commit.c 20.6 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
/*
 * This file is part of UBIFS.
 *
 * Copyright (C) 2006-2008 Nokia Corporation.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License version 2 as published by
 * the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 * more details.
 *
 * You should have received a copy of the GNU General Public License along with
 * this program; if not, write to the Free Software Foundation, Inc., 51
 * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
 *
 * Authors: Adrian Hunter
 *          Artem Bityutskiy (Битюцкий Артём)
 */

/*
 * This file implements functions that manage the running of the commit process.
 * Each affected module has its own functions to accomplish their part in the
 * commit and those functions are called here.
 *
 * The commit is the process whereby all updates to the index and LEB properties
 * are written out together and the journal becomes empty. This keeps the
 * file system consistent - at all times the state can be recreated by reading
 * the index and LEB properties and then replaying the journal.
 *
 * The commit is split into two parts named "commit start" and "commit end".
 * During commit start, the commit process has exclusive access to the journal
 * by holding the commit semaphore down for writing. As few I/O operations as
 * possible are performed during commit start, instead the nodes that are to be
 * written are merely identified. During commit end, the commit semaphore is no
 * longer held and the journal is again in operation, allowing users to continue
 * to use the file system while the bulk of the commit I/O is performed. The
 * purpose of this two-step approach is to prevent the commit from causing any
 * latency blips. Note that in any case, the commit does not prevent lookups
 * (as permitted by the TNC mutex), or access to VFS data structures e.g. page
 * cache.
 */

#include <linux/freezer.h>
#include <linux/kthread.h>
48
#include <linux/slab.h>
49 50
#include "ubifs.h"

51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100
/*
 * nothing_to_commit - check if there is nothing to commit.
 * @c: UBIFS file-system description object
 *
 * This is a helper function which checks if there is anything to commit. It is
 * used as an optimization to avoid starting the commit if it is not really
 * necessary. Indeed, the commit operation always assumes flash I/O (e.g.,
 * writing the commit start node to the log), and it is better to avoid doing
 * this unnecessarily. E.g., 'ubifs_sync_fs()' runs the commit, but if there is
 * nothing to commit, it is more optimal to avoid any flash I/O.
 *
 * This function has to be called with @c->commit_sem locked for writing -
 * this function does not take LPT/TNC locks because the @c->commit_sem
 * guarantees that we have exclusive access to the TNC and LPT data structures.
 *
 * This function returns %1 if there is nothing to commit and %0 otherwise.
 */
static int nothing_to_commit(struct ubifs_info *c)
{
	/*
	 * During mounting or remounting from R/O mode to R/W mode we may
	 * commit for various recovery-related reasons.
	 */
	if (c->mounting || c->remounting_rw)
		return 0;

	/*
	 * If the root TNC node is dirty, we definitely have something to
	 * commit.
	 */
	if (c->zroot.znode && test_bit(DIRTY_ZNODE, &c->zroot.znode->flags))
		return 0;

	/*
	 * Even though the TNC is clean, the LPT tree may have dirty nodes. For
	 * example, this may happen if the budgeting subsystem invoked GC to
	 * make some free space, and the GC found an LEB with only dirty and
	 * free space. In this case GC would just change the lprops of this
	 * LEB (by turning all space into free space) and unmap it.
	 */
	if (c->nroot && test_bit(DIRTY_CNODE, &c->nroot->flags))
		return 0;

	ubifs_assert(atomic_long_read(&c->dirty_zn_cnt) == 0);
	ubifs_assert(c->dirty_pn_cnt == 0);
	ubifs_assert(c->dirty_nn_cnt == 0);

	return 1;
}

101 102 103 104 105 106 107 108 109 110 111 112 113 114 115
/**
 * do_commit - commit the journal.
 * @c: UBIFS file-system description object
 *
 * This function implements UBIFS commit. It has to be called with commit lock
 * locked. Returns zero in case of success and a negative error code in case of
 * failure.
 */
static int do_commit(struct ubifs_info *c)
{
	int err, new_ltail_lnum, old_ltail_lnum, i;
	struct ubifs_zbranch zroot;
	struct ubifs_lp_stats lst;

	dbg_cmt("start");
116
	ubifs_assert(!c->ro_media && !c->ro_mount);
117 118

	if (c->ro_error) {
119 120 121 122
		err = -EROFS;
		goto out_up;
	}

123 124 125 126 127 128
	if (nothing_to_commit(c)) {
		up_write(&c->commit_sem);
		err = 0;
		goto out_cancel;
	}

129 130 131 132 133 134 135
	/* Sync all write buffers (necessary for recovery) */
	for (i = 0; i < c->jhead_cnt; i++) {
		err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
		if (err)
			goto out_up;
	}

136
	c->cmt_no += 1;
137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
	err = ubifs_gc_start_commit(c);
	if (err)
		goto out_up;
	err = dbg_check_lprops(c);
	if (err)
		goto out_up;
	err = ubifs_log_start_commit(c, &new_ltail_lnum);
	if (err)
		goto out_up;
	err = ubifs_tnc_start_commit(c, &zroot);
	if (err)
		goto out_up;
	err = ubifs_lpt_start_commit(c);
	if (err)
		goto out_up;
	err = ubifs_orphan_start_commit(c);
	if (err)
		goto out_up;

	ubifs_get_lp_stats(c, &lst);

	up_write(&c->commit_sem);

	err = ubifs_tnc_end_commit(c);
	if (err)
		goto out;
	err = ubifs_lpt_end_commit(c);
	if (err)
		goto out;
	err = ubifs_orphan_end_commit(c);
	if (err)
		goto out;
	old_ltail_lnum = c->ltail_lnum;
	err = ubifs_log_end_commit(c, new_ltail_lnum);
	if (err)
		goto out;
	err = dbg_check_old_index(c, &zroot);
	if (err)
		goto out;

	mutex_lock(&c->mst_mutex);
178
	c->mst_node->cmt_no      = cpu_to_le64(c->cmt_no);
179 180 181 182 183 184
	c->mst_node->log_lnum    = cpu_to_le32(new_ltail_lnum);
	c->mst_node->root_lnum   = cpu_to_le32(zroot.lnum);
	c->mst_node->root_offs   = cpu_to_le32(zroot.offs);
	c->mst_node->root_len    = cpu_to_le32(zroot.len);
	c->mst_node->ihead_lnum  = cpu_to_le32(c->ihead_lnum);
	c->mst_node->ihead_offs  = cpu_to_le32(c->ihead_offs);
185
	c->mst_node->index_size  = cpu_to_le64(c->bi.old_idx_sz);
186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220
	c->mst_node->lpt_lnum    = cpu_to_le32(c->lpt_lnum);
	c->mst_node->lpt_offs    = cpu_to_le32(c->lpt_offs);
	c->mst_node->nhead_lnum  = cpu_to_le32(c->nhead_lnum);
	c->mst_node->nhead_offs  = cpu_to_le32(c->nhead_offs);
	c->mst_node->ltab_lnum   = cpu_to_le32(c->ltab_lnum);
	c->mst_node->ltab_offs   = cpu_to_le32(c->ltab_offs);
	c->mst_node->lsave_lnum  = cpu_to_le32(c->lsave_lnum);
	c->mst_node->lsave_offs  = cpu_to_le32(c->lsave_offs);
	c->mst_node->lscan_lnum  = cpu_to_le32(c->lscan_lnum);
	c->mst_node->empty_lebs  = cpu_to_le32(lst.empty_lebs);
	c->mst_node->idx_lebs    = cpu_to_le32(lst.idx_lebs);
	c->mst_node->total_free  = cpu_to_le64(lst.total_free);
	c->mst_node->total_dirty = cpu_to_le64(lst.total_dirty);
	c->mst_node->total_used  = cpu_to_le64(lst.total_used);
	c->mst_node->total_dead  = cpu_to_le64(lst.total_dead);
	c->mst_node->total_dark  = cpu_to_le64(lst.total_dark);
	if (c->no_orphs)
		c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
	else
		c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_NO_ORPHS);
	err = ubifs_write_master(c);
	mutex_unlock(&c->mst_mutex);
	if (err)
		goto out;

	err = ubifs_log_post_commit(c, old_ltail_lnum);
	if (err)
		goto out;
	err = ubifs_gc_end_commit(c);
	if (err)
		goto out;
	err = ubifs_lpt_post_commit(c);
	if (err)
		goto out;

221
out_cancel:
222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295
	spin_lock(&c->cs_lock);
	c->cmt_state = COMMIT_RESTING;
	wake_up(&c->cmt_wq);
	dbg_cmt("commit end");
	spin_unlock(&c->cs_lock);
	return 0;

out_up:
	up_write(&c->commit_sem);
out:
	ubifs_err("commit failed, error %d", err);
	spin_lock(&c->cs_lock);
	c->cmt_state = COMMIT_BROKEN;
	wake_up(&c->cmt_wq);
	spin_unlock(&c->cs_lock);
	ubifs_ro_mode(c, err);
	return err;
}

/**
 * run_bg_commit - run background commit if it is needed.
 * @c: UBIFS file-system description object
 *
 * This function runs background commit if it is needed. Returns zero in case
 * of success and a negative error code in case of failure.
 */
static int run_bg_commit(struct ubifs_info *c)
{
	spin_lock(&c->cs_lock);
	/*
	 * Run background commit only if background commit was requested or if
	 * commit is required.
	 */
	if (c->cmt_state != COMMIT_BACKGROUND &&
	    c->cmt_state != COMMIT_REQUIRED)
		goto out;
	spin_unlock(&c->cs_lock);

	down_write(&c->commit_sem);
	spin_lock(&c->cs_lock);
	if (c->cmt_state == COMMIT_REQUIRED)
		c->cmt_state = COMMIT_RUNNING_REQUIRED;
	else if (c->cmt_state == COMMIT_BACKGROUND)
		c->cmt_state = COMMIT_RUNNING_BACKGROUND;
	else
		goto out_cmt_unlock;
	spin_unlock(&c->cs_lock);

	return do_commit(c);

out_cmt_unlock:
	up_write(&c->commit_sem);
out:
	spin_unlock(&c->cs_lock);
	return 0;
}

/**
 * ubifs_bg_thread - UBIFS background thread function.
 * @info: points to the file-system description object
 *
 * This function implements various file-system background activities:
 * o when a write-buffer timer expires it synchronizes the appropriate
 *   write-buffer;
 * o when the journal is about to be full, it starts in-advance commit.
 *
 * Note, other stuff like background garbage collection may be added here in
 * future.
 */
int ubifs_bg_thread(void *info)
{
	int err;
	struct ubifs_info *c = info;

A
Artem Bityutskiy 已提交
296 297
	dbg_msg("background thread \"%s\" started, PID %d",
		c->bgt_name, current->pid);
298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420
	set_freezable();

	while (1) {
		if (kthread_should_stop())
			break;

		if (try_to_freeze())
			continue;

		set_current_state(TASK_INTERRUPTIBLE);
		/* Check if there is something to do */
		if (!c->need_bgt) {
			/*
			 * Nothing prevents us from going sleep now and
			 * be never woken up and block the task which
			 * could wait in 'kthread_stop()' forever.
			 */
			if (kthread_should_stop())
				break;
			schedule();
			continue;
		} else
			__set_current_state(TASK_RUNNING);

		c->need_bgt = 0;
		err = ubifs_bg_wbufs_sync(c);
		if (err)
			ubifs_ro_mode(c, err);

		run_bg_commit(c);
		cond_resched();
	}

	dbg_msg("background thread \"%s\" stops", c->bgt_name);
	return 0;
}

/**
 * ubifs_commit_required - set commit state to "required".
 * @c: UBIFS file-system description object
 *
 * This function is called if a commit is required but cannot be done from the
 * calling function, so it is just flagged instead.
 */
void ubifs_commit_required(struct ubifs_info *c)
{
	spin_lock(&c->cs_lock);
	switch (c->cmt_state) {
	case COMMIT_RESTING:
	case COMMIT_BACKGROUND:
		dbg_cmt("old: %s, new: %s", dbg_cstate(c->cmt_state),
			dbg_cstate(COMMIT_REQUIRED));
		c->cmt_state = COMMIT_REQUIRED;
		break;
	case COMMIT_RUNNING_BACKGROUND:
		dbg_cmt("old: %s, new: %s", dbg_cstate(c->cmt_state),
			dbg_cstate(COMMIT_RUNNING_REQUIRED));
		c->cmt_state = COMMIT_RUNNING_REQUIRED;
		break;
	case COMMIT_REQUIRED:
	case COMMIT_RUNNING_REQUIRED:
	case COMMIT_BROKEN:
		break;
	}
	spin_unlock(&c->cs_lock);
}

/**
 * ubifs_request_bg_commit - notify the background thread to do a commit.
 * @c: UBIFS file-system description object
 *
 * This function is called if the journal is full enough to make a commit
 * worthwhile, so background thread is kicked to start it.
 */
void ubifs_request_bg_commit(struct ubifs_info *c)
{
	spin_lock(&c->cs_lock);
	if (c->cmt_state == COMMIT_RESTING) {
		dbg_cmt("old: %s, new: %s", dbg_cstate(c->cmt_state),
			dbg_cstate(COMMIT_BACKGROUND));
		c->cmt_state = COMMIT_BACKGROUND;
		spin_unlock(&c->cs_lock);
		ubifs_wake_up_bgt(c);
	} else
		spin_unlock(&c->cs_lock);
}

/**
 * wait_for_commit - wait for commit.
 * @c: UBIFS file-system description object
 *
 * This function sleeps until the commit operation is no longer running.
 */
static int wait_for_commit(struct ubifs_info *c)
{
	dbg_cmt("pid %d goes sleep", current->pid);

	/*
	 * The following sleeps if the condition is false, and will be woken
	 * when the commit ends. It is possible, although very unlikely, that we
	 * will wake up and see the subsequent commit running, rather than the
	 * one we were waiting for, and go back to sleep.  However, we will be
	 * woken again, so there is no danger of sleeping forever.
	 */
	wait_event(c->cmt_wq, c->cmt_state != COMMIT_RUNNING_BACKGROUND &&
			      c->cmt_state != COMMIT_RUNNING_REQUIRED);
	dbg_cmt("commit finished, pid %d woke up", current->pid);
	return 0;
}

/**
 * ubifs_run_commit - run or wait for commit.
 * @c: UBIFS file-system description object
 *
 * This function runs commit and returns zero in case of success and a negative
 * error code in case of failure.
 */
int ubifs_run_commit(struct ubifs_info *c)
{
	int err = 0;

	spin_lock(&c->cs_lock);
	if (c->cmt_state == COMMIT_BROKEN) {
421
		err = -EROFS;
422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446
		goto out;
	}

	if (c->cmt_state == COMMIT_RUNNING_BACKGROUND)
		/*
		 * We set the commit state to 'running required' to indicate
		 * that we want it to complete as quickly as possible.
		 */
		c->cmt_state = COMMIT_RUNNING_REQUIRED;

	if (c->cmt_state == COMMIT_RUNNING_REQUIRED) {
		spin_unlock(&c->cs_lock);
		return wait_for_commit(c);
	}
	spin_unlock(&c->cs_lock);

	/* Ok, the commit is indeed needed */

	down_write(&c->commit_sem);
	spin_lock(&c->cs_lock);
	/*
	 * Since we unlocked 'c->cs_lock', the state may have changed, so
	 * re-check it.
	 */
	if (c->cmt_state == COMMIT_BROKEN) {
447
		err = -EROFS;
448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531
		goto out_cmt_unlock;
	}

	if (c->cmt_state == COMMIT_RUNNING_BACKGROUND)
		c->cmt_state = COMMIT_RUNNING_REQUIRED;

	if (c->cmt_state == COMMIT_RUNNING_REQUIRED) {
		up_write(&c->commit_sem);
		spin_unlock(&c->cs_lock);
		return wait_for_commit(c);
	}
	c->cmt_state = COMMIT_RUNNING_REQUIRED;
	spin_unlock(&c->cs_lock);

	err = do_commit(c);
	return err;

out_cmt_unlock:
	up_write(&c->commit_sem);
out:
	spin_unlock(&c->cs_lock);
	return err;
}

/**
 * ubifs_gc_should_commit - determine if it is time for GC to run commit.
 * @c: UBIFS file-system description object
 *
 * This function is called by garbage collection to determine if commit should
 * be run. If commit state is @COMMIT_BACKGROUND, which means that the journal
 * is full enough to start commit, this function returns true. It is not
 * absolutely necessary to commit yet, but it feels like this should be better
 * then to keep doing GC. This function returns %1 if GC has to initiate commit
 * and %0 if not.
 */
int ubifs_gc_should_commit(struct ubifs_info *c)
{
	int ret = 0;

	spin_lock(&c->cs_lock);
	if (c->cmt_state == COMMIT_BACKGROUND) {
		dbg_cmt("commit required now");
		c->cmt_state = COMMIT_REQUIRED;
	} else
		dbg_cmt("commit not requested");
	if (c->cmt_state == COMMIT_REQUIRED)
		ret = 1;
	spin_unlock(&c->cs_lock);
	return ret;
}

#ifdef CONFIG_UBIFS_FS_DEBUG

/**
 * struct idx_node - hold index nodes during index tree traversal.
 * @list: list
 * @iip: index in parent (slot number of this indexing node in the parent
 *       indexing node)
 * @upper_key: all keys in this indexing node have to be less or equivalent to
 *             this key
 * @idx: index node (8-byte aligned because all node structures must be 8-byte
 *       aligned)
 */
struct idx_node {
	struct list_head list;
	int iip;
	union ubifs_key upper_key;
	struct ubifs_idx_node idx __attribute__((aligned(8)));
};

/**
 * dbg_old_index_check_init - get information for the next old index check.
 * @c: UBIFS file-system description object
 * @zroot: root of the index
 *
 * This function records information about the index that will be needed for the
 * next old index check i.e. 'dbg_check_old_index()'.
 *
 * This function returns %0 on success and a negative error code on failure.
 */
int dbg_old_index_check_init(struct ubifs_info *c, struct ubifs_zbranch *zroot)
{
	struct ubifs_idx_node *idx;
	int lnum, offs, len, err = 0;
532
	struct ubifs_debug_info *d = c->dbg;
533

534 535 536 537
	d->old_zroot = *zroot;
	lnum = d->old_zroot.lnum;
	offs = d->old_zroot.offs;
	len = d->old_zroot.len;
538 539 540 541 542 543 544 545 546

	idx = kmalloc(c->max_idx_node_sz, GFP_NOFS);
	if (!idx)
		return -ENOMEM;

	err = ubifs_read_node(c, idx, UBIFS_IDX_NODE, len, lnum, offs);
	if (err)
		goto out;

547 548
	d->old_zroot_level = le16_to_cpu(idx->level);
	d->old_zroot_sqnum = le64_to_cpu(idx->ch.sqnum);
549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570
out:
	kfree(idx);
	return err;
}

/**
 * dbg_check_old_index - check the old copy of the index.
 * @c: UBIFS file-system description object
 * @zroot: root of the new index
 *
 * In order to be able to recover from an unclean unmount, a complete copy of
 * the index must exist on flash. This is the "old" index. The commit process
 * must write the "new" index to flash without overwriting or destroying any
 * part of the old index. This function is run at commit end in order to check
 * that the old index does indeed exist completely intact.
 *
 * This function returns %0 on success and a negative error code on failure.
 */
int dbg_check_old_index(struct ubifs_info *c, struct ubifs_zbranch *zroot)
{
	int lnum, offs, len, err = 0, uninitialized_var(last_level), child_cnt;
	int first = 1, iip;
571
	struct ubifs_debug_info *d = c->dbg;
572
	union ubifs_key uninitialized_var(lower_key), upper_key, l_key, u_key;
573 574 575 576 577 578 579
	unsigned long long uninitialized_var(last_sqnum);
	struct ubifs_idx_node *idx;
	struct list_head list;
	struct idx_node *i;
	size_t sz;

	if (!(ubifs_chk_flags & UBIFS_CHK_OLD_IDX))
580
		return 0;
581 582 583 584 585 586 587

	INIT_LIST_HEAD(&list);

	sz = sizeof(struct idx_node) + ubifs_idx_node_sz(c, c->fanout) -
	     UBIFS_IDX_NODE_SZ;

	/* Start at the old zroot */
588 589 590
	lnum = d->old_zroot.lnum;
	offs = d->old_zroot.offs;
	len = d->old_zroot.len;
591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622
	iip = 0;

	/*
	 * Traverse the index tree preorder depth-first i.e. do a node and then
	 * its subtrees from left to right.
	 */
	while (1) {
		struct ubifs_branch *br;

		/* Get the next index node */
		i = kmalloc(sz, GFP_NOFS);
		if (!i) {
			err = -ENOMEM;
			goto out_free;
		}
		i->iip = iip;
		/* Keep the index nodes on our path in a linked list */
		list_add_tail(&i->list, &list);
		/* Read the index node */
		idx = &i->idx;
		err = ubifs_read_node(c, idx, UBIFS_IDX_NODE, len, lnum, offs);
		if (err)
			goto out_free;
		/* Validate index node */
		child_cnt = le16_to_cpu(idx->child_cnt);
		if (child_cnt < 1 || child_cnt > c->fanout) {
			err = 1;
			goto out_dump;
		}
		if (first) {
			first = 0;
			/* Check root level and sqnum */
623
			if (le16_to_cpu(idx->level) != d->old_zroot_level) {
624 625 626
				err = 2;
				goto out_dump;
			}
627
			if (le64_to_cpu(idx->ch.sqnum) != d->old_zroot_sqnum) {
628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738
				err = 3;
				goto out_dump;
			}
			/* Set last values as though root had a parent */
			last_level = le16_to_cpu(idx->level) + 1;
			last_sqnum = le64_to_cpu(idx->ch.sqnum) + 1;
			key_read(c, ubifs_idx_key(c, idx), &lower_key);
			highest_ino_key(c, &upper_key, INUM_WATERMARK);
		}
		key_copy(c, &upper_key, &i->upper_key);
		if (le16_to_cpu(idx->level) != last_level - 1) {
			err = 3;
			goto out_dump;
		}
		/*
		 * The index is always written bottom up hence a child's sqnum
		 * is always less than the parents.
		 */
		if (le64_to_cpu(idx->ch.sqnum) >= last_sqnum) {
			err = 4;
			goto out_dump;
		}
		/* Check key range */
		key_read(c, ubifs_idx_key(c, idx), &l_key);
		br = ubifs_idx_branch(c, idx, child_cnt - 1);
		key_read(c, &br->key, &u_key);
		if (keys_cmp(c, &lower_key, &l_key) > 0) {
			err = 5;
			goto out_dump;
		}
		if (keys_cmp(c, &upper_key, &u_key) < 0) {
			err = 6;
			goto out_dump;
		}
		if (keys_cmp(c, &upper_key, &u_key) == 0)
			if (!is_hash_key(c, &u_key)) {
				err = 7;
				goto out_dump;
			}
		/* Go to next index node */
		if (le16_to_cpu(idx->level) == 0) {
			/* At the bottom, so go up until can go right */
			while (1) {
				/* Drop the bottom of the list */
				list_del(&i->list);
				kfree(i);
				/* No more list means we are done */
				if (list_empty(&list))
					goto out;
				/* Look at the new bottom */
				i = list_entry(list.prev, struct idx_node,
					       list);
				idx = &i->idx;
				/* Can we go right */
				if (iip + 1 < le16_to_cpu(idx->child_cnt)) {
					iip = iip + 1;
					break;
				} else
					/* Nope, so go up again */
					iip = i->iip;
			}
		} else
			/* Go down left */
			iip = 0;
		/*
		 * We have the parent in 'idx' and now we set up for reading the
		 * child pointed to by slot 'iip'.
		 */
		last_level = le16_to_cpu(idx->level);
		last_sqnum = le64_to_cpu(idx->ch.sqnum);
		br = ubifs_idx_branch(c, idx, iip);
		lnum = le32_to_cpu(br->lnum);
		offs = le32_to_cpu(br->offs);
		len = le32_to_cpu(br->len);
		key_read(c, &br->key, &lower_key);
		if (iip + 1 < le16_to_cpu(idx->child_cnt)) {
			br = ubifs_idx_branch(c, idx, iip + 1);
			key_read(c, &br->key, &upper_key);
		} else
			key_copy(c, &i->upper_key, &upper_key);
	}
out:
	err = dbg_old_index_check_init(c, zroot);
	if (err)
		goto out_free;

	return 0;

out_dump:
	dbg_err("dumping index node (iip=%d)", i->iip);
	dbg_dump_node(c, idx);
	list_del(&i->list);
	kfree(i);
	if (!list_empty(&list)) {
		i = list_entry(list.prev, struct idx_node, list);
		dbg_err("dumping parent index node");
		dbg_dump_node(c, &i->idx);
	}
out_free:
	while (!list_empty(&list)) {
		i = list_entry(list.next, struct idx_node, list);
		list_del(&i->list);
		kfree(i);
	}
	ubifs_err("failed, error %d", err);
	if (err > 0)
		err = -EINVAL;
	return err;
}

#endif /* CONFIG_UBIFS_FS_DEBUG */