2123                          */
2124                         if ((cursor->flags & HAMMER_CURSOR_DISKEOF) == 0) {
2125                                 cursor->flags |= HAMMER_CURSOR_RETEST;
2126                                 cursor->flags &= ~HAMMER_CURSOR_ATEDISK;
2127                         }
2128                 }
2129                 if (error == 0) {
2130                         switch(data_offset & HAMMER_OFF_ZONE_MASK) {
2131                         case HAMMER_ZONE_LARGE_DATA:
2132                         case HAMMER_ZONE_SMALL_DATA:
2133                         case HAMMER_ZONE_META:
2134                                 hammer_blockmap_free(trans,
2135                                                      data_offset, data_len);
2136                                 break;
2137                         default:
2138                                 break;
2139                         }
2140                 }
2141         }
2142
2143         /*
2144          * Track inode count and next_tid.  This is used by the mirroring
2145          * and PFS code.  icount can be negative, zero, or positive.
2146          */
2147         if (error == 0 && track) {
2148                 if (icount) {
2149                         hammer_modify_volume_field(trans, trans->rootvol,
2150                                                    vol0_stat_inodes);
2151                         trans->rootvol->ondisk->vol0_stat_inodes += icount;
2152                         hammer_modify_volume_done(trans->rootvol);
2153                 }
2154                 if (trans->rootvol->ondisk->vol0_next_tid < delete_tid) {
2155                         hammer_modify_volume(trans, trans->rootvol, NULL, 0);
2156                         trans->rootvol->ondisk->vol0_next_tid = delete_tid;
2157                         hammer_modify_volume_done(trans->rootvol);
2158                 }
2159         }
2160
2161         /*
2162          * mirror_tid propagation occurs if the node's mirror_tid had to be
2163          * updated while adjusting the delete_tid.
2164          *
2165          * This occurs when deleting even in nohistory mode, but does not
2166          * occur when pruning an already-deleted node.
2167          *
2168          * cursor->ip is NULL when called from the pruning, mirroring,
2169          * and pfs code.  If non-NULL propagation will be conditionalized
2170          * on whether the PFS is in no-history mode or not.
2171          */
2172         if (doprop) {
2173                 if (cursor->ip)
2174                         hammer_btree_do_propagation(cursor, cursor->ip->pfsm, leaf);
2175                 else
2176                         hammer_btree_do_propagation(cursor, NULL, leaf);
2177         }
2178         hammer_sync_unlock(trans);
2179         return (error);
2180 }
2181
2182 /*
2183  * Determine whether we can remove a directory.  This routine checks whether
2184  * a directory is empty or not and enforces flush connectivity.
2185  *
2186  * Flush connectivity requires that we block if the target directory is
2187  * currently flushing, otherwise it may not end up in the same flush group.
2188  *
2189  * Returns 0 on success, ENOTEMPTY or EDEADLK (or other errors) on failure.
2190  */
2191 int
2192 hammer_ip_check_directory_empty(hammer_transaction_t trans, hammer_inode_t ip)
2193 {
2194         struct hammer_cursor cursor;
2195         int error;
2196
2197         /*
2198          * Check directory empty
2199          */
2200         hammer_init_cursor(trans, &cursor, &ip->cache[1], ip);
2201
2202         cursor.key_beg.localization = ip->obj_localization +
2203                                       HAMMER_LOCALIZE_MISC;
2204         cursor.key_beg.obj_id = ip->obj_id;
2205         cursor.key_beg.create_tid = 0;
2206         cursor.key_beg.delete_tid = 0;
2207         cursor.key_beg.obj_type = 0;
2208         cursor.key_beg.rec_type = HAMMER_RECTYPE_INODE + 1;
2209         cursor.key_beg.key = HAMMER_MIN_KEY;
2210
2211         cursor.key_end = cursor.key_beg;
2212         cursor.key_end.rec_type = 0xFFFF;
2213         cursor.key_end.key = HAMMER_MAX_KEY;
2214
2215         cursor.asof = ip->obj_asof;
2216         cursor.flags |= HAMMER_CURSOR_END_INCLUSIVE | HAMMER_CURSOR_ASOF;
2217
2218         error = hammer_ip_first(&cursor);
2219         if (error == ENOENT)
2220                 error = 0;
2221         else if (error == 0)
2222                 error = ENOTEMPTY;
2223         hammer_done_cursor(&cursor);
2224         return(error);
2225 }
2226