blob: f0f5f15d384ea311e730ae73f3fe3900b0a0eca9 [file] [log] [blame]
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001/*
2 * This file is part of UBIFS.
3 *
4 * Copyright (C) 2006-2008 Nokia Corporation.
5 *
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License version 2 as published by
8 * the Free Software Foundation.
9 *
10 * This program is distributed in the hope that it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
13 * more details.
14 *
15 * You should have received a copy of the GNU General Public License along with
16 * this program; if not, write to the Free Software Foundation, Inc., 51
17 * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
18 *
19 * Authors: Adrian Hunter
20 * Artem Bityutskiy (Битюцкий Артём)
21 */
22
23/*
24 * This file implements garbage collection. The procedure for garbage collection
25 * is different depending on whether a LEB as an index LEB (contains index
26 * nodes) or not. For non-index LEBs, garbage collection finds a LEB which
27 * contains a lot of dirty space (obsolete nodes), and copies the non-obsolete
28 * nodes to the journal, at which point the garbage-collected LEB is free to be
29 * reused. For index LEBs, garbage collection marks the non-obsolete index nodes
30 * dirty in the TNC, and after the next commit, the garbage-collected LEB is
31 * to be reused. Garbage collection will cause the number of dirty index nodes
32 * to grow, however sufficient space is reserved for the index to ensure the
33 * commit will never run out of space.
Artem Bityutskiy70782022009-01-19 19:57:27 +020034 *
35 * Notes about dead watermark. At current UBIFS implementation we assume that
36 * LEBs which have less than @c->dead_wm bytes of free + dirty space are full
37 * and not worth garbage-collecting. The dead watermark is one min. I/O unit
38 * size, or min. UBIFS node size, depending on what is greater. Indeed, UBIFS
39 * Garbage Collector has to synchronize the GC head's write buffer before
40 * returning, so this is about wasting one min. I/O unit. However, UBIFS GC can
41 * actually reclaim even very small pieces of dirty space by garbage collecting
42 * enough dirty LEBs, but we do not bother doing this at this implementation.
43 *
44 * Notes about dark watermark. The results of GC work depends on how big are
45 * the UBIFS nodes GC deals with. Large nodes make GC waste more space. Indeed,
46 * if GC move data from LEB A to LEB B and nodes in LEB A are large, GC would
47 * have to waste large pieces of free space at the end of LEB B, because nodes
48 * from LEB A would not fit. And the worst situation is when all nodes are of
49 * maximum size. So dark watermark is the amount of free + dirty space in LEB
Artem Bityutskiyf10770f2009-03-08 15:13:00 +020050 * which are guaranteed to be reclaimable. If LEB has less space, the GC might
Artem Bityutskiy70782022009-01-19 19:57:27 +020051 * be unable to reclaim it. So, LEBs with free + dirty greater than dark
52 * watermark are "good" LEBs from GC's point of few. The other LEBs are not so
53 * good, and GC takes extra care when moving them.
Artem Bityutskiy1e517642008-07-14 19:08:37 +030054 */
55
56#include <linux/pagemap.h>
57#include "ubifs.h"
58
59/*
Frederik Schwarzer025dfda2008-10-16 19:02:37 +020060 * GC may need to move more than one LEB to make progress. The below constants
Artem Bityutskiy1e517642008-07-14 19:08:37 +030061 * define "soft" and "hard" limits on the number of LEBs the garbage collector
62 * may move.
63 */
64#define SOFT_LEBS_LIMIT 4
65#define HARD_LEBS_LIMIT 32
66
67/**
68 * switch_gc_head - switch the garbage collection journal head.
69 * @c: UBIFS file-system description object
70 * @buf: buffer to write
71 * @len: length of the buffer to write
72 * @lnum: LEB number written is returned here
73 * @offs: offset written is returned here
74 *
75 * This function switch the GC head to the next LEB which is reserved in
76 * @c->gc_lnum. Returns %0 in case of success, %-EAGAIN if commit is required,
77 * and other negative error code in case of failures.
78 */
79static int switch_gc_head(struct ubifs_info *c)
80{
81 int err, gc_lnum = c->gc_lnum;
82 struct ubifs_wbuf *wbuf = &c->jheads[GCHD].wbuf;
83
84 ubifs_assert(gc_lnum != -1);
85 dbg_gc("switch GC head from LEB %d:%d to LEB %d (waste %d bytes)",
86 wbuf->lnum, wbuf->offs + wbuf->used, gc_lnum,
87 c->leb_size - wbuf->offs - wbuf->used);
88
89 err = ubifs_wbuf_sync_nolock(wbuf);
90 if (err)
91 return err;
92
93 /*
94 * The GC write-buffer was synchronized, we may safely unmap
95 * 'c->gc_lnum'.
96 */
97 err = ubifs_leb_unmap(c, gc_lnum);
98 if (err)
99 return err;
100
101 err = ubifs_add_bud_to_log(c, GCHD, gc_lnum, 0);
102 if (err)
103 return err;
104
105 c->gc_lnum = -1;
106 err = ubifs_wbuf_seek_nolock(wbuf, gc_lnum, 0, UBI_LONGTERM);
107 return err;
108}
109
110/**
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200111 * list_sort - sort a list.
112 * @priv: private data, passed to @cmp
113 * @head: the list to sort
114 * @cmp: the elements comparison function
Adrian Hunter46773be2008-09-11 12:57:49 +0300115 *
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200116 * This function has been implemented by Mark J Roberts <mjr@znex.org>. It
117 * implements "merge sort" which has O(nlog(n)) complexity. The list is sorted
118 * in ascending order.
119 *
120 * The comparison function @cmp is supposed to return a negative value if @a is
121 * than @b, and a positive value if @a is greater than @b. If @a and @b are
122 * equivalent, then it does not matter what this function returns.
Adrian Hunter46773be2008-09-11 12:57:49 +0300123 */
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200124static void list_sort(void *priv, struct list_head *head,
125 int (*cmp)(void *priv, struct list_head *a,
126 struct list_head *b))
Adrian Hunter46773be2008-09-11 12:57:49 +0300127{
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200128 struct list_head *p, *q, *e, *list, *tail, *oldhead;
129 int insize, nmerges, psize, qsize, i;
Adrian Hunter46773be2008-09-11 12:57:49 +0300130
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200131 if (list_empty(head))
132 return;
133
134 list = head->next;
135 list_del(head);
136 insize = 1;
137 for (;;) {
138 p = oldhead = list;
139 list = tail = NULL;
140 nmerges = 0;
141
142 while (p) {
143 nmerges++;
144 q = p;
145 psize = 0;
146 for (i = 0; i < insize; i++) {
147 psize++;
148 q = q->next == oldhead ? NULL : q->next;
149 if (!q)
150 break;
151 }
152
153 qsize = insize;
154 while (psize > 0 || (qsize > 0 && q)) {
155 if (!psize) {
156 e = q;
157 q = q->next;
158 qsize--;
159 if (q == oldhead)
160 q = NULL;
161 } else if (!qsize || !q) {
162 e = p;
163 p = p->next;
164 psize--;
165 if (p == oldhead)
166 p = NULL;
167 } else if (cmp(priv, p, q) <= 0) {
168 e = p;
169 p = p->next;
170 psize--;
171 if (p == oldhead)
172 p = NULL;
173 } else {
174 e = q;
175 q = q->next;
176 qsize--;
177 if (q == oldhead)
178 q = NULL;
179 }
180 if (tail)
181 tail->next = e;
182 else
183 list = e;
184 e->prev = tail;
185 tail = e;
186 }
187 p = q;
188 }
189
190 tail->next = list;
191 list->prev = tail;
192
193 if (nmerges <= 1)
Adrian Hunter46773be2008-09-11 12:57:49 +0300194 break;
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200195
196 insize *= 2;
Adrian Hunter46773be2008-09-11 12:57:49 +0300197 }
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200198
199 head->next = list;
200 head->prev = list->prev;
201 list->prev->next = head;
202 list->prev = head;
Adrian Hunter46773be2008-09-11 12:57:49 +0300203}
204
205/**
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200206 * data_nodes_cmp - compare 2 data nodes.
207 * @priv: UBIFS file-system description object
208 * @a: first data node
209 * @a: second data node
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300210 *
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200211 * This function compares data nodes @a and @b. Returns %1 if @a has greater
212 * inode or block number, and %-1 otherwise.
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300213 */
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200214int data_nodes_cmp(void *priv, struct list_head *a, struct list_head *b)
215{
216 ino_t inuma, inumb;
217 struct ubifs_info *c = priv;
218 struct ubifs_scan_node *sa, *sb;
219
220 cond_resched();
221 sa = list_entry(a, struct ubifs_scan_node, list);
222 sb = list_entry(b, struct ubifs_scan_node, list);
223 ubifs_assert(key_type(c, &sa->key) == UBIFS_DATA_KEY);
224 ubifs_assert(key_type(c, &sb->key) == UBIFS_DATA_KEY);
225
226 inuma = key_inum(c, &sa->key);
227 inumb = key_inum(c, &sb->key);
228
229 if (inuma == inumb) {
230 unsigned int blka = key_block(c, &sa->key);
231 unsigned int blkb = key_block(c, &sb->key);
232
233 if (blka <= blkb)
234 return -1;
235 } else if (inuma <= inumb)
236 return -1;
237
238 return 1;
239}
240
241/*
242 * nondata_nodes_cmp - compare 2 non-data nodes.
243 * @priv: UBIFS file-system description object
244 * @a: first node
245 * @a: second node
246 *
247 * This function compares nodes @a and @b. It makes sure that inode nodes go
248 * first and sorted by length in descending order. Directory entry nodes go
249 * after inode nodes and are sorted in ascending hash valuer order.
250 */
251int nondata_nodes_cmp(void *priv, struct list_head *a, struct list_head *b)
252{
253 int typea, typeb;
254 ino_t inuma, inumb;
255 struct ubifs_info *c = priv;
256 struct ubifs_scan_node *sa, *sb;
257
258 cond_resched();
259 sa = list_entry(a, struct ubifs_scan_node, list);
260 sb = list_entry(b, struct ubifs_scan_node, list);
261 typea = key_type(c, &sa->key);
262 typeb = key_type(c, &sb->key);
263 ubifs_assert(typea != UBIFS_DATA_KEY && typeb != UBIFS_DATA_KEY);
264
265 /* Inodes go before directory entries */
266 if (typea == UBIFS_INO_KEY) {
267 if (typeb == UBIFS_INO_KEY)
268 return sb->len - sa->len;
269 return -1;
270 }
271 if (typeb == UBIFS_INO_KEY)
272 return 1;
273
274 ubifs_assert(typea == UBIFS_DENT_KEY && typeb == UBIFS_DENT_KEY);
275 inuma = key_inum(c, &sa->key);
276 inumb = key_inum(c, &sb->key);
277
278 if (inuma == inumb) {
279 uint32_t hasha = key_hash(c, &sa->key);
280 uint32_t hashb = key_hash(c, &sb->key);
281
282 if (hasha <= hashb)
283 return -1;
284 } else if (inuma <= inumb)
285 return -1;
286
287 return 1;
288}
289
290/**
291 * sort_nodes - sort nodes for GC.
292 * @c: UBIFS file-system description object
293 * @sleb: describes nodes to sort and contains the result on exit
294 * @nondata: contains non-data nodes on exit
295 * @min: minimum node size is returned here
296 *
297 * This function sorts the list of inodes to garbage collect. First of all, it
298 * kills obsolete nodes and separates data and non-data nodes to the
299 * @sleb->nodes and @nondata lists correspondingly.
300 *
301 * Data nodes are then sorted in block number order - this is important for
302 * bulk-read; data nodes with lower inode number go before data nodes with
303 * higher inode number, and data nodes with lower block number go before data
304 * nodes with higher block number;
305 *
306 * Non-data nodes are sorted as follows.
307 * o First go inode nodes - they are sorted in descending length order.
308 * o Then go directory entry nodes - they are sorted in hash order, which
309 * should supposedly optimize 'readdir()'. Direntry nodes with lower parent
310 * inode number go before direntry nodes with higher parent inode number,
311 * and direntry nodes with lower name hash values go before direntry nodes
312 * with higher name hash values.
313 *
314 * This function returns zero in case of success and a negative error code in
315 * case of failure.
316 */
317static int sort_nodes(struct ubifs_info *c, struct ubifs_scan_leb *sleb,
318 struct list_head *nondata, int *min)
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300319{
320 struct ubifs_scan_node *snod, *tmp;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300321
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200322 *min = INT_MAX;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300323
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200324 /* Separate data nodes and non-data nodes */
325 list_for_each_entry_safe(snod, tmp, &sleb->nodes, list) {
326 int err;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300327
328 ubifs_assert(snod->type != UBIFS_IDX_NODE);
329 ubifs_assert(snod->type != UBIFS_REF_NODE);
330 ubifs_assert(snod->type != UBIFS_CS_NODE);
331
332 err = ubifs_tnc_has_node(c, &snod->key, 0, sleb->lnum,
333 snod->offs, 0);
334 if (err < 0)
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200335 return err;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300336
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300337 if (!err) {
338 /* The node is obsolete, remove it from the list */
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200339 list_del(&snod->list);
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300340 kfree(snod);
341 continue;
342 }
343
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200344 if (snod->len < *min)
345 *min = snod->len;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300346
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200347 if (key_type(c, &snod->key) != UBIFS_DATA_KEY)
348 list_move_tail(&snod->list, nondata);
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300349 }
350
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200351 /* Sort data and non-data nodes */
352 list_sort(c, &sleb->nodes, &data_nodes_cmp);
353 list_sort(c, nondata, &nondata_nodes_cmp);
354 return 0;
355}
356
357/**
358 * move_node - move a node.
359 * @c: UBIFS file-system description object
360 * @sleb: describes the LEB to move nodes from
361 * @snod: the mode to move
362 * @wbuf: write-buffer to move node to
363 *
364 * This function moves node @snod to @wbuf, changes TNC correspondingly, and
365 * destroys @snod. Returns zero in case of success and a negative error code in
366 * case of failure.
367 */
368static int move_node(struct ubifs_info *c, struct ubifs_scan_leb *sleb,
369 struct ubifs_scan_node *snod, struct ubifs_wbuf *wbuf)
370{
371 int err, new_lnum = wbuf->lnum, new_offs = wbuf->offs + wbuf->used;
372
373 cond_resched();
374 err = ubifs_wbuf_write_nolock(wbuf, snod->node, snod->len);
375 if (err)
376 return err;
377
378 err = ubifs_tnc_replace(c, &snod->key, sleb->lnum,
379 snod->offs, new_lnum, new_offs,
380 snod->len);
381 list_del(&snod->list);
382 kfree(snod);
383 return err;
384}
385
386/**
387 * move_nodes - move nodes.
388 * @c: UBIFS file-system description object
389 * @sleb: describes the LEB to move nodes from
390 *
391 * This function moves valid nodes from data LEB described by @sleb to the GC
392 * journal head. This function returns zero in case of success, %-EAGAIN if
393 * commit is required, and other negative error codes in case of other
394 * failures.
395 */
396static int move_nodes(struct ubifs_info *c, struct ubifs_scan_leb *sleb)
397{
398 int err, min;
399 LIST_HEAD(nondata);
400 struct ubifs_wbuf *wbuf = &c->jheads[GCHD].wbuf;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300401
402 if (wbuf->lnum == -1) {
403 /*
404 * The GC journal head is not set, because it is the first GC
405 * invocation since mount.
406 */
407 err = switch_gc_head(c);
408 if (err)
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200409 return err;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300410 }
411
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200412 err = sort_nodes(c, sleb, &nondata, &min);
413 if (err)
414 goto out;
415
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300416 /* Write nodes to their new location. Use the first-fit strategy */
417 while (1) {
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200418 int avail;
419 struct ubifs_scan_node *snod, *tmp;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300420
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200421 /* Move data nodes */
422 list_for_each_entry_safe(snod, tmp, &sleb->nodes, list) {
423 avail = c->leb_size - wbuf->offs - wbuf->used;
424 if (snod->len > avail)
425 /*
426 * Do not skip data nodes in order to optimize
427 * bulk-read.
428 */
429 break;
430
431 err = move_node(c, sleb, snod, wbuf);
432 if (err)
433 goto out;
434 }
435
436 /* Move non-data nodes */
437 list_for_each_entry_safe(snod, tmp, &nondata, list) {
438 avail = c->leb_size - wbuf->offs - wbuf->used;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300439 if (avail < min)
440 break;
441
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200442 if (snod->len > avail) {
443 /*
444 * Keep going only if this is an inode with
445 * some data. Otherwise stop and switch the GC
446 * head. IOW, we assume that data-less inode
447 * nodes and direntry nodes are roughly of the
448 * same size.
449 */
450 if (key_type(c, &snod->key) == UBIFS_DENT_KEY ||
451 snod->len == UBIFS_INO_NODE_SZ)
452 break;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300453 continue;
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200454 }
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300455
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200456 err = move_node(c, sleb, snod, wbuf);
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300457 if (err)
458 goto out;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300459 }
460
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200461 if (list_empty(&sleb->nodes) && list_empty(&nondata))
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300462 break;
463
464 /*
465 * Waste the rest of the space in the LEB and switch to the
466 * next LEB.
467 */
468 err = switch_gc_head(c);
469 if (err)
470 goto out;
471 }
472
473 return 0;
474
475out:
Artem Bityutskiyf10770f2009-03-08 15:13:00 +0200476 list_splice_tail(&nondata, &sleb->nodes);
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300477 return err;
478}
479
480/**
481 * gc_sync_wbufs - sync write-buffers for GC.
482 * @c: UBIFS file-system description object
483 *
484 * We must guarantee that obsoleting nodes are on flash. Unfortunately they may
485 * be in a write-buffer instead. That is, a node could be written to a
486 * write-buffer, obsoleting another node in a LEB that is GC'd. If that LEB is
487 * erased before the write-buffer is sync'd and then there is an unclean
488 * unmount, then an existing node is lost. To avoid this, we sync all
489 * write-buffers.
490 *
491 * This function returns %0 on success or a negative error code on failure.
492 */
493static int gc_sync_wbufs(struct ubifs_info *c)
494{
495 int err, i;
496
497 for (i = 0; i < c->jhead_cnt; i++) {
498 if (i == GCHD)
499 continue;
500 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
501 if (err)
502 return err;
503 }
504 return 0;
505}
506
507/**
508 * ubifs_garbage_collect_leb - garbage-collect a logical eraseblock.
509 * @c: UBIFS file-system description object
510 * @lp: describes the LEB to garbage collect
511 *
512 * This function garbage-collects an LEB and returns one of the @LEB_FREED,
513 * @LEB_RETAINED, etc positive codes in case of success, %-EAGAIN if commit is
514 * required, and other negative error codes in case of failures.
515 */
516int ubifs_garbage_collect_leb(struct ubifs_info *c, struct ubifs_lprops *lp)
517{
518 struct ubifs_scan_leb *sleb;
519 struct ubifs_scan_node *snod;
520 struct ubifs_wbuf *wbuf = &c->jheads[GCHD].wbuf;
521 int err = 0, lnum = lp->lnum;
522
523 ubifs_assert(c->gc_lnum != -1 || wbuf->offs + wbuf->used == 0 ||
524 c->need_recovery);
525 ubifs_assert(c->gc_lnum != lnum);
526 ubifs_assert(wbuf->lnum != lnum);
527
528 /*
529 * We scan the entire LEB even though we only really need to scan up to
530 * (c->leb_size - lp->free).
531 */
532 sleb = ubifs_scan(c, lnum, 0, c->sbuf);
533 if (IS_ERR(sleb))
534 return PTR_ERR(sleb);
535
536 ubifs_assert(!list_empty(&sleb->nodes));
537 snod = list_entry(sleb->nodes.next, struct ubifs_scan_node, list);
538
539 if (snod->type == UBIFS_IDX_NODE) {
540 struct ubifs_gced_idx_leb *idx_gc;
541
542 dbg_gc("indexing LEB %d (free %d, dirty %d)",
543 lnum, lp->free, lp->dirty);
544 list_for_each_entry(snod, &sleb->nodes, list) {
545 struct ubifs_idx_node *idx = snod->node;
546 int level = le16_to_cpu(idx->level);
547
548 ubifs_assert(snod->type == UBIFS_IDX_NODE);
549 key_read(c, ubifs_idx_key(c, idx), &snod->key);
550 err = ubifs_dirty_idx_node(c, &snod->key, level, lnum,
551 snod->offs);
552 if (err)
553 goto out;
554 }
555
556 idx_gc = kmalloc(sizeof(struct ubifs_gced_idx_leb), GFP_NOFS);
557 if (!idx_gc) {
558 err = -ENOMEM;
559 goto out;
560 }
561
562 idx_gc->lnum = lnum;
563 idx_gc->unmap = 0;
564 list_add(&idx_gc->list, &c->idx_gc);
565
566 /*
567 * Don't release the LEB until after the next commit, because
Adrian Hunter227c75c2009-01-29 11:53:51 +0200568 * it may contain data which is needed for recovery. So
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300569 * although we freed this LEB, it will become usable only after
570 * the commit.
571 */
572 err = ubifs_change_one_lp(c, lnum, c->leb_size, 0, 0,
573 LPROPS_INDEX, 1);
574 if (err)
575 goto out;
576 err = LEB_FREED_IDX;
577 } else {
578 dbg_gc("data LEB %d (free %d, dirty %d)",
579 lnum, lp->free, lp->dirty);
580
581 err = move_nodes(c, sleb);
582 if (err)
Adrian Hunter6dcfac42008-09-12 12:27:47 +0300583 goto out_inc_seq;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300584
585 err = gc_sync_wbufs(c);
586 if (err)
Adrian Hunter6dcfac42008-09-12 12:27:47 +0300587 goto out_inc_seq;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300588
589 err = ubifs_change_one_lp(c, lnum, c->leb_size, 0, 0, 0, 0);
590 if (err)
Adrian Hunter6dcfac42008-09-12 12:27:47 +0300591 goto out_inc_seq;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300592
Adrian Hunter601c0bc2008-08-22 14:23:35 +0300593 /* Allow for races with TNC */
594 c->gced_lnum = lnum;
595 smp_wmb();
596 c->gc_seq += 1;
597 smp_wmb();
598
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300599 if (c->gc_lnum == -1) {
600 c->gc_lnum = lnum;
601 err = LEB_RETAINED;
602 } else {
603 err = ubifs_wbuf_sync_nolock(wbuf);
604 if (err)
605 goto out;
606
607 err = ubifs_leb_unmap(c, lnum);
608 if (err)
609 goto out;
610
611 err = LEB_FREED;
612 }
613 }
614
615out:
616 ubifs_scan_destroy(sleb);
617 return err;
Adrian Hunter6dcfac42008-09-12 12:27:47 +0300618
619out_inc_seq:
620 /* We may have moved at least some nodes so allow for races with TNC */
621 c->gced_lnum = lnum;
622 smp_wmb();
623 c->gc_seq += 1;
624 smp_wmb();
625 goto out;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300626}
627
628/**
629 * ubifs_garbage_collect - UBIFS garbage collector.
630 * @c: UBIFS file-system description object
631 * @anyway: do GC even if there are free LEBs
632 *
633 * This function does out-of-place garbage collection. The return codes are:
634 * o positive LEB number if the LEB has been freed and may be used;
635 * o %-EAGAIN if the caller has to run commit;
636 * o %-ENOSPC if GC failed to make any progress;
637 * o other negative error codes in case of other errors.
638 *
639 * Garbage collector writes data to the journal when GC'ing data LEBs, and just
640 * marking indexing nodes dirty when GC'ing indexing LEBs. Thus, at some point
641 * commit may be required. But commit cannot be run from inside GC, because the
642 * caller might be holding the commit lock, so %-EAGAIN is returned instead;
643 * And this error code means that the caller has to run commit, and re-run GC
644 * if there is still no free space.
645 *
646 * There are many reasons why this function may return %-EAGAIN:
647 * o the log is full and there is no space to write an LEB reference for
648 * @c->gc_lnum;
649 * o the journal is too large and exceeds size limitations;
650 * o GC moved indexing LEBs, but they can be used only after the commit;
651 * o the shrinker fails to find clean znodes to free and requests the commit;
652 * o etc.
653 *
654 * Note, if the file-system is close to be full, this function may return
655 * %-EAGAIN infinitely, so the caller has to limit amount of re-invocations of
656 * the function. E.g., this happens if the limits on the journal size are too
657 * tough and GC writes too much to the journal before an LEB is freed. This
658 * might also mean that the journal is too large, and the TNC becomes to big,
659 * so that the shrinker is constantly called, finds not clean znodes to free,
660 * and requests commit. Well, this may also happen if the journal is all right,
661 * but another kernel process consumes too much memory. Anyway, infinite
662 * %-EAGAIN may happen, but in some extreme/misconfiguration cases.
663 */
664int ubifs_garbage_collect(struct ubifs_info *c, int anyway)
665{
666 int i, err, ret, min_space = c->dead_wm;
667 struct ubifs_lprops lp;
668 struct ubifs_wbuf *wbuf = &c->jheads[GCHD].wbuf;
669
670 ubifs_assert_cmt_locked(c);
671
672 if (ubifs_gc_should_commit(c))
673 return -EAGAIN;
674
675 mutex_lock_nested(&wbuf->io_mutex, wbuf->jhead);
676
677 if (c->ro_media) {
678 ret = -EROFS;
679 goto out_unlock;
680 }
681
682 /* We expect the write-buffer to be empty on entry */
683 ubifs_assert(!wbuf->used);
684
685 for (i = 0; ; i++) {
686 int space_before = c->leb_size - wbuf->offs - wbuf->used;
687 int space_after;
688
689 cond_resched();
690
691 /* Give the commit an opportunity to run */
692 if (ubifs_gc_should_commit(c)) {
693 ret = -EAGAIN;
694 break;
695 }
696
697 if (i > SOFT_LEBS_LIMIT && !list_empty(&c->idx_gc)) {
698 /*
699 * We've done enough iterations. Indexing LEBs were
700 * moved and will be available after the commit.
701 */
702 dbg_gc("soft limit, some index LEBs GC'ed, -EAGAIN");
703 ubifs_commit_required(c);
704 ret = -EAGAIN;
705 break;
706 }
707
708 if (i > HARD_LEBS_LIMIT) {
709 /*
710 * We've moved too many LEBs and have not made
711 * progress, give up.
712 */
713 dbg_gc("hard limit, -ENOSPC");
714 ret = -ENOSPC;
715 break;
716 }
717
718 /*
719 * Empty and freeable LEBs can turn up while we waited for
720 * the wbuf lock, or while we have been running GC. In that
721 * case, we should just return one of those instead of
722 * continuing to GC dirty LEBs. Hence we request
723 * 'ubifs_find_dirty_leb()' to return an empty LEB if it can.
724 */
725 ret = ubifs_find_dirty_leb(c, &lp, min_space, anyway ? 0 : 1);
726 if (ret) {
727 if (ret == -ENOSPC)
728 dbg_gc("no more dirty LEBs");
729 break;
730 }
731
732 dbg_gc("found LEB %d: free %d, dirty %d, sum %d "
733 "(min. space %d)", lp.lnum, lp.free, lp.dirty,
734 lp.free + lp.dirty, min_space);
735
736 if (lp.free + lp.dirty == c->leb_size) {
737 /* An empty LEB was returned */
738 dbg_gc("LEB %d is free, return it", lp.lnum);
739 /*
740 * ubifs_find_dirty_leb() doesn't return freeable index
741 * LEBs.
742 */
743 ubifs_assert(!(lp.flags & LPROPS_INDEX));
744 if (lp.free != c->leb_size) {
745 /*
746 * Write buffers must be sync'd before
747 * unmapping freeable LEBs, because one of them
748 * may contain data which obsoletes something
749 * in 'lp.pnum'.
750 */
751 ret = gc_sync_wbufs(c);
752 if (ret)
753 goto out;
754 ret = ubifs_change_one_lp(c, lp.lnum,
755 c->leb_size, 0, 0, 0,
756 0);
757 if (ret)
758 goto out;
759 }
760 ret = ubifs_leb_unmap(c, lp.lnum);
761 if (ret)
762 goto out;
763 ret = lp.lnum;
764 break;
765 }
766
767 space_before = c->leb_size - wbuf->offs - wbuf->used;
768 if (wbuf->lnum == -1)
769 space_before = 0;
770
771 ret = ubifs_garbage_collect_leb(c, &lp);
772 if (ret < 0) {
773 if (ret == -EAGAIN || ret == -ENOSPC) {
774 /*
775 * These codes are not errors, so we have to
776 * return the LEB to lprops. But if the
777 * 'ubifs_return_leb()' function fails, its
778 * failure code is propagated to the caller
779 * instead of the original '-EAGAIN' or
780 * '-ENOSPC'.
781 */
782 err = ubifs_return_leb(c, lp.lnum);
783 if (err)
784 ret = err;
785 break;
786 }
787 goto out;
788 }
789
790 if (ret == LEB_FREED) {
791 /* An LEB has been freed and is ready for use */
792 dbg_gc("LEB %d freed, return", lp.lnum);
793 ret = lp.lnum;
794 break;
795 }
796
797 if (ret == LEB_FREED_IDX) {
798 /*
799 * This was an indexing LEB and it cannot be
800 * immediately used. And instead of requesting the
801 * commit straight away, we try to garbage collect some
802 * more.
803 */
804 dbg_gc("indexing LEB %d freed, continue", lp.lnum);
805 continue;
806 }
807
808 ubifs_assert(ret == LEB_RETAINED);
809 space_after = c->leb_size - wbuf->offs - wbuf->used;
810 dbg_gc("LEB %d retained, freed %d bytes", lp.lnum,
811 space_after - space_before);
812
813 if (space_after > space_before) {
814 /* GC makes progress, keep working */
815 min_space >>= 1;
816 if (min_space < c->dead_wm)
817 min_space = c->dead_wm;
818 continue;
819 }
820
821 dbg_gc("did not make progress");
822
823 /*
824 * GC moved an LEB bud have not done any progress. This means
825 * that the previous GC head LEB contained too few free space
826 * and the LEB which was GC'ed contained only large nodes which
827 * did not fit that space.
828 *
829 * We can do 2 things:
830 * 1. pick another LEB in a hope it'll contain a small node
831 * which will fit the space we have at the end of current GC
832 * head LEB, but there is no guarantee, so we try this out
833 * unless we have already been working for too long;
834 * 2. request an LEB with more dirty space, which will force
835 * 'ubifs_find_dirty_leb()' to start scanning the lprops
836 * table, instead of just picking one from the heap
837 * (previously it already picked the dirtiest LEB).
838 */
839 if (i < SOFT_LEBS_LIMIT) {
840 dbg_gc("try again");
841 continue;
842 }
843
844 min_space <<= 1;
845 if (min_space > c->dark_wm)
846 min_space = c->dark_wm;
847 dbg_gc("set min. space to %d", min_space);
848 }
849
850 if (ret == -ENOSPC && !list_empty(&c->idx_gc)) {
851 dbg_gc("no space, some index LEBs GC'ed, -EAGAIN");
852 ubifs_commit_required(c);
853 ret = -EAGAIN;
854 }
855
856 err = ubifs_wbuf_sync_nolock(wbuf);
857 if (!err)
858 err = ubifs_leb_unmap(c, c->gc_lnum);
859 if (err) {
860 ret = err;
861 goto out;
862 }
863out_unlock:
864 mutex_unlock(&wbuf->io_mutex);
865 return ret;
866
867out:
868 ubifs_assert(ret < 0);
869 ubifs_assert(ret != -ENOSPC && ret != -EAGAIN);
870 ubifs_ro_mode(c, ret);
871 ubifs_wbuf_sync_nolock(wbuf);
872 mutex_unlock(&wbuf->io_mutex);
873 ubifs_return_leb(c, lp.lnum);
874 return ret;
875}
876
877/**
878 * ubifs_gc_start_commit - garbage collection at start of commit.
879 * @c: UBIFS file-system description object
880 *
881 * If a LEB has only dirty and free space, then we may safely unmap it and make
882 * it free. Note, we cannot do this with indexing LEBs because dirty space may
883 * correspond index nodes that are required for recovery. In that case, the
884 * LEB cannot be unmapped until after the next commit.
885 *
886 * This function returns %0 upon success and a negative error code upon failure.
887 */
888int ubifs_gc_start_commit(struct ubifs_info *c)
889{
890 struct ubifs_gced_idx_leb *idx_gc;
891 const struct ubifs_lprops *lp;
892 int err = 0, flags;
893
894 ubifs_get_lprops(c);
895
896 /*
897 * Unmap (non-index) freeable LEBs. Note that recovery requires that all
898 * wbufs are sync'd before this, which is done in 'do_commit()'.
899 */
900 while (1) {
901 lp = ubifs_fast_find_freeable(c);
Hirofumi Nakagawa8d47aef2008-08-21 17:16:40 +0300902 if (IS_ERR(lp)) {
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300903 err = PTR_ERR(lp);
904 goto out;
905 }
906 if (!lp)
907 break;
908 ubifs_assert(!(lp->flags & LPROPS_TAKEN));
909 ubifs_assert(!(lp->flags & LPROPS_INDEX));
910 err = ubifs_leb_unmap(c, lp->lnum);
911 if (err)
912 goto out;
913 lp = ubifs_change_lp(c, lp, c->leb_size, 0, lp->flags, 0);
Hirofumi Nakagawa8d47aef2008-08-21 17:16:40 +0300914 if (IS_ERR(lp)) {
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300915 err = PTR_ERR(lp);
916 goto out;
917 }
918 ubifs_assert(!(lp->flags & LPROPS_TAKEN));
919 ubifs_assert(!(lp->flags & LPROPS_INDEX));
920 }
921
922 /* Mark GC'd index LEBs OK to unmap after this commit finishes */
923 list_for_each_entry(idx_gc, &c->idx_gc, list)
924 idx_gc->unmap = 1;
925
926 /* Record index freeable LEBs for unmapping after commit */
927 while (1) {
928 lp = ubifs_fast_find_frdi_idx(c);
Hirofumi Nakagawa8d47aef2008-08-21 17:16:40 +0300929 if (IS_ERR(lp)) {
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300930 err = PTR_ERR(lp);
931 goto out;
932 }
933 if (!lp)
934 break;
935 idx_gc = kmalloc(sizeof(struct ubifs_gced_idx_leb), GFP_NOFS);
936 if (!idx_gc) {
937 err = -ENOMEM;
938 goto out;
939 }
940 ubifs_assert(!(lp->flags & LPROPS_TAKEN));
941 ubifs_assert(lp->flags & LPROPS_INDEX);
942 /* Don't release the LEB until after the next commit */
943 flags = (lp->flags | LPROPS_TAKEN) ^ LPROPS_INDEX;
944 lp = ubifs_change_lp(c, lp, c->leb_size, 0, flags, 1);
Hirofumi Nakagawa8d47aef2008-08-21 17:16:40 +0300945 if (IS_ERR(lp)) {
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300946 err = PTR_ERR(lp);
947 kfree(idx_gc);
948 goto out;
949 }
950 ubifs_assert(lp->flags & LPROPS_TAKEN);
951 ubifs_assert(!(lp->flags & LPROPS_INDEX));
952 idx_gc->lnum = lp->lnum;
953 idx_gc->unmap = 1;
954 list_add(&idx_gc->list, &c->idx_gc);
955 }
956out:
957 ubifs_release_lprops(c);
958 return err;
959}
960
961/**
962 * ubifs_gc_end_commit - garbage collection at end of commit.
963 * @c: UBIFS file-system description object
964 *
965 * This function completes out-of-place garbage collection of index LEBs.
966 */
967int ubifs_gc_end_commit(struct ubifs_info *c)
968{
969 struct ubifs_gced_idx_leb *idx_gc, *tmp;
970 struct ubifs_wbuf *wbuf;
971 int err = 0;
972
973 wbuf = &c->jheads[GCHD].wbuf;
974 mutex_lock_nested(&wbuf->io_mutex, wbuf->jhead);
975 list_for_each_entry_safe(idx_gc, tmp, &c->idx_gc, list)
976 if (idx_gc->unmap) {
977 dbg_gc("LEB %d", idx_gc->lnum);
978 err = ubifs_leb_unmap(c, idx_gc->lnum);
979 if (err)
980 goto out;
981 err = ubifs_change_one_lp(c, idx_gc->lnum, LPROPS_NC,
982 LPROPS_NC, 0, LPROPS_TAKEN, -1);
983 if (err)
984 goto out;
985 list_del(&idx_gc->list);
986 kfree(idx_gc);
987 }
988out:
989 mutex_unlock(&wbuf->io_mutex);
990 return err;
991}
992
993/**
994 * ubifs_destroy_idx_gc - destroy idx_gc list.
995 * @c: UBIFS file-system description object
996 *
Adrian Hunterb466f172009-01-29 12:59:33 +0200997 * This function destroys the @c->idx_gc list. It is called when unmounting
998 * so locks are not needed. Returns zero in case of success and a negative
999 * error code in case of failure.
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001000 */
Adrian Hunterb466f172009-01-29 12:59:33 +02001001void ubifs_destroy_idx_gc(struct ubifs_info *c)
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001002{
1003 while (!list_empty(&c->idx_gc)) {
1004 struct ubifs_gced_idx_leb *idx_gc;
1005
1006 idx_gc = list_entry(c->idx_gc.next, struct ubifs_gced_idx_leb,
1007 list);
Adrian Hunterb466f172009-01-29 12:59:33 +02001008 c->idx_gc_cnt -= 1;
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001009 list_del(&idx_gc->list);
1010 kfree(idx_gc);
1011 }
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001012}
1013
1014/**
1015 * ubifs_get_idx_gc_leb - get a LEB from GC'd index LEB list.
1016 * @c: UBIFS file-system description object
1017 *
1018 * Called during start commit so locks are not needed.
1019 */
1020int ubifs_get_idx_gc_leb(struct ubifs_info *c)
1021{
1022 struct ubifs_gced_idx_leb *idx_gc;
1023 int lnum;
1024
1025 if (list_empty(&c->idx_gc))
1026 return -ENOSPC;
1027 idx_gc = list_entry(c->idx_gc.next, struct ubifs_gced_idx_leb, list);
1028 lnum = idx_gc->lnum;
1029 /* c->idx_gc_cnt is updated by the caller when lprops are updated */
1030 list_del(&idx_gc->list);
1031 kfree(idx_gc);
1032 return lnum;
1033}