xref: /openbmc/linux/fs/jffs2/build.c (revision 8fb870df)
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 (child_ic->nlink++ && fd->type == DT_DIR) {
72 			JFFS2_ERROR("child dir \"%s\" (ino #%u) of dir ino #%u appears to be a hard link\n",
73 				fd->name, fd->ino, ic->ino);
74 			/* TODO: What do we do about it? */
75 		}
76 		dbg_fsbuild("increased nlink for child \"%s\" (ino #%u)\n", fd->name, fd->ino);
77 		/* Can't free scan_dents so far. We might need them in pass 2 */
78 	}
79 }
80 
81 /* Scan plan:
82  - Scan physical nodes. Build map of inodes/dirents. Allocate inocaches as we go
83  - Scan directory tree from top down, setting nlink in inocaches
84  - Scan inocaches for inodes with nlink==0
85 */
86 static int jffs2_build_filesystem(struct jffs2_sb_info *c)
87 {
88 	int ret;
89 	int i;
90 	struct jffs2_inode_cache *ic;
91 	struct jffs2_full_dirent *fd;
92 	struct jffs2_full_dirent *dead_fds = NULL;
93 
94 	dbg_fsbuild("build FS data structures\n");
95 
96 	/* First, scan the medium and build all the inode caches with
97 	   lists of physical nodes */
98 
99 	c->flags |= JFFS2_SB_FLAG_SCANNING;
100 	ret = jffs2_scan_medium(c);
101 	c->flags &= ~JFFS2_SB_FLAG_SCANNING;
102 	if (ret)
103 		goto exit;
104 
105 	dbg_fsbuild("scanned flash completely\n");
106 	jffs2_dbg_dump_block_lists_nolock(c);
107 
108 	dbg_fsbuild("pass 1 starting\n");
109 	c->flags |= JFFS2_SB_FLAG_BUILDING;
110 	/* Now scan the directory tree, increasing nlink according to every dirent found. */
111 	for_each_inode(i, c, ic) {
112 		if (ic->scan_dents) {
113 			jffs2_build_inode_pass1(c, ic);
114 			cond_resched();
115 		}
116 	}
117 
118 	dbg_fsbuild("pass 1 complete\n");
119 
120 	/* Next, scan for inodes with nlink == 0 and remove them. If
121 	   they were directories, then decrement the nlink of their
122 	   children too, and repeat the scan. As that's going to be
123 	   a fairly uncommon occurrence, it's not so evil to do it this
124 	   way. Recursion bad. */
125 	dbg_fsbuild("pass 2 starting\n");
126 
127 	for_each_inode(i, c, ic) {
128 		if (ic->nlink)
129 			continue;
130 
131 		jffs2_build_remove_unlinked_inode(c, ic, &dead_fds);
132 		cond_resched();
133 	}
134 
135 	dbg_fsbuild("pass 2a starting\n");
136 
137 	while (dead_fds) {
138 		fd = dead_fds;
139 		dead_fds = fd->next;
140 
141 		ic = jffs2_get_ino_cache(c, fd->ino);
142 
143 		if (ic)
144 			jffs2_build_remove_unlinked_inode(c, ic, &dead_fds);
145 		jffs2_free_full_dirent(fd);
146 	}
147 
148 	dbg_fsbuild("pass 2a complete\n");
149 	dbg_fsbuild("freeing temporary data structures\n");
150 
151 	/* Finally, we can scan again and free the dirent structs */
152 	for_each_inode(i, c, ic) {
153 		while(ic->scan_dents) {
154 			fd = ic->scan_dents;
155 			ic->scan_dents = fd->next;
156 			jffs2_free_full_dirent(fd);
157 		}
158 		ic->scan_dents = NULL;
159 		cond_resched();
160 	}
161 	jffs2_build_xattr_subsystem(c);
162 	c->flags &= ~JFFS2_SB_FLAG_BUILDING;
163 
164 	dbg_fsbuild("FS build complete\n");
165 
166 	/* Rotate the lists by some number to ensure wear levelling */
167 	jffs2_rotate_lists(c);
168 
169 	ret = 0;
170 
171 exit:
172 	if (ret) {
173 		for_each_inode(i, c, ic) {
174 			while(ic->scan_dents) {
175 				fd = ic->scan_dents;
176 				ic->scan_dents = fd->next;
177 				jffs2_free_full_dirent(fd);
178 			}
179 		}
180 		jffs2_clear_xattr_subsystem(c);
181 	}
182 
183 	return ret;
184 }
185 
186 static void jffs2_build_remove_unlinked_inode(struct jffs2_sb_info *c,
187 					struct jffs2_inode_cache *ic,
188 					struct jffs2_full_dirent **dead_fds)
189 {
190 	struct jffs2_raw_node_ref *raw;
191 	struct jffs2_full_dirent *fd;
192 
193 	dbg_fsbuild("removing ino #%u with nlink == zero.\n", ic->ino);
194 
195 	raw = ic->nodes;
196 	while (raw != (void *)ic) {
197 		struct jffs2_raw_node_ref *next = raw->next_in_ino;
198 		dbg_fsbuild("obsoleting node at 0x%08x\n", ref_offset(raw));
199 		jffs2_mark_node_obsolete(c, raw);
200 		raw = next;
201 	}
202 
203 	if (ic->scan_dents) {
204 		int whinged = 0;
205 		dbg_fsbuild("inode #%u was a directory which may have children...\n", ic->ino);
206 
207 		while(ic->scan_dents) {
208 			struct jffs2_inode_cache *child_ic;
209 
210 			fd = ic->scan_dents;
211 			ic->scan_dents = fd->next;
212 
213 			if (!fd->ino) {
214 				/* It's a deletion dirent. Ignore it */
215 				dbg_fsbuild("child \"%s\" is a deletion dirent, skipping...\n", fd->name);
216 				jffs2_free_full_dirent(fd);
217 				continue;
218 			}
219 			if (!whinged)
220 				whinged = 1;
221 
222 			dbg_fsbuild("removing child \"%s\", ino #%u\n", fd->name, fd->ino);
223 
224 			child_ic = jffs2_get_ino_cache(c, fd->ino);
225 			if (!child_ic) {
226 				dbg_fsbuild("cannot remove child \"%s\", ino #%u, because it doesn't exist\n",
227 						fd->name, fd->ino);
228 				jffs2_free_full_dirent(fd);
229 				continue;
230 			}
231 
232 			/* Reduce nlink of the child. If it's now zero, stick it on the
233 			   dead_fds list to be cleaned up later. Else just free the fd */
234 
235 			child_ic->nlink--;
236 
237 			if (!child_ic->nlink) {
238 				dbg_fsbuild("inode #%u (\"%s\") has now got zero nlink, adding to dead_fds list.\n",
239 					  fd->ino, fd->name);
240 				fd->next = *dead_fds;
241 				*dead_fds = fd;
242 			} else {
243 				dbg_fsbuild("inode #%u (\"%s\") has now got nlink %d. Ignoring.\n",
244 					  fd->ino, fd->name, child_ic->nlink);
245 				jffs2_free_full_dirent(fd);
246 			}
247 		}
248 	}
249 
250 	/*
251 	   We don't delete the inocache from the hash list and free it yet.
252 	   The erase code will do that, when all the nodes are completely gone.
253 	*/
254 }
255 
256 static void jffs2_calc_trigger_levels(struct jffs2_sb_info *c)
257 {
258 	uint32_t size;
259 
260 	/* Deletion should almost _always_ be allowed. We're fairly
261 	   buggered once we stop allowing people to delete stuff
262 	   because there's not enough free space... */
263 	c->resv_blocks_deletion = 2;
264 
265 	/* Be conservative about how much space we need before we allow writes.
266 	   On top of that which is required for deletia, require an extra 2%
267 	   of the medium to be available, for overhead caused by nodes being
268 	   split across blocks, etc. */
269 
270 	size = c->flash_size / 50; /* 2% of flash size */
271 	size += c->nr_blocks * 100; /* And 100 bytes per eraseblock */
272 	size += c->sector_size - 1; /* ... and round up */
273 
274 	c->resv_blocks_write = c->resv_blocks_deletion + (size / c->sector_size);
275 
276 	/* When do we let the GC thread run in the background */
277 
278 	c->resv_blocks_gctrigger = c->resv_blocks_write + 1;
279 
280 	/* When do we allow garbage collection to merge nodes to make
281 	   long-term progress at the expense of short-term space exhaustion? */
282 	c->resv_blocks_gcmerge = c->resv_blocks_deletion + 1;
283 
284 	/* When do we allow garbage collection to eat from bad blocks rather
285 	   than actually making progress? */
286 	c->resv_blocks_gcbad = 0;//c->resv_blocks_deletion + 2;
287 
288 	/* What number of 'very dirty' eraseblocks do we allow before we
289 	   trigger the GC thread even if we don't _need_ the space. When we
290 	   can't mark nodes obsolete on the medium, the old dirty nodes cause
291 	   performance problems because we have to inspect and discard them. */
292 	c->vdirty_blocks_gctrigger = c->resv_blocks_gcmerge;
293 	if (jffs2_can_mark_obsolete(c))
294 		c->vdirty_blocks_gctrigger *= 10;
295 
296 	/* If there's less than this amount of dirty space, don't bother
297 	   trying to GC to make more space. It'll be a fruitless task */
298 	c->nospc_dirty_size = c->sector_size + (c->flash_size / 100);
299 
300 	dbg_fsbuild("JFFS2 trigger levels (size %d KiB, block size %d KiB, %d blocks)\n",
301 		  c->flash_size / 1024, c->sector_size / 1024, c->nr_blocks);
302 	dbg_fsbuild("Blocks required to allow deletion:    %d (%d KiB)\n",
303 		  c->resv_blocks_deletion, c->resv_blocks_deletion*c->sector_size/1024);
304 	dbg_fsbuild("Blocks required to allow writes:      %d (%d KiB)\n",
305 		  c->resv_blocks_write, c->resv_blocks_write*c->sector_size/1024);
306 	dbg_fsbuild("Blocks required to quiesce GC thread: %d (%d KiB)\n",
307 		  c->resv_blocks_gctrigger, c->resv_blocks_gctrigger*c->sector_size/1024);
308 	dbg_fsbuild("Blocks required to allow GC merges:   %d (%d KiB)\n",
309 		  c->resv_blocks_gcmerge, c->resv_blocks_gcmerge*c->sector_size/1024);
310 	dbg_fsbuild("Blocks required to GC bad blocks:     %d (%d KiB)\n",
311 		  c->resv_blocks_gcbad, c->resv_blocks_gcbad*c->sector_size/1024);
312 	dbg_fsbuild("Amount of dirty space required to GC: %d bytes\n",
313 		  c->nospc_dirty_size);
314 	dbg_fsbuild("Very dirty blocks before GC triggered: %d\n",
315 		  c->vdirty_blocks_gctrigger);
316 }
317 
318 int jffs2_do_mount_fs(struct jffs2_sb_info *c)
319 {
320 	int ret;
321 	int i;
322 	int size;
323 
324 	c->free_size = c->flash_size;
325 	c->nr_blocks = c->flash_size / c->sector_size;
326 	size = sizeof(struct jffs2_eraseblock) * c->nr_blocks;
327 #ifndef __ECOS
328 	if (jffs2_blocks_use_vmalloc(c))
329 		c->blocks = vmalloc(size);
330 	else
331 #endif
332 		c->blocks = kmalloc(size, GFP_KERNEL);
333 	if (!c->blocks)
334 		return -ENOMEM;
335 
336 	memset(c->blocks, 0, size);
337 	for (i=0; i<c->nr_blocks; i++) {
338 		INIT_LIST_HEAD(&c->blocks[i].list);
339 		c->blocks[i].offset = i * c->sector_size;
340 		c->blocks[i].free_size = c->sector_size;
341 	}
342 
343 	INIT_LIST_HEAD(&c->clean_list);
344 	INIT_LIST_HEAD(&c->very_dirty_list);
345 	INIT_LIST_HEAD(&c->dirty_list);
346 	INIT_LIST_HEAD(&c->erasable_list);
347 	INIT_LIST_HEAD(&c->erasing_list);
348 	INIT_LIST_HEAD(&c->erase_pending_list);
349 	INIT_LIST_HEAD(&c->erasable_pending_wbuf_list);
350 	INIT_LIST_HEAD(&c->erase_complete_list);
351 	INIT_LIST_HEAD(&c->free_list);
352 	INIT_LIST_HEAD(&c->bad_list);
353 	INIT_LIST_HEAD(&c->bad_used_list);
354 	c->highest_ino = 1;
355 	c->summary = NULL;
356 
357 	ret = jffs2_sum_init(c);
358 	if (ret)
359 		goto out_free;
360 
361 	if (jffs2_build_filesystem(c)) {
362 		dbg_fsbuild("build_fs failed\n");
363 		jffs2_free_ino_caches(c);
364 		jffs2_free_raw_node_refs(c);
365 		ret = -EIO;
366 		goto out_free;
367 	}
368 
369 	jffs2_calc_trigger_levels(c);
370 
371 	return 0;
372 
373  out_free:
374 #ifndef __ECOS
375 	if (jffs2_blocks_use_vmalloc(c))
376 		vfree(c->blocks);
377 	else
378 #endif
379 		kfree(c->blocks);
380 
381 	return ret;
382 }
383