xref: /openbmc/linux/fs/jffs2/build.c (revision 1d5cfdb0)
1 /*
2  * JFFS2 -- Journalling Flash File System, Version 2.
3  *
4  * Copyright © 2001-2007 Red Hat, Inc.
5  * Copyright © 2004-2010 David Woodhouse <dwmw2@infradead.org>
6  *
7  * Created by David Woodhouse <dwmw2@infradead.org>
8  *
9  * For licensing information, see the file 'LICENCE' in this directory.
10  *
11  */
12 
13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
14 
15 #include <linux/kernel.h>
16 #include <linux/sched.h>
17 #include <linux/slab.h>
18 #include <linux/vmalloc.h>
19 #include <linux/mtd/mtd.h>
20 #include <linux/mm.h> /* kvfree() */
21 #include "nodelist.h"
22 
23 static void jffs2_build_remove_unlinked_inode(struct jffs2_sb_info *,
24 		struct jffs2_inode_cache *, struct jffs2_full_dirent **);
25 
26 static inline struct jffs2_inode_cache *
27 first_inode_chain(int *i, struct jffs2_sb_info *c)
28 {
29 	for (; *i < c->inocache_hashsize; (*i)++) {
30 		if (c->inocache_list[*i])
31 			return c->inocache_list[*i];
32 	}
33 	return NULL;
34 }
35 
36 static inline struct jffs2_inode_cache *
37 next_inode(int *i, struct jffs2_inode_cache *ic, struct jffs2_sb_info *c)
38 {
39 	/* More in this chain? */
40 	if (ic->next)
41 		return ic->next;
42 	(*i)++;
43 	return first_inode_chain(i, c);
44 }
45 
46 #define for_each_inode(i, c, ic)			\
47 	for (i = 0, ic = first_inode_chain(&i, (c));	\
48 	     ic;					\
49 	     ic = next_inode(&i, ic, (c)))
50 
51 
52 static void jffs2_build_inode_pass1(struct jffs2_sb_info *c,
53 				    struct jffs2_inode_cache *ic)
54 {
55 	struct jffs2_full_dirent *fd;
56 
57 	dbg_fsbuild("building directory inode #%u\n", ic->ino);
58 
59 	/* For each child, increase nlink */
60 	for(fd = ic->scan_dents; fd; fd = fd->next) {
61 		struct jffs2_inode_cache *child_ic;
62 		if (!fd->ino)
63 			continue;
64 
65 		/* we can get high latency here with huge directories */
66 
67 		child_ic = jffs2_get_ino_cache(c, fd->ino);
68 		if (!child_ic) {
69 			dbg_fsbuild("child \"%s\" (ino #%u) of dir ino #%u doesn't exist!\n",
70 				  fd->name, fd->ino, ic->ino);
71 			jffs2_mark_node_obsolete(c, fd->raw);
72 			continue;
73 		}
74 
75 		if (fd->type == DT_DIR) {
76 			if (child_ic->pino_nlink) {
77 				JFFS2_ERROR("child dir \"%s\" (ino #%u) of dir ino #%u appears to be a hard link\n",
78 					    fd->name, fd->ino, ic->ino);
79 				/* TODO: What do we do about it? */
80 			} else {
81 				child_ic->pino_nlink = ic->ino;
82 			}
83 		} else
84 			child_ic->pino_nlink++;
85 
86 		dbg_fsbuild("increased nlink for child \"%s\" (ino #%u)\n", fd->name, fd->ino);
87 		/* Can't free scan_dents so far. We might need them in pass 2 */
88 	}
89 }
90 
91 /* Scan plan:
92  - Scan physical nodes. Build map of inodes/dirents. Allocate inocaches as we go
93  - Scan directory tree from top down, setting nlink in inocaches
94  - Scan inocaches for inodes with nlink==0
95 */
96 static int jffs2_build_filesystem(struct jffs2_sb_info *c)
97 {
98 	int ret;
99 	int i;
100 	struct jffs2_inode_cache *ic;
101 	struct jffs2_full_dirent *fd;
102 	struct jffs2_full_dirent *dead_fds = NULL;
103 
104 	dbg_fsbuild("build FS data structures\n");
105 
106 	/* First, scan the medium and build all the inode caches with
107 	   lists of physical nodes */
108 
109 	c->flags |= JFFS2_SB_FLAG_SCANNING;
110 	ret = jffs2_scan_medium(c);
111 	c->flags &= ~JFFS2_SB_FLAG_SCANNING;
112 	if (ret)
113 		goto exit;
114 
115 	dbg_fsbuild("scanned flash completely\n");
116 	jffs2_dbg_dump_block_lists_nolock(c);
117 
118 	dbg_fsbuild("pass 1 starting\n");
119 	c->flags |= JFFS2_SB_FLAG_BUILDING;
120 	/* Now scan the directory tree, increasing nlink according to every dirent found. */
121 	for_each_inode(i, c, ic) {
122 		if (ic->scan_dents) {
123 			jffs2_build_inode_pass1(c, ic);
124 			cond_resched();
125 		}
126 	}
127 
128 	dbg_fsbuild("pass 1 complete\n");
129 
130 	/* Next, scan for inodes with nlink == 0 and remove them. If
131 	   they were directories, then decrement the nlink of their
132 	   children too, and repeat the scan. As that's going to be
133 	   a fairly uncommon occurrence, it's not so evil to do it this
134 	   way. Recursion bad. */
135 	dbg_fsbuild("pass 2 starting\n");
136 
137 	for_each_inode(i, c, ic) {
138 		if (ic->pino_nlink)
139 			continue;
140 
141 		jffs2_build_remove_unlinked_inode(c, ic, &dead_fds);
142 		cond_resched();
143 	}
144 
145 	dbg_fsbuild("pass 2a starting\n");
146 
147 	while (dead_fds) {
148 		fd = dead_fds;
149 		dead_fds = fd->next;
150 
151 		ic = jffs2_get_ino_cache(c, fd->ino);
152 
153 		if (ic)
154 			jffs2_build_remove_unlinked_inode(c, ic, &dead_fds);
155 		jffs2_free_full_dirent(fd);
156 	}
157 
158 	dbg_fsbuild("pass 2a complete\n");
159 	dbg_fsbuild("freeing temporary data structures\n");
160 
161 	/* Finally, we can scan again and free the dirent structs */
162 	for_each_inode(i, c, ic) {
163 		while(ic->scan_dents) {
164 			fd = ic->scan_dents;
165 			ic->scan_dents = fd->next;
166 			jffs2_free_full_dirent(fd);
167 		}
168 		ic->scan_dents = NULL;
169 		cond_resched();
170 	}
171 	jffs2_build_xattr_subsystem(c);
172 	c->flags &= ~JFFS2_SB_FLAG_BUILDING;
173 
174 	dbg_fsbuild("FS build complete\n");
175 
176 	/* Rotate the lists by some number to ensure wear levelling */
177 	jffs2_rotate_lists(c);
178 
179 	ret = 0;
180 
181 exit:
182 	if (ret) {
183 		for_each_inode(i, c, ic) {
184 			while(ic->scan_dents) {
185 				fd = ic->scan_dents;
186 				ic->scan_dents = fd->next;
187 				jffs2_free_full_dirent(fd);
188 			}
189 		}
190 		jffs2_clear_xattr_subsystem(c);
191 	}
192 
193 	return ret;
194 }
195 
196 static void jffs2_build_remove_unlinked_inode(struct jffs2_sb_info *c,
197 					struct jffs2_inode_cache *ic,
198 					struct jffs2_full_dirent **dead_fds)
199 {
200 	struct jffs2_raw_node_ref *raw;
201 	struct jffs2_full_dirent *fd;
202 
203 	dbg_fsbuild("removing ino #%u with nlink == zero.\n", ic->ino);
204 
205 	raw = ic->nodes;
206 	while (raw != (void *)ic) {
207 		struct jffs2_raw_node_ref *next = raw->next_in_ino;
208 		dbg_fsbuild("obsoleting node at 0x%08x\n", ref_offset(raw));
209 		jffs2_mark_node_obsolete(c, raw);
210 		raw = next;
211 	}
212 
213 	if (ic->scan_dents) {
214 		int whinged = 0;
215 		dbg_fsbuild("inode #%u was a directory which may have children...\n", ic->ino);
216 
217 		while(ic->scan_dents) {
218 			struct jffs2_inode_cache *child_ic;
219 
220 			fd = ic->scan_dents;
221 			ic->scan_dents = fd->next;
222 
223 			if (!fd->ino) {
224 				/* It's a deletion dirent. Ignore it */
225 				dbg_fsbuild("child \"%s\" is a deletion dirent, skipping...\n", fd->name);
226 				jffs2_free_full_dirent(fd);
227 				continue;
228 			}
229 			if (!whinged)
230 				whinged = 1;
231 
232 			dbg_fsbuild("removing child \"%s\", ino #%u\n", fd->name, fd->ino);
233 
234 			child_ic = jffs2_get_ino_cache(c, fd->ino);
235 			if (!child_ic) {
236 				dbg_fsbuild("cannot remove child \"%s\", ino #%u, because it doesn't exist\n",
237 						fd->name, fd->ino);
238 				jffs2_free_full_dirent(fd);
239 				continue;
240 			}
241 
242 			/* Reduce nlink of the child. If it's now zero, stick it on the
243 			   dead_fds list to be cleaned up later. Else just free the fd */
244 
245 			if (fd->type == DT_DIR)
246 				child_ic->pino_nlink = 0;
247 			else
248 				child_ic->pino_nlink--;
249 
250 			if (!child_ic->pino_nlink) {
251 				dbg_fsbuild("inode #%u (\"%s\") now has no links; adding to dead_fds list.\n",
252 					  fd->ino, fd->name);
253 				fd->next = *dead_fds;
254 				*dead_fds = fd;
255 			} else {
256 				dbg_fsbuild("inode #%u (\"%s\") has now got nlink %d. Ignoring.\n",
257 					  fd->ino, fd->name, child_ic->pino_nlink);
258 				jffs2_free_full_dirent(fd);
259 			}
260 		}
261 	}
262 
263 	/*
264 	   We don't delete the inocache from the hash list and free it yet.
265 	   The erase code will do that, when all the nodes are completely gone.
266 	*/
267 }
268 
269 static void jffs2_calc_trigger_levels(struct jffs2_sb_info *c)
270 {
271 	uint32_t size;
272 
273 	/* Deletion should almost _always_ be allowed. We're fairly
274 	   buggered once we stop allowing people to delete stuff
275 	   because there's not enough free space... */
276 	c->resv_blocks_deletion = 2;
277 
278 	/* Be conservative about how much space we need before we allow writes.
279 	   On top of that which is required for deletia, require an extra 2%
280 	   of the medium to be available, for overhead caused by nodes being
281 	   split across blocks, etc. */
282 
283 	size = c->flash_size / 50; /* 2% of flash size */
284 	size += c->nr_blocks * 100; /* And 100 bytes per eraseblock */
285 	size += c->sector_size - 1; /* ... and round up */
286 
287 	c->resv_blocks_write = c->resv_blocks_deletion + (size / c->sector_size);
288 
289 	/* When do we let the GC thread run in the background */
290 
291 	c->resv_blocks_gctrigger = c->resv_blocks_write + 1;
292 
293 	/* When do we allow garbage collection to merge nodes to make
294 	   long-term progress at the expense of short-term space exhaustion? */
295 	c->resv_blocks_gcmerge = c->resv_blocks_deletion + 1;
296 
297 	/* When do we allow garbage collection to eat from bad blocks rather
298 	   than actually making progress? */
299 	c->resv_blocks_gcbad = 0;//c->resv_blocks_deletion + 2;
300 
301 	/* What number of 'very dirty' eraseblocks do we allow before we
302 	   trigger the GC thread even if we don't _need_ the space. When we
303 	   can't mark nodes obsolete on the medium, the old dirty nodes cause
304 	   performance problems because we have to inspect and discard them. */
305 	c->vdirty_blocks_gctrigger = c->resv_blocks_gctrigger;
306 	if (jffs2_can_mark_obsolete(c))
307 		c->vdirty_blocks_gctrigger *= 10;
308 
309 	/* If there's less than this amount of dirty space, don't bother
310 	   trying to GC to make more space. It'll be a fruitless task */
311 	c->nospc_dirty_size = c->sector_size + (c->flash_size / 100);
312 
313 	dbg_fsbuild("trigger levels (size %d KiB, block size %d KiB, %d blocks)\n",
314 		    c->flash_size / 1024, c->sector_size / 1024, c->nr_blocks);
315 	dbg_fsbuild("Blocks required to allow deletion:    %d (%d KiB)\n",
316 		  c->resv_blocks_deletion, c->resv_blocks_deletion*c->sector_size/1024);
317 	dbg_fsbuild("Blocks required to allow writes:      %d (%d KiB)\n",
318 		  c->resv_blocks_write, c->resv_blocks_write*c->sector_size/1024);
319 	dbg_fsbuild("Blocks required to quiesce GC thread: %d (%d KiB)\n",
320 		  c->resv_blocks_gctrigger, c->resv_blocks_gctrigger*c->sector_size/1024);
321 	dbg_fsbuild("Blocks required to allow GC merges:   %d (%d KiB)\n",
322 		  c->resv_blocks_gcmerge, c->resv_blocks_gcmerge*c->sector_size/1024);
323 	dbg_fsbuild("Blocks required to GC bad blocks:     %d (%d KiB)\n",
324 		  c->resv_blocks_gcbad, c->resv_blocks_gcbad*c->sector_size/1024);
325 	dbg_fsbuild("Amount of dirty space required to GC: %d bytes\n",
326 		  c->nospc_dirty_size);
327 	dbg_fsbuild("Very dirty blocks before GC triggered: %d\n",
328 		  c->vdirty_blocks_gctrigger);
329 }
330 
331 int jffs2_do_mount_fs(struct jffs2_sb_info *c)
332 {
333 	int ret;
334 	int i;
335 	int size;
336 
337 	c->free_size = c->flash_size;
338 	c->nr_blocks = c->flash_size / c->sector_size;
339 	size = sizeof(struct jffs2_eraseblock) * c->nr_blocks;
340 #ifndef __ECOS
341 	if (jffs2_blocks_use_vmalloc(c))
342 		c->blocks = vzalloc(size);
343 	else
344 #endif
345 		c->blocks = kzalloc(size, GFP_KERNEL);
346 	if (!c->blocks)
347 		return -ENOMEM;
348 
349 	for (i=0; i<c->nr_blocks; i++) {
350 		INIT_LIST_HEAD(&c->blocks[i].list);
351 		c->blocks[i].offset = i * c->sector_size;
352 		c->blocks[i].free_size = c->sector_size;
353 	}
354 
355 	INIT_LIST_HEAD(&c->clean_list);
356 	INIT_LIST_HEAD(&c->very_dirty_list);
357 	INIT_LIST_HEAD(&c->dirty_list);
358 	INIT_LIST_HEAD(&c->erasable_list);
359 	INIT_LIST_HEAD(&c->erasing_list);
360 	INIT_LIST_HEAD(&c->erase_checking_list);
361 	INIT_LIST_HEAD(&c->erase_pending_list);
362 	INIT_LIST_HEAD(&c->erasable_pending_wbuf_list);
363 	INIT_LIST_HEAD(&c->erase_complete_list);
364 	INIT_LIST_HEAD(&c->free_list);
365 	INIT_LIST_HEAD(&c->bad_list);
366 	INIT_LIST_HEAD(&c->bad_used_list);
367 	c->highest_ino = 1;
368 	c->summary = NULL;
369 
370 	ret = jffs2_sum_init(c);
371 	if (ret)
372 		goto out_free;
373 
374 	if (jffs2_build_filesystem(c)) {
375 		dbg_fsbuild("build_fs failed\n");
376 		jffs2_free_ino_caches(c);
377 		jffs2_free_raw_node_refs(c);
378 		ret = -EIO;
379 		goto out_free;
380 	}
381 
382 	jffs2_calc_trigger_levels(c);
383 
384 	return 0;
385 
386  out_free:
387 	kvfree(c->blocks);
388 
389 	return ret;
390 }
391