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
35 /*
36 * HAMMER B-Tree index - cursor support routines
37 */
38 #include "hammer.h"
39
40 static int hammer_load_cursor_parent(hammer_cursor_t cursor, int try_exclusive);
41
42 /*
43 * Initialize a fresh cursor using the B-Tree node cache. If the cache
44 * is not available initialize a fresh cursor at the root of the filesystem.
45 */
46 int
hammer_init_cursor(hammer_transaction_t trans,hammer_cursor_t cursor,hammer_node_cache_t cache,hammer_inode_t ip)47 hammer_init_cursor(hammer_transaction_t trans, hammer_cursor_t cursor,
48 hammer_node_cache_t cache, hammer_inode_t ip)
49 {
50 hammer_volume_t volume;
51 hammer_node_t node;
52 hammer_mount_t hmp;
53 u_int tticks;
54 int error;
55
56 bzero(cursor, sizeof(*cursor));
57
58 cursor->trans = trans;
59 hmp = trans->hmp;
60
61 /*
62 * As the number of inodes queued to the flusher increases we use
63 * time-domain multiplexing to control read vs flush performance.
64 * We have to do it here, before acquiring any ip or node locks,
65 * to avoid deadlocking or excessively delaying the flusher.
66 *
67 * The full time period is hammer_tdmux_ticks, typically 1/5 of
68 * a second.
69 *
70 * inode allocation begins to get restrained at 2/4 the limit
71 * via the "hmrrcm" mechanism in hammer_inode. We want to begin
72 * limiting read activity before that to try to avoid processes
73 * stalling out in "hmrrcm".
74 */
75 tticks = hammer_tdmux_ticks;
76 if (trans->type != HAMMER_TRANS_FLS && tticks &&
77 hmp->count_reclaims > hammer_limit_reclaims / tticks &&
78 hmp->count_reclaims > hammer_autoflush * 2 &&
79 hammer_flusher_running(hmp)) {
80 u_int rticks;
81 u_int xticks;
82 u_int dummy;
83
84 /*
85 * 0 ... xticks ... tticks
86 *
87 * rticks is the calculated position, xticks is the demarc
88 * where values below xticks are reserved for the flusher
89 * and values >= to xticks may be used by the frontend.
90 *
91 * At least one tick is always made available for the
92 * frontend.
93 */
94 rticks = (u_int)ticks % tticks;
95 xticks = hmp->count_reclaims * tticks / hammer_limit_reclaims;
96
97 /*
98 * Ensure rticks and xticks are stable
99 */
100 cpu_ccfence();
101 if (rticks < xticks) {
102 if (hammer_debug_general & 0x0004)
103 hdkprintf("rt %3u, xt %3u, tt %3u\n",
104 rticks, xticks, tticks);
105 tsleep(&dummy, 0, "htdmux", xticks - rticks);
106 }
107 }
108
109 /*
110 * If the cursor operation is on behalf of an inode, lock
111 * the inode.
112 *
113 * When acquiring a shared lock on an inode on which the backend
114 * flusher deadlocked, wait up to hammer_tdmux_ticks (1 second)
115 * for the deadlock to clear.
116 */
117 if ((cursor->ip = ip) != NULL) {
118 ++ip->cursor_ip_refs;
119 if (trans->type == HAMMER_TRANS_FLS) {
120 hammer_lock_ex(&ip->lock);
121 } else {
122 #if 0
123 if (ip->cursor_exclreq_count) {
124 tsleep(&ip->cursor_exclreq_count, 0,
125 "hstag1", hammer_tdmux_ticks);
126 }
127 #endif
128 hammer_lock_sh(&ip->lock);
129 }
130 }
131
132 /*
133 * Step 1 - acquire a locked node from the cache if possible
134 */
135 if (cache && cache->node) {
136 node = hammer_ref_node_safe(trans, cache, &error);
137 if (error == 0) {
138 hammer_lock_sh(&node->lock);
139 if (node->flags & HAMMER_NODE_DELETED) {
140 hammer_unlock(&node->lock);
141 hammer_rel_node(node);
142 node = NULL;
143 }
144 }
145 if (node == NULL)
146 ++hammer_stats_btree_root_iterations;
147 } else {
148 node = NULL;
149 ++hammer_stats_btree_root_iterations;
150 }
151
152 /*
153 * Step 2 - If we couldn't get a node from the cache, get
154 * the one from the root of the filesystem.
155 */
156 while (node == NULL) {
157 volume = hammer_get_root_volume(hmp, &error);
158 if (error)
159 break;
160 node = hammer_get_node(trans, volume->ondisk->vol0_btree_root,
161 0, &error);
162 hammer_rel_volume(volume, 0);
163 if (error)
164 break;
165 /*
166 * When the frontend acquires the root b-tree node while the
167 * backend is deadlocked on it, wait up to hammer_tdmux_ticks
168 * (1 second) for the deadlock to clear.
169 */
170 #if 0
171 if (node->cursor_exclreq_count &&
172 cursor->trans->type != HAMMER_TRANS_FLS) {
173 tsleep(&node->cursor_exclreq_count, 0,
174 "hstag3", hammer_tdmux_ticks);
175 }
176 #endif
177 hammer_lock_sh(&node->lock);
178
179 /*
180 * If someone got in before we could lock the node, retry.
181 */
182 if (node->flags & HAMMER_NODE_DELETED) {
183 hammer_unlock(&node->lock);
184 hammer_rel_node(node);
185 node = NULL;
186 continue;
187 }
188 if (volume->ondisk->vol0_btree_root != node->node_offset) {
189 hammer_unlock(&node->lock);
190 hammer_rel_node(node);
191 node = NULL;
192 continue;
193 }
194 }
195
196 /*
197 * Step 3 - finish initializing the cursor by acquiring the parent
198 */
199 cursor->node = node;
200 if (error == 0)
201 error = hammer_load_cursor_parent(cursor, 0);
202 KKASSERT(error == 0);
203 /* if (error) hammer_done_cursor(cursor); */
204 return(error);
205 }
206
207 /*
208 * Normalize a cursor. Sometimes cursors can be left in a state
209 * where node is NULL. If the cursor is in this state, cursor up.
210 */
211 void
hammer_normalize_cursor(hammer_cursor_t cursor)212 hammer_normalize_cursor(hammer_cursor_t cursor)
213 {
214 if (cursor->node == NULL) {
215 KKASSERT(cursor->parent != NULL);
216 hammer_cursor_up(cursor);
217 }
218 }
219
220
221 /*
222 * We are finished with a cursor. We NULL out various fields as sanity
223 * check, in case the structure is inappropriately used afterwords.
224 */
225 void
hammer_done_cursor(hammer_cursor_t cursor)226 hammer_done_cursor(hammer_cursor_t cursor)
227 {
228 hammer_inode_t ip;
229
230 KKASSERT((cursor->flags & HAMMER_CURSOR_TRACKED) == 0);
231 if (cursor->parent) {
232 hammer_unlock(&cursor->parent->lock);
233 hammer_rel_node(cursor->parent);
234 cursor->parent = NULL;
235 }
236 if (cursor->node) {
237 hammer_unlock(&cursor->node->lock);
238 hammer_rel_node(cursor->node);
239 cursor->node = NULL;
240 }
241 if (cursor->data_buffer) {
242 hammer_rel_buffer(cursor->data_buffer, 0);
243 cursor->data_buffer = NULL;
244 }
245 if ((ip = cursor->ip) != NULL) {
246 KKASSERT(ip->cursor_ip_refs > 0);
247 --ip->cursor_ip_refs;
248 hammer_unlock(&ip->lock);
249 cursor->ip = NULL;
250 }
251 if (cursor->iprec) {
252 hammer_rel_mem_record(cursor->iprec);
253 cursor->iprec = NULL;
254 }
255
256 /*
257 * If we deadlocked this node will be referenced. Do a quick
258 * lock/unlock to wait for the deadlock condition to clear.
259 *
260 * Maintain exclreq_count / wakeup as necessary to notify new
261 * entrants into ip. We continue to hold the fs_token so our
262 * EDEADLK retry loop should get its chance before another thread
263 * steals the lock.
264 */
265 if (cursor->deadlk_node) {
266 #if 0
267 if (ip && cursor->trans->type == HAMMER_TRANS_FLS)
268 ++ip->cursor_exclreq_count;
269 ++cursor->deadlk_node->cursor_exclreq_count;
270 #endif
271 hammer_lock_ex_ident(&cursor->deadlk_node->lock, "hmrdlk");
272 hammer_unlock(&cursor->deadlk_node->lock);
273 #if 0
274 if (--cursor->deadlk_node->cursor_exclreq_count == 0)
275 wakeup(&cursor->deadlk_node->cursor_exclreq_count);
276 if (ip && cursor->trans->type == HAMMER_TRANS_FLS) {
277 if (--ip->cursor_exclreq_count == 0)
278 wakeup(&ip->cursor_exclreq_count);
279 }
280 #endif
281 hammer_rel_node(cursor->deadlk_node);
282 cursor->deadlk_node = NULL;
283 }
284 if (cursor->deadlk_rec) {
285 hammer_wait_mem_record_ident(cursor->deadlk_rec, "hmmdlr");
286 hammer_rel_mem_record(cursor->deadlk_rec);
287 cursor->deadlk_rec = NULL;
288 }
289
290 cursor->data = NULL;
291 cursor->leaf = NULL;
292 cursor->left_bound = NULL;
293 cursor->right_bound = NULL;
294 cursor->trans = NULL;
295 }
296
297 /*
298 * Upgrade cursor->node and cursor->parent to exclusive locks. This
299 * function can return EDEADLK.
300 *
301 * The lock must already be either held shared or already held exclusively
302 * by us.
303 *
304 * We upgrade the parent first as it is the most likely to collide first
305 * with the downward traversal that the frontend typically does.
306 *
307 * If we fail to upgrade the lock and cursor->deadlk_node is NULL,
308 * we add another reference to the node that failed and set
309 * cursor->deadlk_node so hammer_done_cursor() can block on it.
310 */
311 int
hammer_cursor_upgrade(hammer_cursor_t cursor)312 hammer_cursor_upgrade(hammer_cursor_t cursor)
313 {
314 int error;
315
316 if (cursor->parent) {
317 error = hammer_lock_upgrade(&cursor->parent->lock, 1);
318 if (error && cursor->deadlk_node == NULL) {
319 cursor->deadlk_node = cursor->parent;
320 hammer_ref_node(cursor->deadlk_node);
321 }
322 } else {
323 error = 0;
324 }
325 if (error == 0) {
326 error = hammer_lock_upgrade(&cursor->node->lock, 1);
327 if (error && cursor->deadlk_node == NULL) {
328 cursor->deadlk_node = cursor->node;
329 hammer_ref_node(cursor->deadlk_node);
330 }
331 }
332 #if 0
333 error = hammer_lock_upgrade(&cursor->node->lock, 1);
334 if (error && cursor->deadlk_node == NULL) {
335 cursor->deadlk_node = cursor->node;
336 hammer_ref_node(cursor->deadlk_node);
337 } else if (error == 0 && cursor->parent) {
338 error = hammer_lock_upgrade(&cursor->parent->lock, 1);
339 if (error && cursor->deadlk_node == NULL) {
340 cursor->deadlk_node = cursor->parent;
341 hammer_ref_node(cursor->deadlk_node);
342 }
343 }
344 #endif
345 return(error);
346 }
347
348 int
hammer_cursor_upgrade_node(hammer_cursor_t cursor)349 hammer_cursor_upgrade_node(hammer_cursor_t cursor)
350 {
351 int error;
352
353 error = hammer_lock_upgrade(&cursor->node->lock, 1);
354 if (error && cursor->deadlk_node == NULL) {
355 cursor->deadlk_node = cursor->node;
356 hammer_ref_node(cursor->deadlk_node);
357 }
358 return(error);
359 }
360
361 /*
362 * Downgrade cursor->node and cursor->parent to shared locks.
363 */
364 void
hammer_cursor_downgrade(hammer_cursor_t cursor)365 hammer_cursor_downgrade(hammer_cursor_t cursor)
366 {
367 if (hammer_lock_excl_owned(&cursor->node->lock, curthread))
368 hammer_lock_downgrade(&cursor->node->lock, 1);
369 if (cursor->parent &&
370 hammer_lock_excl_owned(&cursor->parent->lock, curthread)) {
371 hammer_lock_downgrade(&cursor->parent->lock, 1);
372 }
373 }
374
375 /*
376 * Upgrade and downgrade pairs of cursors. This is used by the dedup
377 * code which must deal with two cursors. A special function is needed
378 * because some of the nodes may be shared between the two cursors,
379 * resulting in share counts > 1 which will normally cause an upgrade
380 * to fail.
381 */
382 static __noinline
383 int
collect_node(hammer_node_t * array,int * counts,int n,hammer_node_t node)384 collect_node(hammer_node_t *array, int *counts, int n, hammer_node_t node)
385 {
386 int i;
387
388 for (i = 0; i < n; ++i) {
389 if (array[i] == node)
390 break;
391 }
392 if (i == n) {
393 array[i] = node;
394 counts[i] = 1;
395 ++i;
396 } else {
397 ++counts[i];
398 }
399 return(i);
400 }
401
402 int
hammer_cursor_upgrade2(hammer_cursor_t cursor1,hammer_cursor_t cursor2)403 hammer_cursor_upgrade2(hammer_cursor_t cursor1, hammer_cursor_t cursor2)
404 {
405 hammer_node_t nodes[4];
406 int counts[4];
407 int error;
408 int i;
409 int n;
410
411 n = collect_node(nodes, counts, 0, cursor1->node);
412 if (cursor1->parent)
413 n = collect_node(nodes, counts, n, cursor1->parent);
414 n = collect_node(nodes, counts, n, cursor2->node);
415 if (cursor2->parent)
416 n = collect_node(nodes, counts, n, cursor2->parent);
417
418 error = 0;
419 for (i = 0; i < n; ++i) {
420 error = hammer_lock_upgrade(&nodes[i]->lock, counts[i]);
421 if (error)
422 break;
423 }
424 if (error) {
425 while (--i >= 0)
426 hammer_lock_downgrade(&nodes[i]->lock, counts[i]);
427 }
428 return (error);
429 }
430
431 void
hammer_cursor_downgrade2(hammer_cursor_t cursor1,hammer_cursor_t cursor2)432 hammer_cursor_downgrade2(hammer_cursor_t cursor1, hammer_cursor_t cursor2)
433 {
434 hammer_node_t nodes[4];
435 int counts[4];
436 int i;
437 int n;
438
439 n = collect_node(nodes, counts, 0, cursor1->node);
440 if (cursor1->parent)
441 n = collect_node(nodes, counts, n, cursor1->parent);
442 n = collect_node(nodes, counts, n, cursor2->node);
443 if (cursor2->parent)
444 n = collect_node(nodes, counts, n, cursor2->parent);
445
446 for (i = 0; i < n; ++i)
447 hammer_lock_downgrade(&nodes[i]->lock, counts[i]);
448 }
449
450 /*
451 * Seek the cursor to the specified node and index.
452 *
453 * The caller must ref the node prior to calling this routine and release
454 * it after it returns. If the seek succeeds the cursor will gain its own
455 * ref on the node.
456 */
457 int
hammer_cursor_seek(hammer_cursor_t cursor,hammer_node_t node,int index)458 hammer_cursor_seek(hammer_cursor_t cursor, hammer_node_t node, int index)
459 {
460 int error;
461
462 hammer_cursor_downgrade(cursor);
463 error = 0;
464
465 if (cursor->node != node) {
466 hammer_unlock(&cursor->node->lock);
467 hammer_rel_node(cursor->node);
468 cursor->node = node;
469 hammer_ref_node(node);
470 hammer_lock_sh(&node->lock);
471 KKASSERT ((node->flags & HAMMER_NODE_DELETED) == 0);
472
473 if (cursor->parent) {
474 hammer_unlock(&cursor->parent->lock);
475 hammer_rel_node(cursor->parent);
476 cursor->parent = NULL;
477 cursor->parent_index = 0;
478 }
479 error = hammer_load_cursor_parent(cursor, 0);
480 }
481 cursor->index = index;
482 return (error);
483 }
484
485 /*
486 * Load the parent of cursor->node into cursor->parent.
487 */
488 static
489 int
hammer_load_cursor_parent(hammer_cursor_t cursor,int try_exclusive)490 hammer_load_cursor_parent(hammer_cursor_t cursor, int try_exclusive)
491 {
492 hammer_mount_t hmp;
493 hammer_node_t parent;
494 hammer_node_t node;
495 hammer_btree_elm_t elm;
496 int error;
497 int parent_index;
498
499 hmp = cursor->trans->hmp;
500
501 if (cursor->node->ondisk->parent) {
502 node = cursor->node;
503 parent = hammer_btree_get_parent(cursor->trans, node,
504 &parent_index,
505 &error, try_exclusive);
506 if (error == 0) {
507 elm = &parent->ondisk->elms[parent_index];
508 cursor->parent = parent;
509 cursor->parent_index = parent_index;
510 cursor->left_bound = &elm[0].internal.base;
511 cursor->right_bound = &elm[1].internal.base;
512 }
513 } else {
514 cursor->parent = NULL;
515 cursor->parent_index = 0;
516 cursor->left_bound = &hmp->root_btree_beg;
517 cursor->right_bound = &hmp->root_btree_end;
518 error = 0;
519 }
520 return(error);
521 }
522
523 /*
524 * Cursor up to our parent node. Return ENOENT if we are at the root of
525 * the filesystem.
526 */
527 int
hammer_cursor_up(hammer_cursor_t cursor)528 hammer_cursor_up(hammer_cursor_t cursor)
529 {
530 int error;
531
532 hammer_cursor_downgrade(cursor);
533
534 /*
535 * If the parent is NULL we are at the root of the B-Tree and
536 * return ENOENT.
537 */
538 if (cursor->parent == NULL)
539 return (ENOENT);
540
541 /*
542 * Set the node to its parent.
543 */
544 hammer_unlock(&cursor->node->lock);
545 hammer_rel_node(cursor->node);
546 cursor->node = cursor->parent;
547 cursor->index = cursor->parent_index;
548 cursor->parent = NULL;
549 cursor->parent_index = 0;
550
551 error = hammer_load_cursor_parent(cursor, 0);
552 return(error);
553 }
554
555 /*
556 * Special cursor up given a locked cursor. The orignal node is not
557 * unlocked or released and the cursor is not downgraded.
558 *
559 * This function can fail with EDEADLK.
560 *
561 * This function is only run when recursively deleting parent nodes
562 * to get rid of an empty leaf.
563 */
564 int
hammer_cursor_up_locked(hammer_cursor_t cursor)565 hammer_cursor_up_locked(hammer_cursor_t cursor)
566 {
567 hammer_node_t save;
568 int error;
569 int save_index;
570
571 /*
572 * If the parent is NULL we are at the root of the B-Tree and
573 * return ENOENT.
574 */
575 if (cursor->parent == NULL)
576 return (ENOENT);
577
578 save = cursor->node;
579 save_index = cursor->index;
580
581 /*
582 * Set the node to its parent.
583 */
584 cursor->node = cursor->parent;
585 cursor->index = cursor->parent_index;
586 cursor->parent = NULL;
587 cursor->parent_index = 0;
588
589 /*
590 * load the new parent, attempt to exclusively lock it. Note that
591 * we are still holding the old parent (now cursor->node) exclusively
592 * locked.
593 *
594 * This can return EDEADLK. Undo the operation on any error. These
595 * up sequences can occur during iterations so be sure to restore
596 * the index.
597 */
598 error = hammer_load_cursor_parent(cursor, 1);
599 if (error) {
600 cursor->parent = cursor->node;
601 cursor->parent_index = cursor->index;
602 cursor->node = save;
603 cursor->index = save_index;
604 }
605 return(error);
606 }
607
608
609 /*
610 * Cursor down through the current node, which must be an internal node.
611 *
612 * This routine adjusts the cursor and sets index to 0.
613 */
614 int
hammer_cursor_down(hammer_cursor_t cursor)615 hammer_cursor_down(hammer_cursor_t cursor)
616 {
617 hammer_node_t node;
618 hammer_btree_elm_t elm;
619 int error;
620
621 /*
622 * The current node becomes the current parent
623 */
624 hammer_cursor_downgrade(cursor);
625 node = cursor->node;
626 KKASSERT(cursor->index >= 0 && cursor->index < node->ondisk->count);
627 if (cursor->parent) {
628 hammer_unlock(&cursor->parent->lock);
629 hammer_rel_node(cursor->parent);
630 }
631 cursor->parent = node;
632 cursor->parent_index = cursor->index;
633 cursor->node = NULL;
634 cursor->index = 0;
635
636 /*
637 * Extract element to push into at (node,index), set bounds.
638 */
639 elm = &node->ondisk->elms[cursor->parent_index];
640
641 /*
642 * Ok, push down into elm of an internal node.
643 */
644 KKASSERT(node->ondisk->type == HAMMER_BTREE_TYPE_INTERNAL);
645 KKASSERT(elm->internal.subtree_offset != 0);
646 cursor->left_bound = &elm[0].internal.base;
647 cursor->right_bound = &elm[1].internal.base;
648 node = hammer_get_node(cursor->trans,
649 elm->internal.subtree_offset, 0, &error);
650 if (error == 0) {
651 KASSERT(elm->base.btype == node->ondisk->type,
652 ("BTYPE MISMATCH %c %c NODE %p",
653 elm->base.btype, node->ondisk->type, node));
654 if (node->ondisk->parent != cursor->parent->node_offset)
655 hpanic("node %p %016jx vs %016jx",
656 node,
657 (intmax_t)node->ondisk->parent,
658 (intmax_t)cursor->parent->node_offset);
659 KKASSERT(node->ondisk->parent == cursor->parent->node_offset);
660 }
661
662 /*
663 * If no error occured we can lock the new child node. If the
664 * node is deadlock flagged wait up to hammer_tdmux_ticks (1 second)
665 * for the deadlock to clear. Otherwise a large number of concurrent
666 * readers can continuously stall the flusher.
667 *
668 * We specifically do this in the cursor_down() code in order to
669 * deal with frontend top-down searches smashing against bottom-up
670 * flusher-based mirror updates. These collisions typically occur
671 * above the inode in the B-Tree and are not covered by the
672 * ip->cursor_exclreq_count logic.
673 */
674 if (error == 0) {
675 #if 0
676 if (node->cursor_exclreq_count &&
677 cursor->trans->type != HAMMER_TRANS_FLS) {
678 tsleep(&node->cursor_exclreq_count, 0,
679 "hstag2", hammer_tdmux_ticks);
680 }
681 #endif
682 hammer_lock_sh(&node->lock);
683 KKASSERT ((node->flags & HAMMER_NODE_DELETED) == 0);
684 cursor->node = node;
685 cursor->index = 0;
686 }
687 return(error);
688 }
689
690 /************************************************************************
691 * DEADLOCK RECOVERY *
692 ************************************************************************
693 *
694 * These are the new deadlock recovery functions. Currently they are only
695 * used for the mirror propagation and physical node removal cases but
696 * ultimately the intention is to use them for all deadlock recovery
697 * operations.
698 *
699 * WARNING! The contents of the cursor may be modified while unlocked.
700 * passive modifications including adjusting the node, parent,
701 * indexes, and leaf pointer.
702 *
703 * An outright removal of the element the cursor was pointing at
704 * will cause the HAMMER_CURSOR_TRACKED_RIPOUT flag to be set,
705 * which chains to causing the HAMMER_CURSOR_RETEST to be set
706 * when the cursor is locked again.
707 */
708 void
hammer_unlock_cursor(hammer_cursor_t cursor)709 hammer_unlock_cursor(hammer_cursor_t cursor)
710 {
711 hammer_node_t node;
712
713 KKASSERT((cursor->flags & HAMMER_CURSOR_TRACKED) == 0);
714 KKASSERT(cursor->node);
715
716 /*
717 * Release the cursor's locks and track B-Tree operations on node.
718 * While being tracked our cursor can be modified by other threads
719 * and the node may be replaced.
720 */
721 if (cursor->parent) {
722 hammer_unlock(&cursor->parent->lock);
723 hammer_rel_node(cursor->parent);
724 cursor->parent = NULL;
725 }
726 node = cursor->node;
727 cursor->flags |= HAMMER_CURSOR_TRACKED;
728 TAILQ_INSERT_TAIL(&node->cursor_list, cursor, deadlk_entry);
729 hammer_unlock(&node->lock);
730 }
731
732 /*
733 * Get the cursor heated up again. The cursor's node may have
734 * changed and we might have to locate the new parent.
735 *
736 * If the exact element we were on got deleted RIPOUT will be
737 * set and we must clear ATEDISK so an iteration does not skip
738 * the element after it.
739 */
740 int
hammer_lock_cursor(hammer_cursor_t cursor)741 hammer_lock_cursor(hammer_cursor_t cursor)
742 {
743 hammer_node_t node;
744 int error;
745
746 KKASSERT(cursor->flags & HAMMER_CURSOR_TRACKED);
747
748 /*
749 * Relock the node
750 */
751 for (;;) {
752 node = cursor->node;
753 hammer_ref_node(node);
754 hammer_lock_sh(&node->lock);
755 if (cursor->node == node) {
756 hammer_rel_node(node);
757 break;
758 }
759 hammer_unlock(&node->lock);
760 hammer_rel_node(node);
761 }
762
763 /*
764 * Untrack the cursor, clean up, and re-establish the parent node.
765 */
766 TAILQ_REMOVE(&node->cursor_list, cursor, deadlk_entry);
767 cursor->flags &= ~HAMMER_CURSOR_TRACKED;
768
769 /*
770 * If a ripout has occured iterations must re-test the (new)
771 * current element. Clearing ATEDISK prevents the element from
772 * being skipped and RETEST causes it to be re-tested.
773 */
774 if (cursor->flags & HAMMER_CURSOR_TRACKED_RIPOUT) {
775 cursor->flags &= ~HAMMER_CURSOR_TRACKED_RIPOUT;
776 cursor->flags &= ~HAMMER_CURSOR_ATEDISK;
777 cursor->flags |= HAMMER_CURSOR_RETEST;
778 }
779 error = hammer_load_cursor_parent(cursor, 0);
780 return(error);
781 }
782
783 /*
784 * Recover from a deadlocked cursor, tracking any node removals or
785 * replacements. If the cursor's current node is removed by another
786 * thread (via btree_remove()) the cursor will be seeked upwards.
787 *
788 * The caller is working a modifying operation and must be holding the
789 * sync lock (shared). We do not release the sync lock because this
790 * would break atomicy.
791 */
792 int
hammer_recover_cursor(hammer_cursor_t cursor)793 hammer_recover_cursor(hammer_cursor_t cursor)
794 {
795 hammer_transaction_t trans __debugvar;
796 #if 0
797 hammer_inode_t ip;
798 #endif
799 int error;
800
801 hammer_unlock_cursor(cursor);
802
803 #if 0
804 ip = cursor->ip;
805 #endif
806 trans = cursor->trans;
807 KKASSERT(trans->sync_lock_refs > 0);
808
809 /*
810 * Wait for the deadlock to clear.
811 *
812 * Maintain exclreq_count / wakeup as necessary to notify new
813 * entrants into ip. We continue to hold the fs_token so our
814 * EDEADLK retry loop should get its chance before another thread
815 * steals the lock.
816 */
817 if (cursor->deadlk_node) {
818 #if 0
819 if (ip && trans->type == HAMMER_TRANS_FLS)
820 ++ip->cursor_exclreq_count;
821 ++cursor->deadlk_node->cursor_exclreq_count;
822 #endif
823 hammer_lock_ex_ident(&cursor->deadlk_node->lock, "hmrdlk");
824 hammer_unlock(&cursor->deadlk_node->lock);
825 #if 0
826 if (--cursor->deadlk_node->cursor_exclreq_count == 0)
827 wakeup(&cursor->deadlk_node->cursor_exclreq_count);
828 if (ip && trans->type == HAMMER_TRANS_FLS) {
829 if (--ip->cursor_exclreq_count == 0)
830 wakeup(&ip->cursor_exclreq_count);
831 }
832 #endif
833 hammer_rel_node(cursor->deadlk_node);
834 cursor->deadlk_node = NULL;
835 }
836 if (cursor->deadlk_rec) {
837 hammer_wait_mem_record_ident(cursor->deadlk_rec, "hmmdlr");
838 hammer_rel_mem_record(cursor->deadlk_rec);
839 cursor->deadlk_rec = NULL;
840 }
841 error = hammer_lock_cursor(cursor);
842 return(error);
843 }
844
845 /*
846 * Dup ocursor to ncursor. ncursor inherits ocursor's locks and ocursor
847 * is effectively unlocked and becomes tracked. If ocursor was not locked
848 * then ncursor also inherits the tracking.
849 *
850 * After the caller finishes working with ncursor it must be cleaned up
851 * with hammer_done_cursor(), and the caller must re-lock ocursor.
852 */
853 hammer_cursor_t
hammer_push_cursor(hammer_cursor_t ocursor)854 hammer_push_cursor(hammer_cursor_t ocursor)
855 {
856 hammer_cursor_t ncursor;
857 hammer_inode_t ip;
858 hammer_node_t node;
859 hammer_mount_t hmp;
860
861 hmp = ocursor->trans->hmp;
862 ncursor = kmalloc(sizeof(*ncursor), hmp->m_misc, M_WAITOK | M_ZERO);
863 bcopy(ocursor, ncursor, sizeof(*ocursor));
864
865 node = ocursor->node;
866 hammer_ref_node(node);
867 if ((ocursor->flags & HAMMER_CURSOR_TRACKED) == 0) {
868 ocursor->flags |= HAMMER_CURSOR_TRACKED;
869 TAILQ_INSERT_TAIL(&node->cursor_list, ocursor, deadlk_entry);
870 }
871 if (ncursor->parent)
872 ocursor->parent = NULL;
873 ocursor->data_buffer = NULL;
874 ocursor->leaf = NULL;
875 ocursor->data = NULL;
876 if (ncursor->flags & HAMMER_CURSOR_TRACKED)
877 TAILQ_INSERT_TAIL(&node->cursor_list, ncursor, deadlk_entry);
878 if ((ip = ncursor->ip) != NULL) {
879 ++ip->cursor_ip_refs;
880 }
881 if (ncursor->iprec)
882 hammer_ref(&ncursor->iprec->lock);
883 return(ncursor);
884 }
885
886 /*
887 * Destroy ncursor and restore ocursor
888 *
889 * This is a temporary hack for the release. We can't afford to lose
890 * the IP lock until the IP object scan code is able to deal with it,
891 * so have ocursor inherit it back.
892 */
893 void
hammer_pop_cursor(hammer_cursor_t ocursor,hammer_cursor_t ncursor)894 hammer_pop_cursor(hammer_cursor_t ocursor, hammer_cursor_t ncursor)
895 {
896 hammer_mount_t hmp;
897 hammer_inode_t ip;
898
899 hmp = ncursor->trans->hmp;
900 ip = ncursor->ip;
901 ncursor->ip = NULL;
902 if (ip)
903 --ip->cursor_ip_refs;
904 hammer_done_cursor(ncursor);
905 kfree(ncursor, hmp->m_misc);
906 KKASSERT(ocursor->ip == ip);
907 hammer_lock_cursor(ocursor);
908 }
909
910 /*
911 * onode is being replaced by nnode by the reblocking code.
912 */
913 void
hammer_cursor_replaced_node(hammer_node_t onode,hammer_node_t nnode)914 hammer_cursor_replaced_node(hammer_node_t onode, hammer_node_t nnode)
915 {
916 hammer_cursor_t cursor;
917 hammer_node_ondisk_t ondisk;
918 hammer_node_ondisk_t nndisk;
919
920 ondisk = onode->ondisk;
921 nndisk = nnode->ondisk;
922
923 while ((cursor = TAILQ_FIRST(&onode->cursor_list)) != NULL) {
924 TAILQ_REMOVE(&onode->cursor_list, cursor, deadlk_entry);
925 TAILQ_INSERT_TAIL(&nnode->cursor_list, cursor, deadlk_entry);
926 KKASSERT(cursor->node == onode);
927 if (cursor->leaf == &ondisk->elms[cursor->index].leaf)
928 cursor->leaf = &nndisk->elms[cursor->index].leaf;
929 cursor->node = nnode;
930 hammer_ref_node(nnode);
931 hammer_rel_node(onode);
932 }
933 }
934
935 /*
936 * We have removed <node> from the parent and collapsed the parent.
937 *
938 * Cursors in deadlock recovery are seeked upward to the parent so the
939 * btree_remove() recursion works properly even though we have marked
940 * the cursor as requiring a reseek.
941 *
942 * This is the only cursor function which sets HAMMER_CURSOR_ITERATE_CHECK,
943 * meaning the cursor is no longer definitively pointing at an element
944 * within its iteration (if the cursor is being used to iterate). The
945 * iteration code will take this into account instead of asserting if the
946 * cursor is outside the iteration range.
947 */
948 void
hammer_cursor_removed_node(hammer_node_t node,hammer_node_t parent,int index)949 hammer_cursor_removed_node(hammer_node_t node, hammer_node_t parent, int index)
950 {
951 hammer_cursor_t cursor;
952 hammer_node_ondisk_t ondisk;
953
954 KKASSERT(parent != NULL);
955 ondisk = node->ondisk;
956
957 while ((cursor = TAILQ_FIRST(&node->cursor_list)) != NULL) {
958 KKASSERT(cursor->node == node);
959 KKASSERT(cursor->index == 0);
960 TAILQ_REMOVE(&node->cursor_list, cursor, deadlk_entry);
961 TAILQ_INSERT_TAIL(&parent->cursor_list, cursor, deadlk_entry);
962 if (cursor->leaf == &ondisk->elms[cursor->index].leaf)
963 cursor->leaf = NULL;
964 cursor->flags |= HAMMER_CURSOR_TRACKED_RIPOUT;
965 cursor->flags |= HAMMER_CURSOR_ITERATE_CHECK;
966 cursor->node = parent;
967 cursor->index = index;
968 hammer_ref_node(parent);
969 hammer_rel_node(node);
970 }
971 }
972
973 /*
974 * node was split at (onode, index) with elements >= index moved to nnode.
975 */
976 void
hammer_cursor_split_node(hammer_node_t onode,hammer_node_t nnode,int index)977 hammer_cursor_split_node(hammer_node_t onode, hammer_node_t nnode, int index)
978 {
979 hammer_cursor_t cursor;
980 hammer_node_ondisk_t ondisk;
981 hammer_node_ondisk_t nndisk;
982
983 ondisk = onode->ondisk;
984 nndisk = nnode->ondisk;
985
986 again:
987 TAILQ_FOREACH(cursor, &onode->cursor_list, deadlk_entry) {
988 KKASSERT(cursor->node == onode);
989 if (cursor->index < index)
990 continue;
991 TAILQ_REMOVE(&onode->cursor_list, cursor, deadlk_entry);
992 TAILQ_INSERT_TAIL(&nnode->cursor_list, cursor, deadlk_entry);
993 if (cursor->leaf == &ondisk->elms[cursor->index].leaf)
994 cursor->leaf = &nndisk->elms[cursor->index - index].leaf;
995 cursor->node = nnode;
996 cursor->index -= index;
997 hammer_ref_node(nnode);
998 hammer_rel_node(onode);
999 goto again;
1000 }
1001 }
1002
1003 /*
1004 * An element was moved from one node to another or within a node. The
1005 * index may also represent the end of the node (index == numelements).
1006 *
1007 * {oparent,pindex} is the parent node's pointer to onode/oindex.
1008 *
1009 * This is used by the rebalancing code. This is not an insertion or
1010 * deletion and any additional elements, including the degenerate case at
1011 * the end of the node, will be dealt with by additional distinct calls.
1012 */
1013 void
hammer_cursor_moved_element(hammer_node_t oparent,int pindex,hammer_node_t onode,int oindex,hammer_node_t nnode,int nindex)1014 hammer_cursor_moved_element(hammer_node_t oparent, int pindex,
1015 hammer_node_t onode, int oindex,
1016 hammer_node_t nnode, int nindex)
1017 {
1018 hammer_cursor_t cursor;
1019 hammer_node_ondisk_t ondisk;
1020 hammer_node_ondisk_t nndisk;
1021
1022 /*
1023 * Adjust any cursors pointing at the element
1024 */
1025 ondisk = onode->ondisk;
1026 nndisk = nnode->ondisk;
1027 again1:
1028 TAILQ_FOREACH(cursor, &onode->cursor_list, deadlk_entry) {
1029 KKASSERT(cursor->node == onode);
1030 if (cursor->index != oindex)
1031 continue;
1032 TAILQ_REMOVE(&onode->cursor_list, cursor, deadlk_entry);
1033 TAILQ_INSERT_TAIL(&nnode->cursor_list, cursor, deadlk_entry);
1034 if (cursor->leaf == &ondisk->elms[oindex].leaf)
1035 cursor->leaf = &nndisk->elms[nindex].leaf;
1036 cursor->node = nnode;
1037 cursor->index = nindex;
1038 hammer_ref_node(nnode);
1039 hammer_rel_node(onode);
1040 goto again1;
1041 }
1042
1043 /*
1044 * When moving the first element of onode to a different node any
1045 * cursor which is pointing at (oparent,pindex) must be repointed
1046 * to nnode and ATEDISK must be cleared.
1047 *
1048 * This prevents cursors from losing track due to insertions.
1049 * Insertions temporarily release the cursor in order to update
1050 * the mirror_tids. It primarily effects the mirror_write code.
1051 * The other code paths generally only do a single insertion and
1052 * then relookup or drop the cursor.
1053 */
1054 if (onode == nnode || oindex)
1055 return;
1056 ondisk = oparent->ondisk;
1057 again2:
1058 TAILQ_FOREACH(cursor, &oparent->cursor_list, deadlk_entry) {
1059 KKASSERT(cursor->node == oparent);
1060 if (cursor->index != pindex)
1061 continue;
1062 hkprintf("debug: shifted cursor pointing at parent\n"
1063 "parent %016jx:%d onode %016jx:%d nnode %016jx:%d\n",
1064 (intmax_t)oparent->node_offset, pindex,
1065 (intmax_t)onode->node_offset, oindex,
1066 (intmax_t)nnode->node_offset, nindex);
1067 TAILQ_REMOVE(&oparent->cursor_list, cursor, deadlk_entry);
1068 TAILQ_INSERT_TAIL(&nnode->cursor_list, cursor, deadlk_entry);
1069 if (cursor->leaf == &ondisk->elms[oindex].leaf)
1070 cursor->leaf = &nndisk->elms[nindex].leaf;
1071 cursor->node = nnode;
1072 cursor->index = nindex;
1073 cursor->flags &= ~HAMMER_CURSOR_ATEDISK;
1074 hammer_ref_node(nnode);
1075 hammer_rel_node(oparent);
1076 goto again2;
1077 }
1078 }
1079
1080 /*
1081 * The B-Tree element pointing to the specified node was moved from (oparent)
1082 * to (nparent, nindex). We must locate any tracked cursors pointing at
1083 * node and adjust their parent accordingly.
1084 *
1085 * This is used by the rebalancing code when packing elements causes an
1086 * element to shift from one node to another.
1087 */
1088 void
hammer_cursor_parent_changed(hammer_node_t node,hammer_node_t oparent,hammer_node_t nparent,int nindex)1089 hammer_cursor_parent_changed(hammer_node_t node, hammer_node_t oparent,
1090 hammer_node_t nparent, int nindex)
1091 {
1092 hammer_cursor_t cursor;
1093
1094 again:
1095 TAILQ_FOREACH(cursor, &node->cursor_list, deadlk_entry) {
1096 KKASSERT(cursor->node == node);
1097 if (cursor->parent == oparent) {
1098 cursor->parent = nparent;
1099 cursor->parent_index = nindex;
1100 hammer_ref_node(nparent);
1101 hammer_rel_node(oparent);
1102 goto again;
1103 }
1104 }
1105 }
1106
1107 /*
1108 * Deleted element at (node, index)
1109 *
1110 * Shift indexes >= index
1111 */
1112 void
hammer_cursor_deleted_element(hammer_node_t node,int index)1113 hammer_cursor_deleted_element(hammer_node_t node, int index)
1114 {
1115 hammer_cursor_t cursor;
1116 hammer_node_ondisk_t ondisk;
1117
1118 ondisk = node->ondisk;
1119
1120 TAILQ_FOREACH(cursor, &node->cursor_list, deadlk_entry) {
1121 KKASSERT(cursor->node == node);
1122 if (cursor->index == index) {
1123 cursor->flags |= HAMMER_CURSOR_TRACKED_RIPOUT;
1124 if (cursor->leaf == &ondisk->elms[cursor->index].leaf)
1125 cursor->leaf = NULL;
1126 } else if (cursor->index > index) {
1127 if (cursor->leaf == &ondisk->elms[cursor->index].leaf)
1128 cursor->leaf = &ondisk->elms[cursor->index - 1].leaf;
1129 --cursor->index;
1130 }
1131 }
1132 }
1133
1134 /*
1135 * Inserted element at (node, index)
1136 *
1137 * Shift indexes >= index
1138 */
1139 void
hammer_cursor_inserted_element(hammer_node_t node,int index)1140 hammer_cursor_inserted_element(hammer_node_t node, int index)
1141 {
1142 hammer_cursor_t cursor;
1143 hammer_node_ondisk_t ondisk;
1144
1145 ondisk = node->ondisk;
1146
1147 TAILQ_FOREACH(cursor, &node->cursor_list, deadlk_entry) {
1148 KKASSERT(cursor->node == node);
1149 if (cursor->index >= index) {
1150 if (cursor->leaf == &ondisk->elms[cursor->index].leaf)
1151 cursor->leaf = &ondisk->elms[cursor->index + 1].leaf;
1152 ++cursor->index;
1153 }
1154 }
1155 }
1156
1157 /*
1158 * Invalidate the cached data buffer associated with a cursor.
1159 *
1160 * This needs to be done when the underlying block is being freed or
1161 * the referenced buffer can prevent the related buffer cache buffer
1162 * from being properly invalidated.
1163 */
1164 void
hammer_cursor_invalidate_cache(hammer_cursor_t cursor)1165 hammer_cursor_invalidate_cache(hammer_cursor_t cursor)
1166 {
1167 if (cursor->data_buffer) {
1168 hammer_rel_buffer(cursor->data_buffer, 0);
1169 cursor->data_buffer = NULL;
1170 cursor->data = NULL;
1171 }
1172 }
1173
1174