Merge git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-2.6-frv
[linux-2.6] / fs / jffs2 / build.c
1 /*
2  * JFFS2 -- Journalling Flash File System, Version 2.
3  *
4  * Copyright © 2001-2007 Red Hat, Inc.
5  *
6  * Created by David Woodhouse <dwmw2@infradead.org>
7  *
8  * For licensing information, see the file 'LICENCE' in this directory.
9  *
10  */
11
12 #include <linux/kernel.h>
13 #include <linux/sched.h>
14 #include <linux/slab.h>
15 #include <linux/vmalloc.h>
16 #include <linux/mtd/mtd.h>
17 #include "nodelist.h"
18
19 static void jffs2_build_remove_unlinked_inode(struct jffs2_sb_info *,
20                 struct jffs2_inode_cache *, struct jffs2_full_dirent **);
21
22 static inline struct jffs2_inode_cache *
23 first_inode_chain(int *i, struct jffs2_sb_info *c)
24 {
25         for (; *i < INOCACHE_HASHSIZE; (*i)++) {
26                 if (c->inocache_list[*i])
27                         return c->inocache_list[*i];
28         }
29         return NULL;
30 }
31
32 static inline struct jffs2_inode_cache *
33 next_inode(int *i, struct jffs2_inode_cache *ic, struct jffs2_sb_info *c)
34 {
35         /* More in this chain? */
36         if (ic->next)
37                 return ic->next;
38         (*i)++;
39         return first_inode_chain(i, c);
40 }
41
42 #define for_each_inode(i, c, ic)                        \
43         for (i = 0, ic = first_inode_chain(&i, (c));    \
44              ic;                                        \
45              ic = next_inode(&i, ic, (c)))
46
47
48 static void jffs2_build_inode_pass1(struct jffs2_sb_info *c,
49                                     struct jffs2_inode_cache *ic)
50 {
51         struct jffs2_full_dirent *fd;
52
53         dbg_fsbuild("building directory inode #%u\n", ic->ino);
54
55         /* For each child, increase nlink */
56         for(fd = ic->scan_dents; fd; fd = fd->next) {
57                 struct jffs2_inode_cache *child_ic;
58                 if (!fd->ino)
59                         continue;
60
61                 /* we can get high latency here with huge directories */
62
63                 child_ic = jffs2_get_ino_cache(c, fd->ino);
64                 if (!child_ic) {
65                         dbg_fsbuild("child \"%s\" (ino #%u) of dir ino #%u doesn't exist!\n",
66                                   fd->name, fd->ino, ic->ino);
67                         jffs2_mark_node_obsolete(c, fd->raw);
68                         continue;
69                 }
70
71                 if (fd->type == DT_DIR) {
72                         if (child_ic->pino_nlink) {
73                                 JFFS2_ERROR("child dir \"%s\" (ino #%u) of dir ino #%u appears to be a hard link\n",
74                                             fd->name, fd->ino, ic->ino);
75                                 /* TODO: What do we do about it? */
76                         } else {
77                                 child_ic->pino_nlink = ic->ino;
78                         }
79                 } else
80                         child_ic->pino_nlink++;
81
82                 dbg_fsbuild("increased nlink for child \"%s\" (ino #%u)\n", fd->name, fd->ino);
83                 /* Can't free scan_dents so far. We might need them in pass 2 */
84         }
85 }
86
87 /* Scan plan:
88  - Scan physical nodes. Build map of inodes/dirents. Allocate inocaches as we go
89  - Scan directory tree from top down, setting nlink in inocaches
90  - Scan inocaches for inodes with nlink==0
91 */
92 static int jffs2_build_filesystem(struct jffs2_sb_info *c)
93 {
94         int ret;
95         int i;
96         struct jffs2_inode_cache *ic;
97         struct jffs2_full_dirent *fd;
98         struct jffs2_full_dirent *dead_fds = NULL;
99
100         dbg_fsbuild("build FS data structures\n");
101
102         /* First, scan the medium and build all the inode caches with
103            lists of physical nodes */
104
105         c->flags |= JFFS2_SB_FLAG_SCANNING;
106         ret = jffs2_scan_medium(c);
107         c->flags &= ~JFFS2_SB_FLAG_SCANNING;
108         if (ret)
109                 goto exit;
110
111         dbg_fsbuild("scanned flash completely\n");
112         jffs2_dbg_dump_block_lists_nolock(c);
113
114         dbg_fsbuild("pass 1 starting\n");
115         c->flags |= JFFS2_SB_FLAG_BUILDING;
116         /* Now scan the directory tree, increasing nlink according to every dirent found. */
117         for_each_inode(i, c, ic) {
118                 if (ic->scan_dents) {
119                         jffs2_build_inode_pass1(c, ic);
120                         cond_resched();
121                 }
122         }
123
124         dbg_fsbuild("pass 1 complete\n");
125
126         /* Next, scan for inodes with nlink == 0 and remove them. If
127            they were directories, then decrement the nlink of their
128            children too, and repeat the scan. As that's going to be
129            a fairly uncommon occurrence, it's not so evil to do it this
130            way. Recursion bad. */
131         dbg_fsbuild("pass 2 starting\n");
132
133         for_each_inode(i, c, ic) {
134                 if (ic->pino_nlink)
135                         continue;
136
137                 jffs2_build_remove_unlinked_inode(c, ic, &dead_fds);
138                 cond_resched();
139         }
140
141         dbg_fsbuild("pass 2a starting\n");
142
143         while (dead_fds) {
144                 fd = dead_fds;
145                 dead_fds = fd->next;
146
147                 ic = jffs2_get_ino_cache(c, fd->ino);
148
149                 if (ic)
150                         jffs2_build_remove_unlinked_inode(c, ic, &dead_fds);
151                 jffs2_free_full_dirent(fd);
152         }
153
154         dbg_fsbuild("pass 2a complete\n");
155         dbg_fsbuild("freeing temporary data structures\n");
156
157         /* Finally, we can scan again and free the dirent structs */
158         for_each_inode(i, c, ic) {
159                 while(ic->scan_dents) {
160                         fd = ic->scan_dents;
161                         ic->scan_dents = fd->next;
162                         jffs2_free_full_dirent(fd);
163                 }
164                 ic->scan_dents = NULL;
165                 cond_resched();
166         }
167         jffs2_build_xattr_subsystem(c);
168         c->flags &= ~JFFS2_SB_FLAG_BUILDING;
169
170         dbg_fsbuild("FS build complete\n");
171
172         /* Rotate the lists by some number to ensure wear levelling */
173         jffs2_rotate_lists(c);
174
175         ret = 0;
176
177 exit:
178         if (ret) {
179                 for_each_inode(i, c, ic) {
180                         while(ic->scan_dents) {
181                                 fd = ic->scan_dents;
182                                 ic->scan_dents = fd->next;
183                                 jffs2_free_full_dirent(fd);
184                         }
185                 }
186                 jffs2_clear_xattr_subsystem(c);
187         }
188
189         return ret;
190 }
191
192 static void jffs2_build_remove_unlinked_inode(struct jffs2_sb_info *c,
193                                         struct jffs2_inode_cache *ic,
194                                         struct jffs2_full_dirent **dead_fds)
195 {
196         struct jffs2_raw_node_ref *raw;
197         struct jffs2_full_dirent *fd;
198
199         dbg_fsbuild("removing ino #%u with nlink == zero.\n", ic->ino);
200
201         raw = ic->nodes;
202         while (raw != (void *)ic) {
203                 struct jffs2_raw_node_ref *next = raw->next_in_ino;
204                 dbg_fsbuild("obsoleting node at 0x%08x\n", ref_offset(raw));
205                 jffs2_mark_node_obsolete(c, raw);
206                 raw = next;
207         }
208
209         if (ic->scan_dents) {
210                 int whinged = 0;
211                 dbg_fsbuild("inode #%u was a directory which may have children...\n", ic->ino);
212
213                 while(ic->scan_dents) {
214                         struct jffs2_inode_cache *child_ic;
215
216                         fd = ic->scan_dents;
217                         ic->scan_dents = fd->next;
218
219                         if (!fd->ino) {
220                                 /* It's a deletion dirent. Ignore it */
221                                 dbg_fsbuild("child \"%s\" is a deletion dirent, skipping...\n", fd->name);
222                                 jffs2_free_full_dirent(fd);
223                                 continue;
224                         }
225                         if (!whinged)
226                                 whinged = 1;
227
228                         dbg_fsbuild("removing child \"%s\", ino #%u\n", fd->name, fd->ino);
229
230                         child_ic = jffs2_get_ino_cache(c, fd->ino);
231                         if (!child_ic) {
232                                 dbg_fsbuild("cannot remove child \"%s\", ino #%u, because it doesn't exist\n",
233                                                 fd->name, fd->ino);
234                                 jffs2_free_full_dirent(fd);
235                                 continue;
236                         }
237
238                         /* Reduce nlink of the child. If it's now zero, stick it on the
239                            dead_fds list to be cleaned up later. Else just free the fd */
240
241                         if (fd->type == DT_DIR)
242                                 child_ic->pino_nlink = 0;
243                         else
244                                 child_ic->pino_nlink--;
245
246                         if (!child_ic->pino_nlink) {
247                                 dbg_fsbuild("inode #%u (\"%s\") now has no links; adding to dead_fds list.\n",
248                                           fd->ino, fd->name);
249                                 fd->next = *dead_fds;
250                                 *dead_fds = fd;
251                         } else {
252                                 dbg_fsbuild("inode #%u (\"%s\") has now got nlink %d. Ignoring.\n",
253                                           fd->ino, fd->name, child_ic->pino_nlink);
254                                 jffs2_free_full_dirent(fd);
255                         }
256                 }
257         }
258
259         /*
260            We don't delete the inocache from the hash list and free it yet.
261            The erase code will do that, when all the nodes are completely gone.
262         */
263 }
264
265 static void jffs2_calc_trigger_levels(struct jffs2_sb_info *c)
266 {
267         uint32_t size;
268
269         /* Deletion should almost _always_ be allowed. We're fairly
270            buggered once we stop allowing people to delete stuff
271            because there's not enough free space... */
272         c->resv_blocks_deletion = 2;
273
274         /* Be conservative about how much space we need before we allow writes.
275            On top of that which is required for deletia, require an extra 2%
276            of the medium to be available, for overhead caused by nodes being
277            split across blocks, etc. */
278
279         size = c->flash_size / 50; /* 2% of flash size */
280         size += c->nr_blocks * 100; /* And 100 bytes per eraseblock */
281         size += c->sector_size - 1; /* ... and round up */
282
283         c->resv_blocks_write = c->resv_blocks_deletion + (size / c->sector_size);
284
285         /* When do we let the GC thread run in the background */
286
287         c->resv_blocks_gctrigger = c->resv_blocks_write + 1;
288
289         /* When do we allow garbage collection to merge nodes to make
290            long-term progress at the expense of short-term space exhaustion? */
291         c->resv_blocks_gcmerge = c->resv_blocks_deletion + 1;
292
293         /* When do we allow garbage collection to eat from bad blocks rather
294            than actually making progress? */
295         c->resv_blocks_gcbad = 0;//c->resv_blocks_deletion + 2;
296
297         /* What number of 'very dirty' eraseblocks do we allow before we
298            trigger the GC thread even if we don't _need_ the space. When we
299            can't mark nodes obsolete on the medium, the old dirty nodes cause
300            performance problems because we have to inspect and discard them. */
301         c->vdirty_blocks_gctrigger = c->resv_blocks_gctrigger;
302         if (jffs2_can_mark_obsolete(c))
303                 c->vdirty_blocks_gctrigger *= 10;
304
305         /* If there's less than this amount of dirty space, don't bother
306            trying to GC to make more space. It'll be a fruitless task */
307         c->nospc_dirty_size = c->sector_size + (c->flash_size / 100);
308
309         dbg_fsbuild("JFFS2 trigger levels (size %d KiB, block size %d KiB, %d blocks)\n",
310                   c->flash_size / 1024, c->sector_size / 1024, c->nr_blocks);
311         dbg_fsbuild("Blocks required to allow deletion:    %d (%d KiB)\n",
312                   c->resv_blocks_deletion, c->resv_blocks_deletion*c->sector_size/1024);
313         dbg_fsbuild("Blocks required to allow writes:      %d (%d KiB)\n",
314                   c->resv_blocks_write, c->resv_blocks_write*c->sector_size/1024);
315         dbg_fsbuild("Blocks required to quiesce GC thread: %d (%d KiB)\n",
316                   c->resv_blocks_gctrigger, c->resv_blocks_gctrigger*c->sector_size/1024);
317         dbg_fsbuild("Blocks required to allow GC merges:   %d (%d KiB)\n",
318                   c->resv_blocks_gcmerge, c->resv_blocks_gcmerge*c->sector_size/1024);
319         dbg_fsbuild("Blocks required to GC bad blocks:     %d (%d KiB)\n",
320                   c->resv_blocks_gcbad, c->resv_blocks_gcbad*c->sector_size/1024);
321         dbg_fsbuild("Amount of dirty space required to GC: %d bytes\n",
322                   c->nospc_dirty_size);
323         dbg_fsbuild("Very dirty blocks before GC triggered: %d\n",
324                   c->vdirty_blocks_gctrigger);
325 }
326
327 int jffs2_do_mount_fs(struct jffs2_sb_info *c)
328 {
329         int ret;
330         int i;
331         int size;
332
333         c->free_size = c->flash_size;
334         c->nr_blocks = c->flash_size / c->sector_size;
335         size = sizeof(struct jffs2_eraseblock) * c->nr_blocks;
336 #ifndef __ECOS
337         if (jffs2_blocks_use_vmalloc(c))
338                 c->blocks = vmalloc(size);
339         else
340 #endif
341                 c->blocks = kmalloc(size, GFP_KERNEL);
342         if (!c->blocks)
343                 return -ENOMEM;
344
345         memset(c->blocks, 0, size);
346         for (i=0; i<c->nr_blocks; i++) {
347                 INIT_LIST_HEAD(&c->blocks[i].list);
348                 c->blocks[i].offset = i * c->sector_size;
349                 c->blocks[i].free_size = c->sector_size;
350         }
351
352         INIT_LIST_HEAD(&c->clean_list);
353         INIT_LIST_HEAD(&c->very_dirty_list);
354         INIT_LIST_HEAD(&c->dirty_list);
355         INIT_LIST_HEAD(&c->erasable_list);
356         INIT_LIST_HEAD(&c->erasing_list);
357         INIT_LIST_HEAD(&c->erase_checking_list);
358         INIT_LIST_HEAD(&c->erase_pending_list);
359         INIT_LIST_HEAD(&c->erasable_pending_wbuf_list);
360         INIT_LIST_HEAD(&c->erase_complete_list);
361         INIT_LIST_HEAD(&c->free_list);
362         INIT_LIST_HEAD(&c->bad_list);
363         INIT_LIST_HEAD(&c->bad_used_list);
364         c->highest_ino = 1;
365         c->summary = NULL;
366
367         ret = jffs2_sum_init(c);
368         if (ret)
369                 goto out_free;
370
371         if (jffs2_build_filesystem(c)) {
372                 dbg_fsbuild("build_fs failed\n");
373                 jffs2_free_ino_caches(c);
374                 jffs2_free_raw_node_refs(c);
375                 ret = -EIO;
376                 goto out_free;
377         }
378
379         jffs2_calc_trigger_levels(c);
380
381         return 0;
382
383  out_free:
384 #ifndef __ECOS
385         if (jffs2_blocks_use_vmalloc(c))
386                 vfree(c->blocks);
387         else
388 #endif
389                 kfree(c->blocks);
390
391         return ret;
392 }