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