• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
3  */
4 /* Reiserfs block (de)allocator, bitmap-based. */
5 
6 #include <linux/time.h>
7 #include "reiserfs.h"
8 #include <linux/errno.h>
9 #include <linux/buffer_head.h>
10 #include <linux/kernel.h>
11 #include <linux/pagemap.h>
12 #include <linux/vmalloc.h>
13 #include <linux/quotaops.h>
14 #include <linux/seq_file.h>
15 
16 #define PREALLOCATION_SIZE 9
17 
18 /* different reiserfs block allocator options */
19 
20 #define SB_ALLOC_OPTS(s) (REISERFS_SB(s)->s_alloc_options.bits)
21 
22 #define  _ALLOC_concentrating_formatted_nodes 0
23 #define  _ALLOC_displacing_large_files 1
24 #define  _ALLOC_displacing_new_packing_localities 2
25 #define  _ALLOC_old_hashed_relocation 3
26 #define  _ALLOC_new_hashed_relocation 4
27 #define  _ALLOC_skip_busy 5
28 #define  _ALLOC_displace_based_on_dirid 6
29 #define  _ALLOC_hashed_formatted_nodes 7
30 #define  _ALLOC_old_way 8
31 #define  _ALLOC_hundredth_slices 9
32 #define  _ALLOC_dirid_groups 10
33 #define  _ALLOC_oid_groups 11
34 #define  _ALLOC_packing_groups 12
35 
36 #define  concentrating_formatted_nodes(s)	test_bit(_ALLOC_concentrating_formatted_nodes, &SB_ALLOC_OPTS(s))
37 #define  displacing_large_files(s)		test_bit(_ALLOC_displacing_large_files, &SB_ALLOC_OPTS(s))
38 #define  displacing_new_packing_localities(s)	test_bit(_ALLOC_displacing_new_packing_localities, &SB_ALLOC_OPTS(s))
39 
40 #define SET_OPTION(optname) \
41    do { \
42 	reiserfs_info(s, "block allocator option \"%s\" is set", #optname); \
43 	set_bit(_ALLOC_ ## optname , &SB_ALLOC_OPTS(s)); \
44     } while(0)
45 #define TEST_OPTION(optname, s) \
46     test_bit(_ALLOC_ ## optname , &SB_ALLOC_OPTS(s))
47 
get_bit_address(struct super_block * s,b_blocknr_t block,unsigned int * bmap_nr,unsigned int * offset)48 static inline void get_bit_address(struct super_block *s,
49 				   b_blocknr_t block,
50 				   unsigned int *bmap_nr,
51 				   unsigned int *offset)
52 {
53 	/*
54 	 * It is in the bitmap block number equal to the block
55 	 * number divided by the number of bits in a block.
56 	 */
57 	*bmap_nr = block >> (s->s_blocksize_bits + 3);
58 	/* Within that bitmap block it is located at bit offset *offset. */
59 	*offset = block & ((s->s_blocksize << 3) - 1);
60 }
61 
is_reusable(struct super_block * s,b_blocknr_t block,int bit_value)62 int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value)
63 {
64 	unsigned int bmap, offset;
65 	unsigned int bmap_count = reiserfs_bmap_count(s);
66 
67 	if (block == 0 || block >= SB_BLOCK_COUNT(s)) {
68 		reiserfs_error(s, "vs-4010",
69 			       "block number is out of range %lu (%u)",
70 			       block, SB_BLOCK_COUNT(s));
71 		return 0;
72 	}
73 
74 	get_bit_address(s, block, &bmap, &offset);
75 
76 	/*
77 	 * Old format filesystem? Unlikely, but the bitmaps are all
78 	 * up front so we need to account for it.
79 	 */
80 	if (unlikely(test_bit(REISERFS_OLD_FORMAT,
81 			      &REISERFS_SB(s)->s_properties))) {
82 		b_blocknr_t bmap1 = REISERFS_SB(s)->s_sbh->b_blocknr + 1;
83 		if (block >= bmap1 &&
84 		    block <= bmap1 + bmap_count) {
85 			reiserfs_error(s, "vs-4019", "bitmap block %lu(%u) "
86 				       "can't be freed or reused",
87 				       block, bmap_count);
88 			return 0;
89 		}
90 	} else {
91 		if (offset == 0) {
92 			reiserfs_error(s, "vs-4020", "bitmap block %lu(%u) "
93 				       "can't be freed or reused",
94 				       block, bmap_count);
95 			return 0;
96 		}
97 	}
98 
99 	if (bmap >= bmap_count) {
100 		reiserfs_error(s, "vs-4030", "bitmap for requested block "
101 			       "is out of range: block=%lu, bitmap_nr=%u",
102 			       block, bmap);
103 		return 0;
104 	}
105 
106 	if (bit_value == 0 && block == SB_ROOT_BLOCK(s)) {
107 		reiserfs_error(s, "vs-4050", "this is root block (%u), "
108 			       "it must be busy", SB_ROOT_BLOCK(s));
109 		return 0;
110 	}
111 
112 	return 1;
113 }
114 
115 /*
116  * Searches in journal structures for a given block number (bmap, off).
117  * If block is found in reiserfs journal it suggests next free block
118  * candidate to test.
119  */
is_block_in_journal(struct super_block * s,unsigned int bmap,int off,int * next)120 static inline int is_block_in_journal(struct super_block *s, unsigned int bmap,
121 				      int off, int *next)
122 {
123 	b_blocknr_t tmp;
124 
125 	if (reiserfs_in_journal(s, bmap, off, 1, &tmp)) {
126 		if (tmp) {	/* hint supplied */
127 			*next = tmp;
128 			PROC_INFO_INC(s, scan_bitmap.in_journal_hint);
129 		} else {
130 			(*next) = off + 1;  /* inc offset to avoid looping. */
131 			PROC_INFO_INC(s, scan_bitmap.in_journal_nohint);
132 		}
133 		PROC_INFO_INC(s, scan_bitmap.retry);
134 		return 1;
135 	}
136 	return 0;
137 }
138 
139 /*
140  * Searches for a window of zero bits with given minimum and maximum
141  * lengths in one bitmap block
142  */
scan_bitmap_block(struct reiserfs_transaction_handle * th,unsigned int bmap_n,int * beg,int boundary,int min,int max,int unfm)143 static int scan_bitmap_block(struct reiserfs_transaction_handle *th,
144 			     unsigned int bmap_n, int *beg, int boundary,
145 			     int min, int max, int unfm)
146 {
147 	struct super_block *s = th->t_super;
148 	struct reiserfs_bitmap_info *bi = &SB_AP_BITMAP(s)[bmap_n];
149 	struct buffer_head *bh;
150 	int end, next;
151 	int org = *beg;
152 
153 	BUG_ON(!th->t_trans_id);
154 	RFALSE(bmap_n >= reiserfs_bmap_count(s), "Bitmap %u is out of "
155 	       "range (0..%u)", bmap_n, reiserfs_bmap_count(s) - 1);
156 	PROC_INFO_INC(s, scan_bitmap.bmap);
157 
158 	if (!bi) {
159 		reiserfs_error(s, "jdm-4055", "NULL bitmap info pointer "
160 			       "for bitmap %d", bmap_n);
161 		return 0;
162 	}
163 
164 	bh = reiserfs_read_bitmap_block(s, bmap_n);
165 	if (bh == NULL)
166 		return 0;
167 
168 	while (1) {
169 cont:
170 		if (bi->free_count < min) {
171 			brelse(bh);
172 			return 0;	/* No free blocks in this bitmap */
173 		}
174 
175 		/* search for a first zero bit -- beginning of a window */
176 		*beg = reiserfs_find_next_zero_le_bit
177 		    ((unsigned long *)(bh->b_data), boundary, *beg);
178 
179 		/*
180 		 * search for a zero bit fails or the rest of bitmap block
181 		 * cannot contain a zero window of minimum size
182 		 */
183 		if (*beg + min > boundary) {
184 			brelse(bh);
185 			return 0;
186 		}
187 
188 		if (unfm && is_block_in_journal(s, bmap_n, *beg, beg))
189 			continue;
190 		/* first zero bit found; we check next bits */
191 		for (end = *beg + 1;; end++) {
192 			if (end >= *beg + max || end >= boundary
193 			    || reiserfs_test_le_bit(end, bh->b_data)) {
194 				next = end;
195 				break;
196 			}
197 
198 			/*
199 			 * finding the other end of zero bit window requires
200 			 * looking into journal structures (in case of
201 			 * searching for free blocks for unformatted nodes)
202 			 */
203 			if (unfm && is_block_in_journal(s, bmap_n, end, &next))
204 				break;
205 		}
206 
207 		/*
208 		 * now (*beg) points to beginning of zero bits window,
209 		 * (end) points to one bit after the window end
210 		 */
211 
212 		/* found window of proper size */
213 		if (end - *beg >= min) {
214 			int i;
215 			reiserfs_prepare_for_journal(s, bh, 1);
216 			/*
217 			 * try to set all blocks used checking are
218 			 * they still free
219 			 */
220 			for (i = *beg; i < end; i++) {
221 				/* Don't check in journal again. */
222 				if (reiserfs_test_and_set_le_bit
223 				    (i, bh->b_data)) {
224 					/*
225 					 * bit was set by another process while
226 					 * we slept in prepare_for_journal()
227 					 */
228 					PROC_INFO_INC(s, scan_bitmap.stolen);
229 
230 					/*
231 					 * we can continue with smaller set
232 					 * of allocated blocks, if length of
233 					 * this set is more or equal to `min'
234 					 */
235 					if (i >= *beg + min) {
236 						end = i;
237 						break;
238 					}
239 
240 					/*
241 					 * otherwise we clear all bit
242 					 * were set ...
243 					 */
244 					while (--i >= *beg)
245 						reiserfs_clear_le_bit
246 						    (i, bh->b_data);
247 					reiserfs_restore_prepared_buffer(s, bh);
248 					*beg = org;
249 
250 					/*
251 					 * Search again in current block
252 					 * from beginning
253 					 */
254 					goto cont;
255 				}
256 			}
257 			bi->free_count -= (end - *beg);
258 			journal_mark_dirty(th, bh);
259 			brelse(bh);
260 
261 			/* free block count calculation */
262 			reiserfs_prepare_for_journal(s, SB_BUFFER_WITH_SB(s),
263 						     1);
264 			PUT_SB_FREE_BLOCKS(s, SB_FREE_BLOCKS(s) - (end - *beg));
265 			journal_mark_dirty(th, SB_BUFFER_WITH_SB(s));
266 
267 			return end - (*beg);
268 		} else {
269 			*beg = next;
270 		}
271 	}
272 }
273 
bmap_hash_id(struct super_block * s,u32 id)274 static int bmap_hash_id(struct super_block *s, u32 id)
275 {
276 	char *hash_in = NULL;
277 	unsigned long hash;
278 	unsigned bm;
279 
280 	if (id <= 2) {
281 		bm = 1;
282 	} else {
283 		hash_in = (char *)(&id);
284 		hash = keyed_hash(hash_in, 4);
285 		bm = hash % reiserfs_bmap_count(s);
286 		if (!bm)
287 			bm = 1;
288 	}
289 	/* this can only be true when SB_BMAP_NR = 1 */
290 	if (bm >= reiserfs_bmap_count(s))
291 		bm = 0;
292 	return bm;
293 }
294 
295 /*
296  * hashes the id and then returns > 0 if the block group for the
297  * corresponding hash is full
298  */
block_group_used(struct super_block * s,u32 id)299 static inline int block_group_used(struct super_block *s, u32 id)
300 {
301 	int bm = bmap_hash_id(s, id);
302 	struct reiserfs_bitmap_info *info = &SB_AP_BITMAP(s)[bm];
303 
304 	/*
305 	 * If we don't have cached information on this bitmap block, we're
306 	 * going to have to load it later anyway. Loading it here allows us
307 	 * to make a better decision. This favors long-term performance gain
308 	 * with a better on-disk layout vs. a short term gain of skipping the
309 	 * read and potentially having a bad placement.
310 	 */
311 	if (info->free_count == UINT_MAX) {
312 		struct buffer_head *bh = reiserfs_read_bitmap_block(s, bm);
313 		brelse(bh);
314 	}
315 
316 	if (info->free_count > ((s->s_blocksize << 3) * 60 / 100)) {
317 		return 0;
318 	}
319 	return 1;
320 }
321 
322 /*
323  * the packing is returned in disk byte order
324  */
reiserfs_choose_packing(struct inode * dir)325 __le32 reiserfs_choose_packing(struct inode * dir)
326 {
327 	__le32 packing;
328 	if (TEST_OPTION(packing_groups, dir->i_sb)) {
329 		u32 parent_dir = le32_to_cpu(INODE_PKEY(dir)->k_dir_id);
330 		/*
331 		 * some versions of reiserfsck expect packing locality 1 to be
332 		 * special
333 		 */
334 		if (parent_dir == 1 || block_group_used(dir->i_sb, parent_dir))
335 			packing = INODE_PKEY(dir)->k_objectid;
336 		else
337 			packing = INODE_PKEY(dir)->k_dir_id;
338 	} else
339 		packing = INODE_PKEY(dir)->k_objectid;
340 	return packing;
341 }
342 
343 /*
344  * Tries to find contiguous zero bit window (given size) in given region of
345  * bitmap and place new blocks there. Returns number of allocated blocks.
346  */
scan_bitmap(struct reiserfs_transaction_handle * th,b_blocknr_t * start,b_blocknr_t finish,int min,int max,int unfm,sector_t file_block)347 static int scan_bitmap(struct reiserfs_transaction_handle *th,
348 		       b_blocknr_t * start, b_blocknr_t finish,
349 		       int min, int max, int unfm, sector_t file_block)
350 {
351 	int nr_allocated = 0;
352 	struct super_block *s = th->t_super;
353 	unsigned int bm, off;
354 	unsigned int end_bm, end_off;
355 	unsigned int off_max = s->s_blocksize << 3;
356 
357 	BUG_ON(!th->t_trans_id);
358 	PROC_INFO_INC(s, scan_bitmap.call);
359 
360 	/* No point in looking for more free blocks */
361 	if (SB_FREE_BLOCKS(s) <= 0)
362 		return 0;
363 
364 	get_bit_address(s, *start, &bm, &off);
365 	get_bit_address(s, finish, &end_bm, &end_off);
366 	if (bm > reiserfs_bmap_count(s))
367 		return 0;
368 	if (end_bm > reiserfs_bmap_count(s))
369 		end_bm = reiserfs_bmap_count(s);
370 
371 	/*
372 	 * When the bitmap is more than 10% free, anyone can allocate.
373 	 * When it's less than 10% free, only files that already use the
374 	 * bitmap are allowed. Once we pass 80% full, this restriction
375 	 * is lifted.
376 	 *
377 	 * We do this so that files that grow later still have space close to
378 	 * their original allocation. This improves locality, and presumably
379 	 * performance as a result.
380 	 *
381 	 * This is only an allocation policy and does not make up for getting a
382 	 * bad hint. Decent hinting must be implemented for this to work well.
383 	 */
384 	if (TEST_OPTION(skip_busy, s)
385 	    && SB_FREE_BLOCKS(s) > SB_BLOCK_COUNT(s) / 20) {
386 		for (; bm < end_bm; bm++, off = 0) {
387 			if ((off && (!unfm || (file_block != 0)))
388 			    || SB_AP_BITMAP(s)[bm].free_count >
389 			    (s->s_blocksize << 3) / 10)
390 				nr_allocated =
391 				    scan_bitmap_block(th, bm, &off, off_max,
392 						      min, max, unfm);
393 			if (nr_allocated)
394 				goto ret;
395 		}
396 		/* we know from above that start is a reasonable number */
397 		get_bit_address(s, *start, &bm, &off);
398 	}
399 
400 	for (; bm < end_bm; bm++, off = 0) {
401 		nr_allocated =
402 		    scan_bitmap_block(th, bm, &off, off_max, min, max, unfm);
403 		if (nr_allocated)
404 			goto ret;
405 	}
406 
407 	nr_allocated =
408 	    scan_bitmap_block(th, bm, &off, end_off + 1, min, max, unfm);
409 
410 ret:
411 	*start = bm * off_max + off;
412 	return nr_allocated;
413 
414 }
415 
_reiserfs_free_block(struct reiserfs_transaction_handle * th,struct inode * inode,b_blocknr_t block,int for_unformatted)416 static void _reiserfs_free_block(struct reiserfs_transaction_handle *th,
417 				 struct inode *inode, b_blocknr_t block,
418 				 int for_unformatted)
419 {
420 	struct super_block *s = th->t_super;
421 	struct reiserfs_super_block *rs;
422 	struct buffer_head *sbh, *bmbh;
423 	struct reiserfs_bitmap_info *apbi;
424 	unsigned int nr, offset;
425 
426 	BUG_ON(!th->t_trans_id);
427 	PROC_INFO_INC(s, free_block);
428 	rs = SB_DISK_SUPER_BLOCK(s);
429 	sbh = SB_BUFFER_WITH_SB(s);
430 	apbi = SB_AP_BITMAP(s);
431 
432 	get_bit_address(s, block, &nr, &offset);
433 
434 	if (nr >= reiserfs_bmap_count(s)) {
435 		reiserfs_error(s, "vs-4075", "block %lu is out of range",
436 			       block);
437 		return;
438 	}
439 
440 	bmbh = reiserfs_read_bitmap_block(s, nr);
441 	if (!bmbh)
442 		return;
443 
444 	reiserfs_prepare_for_journal(s, bmbh, 1);
445 
446 	/* clear bit for the given block in bit map */
447 	if (!reiserfs_test_and_clear_le_bit(offset, bmbh->b_data)) {
448 		reiserfs_error(s, "vs-4080",
449 			       "block %lu: bit already cleared", block);
450 	}
451 	apbi[nr].free_count++;
452 	journal_mark_dirty(th, bmbh);
453 	brelse(bmbh);
454 
455 	reiserfs_prepare_for_journal(s, sbh, 1);
456 	/* update super block */
457 	set_sb_free_blocks(rs, sb_free_blocks(rs) + 1);
458 
459 	journal_mark_dirty(th, sbh);
460 	if (for_unformatted) {
461 		int depth = reiserfs_write_unlock_nested(s);
462 		dquot_free_block_nodirty(inode, 1);
463 		reiserfs_write_lock_nested(s, depth);
464 	}
465 }
466 
reiserfs_free_block(struct reiserfs_transaction_handle * th,struct inode * inode,b_blocknr_t block,int for_unformatted)467 void reiserfs_free_block(struct reiserfs_transaction_handle *th,
468 			 struct inode *inode, b_blocknr_t block,
469 			 int for_unformatted)
470 {
471 	struct super_block *s = th->t_super;
472 
473 	BUG_ON(!th->t_trans_id);
474 	RFALSE(!s, "vs-4061: trying to free block on nonexistent device");
475 	if (!is_reusable(s, block, 1))
476 		return;
477 
478 	if (block > sb_block_count(REISERFS_SB(s)->s_rs)) {
479 		reiserfs_error(th->t_super, "bitmap-4072",
480 			       "Trying to free block outside file system "
481 			       "boundaries (%lu > %lu)",
482 			       block, sb_block_count(REISERFS_SB(s)->s_rs));
483 		return;
484 	}
485 	/* mark it before we clear it, just in case */
486 	journal_mark_freed(th, s, block);
487 	_reiserfs_free_block(th, inode, block, for_unformatted);
488 }
489 
490 /* preallocated blocks don't need to be run through journal_mark_freed */
reiserfs_free_prealloc_block(struct reiserfs_transaction_handle * th,struct inode * inode,b_blocknr_t block)491 static void reiserfs_free_prealloc_block(struct reiserfs_transaction_handle *th,
492 					 struct inode *inode, b_blocknr_t block)
493 {
494 	BUG_ON(!th->t_trans_id);
495 	RFALSE(!th->t_super,
496 	       "vs-4060: trying to free block on nonexistent device");
497 	if (!is_reusable(th->t_super, block, 1))
498 		return;
499 	_reiserfs_free_block(th, inode, block, 1);
500 }
501 
__discard_prealloc(struct reiserfs_transaction_handle * th,struct reiserfs_inode_info * ei)502 static void __discard_prealloc(struct reiserfs_transaction_handle *th,
503 			       struct reiserfs_inode_info *ei)
504 {
505 	unsigned long save = ei->i_prealloc_block;
506 	int dirty = 0;
507 	struct inode *inode = &ei->vfs_inode;
508 
509 	BUG_ON(!th->t_trans_id);
510 #ifdef CONFIG_REISERFS_CHECK
511 	if (ei->i_prealloc_count < 0)
512 		reiserfs_error(th->t_super, "zam-4001",
513 			       "inode has negative prealloc blocks count.");
514 #endif
515 	while (ei->i_prealloc_count > 0) {
516 		b_blocknr_t block_to_free;
517 
518 		/*
519 		 * reiserfs_free_prealloc_block can drop the write lock,
520 		 * which could allow another caller to free the same block.
521 		 * We can protect against it by modifying the prealloc
522 		 * state before calling it.
523 		 */
524 		block_to_free = ei->i_prealloc_block++;
525 		ei->i_prealloc_count--;
526 		reiserfs_free_prealloc_block(th, inode, block_to_free);
527 		dirty = 1;
528 	}
529 	if (dirty)
530 		reiserfs_update_sd(th, inode);
531 	ei->i_prealloc_block = save;
532 	list_del_init(&ei->i_prealloc_list);
533 }
534 
535 /* FIXME: It should be inline function */
reiserfs_discard_prealloc(struct reiserfs_transaction_handle * th,struct inode * inode)536 void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
537 			       struct inode *inode)
538 {
539 	struct reiserfs_inode_info *ei = REISERFS_I(inode);
540 
541 	BUG_ON(!th->t_trans_id);
542 	if (ei->i_prealloc_count)
543 		__discard_prealloc(th, ei);
544 }
545 
reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle * th)546 void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th)
547 {
548 	struct list_head *plist = &SB_JOURNAL(th->t_super)->j_prealloc_list;
549 
550 	BUG_ON(!th->t_trans_id);
551 	while (!list_empty(plist)) {
552 		struct reiserfs_inode_info *ei;
553 		ei = list_entry(plist->next, struct reiserfs_inode_info,
554 				i_prealloc_list);
555 #ifdef CONFIG_REISERFS_CHECK
556 		if (!ei->i_prealloc_count) {
557 			reiserfs_error(th->t_super, "zam-4001",
558 				       "inode is in prealloc list but has "
559 				       "no preallocated blocks.");
560 		}
561 #endif
562 		__discard_prealloc(th, ei);
563 	}
564 }
565 
reiserfs_init_alloc_options(struct super_block * s)566 void reiserfs_init_alloc_options(struct super_block *s)
567 {
568 	set_bit(_ALLOC_skip_busy, &SB_ALLOC_OPTS(s));
569 	set_bit(_ALLOC_dirid_groups, &SB_ALLOC_OPTS(s));
570 	set_bit(_ALLOC_packing_groups, &SB_ALLOC_OPTS(s));
571 }
572 
573 /* block allocator related options are parsed here */
reiserfs_parse_alloc_options(struct super_block * s,char * options)574 int reiserfs_parse_alloc_options(struct super_block *s, char *options)
575 {
576 	char *this_char, *value;
577 
578 	/* clear default settings */
579 	REISERFS_SB(s)->s_alloc_options.bits = 0;
580 
581 	while ((this_char = strsep(&options, ":")) != NULL) {
582 		if ((value = strchr(this_char, '=')) != NULL)
583 			*value++ = 0;
584 
585 		if (!strcmp(this_char, "concentrating_formatted_nodes")) {
586 			int temp;
587 			SET_OPTION(concentrating_formatted_nodes);
588 			temp = (value
589 				&& *value) ? simple_strtoul(value, &value,
590 							    0) : 10;
591 			if (temp <= 0 || temp > 100) {
592 				REISERFS_SB(s)->s_alloc_options.border = 10;
593 			} else {
594 				REISERFS_SB(s)->s_alloc_options.border =
595 				    100 / temp;
596 			}
597 			continue;
598 		}
599 		if (!strcmp(this_char, "displacing_large_files")) {
600 			SET_OPTION(displacing_large_files);
601 			REISERFS_SB(s)->s_alloc_options.large_file_size =
602 			    (value
603 			     && *value) ? simple_strtoul(value, &value, 0) : 16;
604 			continue;
605 		}
606 		if (!strcmp(this_char, "displacing_new_packing_localities")) {
607 			SET_OPTION(displacing_new_packing_localities);
608 			continue;
609 		}
610 
611 		if (!strcmp(this_char, "old_hashed_relocation")) {
612 			SET_OPTION(old_hashed_relocation);
613 			continue;
614 		}
615 
616 		if (!strcmp(this_char, "new_hashed_relocation")) {
617 			SET_OPTION(new_hashed_relocation);
618 			continue;
619 		}
620 
621 		if (!strcmp(this_char, "dirid_groups")) {
622 			SET_OPTION(dirid_groups);
623 			continue;
624 		}
625 		if (!strcmp(this_char, "oid_groups")) {
626 			SET_OPTION(oid_groups);
627 			continue;
628 		}
629 		if (!strcmp(this_char, "packing_groups")) {
630 			SET_OPTION(packing_groups);
631 			continue;
632 		}
633 		if (!strcmp(this_char, "hashed_formatted_nodes")) {
634 			SET_OPTION(hashed_formatted_nodes);
635 			continue;
636 		}
637 
638 		if (!strcmp(this_char, "skip_busy")) {
639 			SET_OPTION(skip_busy);
640 			continue;
641 		}
642 
643 		if (!strcmp(this_char, "hundredth_slices")) {
644 			SET_OPTION(hundredth_slices);
645 			continue;
646 		}
647 
648 		if (!strcmp(this_char, "old_way")) {
649 			SET_OPTION(old_way);
650 			continue;
651 		}
652 
653 		if (!strcmp(this_char, "displace_based_on_dirid")) {
654 			SET_OPTION(displace_based_on_dirid);
655 			continue;
656 		}
657 
658 		if (!strcmp(this_char, "preallocmin")) {
659 			REISERFS_SB(s)->s_alloc_options.preallocmin =
660 			    (value
661 			     && *value) ? simple_strtoul(value, &value, 0) : 4;
662 			continue;
663 		}
664 
665 		if (!strcmp(this_char, "preallocsize")) {
666 			REISERFS_SB(s)->s_alloc_options.preallocsize =
667 			    (value
668 			     && *value) ? simple_strtoul(value, &value,
669 							 0) :
670 			    PREALLOCATION_SIZE;
671 			continue;
672 		}
673 
674 		reiserfs_warning(s, "zam-4001", "unknown option - %s",
675 				 this_char);
676 		return 1;
677 	}
678 
679 	reiserfs_info(s, "allocator options = [%08x]\n", SB_ALLOC_OPTS(s));
680 	return 0;
681 }
682 
print_sep(struct seq_file * seq,int * first)683 static void print_sep(struct seq_file *seq, int *first)
684 {
685 	if (!*first)
686 		seq_puts(seq, ":");
687 	else
688 		*first = 0;
689 }
690 
show_alloc_options(struct seq_file * seq,struct super_block * s)691 void show_alloc_options(struct seq_file *seq, struct super_block *s)
692 {
693 	int first = 1;
694 
695 	if (SB_ALLOC_OPTS(s) == ((1 << _ALLOC_skip_busy) |
696 		(1 << _ALLOC_dirid_groups) | (1 << _ALLOC_packing_groups)))
697 		return;
698 
699 	seq_puts(seq, ",alloc=");
700 
701 	if (TEST_OPTION(concentrating_formatted_nodes, s)) {
702 		print_sep(seq, &first);
703 		if (REISERFS_SB(s)->s_alloc_options.border != 10) {
704 			seq_printf(seq, "concentrating_formatted_nodes=%d",
705 				100 / REISERFS_SB(s)->s_alloc_options.border);
706 		} else
707 			seq_puts(seq, "concentrating_formatted_nodes");
708 	}
709 	if (TEST_OPTION(displacing_large_files, s)) {
710 		print_sep(seq, &first);
711 		if (REISERFS_SB(s)->s_alloc_options.large_file_size != 16) {
712 			seq_printf(seq, "displacing_large_files=%lu",
713 			    REISERFS_SB(s)->s_alloc_options.large_file_size);
714 		} else
715 			seq_puts(seq, "displacing_large_files");
716 	}
717 	if (TEST_OPTION(displacing_new_packing_localities, s)) {
718 		print_sep(seq, &first);
719 		seq_puts(seq, "displacing_new_packing_localities");
720 	}
721 	if (TEST_OPTION(old_hashed_relocation, s)) {
722 		print_sep(seq, &first);
723 		seq_puts(seq, "old_hashed_relocation");
724 	}
725 	if (TEST_OPTION(new_hashed_relocation, s)) {
726 		print_sep(seq, &first);
727 		seq_puts(seq, "new_hashed_relocation");
728 	}
729 	if (TEST_OPTION(dirid_groups, s)) {
730 		print_sep(seq, &first);
731 		seq_puts(seq, "dirid_groups");
732 	}
733 	if (TEST_OPTION(oid_groups, s)) {
734 		print_sep(seq, &first);
735 		seq_puts(seq, "oid_groups");
736 	}
737 	if (TEST_OPTION(packing_groups, s)) {
738 		print_sep(seq, &first);
739 		seq_puts(seq, "packing_groups");
740 	}
741 	if (TEST_OPTION(hashed_formatted_nodes, s)) {
742 		print_sep(seq, &first);
743 		seq_puts(seq, "hashed_formatted_nodes");
744 	}
745 	if (TEST_OPTION(skip_busy, s)) {
746 		print_sep(seq, &first);
747 		seq_puts(seq, "skip_busy");
748 	}
749 	if (TEST_OPTION(hundredth_slices, s)) {
750 		print_sep(seq, &first);
751 		seq_puts(seq, "hundredth_slices");
752 	}
753 	if (TEST_OPTION(old_way, s)) {
754 		print_sep(seq, &first);
755 		seq_puts(seq, "old_way");
756 	}
757 	if (TEST_OPTION(displace_based_on_dirid, s)) {
758 		print_sep(seq, &first);
759 		seq_puts(seq, "displace_based_on_dirid");
760 	}
761 	if (REISERFS_SB(s)->s_alloc_options.preallocmin != 0) {
762 		print_sep(seq, &first);
763 		seq_printf(seq, "preallocmin=%d",
764 				REISERFS_SB(s)->s_alloc_options.preallocmin);
765 	}
766 	if (REISERFS_SB(s)->s_alloc_options.preallocsize != 17) {
767 		print_sep(seq, &first);
768 		seq_printf(seq, "preallocsize=%d",
769 				REISERFS_SB(s)->s_alloc_options.preallocsize);
770 	}
771 }
772 
new_hashed_relocation(reiserfs_blocknr_hint_t * hint)773 static inline void new_hashed_relocation(reiserfs_blocknr_hint_t * hint)
774 {
775 	char *hash_in;
776 
777 	if (hint->formatted_node) {
778 		hash_in = (char *)&hint->key.k_dir_id;
779 	} else {
780 		if (!hint->inode) {
781 			/*hint->search_start = hint->beg;*/
782 			hash_in = (char *)&hint->key.k_dir_id;
783 		} else
784 		    if (TEST_OPTION(displace_based_on_dirid, hint->th->t_super))
785 			hash_in = (char *)(&INODE_PKEY(hint->inode)->k_dir_id);
786 		else
787 			hash_in =
788 			    (char *)(&INODE_PKEY(hint->inode)->k_objectid);
789 	}
790 
791 	hint->search_start =
792 	    hint->beg + keyed_hash(hash_in, 4) % (hint->end - hint->beg);
793 }
794 
795 /*
796  * Relocation based on dirid, hashing them into a given bitmap block
797  * files. Formatted nodes are unaffected, a separate policy covers them
798  */
dirid_groups(reiserfs_blocknr_hint_t * hint)799 static void dirid_groups(reiserfs_blocknr_hint_t * hint)
800 {
801 	unsigned long hash;
802 	__u32 dirid = 0;
803 	int bm = 0;
804 	struct super_block *sb = hint->th->t_super;
805 
806 	if (hint->inode)
807 		dirid = le32_to_cpu(INODE_PKEY(hint->inode)->k_dir_id);
808 	else if (hint->formatted_node)
809 		dirid = hint->key.k_dir_id;
810 
811 	if (dirid) {
812 		bm = bmap_hash_id(sb, dirid);
813 		hash = bm * (sb->s_blocksize << 3);
814 		/* give a portion of the block group to metadata */
815 		if (hint->inode)
816 			hash += sb->s_blocksize / 2;
817 		hint->search_start = hash;
818 	}
819 }
820 
821 /*
822  * Relocation based on oid, hashing them into a given bitmap block
823  * files. Formatted nodes are unaffected, a separate policy covers them
824  */
oid_groups(reiserfs_blocknr_hint_t * hint)825 static void oid_groups(reiserfs_blocknr_hint_t * hint)
826 {
827 	if (hint->inode) {
828 		unsigned long hash;
829 		__u32 oid;
830 		__u32 dirid;
831 		int bm;
832 
833 		dirid = le32_to_cpu(INODE_PKEY(hint->inode)->k_dir_id);
834 
835 		/*
836 		 * keep the root dir and it's first set of subdirs close to
837 		 * the start of the disk
838 		 */
839 		if (dirid <= 2)
840 			hash = (hint->inode->i_sb->s_blocksize << 3);
841 		else {
842 			oid = le32_to_cpu(INODE_PKEY(hint->inode)->k_objectid);
843 			bm = bmap_hash_id(hint->inode->i_sb, oid);
844 			hash = bm * (hint->inode->i_sb->s_blocksize << 3);
845 		}
846 		hint->search_start = hash;
847 	}
848 }
849 
850 /*
851  * returns 1 if it finds an indirect item and gets valid hint info
852  * from it, otherwise 0
853  */
get_left_neighbor(reiserfs_blocknr_hint_t * hint)854 static int get_left_neighbor(reiserfs_blocknr_hint_t * hint)
855 {
856 	struct treepath *path;
857 	struct buffer_head *bh;
858 	struct item_head *ih;
859 	int pos_in_item;
860 	__le32 *item;
861 	int ret = 0;
862 
863 	/*
864 	 * reiserfs code can call this function w/o pointer to path
865 	 * structure supplied; then we rely on supplied search_start
866 	 */
867 	if (!hint->path)
868 		return 0;
869 
870 	path = hint->path;
871 	bh = get_last_bh(path);
872 	RFALSE(!bh, "green-4002: Illegal path specified to get_left_neighbor");
873 	ih = tp_item_head(path);
874 	pos_in_item = path->pos_in_item;
875 	item = tp_item_body(path);
876 
877 	hint->search_start = bh->b_blocknr;
878 
879 	/*
880 	 * for indirect item: go to left and look for the first non-hole entry
881 	 * in the indirect item
882 	 */
883 	if (!hint->formatted_node && is_indirect_le_ih(ih)) {
884 		if (pos_in_item == I_UNFM_NUM(ih))
885 			pos_in_item--;
886 		while (pos_in_item >= 0) {
887 			int t = get_block_num(item, pos_in_item);
888 			if (t) {
889 				hint->search_start = t;
890 				ret = 1;
891 				break;
892 			}
893 			pos_in_item--;
894 		}
895 	}
896 
897 	/* does result value fit into specified region? */
898 	return ret;
899 }
900 
901 /*
902  * should be, if formatted node, then try to put on first part of the device
903  * specified as number of percent with mount option device, else try to put
904  * on last of device.  This is not to say it is good code to do so,
905  * but the effect should be measured.
906  */
set_border_in_hint(struct super_block * s,reiserfs_blocknr_hint_t * hint)907 static inline void set_border_in_hint(struct super_block *s,
908 				      reiserfs_blocknr_hint_t * hint)
909 {
910 	b_blocknr_t border =
911 	    SB_BLOCK_COUNT(s) / REISERFS_SB(s)->s_alloc_options.border;
912 
913 	if (hint->formatted_node)
914 		hint->end = border - 1;
915 	else
916 		hint->beg = border;
917 }
918 
displace_large_file(reiserfs_blocknr_hint_t * hint)919 static inline void displace_large_file(reiserfs_blocknr_hint_t * hint)
920 {
921 	if (TEST_OPTION(displace_based_on_dirid, hint->th->t_super))
922 		hint->search_start =
923 		    hint->beg +
924 		    keyed_hash((char *)(&INODE_PKEY(hint->inode)->k_dir_id),
925 			       4) % (hint->end - hint->beg);
926 	else
927 		hint->search_start =
928 		    hint->beg +
929 		    keyed_hash((char *)(&INODE_PKEY(hint->inode)->k_objectid),
930 			       4) % (hint->end - hint->beg);
931 }
932 
hash_formatted_node(reiserfs_blocknr_hint_t * hint)933 static inline void hash_formatted_node(reiserfs_blocknr_hint_t * hint)
934 {
935 	char *hash_in;
936 
937 	if (!hint->inode)
938 		hash_in = (char *)&hint->key.k_dir_id;
939 	else if (TEST_OPTION(displace_based_on_dirid, hint->th->t_super))
940 		hash_in = (char *)(&INODE_PKEY(hint->inode)->k_dir_id);
941 	else
942 		hash_in = (char *)(&INODE_PKEY(hint->inode)->k_objectid);
943 
944 	hint->search_start =
945 	    hint->beg + keyed_hash(hash_in, 4) % (hint->end - hint->beg);
946 }
947 
948 static inline int
this_blocknr_allocation_would_make_it_a_large_file(reiserfs_blocknr_hint_t * hint)949 this_blocknr_allocation_would_make_it_a_large_file(reiserfs_blocknr_hint_t *
950 						   hint)
951 {
952 	return hint->block ==
953 	    REISERFS_SB(hint->th->t_super)->s_alloc_options.large_file_size;
954 }
955 
956 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
displace_new_packing_locality(reiserfs_blocknr_hint_t * hint)957 static inline void displace_new_packing_locality(reiserfs_blocknr_hint_t * hint)
958 {
959 	struct in_core_key *key = &hint->key;
960 
961 	hint->th->displace_new_blocks = 0;
962 	hint->search_start =
963 	    hint->beg + keyed_hash((char *)(&key->k_objectid),
964 				   4) % (hint->end - hint->beg);
965 }
966 #endif
967 
old_hashed_relocation(reiserfs_blocknr_hint_t * hint)968 static inline int old_hashed_relocation(reiserfs_blocknr_hint_t * hint)
969 {
970 	b_blocknr_t border;
971 	u32 hash_in;
972 
973 	if (hint->formatted_node || hint->inode == NULL) {
974 		return 0;
975 	}
976 
977 	hash_in = le32_to_cpu((INODE_PKEY(hint->inode))->k_dir_id);
978 	border =
979 	    hint->beg + (u32) keyed_hash(((char *)(&hash_in)),
980 					 4) % (hint->end - hint->beg - 1);
981 	if (border > hint->search_start)
982 		hint->search_start = border;
983 
984 	return 1;
985 }
986 
old_way(reiserfs_blocknr_hint_t * hint)987 static inline int old_way(reiserfs_blocknr_hint_t * hint)
988 {
989 	b_blocknr_t border;
990 
991 	if (hint->formatted_node || hint->inode == NULL) {
992 		return 0;
993 	}
994 
995 	border =
996 	    hint->beg +
997 	    le32_to_cpu(INODE_PKEY(hint->inode)->k_dir_id) % (hint->end -
998 							      hint->beg);
999 	if (border > hint->search_start)
1000 		hint->search_start = border;
1001 
1002 	return 1;
1003 }
1004 
hundredth_slices(reiserfs_blocknr_hint_t * hint)1005 static inline void hundredth_slices(reiserfs_blocknr_hint_t * hint)
1006 {
1007 	struct in_core_key *key = &hint->key;
1008 	b_blocknr_t slice_start;
1009 
1010 	slice_start =
1011 	    (keyed_hash((char *)(&key->k_dir_id), 4) % 100) * (hint->end / 100);
1012 	if (slice_start > hint->search_start
1013 	    || slice_start + (hint->end / 100) <= hint->search_start) {
1014 		hint->search_start = slice_start;
1015 	}
1016 }
1017 
determine_search_start(reiserfs_blocknr_hint_t * hint,int amount_needed)1018 static void determine_search_start(reiserfs_blocknr_hint_t * hint,
1019 				   int amount_needed)
1020 {
1021 	struct super_block *s = hint->th->t_super;
1022 	int unfm_hint;
1023 
1024 	hint->beg = 0;
1025 	hint->end = SB_BLOCK_COUNT(s) - 1;
1026 
1027 	/* This is former border algorithm. Now with tunable border offset */
1028 	if (concentrating_formatted_nodes(s))
1029 		set_border_in_hint(s, hint);
1030 
1031 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1032 	/*
1033 	 * whenever we create a new directory, we displace it.  At first
1034 	 * we will hash for location, later we might look for a moderately
1035 	 * empty place for it
1036 	 */
1037 	if (displacing_new_packing_localities(s)
1038 	    && hint->th->displace_new_blocks) {
1039 		displace_new_packing_locality(hint);
1040 
1041 		/*
1042 		 * we do not continue determine_search_start,
1043 		 * if new packing locality is being displaced
1044 		 */
1045 		return;
1046 	}
1047 #endif
1048 
1049 	/*
1050 	 * all persons should feel encouraged to add more special cases
1051 	 * here and test them
1052 	 */
1053 
1054 	if (displacing_large_files(s) && !hint->formatted_node
1055 	    && this_blocknr_allocation_would_make_it_a_large_file(hint)) {
1056 		displace_large_file(hint);
1057 		return;
1058 	}
1059 
1060 	/*
1061 	 * if none of our special cases is relevant, use the left
1062 	 * neighbor in the tree order of the new node we are allocating for
1063 	 */
1064 	if (hint->formatted_node && TEST_OPTION(hashed_formatted_nodes, s)) {
1065 		hash_formatted_node(hint);
1066 		return;
1067 	}
1068 
1069 	unfm_hint = get_left_neighbor(hint);
1070 
1071 	/*
1072 	 * Mimic old block allocator behaviour, that is if VFS allowed for
1073 	 * preallocation, new blocks are displaced based on directory ID.
1074 	 * Also, if suggested search_start is less than last preallocated
1075 	 * block, we start searching from it, assuming that HDD dataflow
1076 	 * is faster in forward direction
1077 	 */
1078 	if (TEST_OPTION(old_way, s)) {
1079 		if (!hint->formatted_node) {
1080 			if (!reiserfs_hashed_relocation(s))
1081 				old_way(hint);
1082 			else if (!reiserfs_no_unhashed_relocation(s))
1083 				old_hashed_relocation(hint);
1084 
1085 			if (hint->inode
1086 			    && hint->search_start <
1087 			    REISERFS_I(hint->inode)->i_prealloc_block)
1088 				hint->search_start =
1089 				    REISERFS_I(hint->inode)->i_prealloc_block;
1090 		}
1091 		return;
1092 	}
1093 
1094 	/* This is an approach proposed by Hans */
1095 	if (TEST_OPTION(hundredth_slices, s)
1096 	    && !(displacing_large_files(s) && !hint->formatted_node)) {
1097 		hundredth_slices(hint);
1098 		return;
1099 	}
1100 
1101 	/* old_hashed_relocation only works on unformatted */
1102 	if (!unfm_hint && !hint->formatted_node &&
1103 	    TEST_OPTION(old_hashed_relocation, s)) {
1104 		old_hashed_relocation(hint);
1105 	}
1106 
1107 	/* new_hashed_relocation works with both formatted/unformatted nodes */
1108 	if ((!unfm_hint || hint->formatted_node) &&
1109 	    TEST_OPTION(new_hashed_relocation, s)) {
1110 		new_hashed_relocation(hint);
1111 	}
1112 
1113 	/* dirid grouping works only on unformatted nodes */
1114 	if (!unfm_hint && !hint->formatted_node && TEST_OPTION(dirid_groups, s)) {
1115 		dirid_groups(hint);
1116 	}
1117 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1118 	if (hint->formatted_node && TEST_OPTION(dirid_groups, s)) {
1119 		dirid_groups(hint);
1120 	}
1121 #endif
1122 
1123 	/* oid grouping works only on unformatted nodes */
1124 	if (!unfm_hint && !hint->formatted_node && TEST_OPTION(oid_groups, s)) {
1125 		oid_groups(hint);
1126 	}
1127 	return;
1128 }
1129 
determine_prealloc_size(reiserfs_blocknr_hint_t * hint)1130 static int determine_prealloc_size(reiserfs_blocknr_hint_t * hint)
1131 {
1132 	/* make minimum size a mount option and benchmark both ways */
1133 	/* we preallocate blocks only for regular files, specific size */
1134 	/* benchmark preallocating always and see what happens */
1135 
1136 	hint->prealloc_size = 0;
1137 
1138 	if (!hint->formatted_node && hint->preallocate) {
1139 		if (S_ISREG(hint->inode->i_mode) && !IS_PRIVATE(hint->inode)
1140 		    && hint->inode->i_size >=
1141 		    REISERFS_SB(hint->th->t_super)->s_alloc_options.
1142 		    preallocmin * hint->inode->i_sb->s_blocksize)
1143 			hint->prealloc_size =
1144 			    REISERFS_SB(hint->th->t_super)->s_alloc_options.
1145 			    preallocsize - 1;
1146 	}
1147 	return CARRY_ON;
1148 }
1149 
allocate_without_wrapping_disk(reiserfs_blocknr_hint_t * hint,b_blocknr_t * new_blocknrs,b_blocknr_t start,b_blocknr_t finish,int min,int amount_needed,int prealloc_size)1150 static inline int allocate_without_wrapping_disk(reiserfs_blocknr_hint_t * hint,
1151 						 b_blocknr_t * new_blocknrs,
1152 						 b_blocknr_t start,
1153 						 b_blocknr_t finish, int min,
1154 						 int amount_needed,
1155 						 int prealloc_size)
1156 {
1157 	int rest = amount_needed;
1158 	int nr_allocated;
1159 
1160 	while (rest > 0 && start <= finish) {
1161 		nr_allocated = scan_bitmap(hint->th, &start, finish, min,
1162 					   rest + prealloc_size,
1163 					   !hint->formatted_node, hint->block);
1164 
1165 		if (nr_allocated == 0)	/* no new blocks allocated, return */
1166 			break;
1167 
1168 		/* fill free_blocknrs array first */
1169 		while (rest > 0 && nr_allocated > 0) {
1170 			*new_blocknrs++ = start++;
1171 			rest--;
1172 			nr_allocated--;
1173 		}
1174 
1175 		/* do we have something to fill prealloc. array also ? */
1176 		if (nr_allocated > 0) {
1177 			/*
1178 			 * it means prealloc_size was greater that 0 and
1179 			 * we do preallocation
1180 			 */
1181 			list_add(&REISERFS_I(hint->inode)->i_prealloc_list,
1182 				 &SB_JOURNAL(hint->th->t_super)->
1183 				 j_prealloc_list);
1184 			REISERFS_I(hint->inode)->i_prealloc_block = start;
1185 			REISERFS_I(hint->inode)->i_prealloc_count =
1186 			    nr_allocated;
1187 			break;
1188 		}
1189 	}
1190 
1191 	return (amount_needed - rest);
1192 }
1193 
blocknrs_and_prealloc_arrays_from_search_start(reiserfs_blocknr_hint_t * hint,b_blocknr_t * new_blocknrs,int amount_needed)1194 static inline int blocknrs_and_prealloc_arrays_from_search_start
1195     (reiserfs_blocknr_hint_t * hint, b_blocknr_t * new_blocknrs,
1196      int amount_needed) {
1197 	struct super_block *s = hint->th->t_super;
1198 	b_blocknr_t start = hint->search_start;
1199 	b_blocknr_t finish = SB_BLOCK_COUNT(s) - 1;
1200 	int passno = 0;
1201 	int nr_allocated = 0;
1202 	int depth;
1203 
1204 	determine_prealloc_size(hint);
1205 	if (!hint->formatted_node) {
1206 		int quota_ret;
1207 #ifdef REISERQUOTA_DEBUG
1208 		reiserfs_debug(s, REISERFS_DEBUG_CODE,
1209 			       "reiserquota: allocating %d blocks id=%u",
1210 			       amount_needed, hint->inode->i_uid);
1211 #endif
1212 		depth = reiserfs_write_unlock_nested(s);
1213 		quota_ret =
1214 		    dquot_alloc_block_nodirty(hint->inode, amount_needed);
1215 		if (quota_ret) {	/* Quota exceeded? */
1216 			reiserfs_write_lock_nested(s, depth);
1217 			return QUOTA_EXCEEDED;
1218 		}
1219 		if (hint->preallocate && hint->prealloc_size) {
1220 #ifdef REISERQUOTA_DEBUG
1221 			reiserfs_debug(s, REISERFS_DEBUG_CODE,
1222 				       "reiserquota: allocating (prealloc) %d blocks id=%u",
1223 				       hint->prealloc_size, hint->inode->i_uid);
1224 #endif
1225 			quota_ret = dquot_prealloc_block_nodirty(hint->inode,
1226 							 hint->prealloc_size);
1227 			if (quota_ret)
1228 				hint->preallocate = hint->prealloc_size = 0;
1229 		}
1230 		/* for unformatted nodes, force large allocations */
1231 		reiserfs_write_lock_nested(s, depth);
1232 	}
1233 
1234 	do {
1235 		switch (passno++) {
1236 		case 0:	/* Search from hint->search_start to end of disk */
1237 			start = hint->search_start;
1238 			finish = SB_BLOCK_COUNT(s) - 1;
1239 			break;
1240 		case 1:	/* Search from hint->beg to hint->search_start */
1241 			start = hint->beg;
1242 			finish = hint->search_start;
1243 			break;
1244 		case 2:	/* Last chance: Search from 0 to hint->beg */
1245 			start = 0;
1246 			finish = hint->beg;
1247 			break;
1248 		default:
1249 			/* We've tried searching everywhere, not enough space */
1250 			/* Free the blocks */
1251 			if (!hint->formatted_node) {
1252 #ifdef REISERQUOTA_DEBUG
1253 				reiserfs_debug(s, REISERFS_DEBUG_CODE,
1254 					       "reiserquota: freeing (nospace) %d blocks id=%u",
1255 					       amount_needed +
1256 					       hint->prealloc_size -
1257 					       nr_allocated,
1258 					       hint->inode->i_uid);
1259 #endif
1260 				/* Free not allocated blocks */
1261 				depth = reiserfs_write_unlock_nested(s);
1262 				dquot_free_block_nodirty(hint->inode,
1263 					amount_needed + hint->prealloc_size -
1264 					nr_allocated);
1265 				reiserfs_write_lock_nested(s, depth);
1266 			}
1267 			while (nr_allocated--)
1268 				reiserfs_free_block(hint->th, hint->inode,
1269 						    new_blocknrs[nr_allocated],
1270 						    !hint->formatted_node);
1271 
1272 			return NO_DISK_SPACE;
1273 		}
1274 	} while ((nr_allocated += allocate_without_wrapping_disk(hint,
1275 								 new_blocknrs +
1276 								 nr_allocated,
1277 								 start, finish,
1278 								 1,
1279 								 amount_needed -
1280 								 nr_allocated,
1281 								 hint->
1282 								 prealloc_size))
1283 		 < amount_needed);
1284 	if (!hint->formatted_node &&
1285 	    amount_needed + hint->prealloc_size >
1286 	    nr_allocated + REISERFS_I(hint->inode)->i_prealloc_count) {
1287 		/* Some of preallocation blocks were not allocated */
1288 #ifdef REISERQUOTA_DEBUG
1289 		reiserfs_debug(s, REISERFS_DEBUG_CODE,
1290 			       "reiserquota: freeing (failed prealloc) %d blocks id=%u",
1291 			       amount_needed + hint->prealloc_size -
1292 			       nr_allocated -
1293 			       REISERFS_I(hint->inode)->i_prealloc_count,
1294 			       hint->inode->i_uid);
1295 #endif
1296 
1297 		depth = reiserfs_write_unlock_nested(s);
1298 		dquot_free_block_nodirty(hint->inode, amount_needed +
1299 					 hint->prealloc_size - nr_allocated -
1300 					 REISERFS_I(hint->inode)->
1301 					 i_prealloc_count);
1302 		reiserfs_write_lock_nested(s, depth);
1303 	}
1304 
1305 	return CARRY_ON;
1306 }
1307 
1308 /* grab new blocknrs from preallocated list */
1309 /* return amount still needed after using them */
use_preallocated_list_if_available(reiserfs_blocknr_hint_t * hint,b_blocknr_t * new_blocknrs,int amount_needed)1310 static int use_preallocated_list_if_available(reiserfs_blocknr_hint_t * hint,
1311 					      b_blocknr_t * new_blocknrs,
1312 					      int amount_needed)
1313 {
1314 	struct inode *inode = hint->inode;
1315 
1316 	if (REISERFS_I(inode)->i_prealloc_count > 0) {
1317 		while (amount_needed) {
1318 
1319 			*new_blocknrs++ = REISERFS_I(inode)->i_prealloc_block++;
1320 			REISERFS_I(inode)->i_prealloc_count--;
1321 
1322 			amount_needed--;
1323 
1324 			if (REISERFS_I(inode)->i_prealloc_count <= 0) {
1325 				list_del(&REISERFS_I(inode)->i_prealloc_list);
1326 				break;
1327 			}
1328 		}
1329 	}
1330 	/* return amount still needed after using preallocated blocks */
1331 	return amount_needed;
1332 }
1333 
reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t * hint,b_blocknr_t * new_blocknrs,int amount_needed,int reserved_by_us)1334 int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *hint,
1335 			       b_blocknr_t *new_blocknrs,
1336 			       int amount_needed,
1337 			       /* Amount of blocks we have already reserved */
1338 			       int reserved_by_us)
1339 {
1340 	int initial_amount_needed = amount_needed;
1341 	int ret;
1342 	struct super_block *s = hint->th->t_super;
1343 
1344 	/* Check if there is enough space, taking into account reserved space */
1345 	if (SB_FREE_BLOCKS(s) - REISERFS_SB(s)->reserved_blocks <
1346 	    amount_needed - reserved_by_us)
1347 		return NO_DISK_SPACE;
1348 	/* should this be if !hint->inode &&  hint->preallocate? */
1349 	/* do you mean hint->formatted_node can be removed ? - Zam */
1350 	/*
1351 	 * hint->formatted_node cannot be removed because we try to access
1352 	 * inode information here, and there is often no inode associated with
1353 	 * metadata allocations - green
1354 	 */
1355 
1356 	if (!hint->formatted_node && hint->preallocate) {
1357 		amount_needed = use_preallocated_list_if_available
1358 		    (hint, new_blocknrs, amount_needed);
1359 
1360 		/*
1361 		 * We have all the block numbers we need from the
1362 		 * prealloc list
1363 		 */
1364 		if (amount_needed == 0)
1365 			return CARRY_ON;
1366 		new_blocknrs += (initial_amount_needed - amount_needed);
1367 	}
1368 
1369 	/* find search start and save it in hint structure */
1370 	determine_search_start(hint, amount_needed);
1371 	if (hint->search_start >= SB_BLOCK_COUNT(s))
1372 		hint->search_start = SB_BLOCK_COUNT(s) - 1;
1373 
1374 	/* allocation itself; fill new_blocknrs and preallocation arrays */
1375 	ret = blocknrs_and_prealloc_arrays_from_search_start
1376 	    (hint, new_blocknrs, amount_needed);
1377 
1378 	/*
1379 	 * We used prealloc. list to fill (partially) new_blocknrs array.
1380 	 * If final allocation fails we need to return blocks back to
1381 	 * prealloc. list or just free them. -- Zam (I chose second
1382 	 * variant)
1383 	 */
1384 	if (ret != CARRY_ON) {
1385 		while (amount_needed++ < initial_amount_needed) {
1386 			reiserfs_free_block(hint->th, hint->inode,
1387 					    *(--new_blocknrs), 1);
1388 		}
1389 	}
1390 	return ret;
1391 }
1392 
reiserfs_cache_bitmap_metadata(struct super_block * sb,struct buffer_head * bh,struct reiserfs_bitmap_info * info)1393 void reiserfs_cache_bitmap_metadata(struct super_block *sb,
1394                                     struct buffer_head *bh,
1395                                     struct reiserfs_bitmap_info *info)
1396 {
1397 	unsigned long *cur = (unsigned long *)(bh->b_data + bh->b_size);
1398 
1399 	/* The first bit must ALWAYS be 1 */
1400 	if (!reiserfs_test_le_bit(0, (unsigned long *)bh->b_data))
1401 		reiserfs_error(sb, "reiserfs-2025", "bitmap block %lu is "
1402 			       "corrupted: first bit must be 1", bh->b_blocknr);
1403 
1404 	info->free_count = 0;
1405 
1406 	while (--cur >= (unsigned long *)bh->b_data) {
1407 		/* 0 and ~0 are special, we can optimize for them */
1408 		if (*cur == 0)
1409 			info->free_count += BITS_PER_LONG;
1410 		else if (*cur != ~0L)	/* A mix, investigate */
1411 			info->free_count += BITS_PER_LONG - hweight_long(*cur);
1412 	}
1413 }
1414 
reiserfs_read_bitmap_block(struct super_block * sb,unsigned int bitmap)1415 struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb,
1416                                                unsigned int bitmap)
1417 {
1418 	b_blocknr_t block = (sb->s_blocksize << 3) * bitmap;
1419 	struct reiserfs_bitmap_info *info = SB_AP_BITMAP(sb) + bitmap;
1420 	struct buffer_head *bh;
1421 
1422 	/*
1423 	 * Way old format filesystems had the bitmaps packed up front.
1424 	 * I doubt there are any of these left, but just in case...
1425 	 */
1426 	if (unlikely(test_bit(REISERFS_OLD_FORMAT,
1427 			      &REISERFS_SB(sb)->s_properties)))
1428 		block = REISERFS_SB(sb)->s_sbh->b_blocknr + 1 + bitmap;
1429 	else if (bitmap == 0)
1430 		block = (REISERFS_DISK_OFFSET_IN_BYTES >> sb->s_blocksize_bits) + 1;
1431 
1432 	bh = sb_bread(sb, block);
1433 	if (bh == NULL)
1434 		reiserfs_warning(sb, "sh-2029: %s: bitmap block (#%u) "
1435 		                 "reading failed", __func__, block);
1436 	else {
1437 		if (buffer_locked(bh)) {
1438 			int depth;
1439 			PROC_INFO_INC(sb, scan_bitmap.wait);
1440 			depth = reiserfs_write_unlock_nested(sb);
1441 			__wait_on_buffer(bh);
1442 			reiserfs_write_lock_nested(sb, depth);
1443 		}
1444 		BUG_ON(!buffer_uptodate(bh));
1445 		BUG_ON(atomic_read(&bh->b_count) == 0);
1446 
1447 		if (info->free_count == UINT_MAX)
1448 			reiserfs_cache_bitmap_metadata(sb, bh, info);
1449 	}
1450 
1451 	return bh;
1452 }
1453 
reiserfs_init_bitmap_cache(struct super_block * sb)1454 int reiserfs_init_bitmap_cache(struct super_block *sb)
1455 {
1456 	struct reiserfs_bitmap_info *bitmap;
1457 	unsigned int bmap_nr = reiserfs_bmap_count(sb);
1458 
1459 	bitmap = vmalloc(sizeof(*bitmap) * bmap_nr);
1460 	if (bitmap == NULL)
1461 		return -ENOMEM;
1462 
1463 	memset(bitmap, 0xff, sizeof(*bitmap) * bmap_nr);
1464 
1465 	SB_AP_BITMAP(sb) = bitmap;
1466 
1467 	return 0;
1468 }
1469 
reiserfs_free_bitmap_cache(struct super_block * sb)1470 void reiserfs_free_bitmap_cache(struct super_block *sb)
1471 {
1472 	if (SB_AP_BITMAP(sb)) {
1473 		vfree(SB_AP_BITMAP(sb));
1474 		SB_AP_BITMAP(sb) = NULL;
1475 	}
1476 }
1477