xref: /kernel/linux/linux-6.6/fs/sysv/itree.c (revision 62306a36) 
1// SPDX-License-Identifier: GPL-2.0
2/*
3 *  linux/fs/sysv/itree.c
4 *
5 *  Handling of indirect blocks' trees.
6 *  AV, Sep--Dec 2000
7 */
8
9#include <linux/buffer_head.h>
10#include <linux/mount.h>
11#include <linux/string.h>
12#include "sysv.h"
13
14enum {DIRECT = 10, DEPTH = 4};	/* Have triple indirect */
15
16static inline void dirty_indirect(struct buffer_head *bh, struct inode *inode)
17{
18	mark_buffer_dirty_inode(bh, inode);
19	if (IS_SYNC(inode))
20		sync_dirty_buffer(bh);
21}
22
23static int block_to_path(struct inode *inode, long block, int offsets[DEPTH])
24{
25	struct super_block *sb = inode->i_sb;
26	struct sysv_sb_info *sbi = SYSV_SB(sb);
27	int ptrs_bits = sbi->s_ind_per_block_bits;
28	unsigned long	indirect_blocks = sbi->s_ind_per_block,
29			double_blocks = sbi->s_ind_per_block_2;
30	int n = 0;
31
32	if (block < 0) {
33		printk("sysv_block_map: block < 0\n");
34	} else if (block < DIRECT) {
35		offsets[n++] = block;
36	} else if ( (block -= DIRECT) < indirect_blocks) {
37		offsets[n++] = DIRECT;
38		offsets[n++] = block;
39	} else if ((block -= indirect_blocks) < double_blocks) {
40		offsets[n++] = DIRECT+1;
41		offsets[n++] = block >> ptrs_bits;
42		offsets[n++] = block & (indirect_blocks - 1);
43	} else if (((block -= double_blocks) >> (ptrs_bits * 2)) < indirect_blocks) {
44		offsets[n++] = DIRECT+2;
45		offsets[n++] = block >> (ptrs_bits * 2);
46		offsets[n++] = (block >> ptrs_bits) & (indirect_blocks - 1);
47		offsets[n++] = block & (indirect_blocks - 1);
48	} else {
49		/* nothing */;
50	}
51	return n;
52}
53
54static inline int block_to_cpu(struct sysv_sb_info *sbi, sysv_zone_t nr)
55{
56	return sbi->s_block_base + fs32_to_cpu(sbi, nr);
57}
58
59typedef struct {
60	sysv_zone_t     *p;
61	sysv_zone_t     key;
62	struct buffer_head *bh;
63} Indirect;
64
65static DEFINE_RWLOCK(pointers_lock);
66
67static inline void add_chain(Indirect *p, struct buffer_head *bh, sysv_zone_t *v)
68{
69	p->key = *(p->p = v);
70	p->bh = bh;
71}
72
73static inline int verify_chain(Indirect *from, Indirect *to)
74{
75	while (from <= to && from->key == *from->p)
76		from++;
77	return (from > to);
78}
79
80static inline sysv_zone_t *block_end(struct buffer_head *bh)
81{
82	return (sysv_zone_t*)((char*)bh->b_data + bh->b_size);
83}
84
85/*
86 * Requires read_lock(&pointers_lock) or write_lock(&pointers_lock)
87 */
88static Indirect *get_branch(struct inode *inode,
89			    int depth,
90			    int offsets[],
91			    Indirect chain[],
92			    int *err)
93{
94	struct super_block *sb = inode->i_sb;
95	Indirect *p = chain;
96	struct buffer_head *bh;
97
98	*err = 0;
99	add_chain(chain, NULL, SYSV_I(inode)->i_data + *offsets);
100	if (!p->key)
101		goto no_block;
102	while (--depth) {
103		int block = block_to_cpu(SYSV_SB(sb), p->key);
104		bh = sb_bread(sb, block);
105		if (!bh)
106			goto failure;
107		if (!verify_chain(chain, p))
108			goto changed;
109		add_chain(++p, bh, (sysv_zone_t*)bh->b_data + *++offsets);
110		if (!p->key)
111			goto no_block;
112	}
113	return NULL;
114
115changed:
116	brelse(bh);
117	*err = -EAGAIN;
118	goto no_block;
119failure:
120	*err = -EIO;
121no_block:
122	return p;
123}
124
125static int alloc_branch(struct inode *inode,
126			int num,
127			int *offsets,
128			Indirect *branch)
129{
130	int blocksize = inode->i_sb->s_blocksize;
131	int n = 0;
132	int i;
133
134	branch[0].key = sysv_new_block(inode->i_sb);
135	if (branch[0].key) for (n = 1; n < num; n++) {
136		struct buffer_head *bh;
137		int parent;
138		/* Allocate the next block */
139		branch[n].key = sysv_new_block(inode->i_sb);
140		if (!branch[n].key)
141			break;
142		/*
143		 * Get buffer_head for parent block, zero it out and set
144		 * the pointer to new one, then send parent to disk.
145		 */
146		parent = block_to_cpu(SYSV_SB(inode->i_sb), branch[n-1].key);
147		bh = sb_getblk(inode->i_sb, parent);
148		if (!bh) {
149			sysv_free_block(inode->i_sb, branch[n].key);
150			break;
151		}
152		lock_buffer(bh);
153		memset(bh->b_data, 0, blocksize);
154		branch[n].bh = bh;
155		branch[n].p = (sysv_zone_t*) bh->b_data + offsets[n];
156		*branch[n].p = branch[n].key;
157		set_buffer_uptodate(bh);
158		unlock_buffer(bh);
159		dirty_indirect(bh, inode);
160	}
161	if (n == num)
162		return 0;
163
164	/* Allocation failed, free what we already allocated */
165	for (i = 1; i < n; i++)
166		bforget(branch[i].bh);
167	for (i = 0; i < n; i++)
168		sysv_free_block(inode->i_sb, branch[i].key);
169	return -ENOSPC;
170}
171
172static inline int splice_branch(struct inode *inode,
173				Indirect chain[],
174				Indirect *where,
175				int num)
176{
177	int i;
178
179	/* Verify that place we are splicing to is still there and vacant */
180	write_lock(&pointers_lock);
181	if (!verify_chain(chain, where-1) || *where->p)
182		goto changed;
183	*where->p = where->key;
184	write_unlock(&pointers_lock);
185
186	inode_set_ctime_current(inode);
187
188	/* had we spliced it onto indirect block? */
189	if (where->bh)
190		dirty_indirect(where->bh, inode);
191
192	if (IS_SYNC(inode))
193		sysv_sync_inode(inode);
194	else
195		mark_inode_dirty(inode);
196	return 0;
197
198changed:
199	write_unlock(&pointers_lock);
200	for (i = 1; i < num; i++)
201		bforget(where[i].bh);
202	for (i = 0; i < num; i++)
203		sysv_free_block(inode->i_sb, where[i].key);
204	return -EAGAIN;
205}
206
207static int get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create)
208{
209	int err = -EIO;
210	int offsets[DEPTH];
211	Indirect chain[DEPTH];
212	struct super_block *sb = inode->i_sb;
213	Indirect *partial;
214	int left;
215	int depth = block_to_path(inode, iblock, offsets);
216
217	if (depth == 0)
218		goto out;
219
220reread:
221	read_lock(&pointers_lock);
222	partial = get_branch(inode, depth, offsets, chain, &err);
223	read_unlock(&pointers_lock);
224
225	/* Simplest case - block found, no allocation needed */
226	if (!partial) {
227got_it:
228		map_bh(bh_result, sb, block_to_cpu(SYSV_SB(sb),
229					chain[depth-1].key));
230		/* Clean up and exit */
231		partial = chain+depth-1; /* the whole chain */
232		goto cleanup;
233	}
234
235	/* Next simple case - plain lookup or failed read of indirect block */
236	if (!create || err == -EIO) {
237cleanup:
238		while (partial > chain) {
239			brelse(partial->bh);
240			partial--;
241		}
242out:
243		return err;
244	}
245
246	/*
247	 * Indirect block might be removed by truncate while we were
248	 * reading it. Handling of that case (forget what we've got and
249	 * reread) is taken out of the main path.
250	 */
251	if (err == -EAGAIN)
252		goto changed;
253
254	left = (chain + depth) - partial;
255	err = alloc_branch(inode, left, offsets+(partial-chain), partial);
256	if (err)
257		goto cleanup;
258
259	if (splice_branch(inode, chain, partial, left) < 0)
260		goto changed;
261
262	set_buffer_new(bh_result);
263	goto got_it;
264
265changed:
266	while (partial > chain) {
267		brelse(partial->bh);
268		partial--;
269	}
270	goto reread;
271}
272
273static inline int all_zeroes(sysv_zone_t *p, sysv_zone_t *q)
274{
275	while (p < q)
276		if (*p++)
277			return 0;
278	return 1;
279}
280
281static Indirect *find_shared(struct inode *inode,
282				int depth,
283				int offsets[],
284				Indirect chain[],
285				sysv_zone_t *top)
286{
287	Indirect *partial, *p;
288	int k, err;
289
290	*top = 0;
291	for (k = depth; k > 1 && !offsets[k-1]; k--)
292		;
293
294	write_lock(&pointers_lock);
295	partial = get_branch(inode, k, offsets, chain, &err);
296	if (!partial)
297		partial = chain + k-1;
298	/*
299	 * If the branch acquired continuation since we've looked at it -
300	 * fine, it should all survive and (new) top doesn't belong to us.
301	 */
302	if (!partial->key && *partial->p) {
303		write_unlock(&pointers_lock);
304		goto no_top;
305	}
306	for (p=partial; p>chain && all_zeroes((sysv_zone_t*)p->bh->b_data,p->p); p--)
307		;
308	/*
309	 * OK, we've found the last block that must survive. The rest of our
310	 * branch should be detached before unlocking. However, if that rest
311	 * of branch is all ours and does not grow immediately from the inode
312	 * it's easier to cheat and just decrement partial->p.
313	 */
314	if (p == chain + k - 1 && p > chain) {
315		p->p--;
316	} else {
317		*top = *p->p;
318		*p->p = 0;
319	}
320	write_unlock(&pointers_lock);
321
322	while (partial > p) {
323		brelse(partial->bh);
324		partial--;
325	}
326no_top:
327	return partial;
328}
329
330static inline void free_data(struct inode *inode, sysv_zone_t *p, sysv_zone_t *q)
331{
332	for ( ; p < q ; p++) {
333		sysv_zone_t nr = *p;
334		if (nr) {
335			*p = 0;
336			sysv_free_block(inode->i_sb, nr);
337			mark_inode_dirty(inode);
338		}
339	}
340}
341
342static void free_branches(struct inode *inode, sysv_zone_t *p, sysv_zone_t *q, int depth)
343{
344	struct buffer_head * bh;
345	struct super_block *sb = inode->i_sb;
346
347	if (depth--) {
348		for ( ; p < q ; p++) {
349			int block;
350			sysv_zone_t nr = *p;
351			if (!nr)
352				continue;
353			*p = 0;
354			block = block_to_cpu(SYSV_SB(sb), nr);
355			bh = sb_bread(sb, block);
356			if (!bh)
357				continue;
358			free_branches(inode, (sysv_zone_t*)bh->b_data,
359					block_end(bh), depth);
360			bforget(bh);
361			sysv_free_block(sb, nr);
362			mark_inode_dirty(inode);
363		}
364	} else
365		free_data(inode, p, q);
366}
367
368void sysv_truncate (struct inode * inode)
369{
370	sysv_zone_t *i_data = SYSV_I(inode)->i_data;
371	int offsets[DEPTH];
372	Indirect chain[DEPTH];
373	Indirect *partial;
374	sysv_zone_t nr = 0;
375	int n;
376	long iblock;
377	unsigned blocksize;
378
379	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
380	    S_ISLNK(inode->i_mode)))
381		return;
382
383	blocksize = inode->i_sb->s_blocksize;
384	iblock = (inode->i_size + blocksize-1)
385					>> inode->i_sb->s_blocksize_bits;
386
387	block_truncate_page(inode->i_mapping, inode->i_size, get_block);
388
389	n = block_to_path(inode, iblock, offsets);
390	if (n == 0)
391		return;
392
393	if (n == 1) {
394		free_data(inode, i_data+offsets[0], i_data + DIRECT);
395		goto do_indirects;
396	}
397
398	partial = find_shared(inode, n, offsets, chain, &nr);
399	/* Kill the top of shared branch (already detached) */
400	if (nr) {
401		if (partial == chain)
402			mark_inode_dirty(inode);
403		else
404			dirty_indirect(partial->bh, inode);
405		free_branches(inode, &nr, &nr+1, (chain+n-1) - partial);
406	}
407	/* Clear the ends of indirect blocks on the shared branch */
408	while (partial > chain) {
409		free_branches(inode, partial->p + 1, block_end(partial->bh),
410				(chain+n-1) - partial);
411		dirty_indirect(partial->bh, inode);
412		brelse (partial->bh);
413		partial--;
414	}
415do_indirects:
416	/* Kill the remaining (whole) subtrees (== subtrees deeper than...) */
417	while (n < DEPTH) {
418		nr = i_data[DIRECT + n - 1];
419		if (nr) {
420			i_data[DIRECT + n - 1] = 0;
421			mark_inode_dirty(inode);
422			free_branches(inode, &nr, &nr+1, n);
423		}
424		n++;
425	}
426	inode->i_mtime = inode_set_ctime_current(inode);
427	if (IS_SYNC(inode))
428		sysv_sync_inode (inode);
429	else
430		mark_inode_dirty(inode);
431}
432
433static unsigned sysv_nblocks(struct super_block *s, loff_t size)
434{
435	struct sysv_sb_info *sbi = SYSV_SB(s);
436	int ptrs_bits = sbi->s_ind_per_block_bits;
437	unsigned blocks, res, direct = DIRECT, i = DEPTH;
438	blocks = (size + s->s_blocksize - 1) >> s->s_blocksize_bits;
439	res = blocks;
440	while (--i && blocks > direct) {
441		blocks = ((blocks - direct - 1) >> ptrs_bits) + 1;
442		res += blocks;
443		direct = 1;
444	}
445	return res;
446}
447
448int sysv_getattr(struct mnt_idmap *idmap, const struct path *path,
449		 struct kstat *stat, u32 request_mask, unsigned int flags)
450{
451	struct super_block *s = path->dentry->d_sb;
452	generic_fillattr(&nop_mnt_idmap, request_mask, d_inode(path->dentry),
453			 stat);
454	stat->blocks = (s->s_blocksize / 512) * sysv_nblocks(s, stat->size);
455	stat->blksize = s->s_blocksize;
456	return 0;
457}
458
459static int sysv_writepage(struct page *page, struct writeback_control *wbc)
460{
461	return block_write_full_page(page,get_block,wbc);
462}
463
464static int sysv_read_folio(struct file *file, struct folio *folio)
465{
466	return block_read_full_folio(folio, get_block);
467}
468
469int sysv_prepare_chunk(struct page *page, loff_t pos, unsigned len)
470{
471	return __block_write_begin(page, pos, len, get_block);
472}
473
474static void sysv_write_failed(struct address_space *mapping, loff_t to)
475{
476	struct inode *inode = mapping->host;
477
478	if (to > inode->i_size) {
479		truncate_pagecache(inode, inode->i_size);
480		sysv_truncate(inode);
481	}
482}
483
484static int sysv_write_begin(struct file *file, struct address_space *mapping,
485			loff_t pos, unsigned len,
486			struct page **pagep, void **fsdata)
487{
488	int ret;
489
490	ret = block_write_begin(mapping, pos, len, pagep, get_block);
491	if (unlikely(ret))
492		sysv_write_failed(mapping, pos + len);
493
494	return ret;
495}
496
497static sector_t sysv_bmap(struct address_space *mapping, sector_t block)
498{
499	return generic_block_bmap(mapping,block,get_block);
500}
501
502const struct address_space_operations sysv_aops = {
503	.dirty_folio = block_dirty_folio,
504	.invalidate_folio = block_invalidate_folio,
505	.read_folio = sysv_read_folio,
506	.writepage = sysv_writepage,
507	.write_begin = sysv_write_begin,
508	.write_end = generic_write_end,
509	.bmap = sysv_bmap
510};
511