mm: fix exec activate_mm vs TLB shootdown and lazy tlb switching race
[linux/fpc-iii.git] / fs / sysv / itree.c
blobbcb67b0cabe7e5b99561610106fe383066925d9e
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * linux/fs/sysv/itree.c
5 * Handling of indirect blocks' trees.
6 * AV, Sep--Dec 2000
7 */
9 #include <linux/buffer_head.h>
10 #include <linux/mount.h>
11 #include <linux/string.h>
12 #include "sysv.h"
14 enum {DIRECT = 10, DEPTH = 4}; /* Have triple indirect */
16 static inline void dirty_indirect(struct buffer_head *bh, struct inode *inode)
18 mark_buffer_dirty_inode(bh, inode);
19 if (IS_SYNC(inode))
20 sync_dirty_buffer(bh);
23 static int block_to_path(struct inode *inode, long block, int offsets[DEPTH])
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;
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 */;
51 return n;
54 static inline int block_to_cpu(struct sysv_sb_info *sbi, sysv_zone_t nr)
56 return sbi->s_block_base + fs32_to_cpu(sbi, nr);
59 typedef struct {
60 sysv_zone_t *p;
61 sysv_zone_t key;
62 struct buffer_head *bh;
63 } Indirect;
65 static DEFINE_RWLOCK(pointers_lock);
67 static inline void add_chain(Indirect *p, struct buffer_head *bh, sysv_zone_t *v)
69 p->key = *(p->p = v);
70 p->bh = bh;
73 static inline int verify_chain(Indirect *from, Indirect *to)
75 while (from <= to && from->key == *from->p)
76 from++;
77 return (from > to);
80 static inline sysv_zone_t *block_end(struct buffer_head *bh)
82 return (sysv_zone_t*)((char*)bh->b_data + bh->b_size);
86 * Requires read_lock(&pointers_lock) or write_lock(&pointers_lock)
88 static Indirect *get_branch(struct inode *inode,
89 int depth,
90 int offsets[],
91 Indirect chain[],
92 int *err)
94 struct super_block *sb = inode->i_sb;
95 Indirect *p = chain;
96 struct buffer_head *bh;
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;
113 return NULL;
115 changed:
116 brelse(bh);
117 *err = -EAGAIN;
118 goto no_block;
119 failure:
120 *err = -EIO;
121 no_block:
122 return p;
125 static int alloc_branch(struct inode *inode,
126 int num,
127 int *offsets,
128 Indirect *branch)
130 int blocksize = inode->i_sb->s_blocksize;
131 int n = 0;
132 int i;
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;
143 * Get buffer_head for parent block, zero it out and set
144 * the pointer to new one, then send parent to disk.
146 parent = block_to_cpu(SYSV_SB(inode->i_sb), branch[n-1].key);
147 bh = sb_getblk(inode->i_sb, parent);
148 lock_buffer(bh);
149 memset(bh->b_data, 0, blocksize);
150 branch[n].bh = bh;
151 branch[n].p = (sysv_zone_t*) bh->b_data + offsets[n];
152 *branch[n].p = branch[n].key;
153 set_buffer_uptodate(bh);
154 unlock_buffer(bh);
155 dirty_indirect(bh, inode);
157 if (n == num)
158 return 0;
160 /* Allocation failed, free what we already allocated */
161 for (i = 1; i < n; i++)
162 bforget(branch[i].bh);
163 for (i = 0; i < n; i++)
164 sysv_free_block(inode->i_sb, branch[i].key);
165 return -ENOSPC;
168 static inline int splice_branch(struct inode *inode,
169 Indirect chain[],
170 Indirect *where,
171 int num)
173 int i;
175 /* Verify that place we are splicing to is still there and vacant */
176 write_lock(&pointers_lock);
177 if (!verify_chain(chain, where-1) || *where->p)
178 goto changed;
179 *where->p = where->key;
180 write_unlock(&pointers_lock);
182 inode->i_ctime = current_time(inode);
184 /* had we spliced it onto indirect block? */
185 if (where->bh)
186 dirty_indirect(where->bh, inode);
188 if (IS_SYNC(inode))
189 sysv_sync_inode(inode);
190 else
191 mark_inode_dirty(inode);
192 return 0;
194 changed:
195 write_unlock(&pointers_lock);
196 for (i = 1; i < num; i++)
197 bforget(where[i].bh);
198 for (i = 0; i < num; i++)
199 sysv_free_block(inode->i_sb, where[i].key);
200 return -EAGAIN;
203 static int get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create)
205 int err = -EIO;
206 int offsets[DEPTH];
207 Indirect chain[DEPTH];
208 struct super_block *sb = inode->i_sb;
209 Indirect *partial;
210 int left;
211 int depth = block_to_path(inode, iblock, offsets);
213 if (depth == 0)
214 goto out;
216 reread:
217 read_lock(&pointers_lock);
218 partial = get_branch(inode, depth, offsets, chain, &err);
219 read_unlock(&pointers_lock);
221 /* Simplest case - block found, no allocation needed */
222 if (!partial) {
223 got_it:
224 map_bh(bh_result, sb, block_to_cpu(SYSV_SB(sb),
225 chain[depth-1].key));
226 /* Clean up and exit */
227 partial = chain+depth-1; /* the whole chain */
228 goto cleanup;
231 /* Next simple case - plain lookup or failed read of indirect block */
232 if (!create || err == -EIO) {
233 cleanup:
234 while (partial > chain) {
235 brelse(partial->bh);
236 partial--;
238 out:
239 return err;
243 * Indirect block might be removed by truncate while we were
244 * reading it. Handling of that case (forget what we've got and
245 * reread) is taken out of the main path.
247 if (err == -EAGAIN)
248 goto changed;
250 left = (chain + depth) - partial;
251 err = alloc_branch(inode, left, offsets+(partial-chain), partial);
252 if (err)
253 goto cleanup;
255 if (splice_branch(inode, chain, partial, left) < 0)
256 goto changed;
258 set_buffer_new(bh_result);
259 goto got_it;
261 changed:
262 while (partial > chain) {
263 brelse(partial->bh);
264 partial--;
266 goto reread;
269 static inline int all_zeroes(sysv_zone_t *p, sysv_zone_t *q)
271 while (p < q)
272 if (*p++)
273 return 0;
274 return 1;
277 static Indirect *find_shared(struct inode *inode,
278 int depth,
279 int offsets[],
280 Indirect chain[],
281 sysv_zone_t *top)
283 Indirect *partial, *p;
284 int k, err;
286 *top = 0;
287 for (k = depth; k > 1 && !offsets[k-1]; k--)
290 write_lock(&pointers_lock);
291 partial = get_branch(inode, k, offsets, chain, &err);
292 if (!partial)
293 partial = chain + k-1;
295 * If the branch acquired continuation since we've looked at it -
296 * fine, it should all survive and (new) top doesn't belong to us.
298 if (!partial->key && *partial->p) {
299 write_unlock(&pointers_lock);
300 goto no_top;
302 for (p=partial; p>chain && all_zeroes((sysv_zone_t*)p->bh->b_data,p->p); p--)
305 * OK, we've found the last block that must survive. The rest of our
306 * branch should be detached before unlocking. However, if that rest
307 * of branch is all ours and does not grow immediately from the inode
308 * it's easier to cheat and just decrement partial->p.
310 if (p == chain + k - 1 && p > chain) {
311 p->p--;
312 } else {
313 *top = *p->p;
314 *p->p = 0;
316 write_unlock(&pointers_lock);
318 while (partial > p) {
319 brelse(partial->bh);
320 partial--;
322 no_top:
323 return partial;
326 static inline void free_data(struct inode *inode, sysv_zone_t *p, sysv_zone_t *q)
328 for ( ; p < q ; p++) {
329 sysv_zone_t nr = *p;
330 if (nr) {
331 *p = 0;
332 sysv_free_block(inode->i_sb, nr);
333 mark_inode_dirty(inode);
338 static void free_branches(struct inode *inode, sysv_zone_t *p, sysv_zone_t *q, int depth)
340 struct buffer_head * bh;
341 struct super_block *sb = inode->i_sb;
343 if (depth--) {
344 for ( ; p < q ; p++) {
345 int block;
346 sysv_zone_t nr = *p;
347 if (!nr)
348 continue;
349 *p = 0;
350 block = block_to_cpu(SYSV_SB(sb), nr);
351 bh = sb_bread(sb, block);
352 if (!bh)
353 continue;
354 free_branches(inode, (sysv_zone_t*)bh->b_data,
355 block_end(bh), depth);
356 bforget(bh);
357 sysv_free_block(sb, nr);
358 mark_inode_dirty(inode);
360 } else
361 free_data(inode, p, q);
364 void sysv_truncate (struct inode * inode)
366 sysv_zone_t *i_data = SYSV_I(inode)->i_data;
367 int offsets[DEPTH];
368 Indirect chain[DEPTH];
369 Indirect *partial;
370 sysv_zone_t nr = 0;
371 int n;
372 long iblock;
373 unsigned blocksize;
375 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
376 S_ISLNK(inode->i_mode)))
377 return;
379 blocksize = inode->i_sb->s_blocksize;
380 iblock = (inode->i_size + blocksize-1)
381 >> inode->i_sb->s_blocksize_bits;
383 block_truncate_page(inode->i_mapping, inode->i_size, get_block);
385 n = block_to_path(inode, iblock, offsets);
386 if (n == 0)
387 return;
389 if (n == 1) {
390 free_data(inode, i_data+offsets[0], i_data + DIRECT);
391 goto do_indirects;
394 partial = find_shared(inode, n, offsets, chain, &nr);
395 /* Kill the top of shared branch (already detached) */
396 if (nr) {
397 if (partial == chain)
398 mark_inode_dirty(inode);
399 else
400 dirty_indirect(partial->bh, inode);
401 free_branches(inode, &nr, &nr+1, (chain+n-1) - partial);
403 /* Clear the ends of indirect blocks on the shared branch */
404 while (partial > chain) {
405 free_branches(inode, partial->p + 1, block_end(partial->bh),
406 (chain+n-1) - partial);
407 dirty_indirect(partial->bh, inode);
408 brelse (partial->bh);
409 partial--;
411 do_indirects:
412 /* Kill the remaining (whole) subtrees (== subtrees deeper than...) */
413 while (n < DEPTH) {
414 nr = i_data[DIRECT + n - 1];
415 if (nr) {
416 i_data[DIRECT + n - 1] = 0;
417 mark_inode_dirty(inode);
418 free_branches(inode, &nr, &nr+1, n);
420 n++;
422 inode->i_mtime = inode->i_ctime = current_time(inode);
423 if (IS_SYNC(inode))
424 sysv_sync_inode (inode);
425 else
426 mark_inode_dirty(inode);
429 static unsigned sysv_nblocks(struct super_block *s, loff_t size)
431 struct sysv_sb_info *sbi = SYSV_SB(s);
432 int ptrs_bits = sbi->s_ind_per_block_bits;
433 unsigned blocks, res, direct = DIRECT, i = DEPTH;
434 blocks = (size + s->s_blocksize - 1) >> s->s_blocksize_bits;
435 res = blocks;
436 while (--i && blocks > direct) {
437 blocks = ((blocks - direct - 1) >> ptrs_bits) + 1;
438 res += blocks;
439 direct = 1;
441 return blocks;
444 int sysv_getattr(const struct path *path, struct kstat *stat,
445 u32 request_mask, unsigned int flags)
447 struct super_block *s = path->dentry->d_sb;
448 generic_fillattr(d_inode(path->dentry), stat);
449 stat->blocks = (s->s_blocksize / 512) * sysv_nblocks(s, stat->size);
450 stat->blksize = s->s_blocksize;
451 return 0;
454 static int sysv_writepage(struct page *page, struct writeback_control *wbc)
456 return block_write_full_page(page,get_block,wbc);
459 static int sysv_readpage(struct file *file, struct page *page)
461 return block_read_full_page(page,get_block);
464 int sysv_prepare_chunk(struct page *page, loff_t pos, unsigned len)
466 return __block_write_begin(page, pos, len, get_block);
469 static void sysv_write_failed(struct address_space *mapping, loff_t to)
471 struct inode *inode = mapping->host;
473 if (to > inode->i_size) {
474 truncate_pagecache(inode, inode->i_size);
475 sysv_truncate(inode);
479 static int sysv_write_begin(struct file *file, struct address_space *mapping,
480 loff_t pos, unsigned len, unsigned flags,
481 struct page **pagep, void **fsdata)
483 int ret;
485 ret = block_write_begin(mapping, pos, len, flags, pagep, get_block);
486 if (unlikely(ret))
487 sysv_write_failed(mapping, pos + len);
489 return ret;
492 static sector_t sysv_bmap(struct address_space *mapping, sector_t block)
494 return generic_block_bmap(mapping,block,get_block);
497 const struct address_space_operations sysv_aops = {
498 .readpage = sysv_readpage,
499 .writepage = sysv_writepage,
500 .write_begin = sysv_write_begin,
501 .write_end = generic_write_end,
502 .bmap = sysv_bmap