xref: /openbmc/linux/fs/ubifs/gc.c (revision b830f94f)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * This file is part of UBIFS.
4  *
5  * Copyright (C) 2006-2008 Nokia Corporation.
6  *
7  * Authors: Adrian Hunter
8  *          Artem Bityutskiy (Битюцкий Артём)
9  */
10 
11 /*
12  * This file implements garbage collection. The procedure for garbage collection
13  * is different depending on whether a LEB as an index LEB (contains index
14  * nodes) or not. For non-index LEBs, garbage collection finds a LEB which
15  * contains a lot of dirty space (obsolete nodes), and copies the non-obsolete
16  * nodes to the journal, at which point the garbage-collected LEB is free to be
17  * reused. For index LEBs, garbage collection marks the non-obsolete index nodes
18  * dirty in the TNC, and after the next commit, the garbage-collected LEB is
19  * to be reused. Garbage collection will cause the number of dirty index nodes
20  * to grow, however sufficient space is reserved for the index to ensure the
21  * commit will never run out of space.
22  *
23  * Notes about dead watermark. At current UBIFS implementation we assume that
24  * LEBs which have less than @c->dead_wm bytes of free + dirty space are full
25  * and not worth garbage-collecting. The dead watermark is one min. I/O unit
26  * size, or min. UBIFS node size, depending on what is greater. Indeed, UBIFS
27  * Garbage Collector has to synchronize the GC head's write buffer before
28  * returning, so this is about wasting one min. I/O unit. However, UBIFS GC can
29  * actually reclaim even very small pieces of dirty space by garbage collecting
30  * enough dirty LEBs, but we do not bother doing this at this implementation.
31  *
32  * Notes about dark watermark. The results of GC work depends on how big are
33  * the UBIFS nodes GC deals with. Large nodes make GC waste more space. Indeed,
34  * if GC move data from LEB A to LEB B and nodes in LEB A are large, GC would
35  * have to waste large pieces of free space at the end of LEB B, because nodes
36  * from LEB A would not fit. And the worst situation is when all nodes are of
37  * maximum size. So dark watermark is the amount of free + dirty space in LEB
38  * which are guaranteed to be reclaimable. If LEB has less space, the GC might
39  * be unable to reclaim it. So, LEBs with free + dirty greater than dark
40  * watermark are "good" LEBs from GC's point of view. The other LEBs are not so
41  * good, and GC takes extra care when moving them.
42  */
43 
44 #include <linux/slab.h>
45 #include <linux/pagemap.h>
46 #include <linux/list_sort.h>
47 #include "ubifs.h"
48 
49 /*
50  * GC may need to move more than one LEB to make progress. The below constants
51  * define "soft" and "hard" limits on the number of LEBs the garbage collector
52  * may move.
53  */
54 #define SOFT_LEBS_LIMIT 4
55 #define HARD_LEBS_LIMIT 32
56 
57 /**
58  * switch_gc_head - switch the garbage collection journal head.
59  * @c: UBIFS file-system description object
60  * @buf: buffer to write
61  * @len: length of the buffer to write
62  * @lnum: LEB number written is returned here
63  * @offs: offset written is returned here
64  *
65  * This function switch the GC head to the next LEB which is reserved in
66  * @c->gc_lnum. Returns %0 in case of success, %-EAGAIN if commit is required,
67  * and other negative error code in case of failures.
68  */
69 static int switch_gc_head(struct ubifs_info *c)
70 {
71 	int err, gc_lnum = c->gc_lnum;
72 	struct ubifs_wbuf *wbuf = &c->jheads[GCHD].wbuf;
73 
74 	ubifs_assert(c, gc_lnum != -1);
75 	dbg_gc("switch GC head from LEB %d:%d to LEB %d (waste %d bytes)",
76 	       wbuf->lnum, wbuf->offs + wbuf->used, gc_lnum,
77 	       c->leb_size - wbuf->offs - wbuf->used);
78 
79 	err = ubifs_wbuf_sync_nolock(wbuf);
80 	if (err)
81 		return err;
82 
83 	/*
84 	 * The GC write-buffer was synchronized, we may safely unmap
85 	 * 'c->gc_lnum'.
86 	 */
87 	err = ubifs_leb_unmap(c, gc_lnum);
88 	if (err)
89 		return err;
90 
91 	err = ubifs_add_bud_to_log(c, GCHD, gc_lnum, 0);
92 	if (err)
93 		return err;
94 
95 	c->gc_lnum = -1;
96 	err = ubifs_wbuf_seek_nolock(wbuf, gc_lnum, 0);
97 	return err;
98 }
99 
100 /**
101  * data_nodes_cmp - compare 2 data nodes.
102  * @priv: UBIFS file-system description object
103  * @a: first data node
104  * @b: second data node
105  *
106  * This function compares data nodes @a and @b. Returns %1 if @a has greater
107  * inode or block number, and %-1 otherwise.
108  */
109 static int data_nodes_cmp(void *priv, struct list_head *a, struct list_head *b)
110 {
111 	ino_t inuma, inumb;
112 	struct ubifs_info *c = priv;
113 	struct ubifs_scan_node *sa, *sb;
114 
115 	cond_resched();
116 	if (a == b)
117 		return 0;
118 
119 	sa = list_entry(a, struct ubifs_scan_node, list);
120 	sb = list_entry(b, struct ubifs_scan_node, list);
121 
122 	ubifs_assert(c, key_type(c, &sa->key) == UBIFS_DATA_KEY);
123 	ubifs_assert(c, key_type(c, &sb->key) == UBIFS_DATA_KEY);
124 	ubifs_assert(c, sa->type == UBIFS_DATA_NODE);
125 	ubifs_assert(c, sb->type == UBIFS_DATA_NODE);
126 
127 	inuma = key_inum(c, &sa->key);
128 	inumb = key_inum(c, &sb->key);
129 
130 	if (inuma == inumb) {
131 		unsigned int blka = key_block(c, &sa->key);
132 		unsigned int blkb = key_block(c, &sb->key);
133 
134 		if (blka <= blkb)
135 			return -1;
136 	} else if (inuma <= inumb)
137 		return -1;
138 
139 	return 1;
140 }
141 
142 /*
143  * nondata_nodes_cmp - compare 2 non-data nodes.
144  * @priv: UBIFS file-system description object
145  * @a: first node
146  * @a: second node
147  *
148  * This function compares nodes @a and @b. It makes sure that inode nodes go
149  * first and sorted by length in descending order. Directory entry nodes go
150  * after inode nodes and are sorted in ascending hash valuer order.
151  */
152 static int nondata_nodes_cmp(void *priv, struct list_head *a,
153 			     struct list_head *b)
154 {
155 	ino_t inuma, inumb;
156 	struct ubifs_info *c = priv;
157 	struct ubifs_scan_node *sa, *sb;
158 
159 	cond_resched();
160 	if (a == b)
161 		return 0;
162 
163 	sa = list_entry(a, struct ubifs_scan_node, list);
164 	sb = list_entry(b, struct ubifs_scan_node, list);
165 
166 	ubifs_assert(c, key_type(c, &sa->key) != UBIFS_DATA_KEY &&
167 		     key_type(c, &sb->key) != UBIFS_DATA_KEY);
168 	ubifs_assert(c, sa->type != UBIFS_DATA_NODE &&
169 		     sb->type != UBIFS_DATA_NODE);
170 
171 	/* Inodes go before directory entries */
172 	if (sa->type == UBIFS_INO_NODE) {
173 		if (sb->type == UBIFS_INO_NODE)
174 			return sb->len - sa->len;
175 		return -1;
176 	}
177 	if (sb->type == UBIFS_INO_NODE)
178 		return 1;
179 
180 	ubifs_assert(c, key_type(c, &sa->key) == UBIFS_DENT_KEY ||
181 		     key_type(c, &sa->key) == UBIFS_XENT_KEY);
182 	ubifs_assert(c, key_type(c, &sb->key) == UBIFS_DENT_KEY ||
183 		     key_type(c, &sb->key) == UBIFS_XENT_KEY);
184 	ubifs_assert(c, sa->type == UBIFS_DENT_NODE ||
185 		     sa->type == UBIFS_XENT_NODE);
186 	ubifs_assert(c, sb->type == UBIFS_DENT_NODE ||
187 		     sb->type == UBIFS_XENT_NODE);
188 
189 	inuma = key_inum(c, &sa->key);
190 	inumb = key_inum(c, &sb->key);
191 
192 	if (inuma == inumb) {
193 		uint32_t hasha = key_hash(c, &sa->key);
194 		uint32_t hashb = key_hash(c, &sb->key);
195 
196 		if (hasha <= hashb)
197 			return -1;
198 	} else if (inuma <= inumb)
199 		return -1;
200 
201 	return 1;
202 }
203 
204 /**
205  * sort_nodes - sort nodes for GC.
206  * @c: UBIFS file-system description object
207  * @sleb: describes nodes to sort and contains the result on exit
208  * @nondata: contains non-data nodes on exit
209  * @min: minimum node size is returned here
210  *
211  * This function sorts the list of inodes to garbage collect. First of all, it
212  * kills obsolete nodes and separates data and non-data nodes to the
213  * @sleb->nodes and @nondata lists correspondingly.
214  *
215  * Data nodes are then sorted in block number order - this is important for
216  * bulk-read; data nodes with lower inode number go before data nodes with
217  * higher inode number, and data nodes with lower block number go before data
218  * nodes with higher block number;
219  *
220  * Non-data nodes are sorted as follows.
221  *   o First go inode nodes - they are sorted in descending length order.
222  *   o Then go directory entry nodes - they are sorted in hash order, which
223  *     should supposedly optimize 'readdir()'. Direntry nodes with lower parent
224  *     inode number go before direntry nodes with higher parent inode number,
225  *     and direntry nodes with lower name hash values go before direntry nodes
226  *     with higher name hash values.
227  *
228  * This function returns zero in case of success and a negative error code in
229  * case of failure.
230  */
231 static int sort_nodes(struct ubifs_info *c, struct ubifs_scan_leb *sleb,
232 		      struct list_head *nondata, int *min)
233 {
234 	int err;
235 	struct ubifs_scan_node *snod, *tmp;
236 
237 	*min = INT_MAX;
238 
239 	/* Separate data nodes and non-data nodes */
240 	list_for_each_entry_safe(snod, tmp, &sleb->nodes, list) {
241 		ubifs_assert(c, snod->type == UBIFS_INO_NODE  ||
242 			     snod->type == UBIFS_DATA_NODE ||
243 			     snod->type == UBIFS_DENT_NODE ||
244 			     snod->type == UBIFS_XENT_NODE ||
245 			     snod->type == UBIFS_TRUN_NODE ||
246 			     snod->type == UBIFS_AUTH_NODE);
247 
248 		if (snod->type != UBIFS_INO_NODE  &&
249 		    snod->type != UBIFS_DATA_NODE &&
250 		    snod->type != UBIFS_DENT_NODE &&
251 		    snod->type != UBIFS_XENT_NODE) {
252 			/* Probably truncation node, zap it */
253 			list_del(&snod->list);
254 			kfree(snod);
255 			continue;
256 		}
257 
258 		ubifs_assert(c, key_type(c, &snod->key) == UBIFS_DATA_KEY ||
259 			     key_type(c, &snod->key) == UBIFS_INO_KEY  ||
260 			     key_type(c, &snod->key) == UBIFS_DENT_KEY ||
261 			     key_type(c, &snod->key) == UBIFS_XENT_KEY);
262 
263 		err = ubifs_tnc_has_node(c, &snod->key, 0, sleb->lnum,
264 					 snod->offs, 0);
265 		if (err < 0)
266 			return err;
267 
268 		if (!err) {
269 			/* The node is obsolete, remove it from the list */
270 			list_del(&snod->list);
271 			kfree(snod);
272 			continue;
273 		}
274 
275 		if (snod->len < *min)
276 			*min = snod->len;
277 
278 		if (key_type(c, &snod->key) != UBIFS_DATA_KEY)
279 			list_move_tail(&snod->list, nondata);
280 	}
281 
282 	/* Sort data and non-data nodes */
283 	list_sort(c, &sleb->nodes, &data_nodes_cmp);
284 	list_sort(c, nondata, &nondata_nodes_cmp);
285 
286 	err = dbg_check_data_nodes_order(c, &sleb->nodes);
287 	if (err)
288 		return err;
289 	err = dbg_check_nondata_nodes_order(c, nondata);
290 	if (err)
291 		return err;
292 	return 0;
293 }
294 
295 /**
296  * move_node - move a node.
297  * @c: UBIFS file-system description object
298  * @sleb: describes the LEB to move nodes from
299  * @snod: the mode to move
300  * @wbuf: write-buffer to move node to
301  *
302  * This function moves node @snod to @wbuf, changes TNC correspondingly, and
303  * destroys @snod. Returns zero in case of success and a negative error code in
304  * case of failure.
305  */
306 static int move_node(struct ubifs_info *c, struct ubifs_scan_leb *sleb,
307 		     struct ubifs_scan_node *snod, struct ubifs_wbuf *wbuf)
308 {
309 	int err, new_lnum = wbuf->lnum, new_offs = wbuf->offs + wbuf->used;
310 
311 	cond_resched();
312 	err = ubifs_wbuf_write_nolock(wbuf, snod->node, snod->len);
313 	if (err)
314 		return err;
315 
316 	err = ubifs_tnc_replace(c, &snod->key, sleb->lnum,
317 				snod->offs, new_lnum, new_offs,
318 				snod->len);
319 	list_del(&snod->list);
320 	kfree(snod);
321 	return err;
322 }
323 
324 /**
325  * move_nodes - move nodes.
326  * @c: UBIFS file-system description object
327  * @sleb: describes the LEB to move nodes from
328  *
329  * This function moves valid nodes from data LEB described by @sleb to the GC
330  * journal head. This function returns zero in case of success, %-EAGAIN if
331  * commit is required, and other negative error codes in case of other
332  * failures.
333  */
334 static int move_nodes(struct ubifs_info *c, struct ubifs_scan_leb *sleb)
335 {
336 	int err, min;
337 	LIST_HEAD(nondata);
338 	struct ubifs_wbuf *wbuf = &c->jheads[GCHD].wbuf;
339 
340 	if (wbuf->lnum == -1) {
341 		/*
342 		 * The GC journal head is not set, because it is the first GC
343 		 * invocation since mount.
344 		 */
345 		err = switch_gc_head(c);
346 		if (err)
347 			return err;
348 	}
349 
350 	err = sort_nodes(c, sleb, &nondata, &min);
351 	if (err)
352 		goto out;
353 
354 	/* Write nodes to their new location. Use the first-fit strategy */
355 	while (1) {
356 		int avail, moved = 0;
357 		struct ubifs_scan_node *snod, *tmp;
358 
359 		/* Move data nodes */
360 		list_for_each_entry_safe(snod, tmp, &sleb->nodes, list) {
361 			avail = c->leb_size - wbuf->offs - wbuf->used -
362 					ubifs_auth_node_sz(c);
363 			if  (snod->len > avail)
364 				/*
365 				 * Do not skip data nodes in order to optimize
366 				 * bulk-read.
367 				 */
368 				break;
369 
370 			err = ubifs_shash_update(c, c->jheads[GCHD].log_hash,
371 						 snod->node, snod->len);
372 			if (err)
373 				goto out;
374 
375 			err = move_node(c, sleb, snod, wbuf);
376 			if (err)
377 				goto out;
378 			moved = 1;
379 		}
380 
381 		/* Move non-data nodes */
382 		list_for_each_entry_safe(snod, tmp, &nondata, list) {
383 			avail = c->leb_size - wbuf->offs - wbuf->used -
384 					ubifs_auth_node_sz(c);
385 			if (avail < min)
386 				break;
387 
388 			if  (snod->len > avail) {
389 				/*
390 				 * Keep going only if this is an inode with
391 				 * some data. Otherwise stop and switch the GC
392 				 * head. IOW, we assume that data-less inode
393 				 * nodes and direntry nodes are roughly of the
394 				 * same size.
395 				 */
396 				if (key_type(c, &snod->key) == UBIFS_DENT_KEY ||
397 				    snod->len == UBIFS_INO_NODE_SZ)
398 					break;
399 				continue;
400 			}
401 
402 			err = ubifs_shash_update(c, c->jheads[GCHD].log_hash,
403 						 snod->node, snod->len);
404 			if (err)
405 				goto out;
406 
407 			err = move_node(c, sleb, snod, wbuf);
408 			if (err)
409 				goto out;
410 			moved = 1;
411 		}
412 
413 		if (ubifs_authenticated(c) && moved) {
414 			struct ubifs_auth_node *auth;
415 
416 			auth = kmalloc(ubifs_auth_node_sz(c), GFP_NOFS);
417 			if (!auth) {
418 				err = -ENOMEM;
419 				goto out;
420 			}
421 
422 			err = ubifs_prepare_auth_node(c, auth,
423 						c->jheads[GCHD].log_hash);
424 			if (err) {
425 				kfree(auth);
426 				goto out;
427 			}
428 
429 			err = ubifs_wbuf_write_nolock(wbuf, auth,
430 						      ubifs_auth_node_sz(c));
431 			if (err) {
432 				kfree(auth);
433 				goto out;
434 			}
435 
436 			ubifs_add_dirt(c, wbuf->lnum, ubifs_auth_node_sz(c));
437 		}
438 
439 		if (list_empty(&sleb->nodes) && list_empty(&nondata))
440 			break;
441 
442 		/*
443 		 * Waste the rest of the space in the LEB and switch to the
444 		 * next LEB.
445 		 */
446 		err = switch_gc_head(c);
447 		if (err)
448 			goto out;
449 	}
450 
451 	return 0;
452 
453 out:
454 	list_splice_tail(&nondata, &sleb->nodes);
455 	return err;
456 }
457 
458 /**
459  * gc_sync_wbufs - sync write-buffers for GC.
460  * @c: UBIFS file-system description object
461  *
462  * We must guarantee that obsoleting nodes are on flash. Unfortunately they may
463  * be in a write-buffer instead. That is, a node could be written to a
464  * write-buffer, obsoleting another node in a LEB that is GC'd. If that LEB is
465  * erased before the write-buffer is sync'd and then there is an unclean
466  * unmount, then an existing node is lost. To avoid this, we sync all
467  * write-buffers.
468  *
469  * This function returns %0 on success or a negative error code on failure.
470  */
471 static int gc_sync_wbufs(struct ubifs_info *c)
472 {
473 	int err, i;
474 
475 	for (i = 0; i < c->jhead_cnt; i++) {
476 		if (i == GCHD)
477 			continue;
478 		err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
479 		if (err)
480 			return err;
481 	}
482 	return 0;
483 }
484 
485 /**
486  * ubifs_garbage_collect_leb - garbage-collect a logical eraseblock.
487  * @c: UBIFS file-system description object
488  * @lp: describes the LEB to garbage collect
489  *
490  * This function garbage-collects an LEB and returns one of the @LEB_FREED,
491  * @LEB_RETAINED, etc positive codes in case of success, %-EAGAIN if commit is
492  * required, and other negative error codes in case of failures.
493  */
494 int ubifs_garbage_collect_leb(struct ubifs_info *c, struct ubifs_lprops *lp)
495 {
496 	struct ubifs_scan_leb *sleb;
497 	struct ubifs_scan_node *snod;
498 	struct ubifs_wbuf *wbuf = &c->jheads[GCHD].wbuf;
499 	int err = 0, lnum = lp->lnum;
500 
501 	ubifs_assert(c, c->gc_lnum != -1 || wbuf->offs + wbuf->used == 0 ||
502 		     c->need_recovery);
503 	ubifs_assert(c, c->gc_lnum != lnum);
504 	ubifs_assert(c, wbuf->lnum != lnum);
505 
506 	if (lp->free + lp->dirty == c->leb_size) {
507 		/* Special case - a free LEB  */
508 		dbg_gc("LEB %d is free, return it", lp->lnum);
509 		ubifs_assert(c, !(lp->flags & LPROPS_INDEX));
510 
511 		if (lp->free != c->leb_size) {
512 			/*
513 			 * Write buffers must be sync'd before unmapping
514 			 * freeable LEBs, because one of them may contain data
515 			 * which obsoletes something in 'lp->lnum'.
516 			 */
517 			err = gc_sync_wbufs(c);
518 			if (err)
519 				return err;
520 			err = ubifs_change_one_lp(c, lp->lnum, c->leb_size,
521 						  0, 0, 0, 0);
522 			if (err)
523 				return err;
524 		}
525 		err = ubifs_leb_unmap(c, lp->lnum);
526 		if (err)
527 			return err;
528 
529 		if (c->gc_lnum == -1) {
530 			c->gc_lnum = lnum;
531 			return LEB_RETAINED;
532 		}
533 
534 		return LEB_FREED;
535 	}
536 
537 	/*
538 	 * We scan the entire LEB even though we only really need to scan up to
539 	 * (c->leb_size - lp->free).
540 	 */
541 	sleb = ubifs_scan(c, lnum, 0, c->sbuf, 0);
542 	if (IS_ERR(sleb))
543 		return PTR_ERR(sleb);
544 
545 	ubifs_assert(c, !list_empty(&sleb->nodes));
546 	snod = list_entry(sleb->nodes.next, struct ubifs_scan_node, list);
547 
548 	if (snod->type == UBIFS_IDX_NODE) {
549 		struct ubifs_gced_idx_leb *idx_gc;
550 
551 		dbg_gc("indexing LEB %d (free %d, dirty %d)",
552 		       lnum, lp->free, lp->dirty);
553 		list_for_each_entry(snod, &sleb->nodes, list) {
554 			struct ubifs_idx_node *idx = snod->node;
555 			int level = le16_to_cpu(idx->level);
556 
557 			ubifs_assert(c, snod->type == UBIFS_IDX_NODE);
558 			key_read(c, ubifs_idx_key(c, idx), &snod->key);
559 			err = ubifs_dirty_idx_node(c, &snod->key, level, lnum,
560 						   snod->offs);
561 			if (err)
562 				goto out;
563 		}
564 
565 		idx_gc = kmalloc(sizeof(struct ubifs_gced_idx_leb), GFP_NOFS);
566 		if (!idx_gc) {
567 			err = -ENOMEM;
568 			goto out;
569 		}
570 
571 		idx_gc->lnum = lnum;
572 		idx_gc->unmap = 0;
573 		list_add(&idx_gc->list, &c->idx_gc);
574 
575 		/*
576 		 * Don't release the LEB until after the next commit, because
577 		 * it may contain data which is needed for recovery. So
578 		 * although we freed this LEB, it will become usable only after
579 		 * the commit.
580 		 */
581 		err = ubifs_change_one_lp(c, lnum, c->leb_size, 0, 0,
582 					  LPROPS_INDEX, 1);
583 		if (err)
584 			goto out;
585 		err = LEB_FREED_IDX;
586 	} else {
587 		dbg_gc("data LEB %d (free %d, dirty %d)",
588 		       lnum, lp->free, lp->dirty);
589 
590 		err = move_nodes(c, sleb);
591 		if (err)
592 			goto out_inc_seq;
593 
594 		err = gc_sync_wbufs(c);
595 		if (err)
596 			goto out_inc_seq;
597 
598 		err = ubifs_change_one_lp(c, lnum, c->leb_size, 0, 0, 0, 0);
599 		if (err)
600 			goto out_inc_seq;
601 
602 		/* Allow for races with TNC */
603 		c->gced_lnum = lnum;
604 		smp_wmb();
605 		c->gc_seq += 1;
606 		smp_wmb();
607 
608 		if (c->gc_lnum == -1) {
609 			c->gc_lnum = lnum;
610 			err = LEB_RETAINED;
611 		} else {
612 			err = ubifs_wbuf_sync_nolock(wbuf);
613 			if (err)
614 				goto out;
615 
616 			err = ubifs_leb_unmap(c, lnum);
617 			if (err)
618 				goto out;
619 
620 			err = LEB_FREED;
621 		}
622 	}
623 
624 out:
625 	ubifs_scan_destroy(sleb);
626 	return err;
627 
628 out_inc_seq:
629 	/* We may have moved at least some nodes so allow for races with TNC */
630 	c->gced_lnum = lnum;
631 	smp_wmb();
632 	c->gc_seq += 1;
633 	smp_wmb();
634 	goto out;
635 }
636 
637 /**
638  * ubifs_garbage_collect - UBIFS garbage collector.
639  * @c: UBIFS file-system description object
640  * @anyway: do GC even if there are free LEBs
641  *
642  * This function does out-of-place garbage collection. The return codes are:
643  *   o positive LEB number if the LEB has been freed and may be used;
644  *   o %-EAGAIN if the caller has to run commit;
645  *   o %-ENOSPC if GC failed to make any progress;
646  *   o other negative error codes in case of other errors.
647  *
648  * Garbage collector writes data to the journal when GC'ing data LEBs, and just
649  * marking indexing nodes dirty when GC'ing indexing LEBs. Thus, at some point
650  * commit may be required. But commit cannot be run from inside GC, because the
651  * caller might be holding the commit lock, so %-EAGAIN is returned instead;
652  * And this error code means that the caller has to run commit, and re-run GC
653  * if there is still no free space.
654  *
655  * There are many reasons why this function may return %-EAGAIN:
656  * o the log is full and there is no space to write an LEB reference for
657  *   @c->gc_lnum;
658  * o the journal is too large and exceeds size limitations;
659  * o GC moved indexing LEBs, but they can be used only after the commit;
660  * o the shrinker fails to find clean znodes to free and requests the commit;
661  * o etc.
662  *
663  * Note, if the file-system is close to be full, this function may return
664  * %-EAGAIN infinitely, so the caller has to limit amount of re-invocations of
665  * the function. E.g., this happens if the limits on the journal size are too
666  * tough and GC writes too much to the journal before an LEB is freed. This
667  * might also mean that the journal is too large, and the TNC becomes to big,
668  * so that the shrinker is constantly called, finds not clean znodes to free,
669  * and requests commit. Well, this may also happen if the journal is all right,
670  * but another kernel process consumes too much memory. Anyway, infinite
671  * %-EAGAIN may happen, but in some extreme/misconfiguration cases.
672  */
673 int ubifs_garbage_collect(struct ubifs_info *c, int anyway)
674 {
675 	int i, err, ret, min_space = c->dead_wm;
676 	struct ubifs_lprops lp;
677 	struct ubifs_wbuf *wbuf = &c->jheads[GCHD].wbuf;
678 
679 	ubifs_assert_cmt_locked(c);
680 	ubifs_assert(c, !c->ro_media && !c->ro_mount);
681 
682 	if (ubifs_gc_should_commit(c))
683 		return -EAGAIN;
684 
685 	mutex_lock_nested(&wbuf->io_mutex, wbuf->jhead);
686 
687 	if (c->ro_error) {
688 		ret = -EROFS;
689 		goto out_unlock;
690 	}
691 
692 	/* We expect the write-buffer to be empty on entry */
693 	ubifs_assert(c, !wbuf->used);
694 
695 	for (i = 0; ; i++) {
696 		int space_before, space_after;
697 
698 		cond_resched();
699 
700 		/* Give the commit an opportunity to run */
701 		if (ubifs_gc_should_commit(c)) {
702 			ret = -EAGAIN;
703 			break;
704 		}
705 
706 		if (i > SOFT_LEBS_LIMIT && !list_empty(&c->idx_gc)) {
707 			/*
708 			 * We've done enough iterations. Indexing LEBs were
709 			 * moved and will be available after the commit.
710 			 */
711 			dbg_gc("soft limit, some index LEBs GC'ed, -EAGAIN");
712 			ubifs_commit_required(c);
713 			ret = -EAGAIN;
714 			break;
715 		}
716 
717 		if (i > HARD_LEBS_LIMIT) {
718 			/*
719 			 * We've moved too many LEBs and have not made
720 			 * progress, give up.
721 			 */
722 			dbg_gc("hard limit, -ENOSPC");
723 			ret = -ENOSPC;
724 			break;
725 		}
726 
727 		/*
728 		 * Empty and freeable LEBs can turn up while we waited for
729 		 * the wbuf lock, or while we have been running GC. In that
730 		 * case, we should just return one of those instead of
731 		 * continuing to GC dirty LEBs. Hence we request
732 		 * 'ubifs_find_dirty_leb()' to return an empty LEB if it can.
733 		 */
734 		ret = ubifs_find_dirty_leb(c, &lp, min_space, anyway ? 0 : 1);
735 		if (ret) {
736 			if (ret == -ENOSPC)
737 				dbg_gc("no more dirty LEBs");
738 			break;
739 		}
740 
741 		dbg_gc("found LEB %d: free %d, dirty %d, sum %d (min. space %d)",
742 		       lp.lnum, lp.free, lp.dirty, lp.free + lp.dirty,
743 		       min_space);
744 
745 		space_before = c->leb_size - wbuf->offs - wbuf->used;
746 		if (wbuf->lnum == -1)
747 			space_before = 0;
748 
749 		ret = ubifs_garbage_collect_leb(c, &lp);
750 		if (ret < 0) {
751 			if (ret == -EAGAIN) {
752 				/*
753 				 * This is not error, so we have to return the
754 				 * LEB to lprops. But if 'ubifs_return_leb()'
755 				 * fails, its failure code is propagated to the
756 				 * caller instead of the original '-EAGAIN'.
757 				 */
758 				err = ubifs_return_leb(c, lp.lnum);
759 				if (err)
760 					ret = err;
761 				break;
762 			}
763 			goto out;
764 		}
765 
766 		if (ret == LEB_FREED) {
767 			/* An LEB has been freed and is ready for use */
768 			dbg_gc("LEB %d freed, return", lp.lnum);
769 			ret = lp.lnum;
770 			break;
771 		}
772 
773 		if (ret == LEB_FREED_IDX) {
774 			/*
775 			 * This was an indexing LEB and it cannot be
776 			 * immediately used. And instead of requesting the
777 			 * commit straight away, we try to garbage collect some
778 			 * more.
779 			 */
780 			dbg_gc("indexing LEB %d freed, continue", lp.lnum);
781 			continue;
782 		}
783 
784 		ubifs_assert(c, ret == LEB_RETAINED);
785 		space_after = c->leb_size - wbuf->offs - wbuf->used;
786 		dbg_gc("LEB %d retained, freed %d bytes", lp.lnum,
787 		       space_after - space_before);
788 
789 		if (space_after > space_before) {
790 			/* GC makes progress, keep working */
791 			min_space >>= 1;
792 			if (min_space < c->dead_wm)
793 				min_space = c->dead_wm;
794 			continue;
795 		}
796 
797 		dbg_gc("did not make progress");
798 
799 		/*
800 		 * GC moved an LEB bud have not done any progress. This means
801 		 * that the previous GC head LEB contained too few free space
802 		 * and the LEB which was GC'ed contained only large nodes which
803 		 * did not fit that space.
804 		 *
805 		 * We can do 2 things:
806 		 * 1. pick another LEB in a hope it'll contain a small node
807 		 *    which will fit the space we have at the end of current GC
808 		 *    head LEB, but there is no guarantee, so we try this out
809 		 *    unless we have already been working for too long;
810 		 * 2. request an LEB with more dirty space, which will force
811 		 *    'ubifs_find_dirty_leb()' to start scanning the lprops
812 		 *    table, instead of just picking one from the heap
813 		 *    (previously it already picked the dirtiest LEB).
814 		 */
815 		if (i < SOFT_LEBS_LIMIT) {
816 			dbg_gc("try again");
817 			continue;
818 		}
819 
820 		min_space <<= 1;
821 		if (min_space > c->dark_wm)
822 			min_space = c->dark_wm;
823 		dbg_gc("set min. space to %d", min_space);
824 	}
825 
826 	if (ret == -ENOSPC && !list_empty(&c->idx_gc)) {
827 		dbg_gc("no space, some index LEBs GC'ed, -EAGAIN");
828 		ubifs_commit_required(c);
829 		ret = -EAGAIN;
830 	}
831 
832 	err = ubifs_wbuf_sync_nolock(wbuf);
833 	if (!err)
834 		err = ubifs_leb_unmap(c, c->gc_lnum);
835 	if (err) {
836 		ret = err;
837 		goto out;
838 	}
839 out_unlock:
840 	mutex_unlock(&wbuf->io_mutex);
841 	return ret;
842 
843 out:
844 	ubifs_assert(c, ret < 0);
845 	ubifs_assert(c, ret != -ENOSPC && ret != -EAGAIN);
846 	ubifs_wbuf_sync_nolock(wbuf);
847 	ubifs_ro_mode(c, ret);
848 	mutex_unlock(&wbuf->io_mutex);
849 	ubifs_return_leb(c, lp.lnum);
850 	return ret;
851 }
852 
853 /**
854  * ubifs_gc_start_commit - garbage collection at start of commit.
855  * @c: UBIFS file-system description object
856  *
857  * If a LEB has only dirty and free space, then we may safely unmap it and make
858  * it free.  Note, we cannot do this with indexing LEBs because dirty space may
859  * correspond index nodes that are required for recovery.  In that case, the
860  * LEB cannot be unmapped until after the next commit.
861  *
862  * This function returns %0 upon success and a negative error code upon failure.
863  */
864 int ubifs_gc_start_commit(struct ubifs_info *c)
865 {
866 	struct ubifs_gced_idx_leb *idx_gc;
867 	const struct ubifs_lprops *lp;
868 	int err = 0, flags;
869 
870 	ubifs_get_lprops(c);
871 
872 	/*
873 	 * Unmap (non-index) freeable LEBs. Note that recovery requires that all
874 	 * wbufs are sync'd before this, which is done in 'do_commit()'.
875 	 */
876 	while (1) {
877 		lp = ubifs_fast_find_freeable(c);
878 		if (!lp)
879 			break;
880 		ubifs_assert(c, !(lp->flags & LPROPS_TAKEN));
881 		ubifs_assert(c, !(lp->flags & LPROPS_INDEX));
882 		err = ubifs_leb_unmap(c, lp->lnum);
883 		if (err)
884 			goto out;
885 		lp = ubifs_change_lp(c, lp, c->leb_size, 0, lp->flags, 0);
886 		if (IS_ERR(lp)) {
887 			err = PTR_ERR(lp);
888 			goto out;
889 		}
890 		ubifs_assert(c, !(lp->flags & LPROPS_TAKEN));
891 		ubifs_assert(c, !(lp->flags & LPROPS_INDEX));
892 	}
893 
894 	/* Mark GC'd index LEBs OK to unmap after this commit finishes */
895 	list_for_each_entry(idx_gc, &c->idx_gc, list)
896 		idx_gc->unmap = 1;
897 
898 	/* Record index freeable LEBs for unmapping after commit */
899 	while (1) {
900 		lp = ubifs_fast_find_frdi_idx(c);
901 		if (IS_ERR(lp)) {
902 			err = PTR_ERR(lp);
903 			goto out;
904 		}
905 		if (!lp)
906 			break;
907 		idx_gc = kmalloc(sizeof(struct ubifs_gced_idx_leb), GFP_NOFS);
908 		if (!idx_gc) {
909 			err = -ENOMEM;
910 			goto out;
911 		}
912 		ubifs_assert(c, !(lp->flags & LPROPS_TAKEN));
913 		ubifs_assert(c, lp->flags & LPROPS_INDEX);
914 		/* Don't release the LEB until after the next commit */
915 		flags = (lp->flags | LPROPS_TAKEN) ^ LPROPS_INDEX;
916 		lp = ubifs_change_lp(c, lp, c->leb_size, 0, flags, 1);
917 		if (IS_ERR(lp)) {
918 			err = PTR_ERR(lp);
919 			kfree(idx_gc);
920 			goto out;
921 		}
922 		ubifs_assert(c, lp->flags & LPROPS_TAKEN);
923 		ubifs_assert(c, !(lp->flags & LPROPS_INDEX));
924 		idx_gc->lnum = lp->lnum;
925 		idx_gc->unmap = 1;
926 		list_add(&idx_gc->list, &c->idx_gc);
927 	}
928 out:
929 	ubifs_release_lprops(c);
930 	return err;
931 }
932 
933 /**
934  * ubifs_gc_end_commit - garbage collection at end of commit.
935  * @c: UBIFS file-system description object
936  *
937  * This function completes out-of-place garbage collection of index LEBs.
938  */
939 int ubifs_gc_end_commit(struct ubifs_info *c)
940 {
941 	struct ubifs_gced_idx_leb *idx_gc, *tmp;
942 	struct ubifs_wbuf *wbuf;
943 	int err = 0;
944 
945 	wbuf = &c->jheads[GCHD].wbuf;
946 	mutex_lock_nested(&wbuf->io_mutex, wbuf->jhead);
947 	list_for_each_entry_safe(idx_gc, tmp, &c->idx_gc, list)
948 		if (idx_gc->unmap) {
949 			dbg_gc("LEB %d", idx_gc->lnum);
950 			err = ubifs_leb_unmap(c, idx_gc->lnum);
951 			if (err)
952 				goto out;
953 			err = ubifs_change_one_lp(c, idx_gc->lnum, LPROPS_NC,
954 					  LPROPS_NC, 0, LPROPS_TAKEN, -1);
955 			if (err)
956 				goto out;
957 			list_del(&idx_gc->list);
958 			kfree(idx_gc);
959 		}
960 out:
961 	mutex_unlock(&wbuf->io_mutex);
962 	return err;
963 }
964 
965 /**
966  * ubifs_destroy_idx_gc - destroy idx_gc list.
967  * @c: UBIFS file-system description object
968  *
969  * This function destroys the @c->idx_gc list. It is called when unmounting
970  * so locks are not needed. Returns zero in case of success and a negative
971  * error code in case of failure.
972  */
973 void ubifs_destroy_idx_gc(struct ubifs_info *c)
974 {
975 	while (!list_empty(&c->idx_gc)) {
976 		struct ubifs_gced_idx_leb *idx_gc;
977 
978 		idx_gc = list_entry(c->idx_gc.next, struct ubifs_gced_idx_leb,
979 				    list);
980 		c->idx_gc_cnt -= 1;
981 		list_del(&idx_gc->list);
982 		kfree(idx_gc);
983 	}
984 }
985 
986 /**
987  * ubifs_get_idx_gc_leb - get a LEB from GC'd index LEB list.
988  * @c: UBIFS file-system description object
989  *
990  * Called during start commit so locks are not needed.
991  */
992 int ubifs_get_idx_gc_leb(struct ubifs_info *c)
993 {
994 	struct ubifs_gced_idx_leb *idx_gc;
995 	int lnum;
996 
997 	if (list_empty(&c->idx_gc))
998 		return -ENOSPC;
999 	idx_gc = list_entry(c->idx_gc.next, struct ubifs_gced_idx_leb, list);
1000 	lnum = idx_gc->lnum;
1001 	/* c->idx_gc_cnt is updated by the caller when lprops are updated */
1002 	list_del(&idx_gc->list);
1003 	kfree(idx_gc);
1004 	return lnum;
1005 }
1006