xref: /openbmc/linux/fs/ext2/balloc.c (revision 54525552)
1 /*
2  *  linux/fs/ext2/balloc.c
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  Enhanced block allocation by Stephen Tweedie (sct@redhat.com), 1993
10  *  Big-endian to little-endian byte-swapping/bitmaps by
11  *        David S. Miller (davem@caip.rutgers.edu), 1995
12  */
13 
14 #include "ext2.h"
15 #include <linux/quotaops.h>
16 #include <linux/slab.h>
17 #include <linux/sched.h>
18 #include <linux/buffer_head.h>
19 #include <linux/capability.h>
20 
21 /*
22  * balloc.c contains the blocks allocation and deallocation routines
23  */
24 
25 /*
26  * The free blocks are managed by bitmaps.  A file system contains several
27  * blocks groups.  Each group contains 1 bitmap block for blocks, 1 bitmap
28  * block for inodes, N blocks for the inode table and data blocks.
29  *
30  * The file system contains group descriptors which are located after the
31  * super block.  Each descriptor contains the number of the bitmap block and
32  * the free blocks count in the block.  The descriptors are loaded in memory
33  * when a file system is mounted (see ext2_fill_super).
34  */
35 
36 
37 #define in_range(b, first, len)	((b) >= (first) && (b) <= (first) + (len) - 1)
38 
39 struct ext2_group_desc * ext2_get_group_desc(struct super_block * sb,
40 					     unsigned int block_group,
41 					     struct buffer_head ** bh)
42 {
43 	unsigned long group_desc;
44 	unsigned long offset;
45 	struct ext2_group_desc * desc;
46 	struct ext2_sb_info *sbi = EXT2_SB(sb);
47 
48 	if (block_group >= sbi->s_groups_count) {
49 		ext2_error (sb, "ext2_get_group_desc",
50 			    "block_group >= groups_count - "
51 			    "block_group = %d, groups_count = %lu",
52 			    block_group, sbi->s_groups_count);
53 
54 		return NULL;
55 	}
56 
57 	group_desc = block_group >> EXT2_DESC_PER_BLOCK_BITS(sb);
58 	offset = block_group & (EXT2_DESC_PER_BLOCK(sb) - 1);
59 	if (!sbi->s_group_desc[group_desc]) {
60 		ext2_error (sb, "ext2_get_group_desc",
61 			    "Group descriptor not loaded - "
62 			    "block_group = %d, group_desc = %lu, desc = %lu",
63 			     block_group, group_desc, offset);
64 		return NULL;
65 	}
66 
67 	desc = (struct ext2_group_desc *) sbi->s_group_desc[group_desc]->b_data;
68 	if (bh)
69 		*bh = sbi->s_group_desc[group_desc];
70 	return desc + offset;
71 }
72 
73 static int ext2_valid_block_bitmap(struct super_block *sb,
74 					struct ext2_group_desc *desc,
75 					unsigned int block_group,
76 					struct buffer_head *bh)
77 {
78 	ext2_grpblk_t offset;
79 	ext2_grpblk_t next_zero_bit;
80 	ext2_fsblk_t bitmap_blk;
81 	ext2_fsblk_t group_first_block;
82 
83 	group_first_block = ext2_group_first_block_no(sb, block_group);
84 
85 	/* check whether block bitmap block number is set */
86 	bitmap_blk = le32_to_cpu(desc->bg_block_bitmap);
87 	offset = bitmap_blk - group_first_block;
88 	if (!ext2_test_bit(offset, bh->b_data))
89 		/* bad block bitmap */
90 		goto err_out;
91 
92 	/* check whether the inode bitmap block number is set */
93 	bitmap_blk = le32_to_cpu(desc->bg_inode_bitmap);
94 	offset = bitmap_blk - group_first_block;
95 	if (!ext2_test_bit(offset, bh->b_data))
96 		/* bad block bitmap */
97 		goto err_out;
98 
99 	/* check whether the inode table block number is set */
100 	bitmap_blk = le32_to_cpu(desc->bg_inode_table);
101 	offset = bitmap_blk - group_first_block;
102 	next_zero_bit = ext2_find_next_zero_bit(bh->b_data,
103 				offset + EXT2_SB(sb)->s_itb_per_group,
104 				offset);
105 	if (next_zero_bit >= offset + EXT2_SB(sb)->s_itb_per_group)
106 		/* good bitmap for inode tables */
107 		return 1;
108 
109 err_out:
110 	ext2_error(sb, __func__,
111 			"Invalid block bitmap - "
112 			"block_group = %d, block = %lu",
113 			block_group, bitmap_blk);
114 	return 0;
115 }
116 
117 /*
118  * Read the bitmap for a given block_group,and validate the
119  * bits for block/inode/inode tables are set in the bitmaps
120  *
121  * Return buffer_head on success or NULL in case of failure.
122  */
123 static struct buffer_head *
124 read_block_bitmap(struct super_block *sb, unsigned int block_group)
125 {
126 	struct ext2_group_desc * desc;
127 	struct buffer_head * bh = NULL;
128 	ext2_fsblk_t bitmap_blk;
129 
130 	desc = ext2_get_group_desc(sb, block_group, NULL);
131 	if (!desc)
132 		return NULL;
133 	bitmap_blk = le32_to_cpu(desc->bg_block_bitmap);
134 	bh = sb_getblk(sb, bitmap_blk);
135 	if (unlikely(!bh)) {
136 		ext2_error(sb, __func__,
137 			    "Cannot read block bitmap - "
138 			    "block_group = %d, block_bitmap = %u",
139 			    block_group, le32_to_cpu(desc->bg_block_bitmap));
140 		return NULL;
141 	}
142 	if (likely(bh_uptodate_or_lock(bh)))
143 		return bh;
144 
145 	if (bh_submit_read(bh) < 0) {
146 		brelse(bh);
147 		ext2_error(sb, __func__,
148 			    "Cannot read block bitmap - "
149 			    "block_group = %d, block_bitmap = %u",
150 			    block_group, le32_to_cpu(desc->bg_block_bitmap));
151 		return NULL;
152 	}
153 
154 	ext2_valid_block_bitmap(sb, desc, block_group, bh);
155 	/*
156 	 * file system mounted not to panic on error, continue with corrupt
157 	 * bitmap
158 	 */
159 	return bh;
160 }
161 
162 static void release_blocks(struct super_block *sb, int count)
163 {
164 	if (count) {
165 		struct ext2_sb_info *sbi = EXT2_SB(sb);
166 
167 		percpu_counter_add(&sbi->s_freeblocks_counter, count);
168 		sb->s_dirt = 1;
169 	}
170 }
171 
172 static void group_adjust_blocks(struct super_block *sb, int group_no,
173 	struct ext2_group_desc *desc, struct buffer_head *bh, int count)
174 {
175 	if (count) {
176 		struct ext2_sb_info *sbi = EXT2_SB(sb);
177 		unsigned free_blocks;
178 
179 		spin_lock(sb_bgl_lock(sbi, group_no));
180 		free_blocks = le16_to_cpu(desc->bg_free_blocks_count);
181 		desc->bg_free_blocks_count = cpu_to_le16(free_blocks + count);
182 		spin_unlock(sb_bgl_lock(sbi, group_no));
183 		sb->s_dirt = 1;
184 		mark_buffer_dirty(bh);
185 	}
186 }
187 
188 /*
189  * The reservation window structure operations
190  * --------------------------------------------
191  * Operations include:
192  * dump, find, add, remove, is_empty, find_next_reservable_window, etc.
193  *
194  * We use a red-black tree to represent per-filesystem reservation
195  * windows.
196  *
197  */
198 
199 /**
200  * __rsv_window_dump() -- Dump the filesystem block allocation reservation map
201  * @rb_root:		root of per-filesystem reservation rb tree
202  * @verbose:		verbose mode
203  * @fn:			function which wishes to dump the reservation map
204  *
205  * If verbose is turned on, it will print the whole block reservation
206  * windows(start, end). Otherwise, it will only print out the "bad" windows,
207  * those windows that overlap with their immediate neighbors.
208  */
209 #if 1
210 static void __rsv_window_dump(struct rb_root *root, int verbose,
211 			      const char *fn)
212 {
213 	struct rb_node *n;
214 	struct ext2_reserve_window_node *rsv, *prev;
215 	int bad;
216 
217 restart:
218 	n = rb_first(root);
219 	bad = 0;
220 	prev = NULL;
221 
222 	printk("Block Allocation Reservation Windows Map (%s):\n", fn);
223 	while (n) {
224 		rsv = rb_entry(n, struct ext2_reserve_window_node, rsv_node);
225 		if (verbose)
226 			printk("reservation window 0x%p "
227 				"start: %lu, end: %lu\n",
228 				rsv, rsv->rsv_start, rsv->rsv_end);
229 		if (rsv->rsv_start && rsv->rsv_start >= rsv->rsv_end) {
230 			printk("Bad reservation %p (start >= end)\n",
231 			       rsv);
232 			bad = 1;
233 		}
234 		if (prev && prev->rsv_end >= rsv->rsv_start) {
235 			printk("Bad reservation %p (prev->end >= start)\n",
236 			       rsv);
237 			bad = 1;
238 		}
239 		if (bad) {
240 			if (!verbose) {
241 				printk("Restarting reservation walk in verbose mode\n");
242 				verbose = 1;
243 				goto restart;
244 			}
245 		}
246 		n = rb_next(n);
247 		prev = rsv;
248 	}
249 	printk("Window map complete.\n");
250 	BUG_ON(bad);
251 }
252 #define rsv_window_dump(root, verbose) \
253 	__rsv_window_dump((root), (verbose), __func__)
254 #else
255 #define rsv_window_dump(root, verbose) do {} while (0)
256 #endif
257 
258 /**
259  * goal_in_my_reservation()
260  * @rsv:		inode's reservation window
261  * @grp_goal:		given goal block relative to the allocation block group
262  * @group:		the current allocation block group
263  * @sb:			filesystem super block
264  *
265  * Test if the given goal block (group relative) is within the file's
266  * own block reservation window range.
267  *
268  * If the reservation window is outside the goal allocation group, return 0;
269  * grp_goal (given goal block) could be -1, which means no specific
270  * goal block. In this case, always return 1.
271  * If the goal block is within the reservation window, return 1;
272  * otherwise, return 0;
273  */
274 static int
275 goal_in_my_reservation(struct ext2_reserve_window *rsv, ext2_grpblk_t grp_goal,
276 			unsigned int group, struct super_block * sb)
277 {
278 	ext2_fsblk_t group_first_block, group_last_block;
279 
280 	group_first_block = ext2_group_first_block_no(sb, group);
281 	group_last_block = group_first_block + EXT2_BLOCKS_PER_GROUP(sb) - 1;
282 
283 	if ((rsv->_rsv_start > group_last_block) ||
284 	    (rsv->_rsv_end < group_first_block))
285 		return 0;
286 	if ((grp_goal >= 0) && ((grp_goal + group_first_block < rsv->_rsv_start)
287 		|| (grp_goal + group_first_block > rsv->_rsv_end)))
288 		return 0;
289 	return 1;
290 }
291 
292 /**
293  * search_reserve_window()
294  * @rb_root:		root of reservation tree
295  * @goal:		target allocation block
296  *
297  * Find the reserved window which includes the goal, or the previous one
298  * if the goal is not in any window.
299  * Returns NULL if there are no windows or if all windows start after the goal.
300  */
301 static struct ext2_reserve_window_node *
302 search_reserve_window(struct rb_root *root, ext2_fsblk_t goal)
303 {
304 	struct rb_node *n = root->rb_node;
305 	struct ext2_reserve_window_node *rsv;
306 
307 	if (!n)
308 		return NULL;
309 
310 	do {
311 		rsv = rb_entry(n, struct ext2_reserve_window_node, rsv_node);
312 
313 		if (goal < rsv->rsv_start)
314 			n = n->rb_left;
315 		else if (goal > rsv->rsv_end)
316 			n = n->rb_right;
317 		else
318 			return rsv;
319 	} while (n);
320 	/*
321 	 * We've fallen off the end of the tree: the goal wasn't inside
322 	 * any particular node.  OK, the previous node must be to one
323 	 * side of the interval containing the goal.  If it's the RHS,
324 	 * we need to back up one.
325 	 */
326 	if (rsv->rsv_start > goal) {
327 		n = rb_prev(&rsv->rsv_node);
328 		rsv = rb_entry(n, struct ext2_reserve_window_node, rsv_node);
329 	}
330 	return rsv;
331 }
332 
333 /*
334  * ext2_rsv_window_add() -- Insert a window to the block reservation rb tree.
335  * @sb:			super block
336  * @rsv:		reservation window to add
337  *
338  * Must be called with rsv_lock held.
339  */
340 void ext2_rsv_window_add(struct super_block *sb,
341 		    struct ext2_reserve_window_node *rsv)
342 {
343 	struct rb_root *root = &EXT2_SB(sb)->s_rsv_window_root;
344 	struct rb_node *node = &rsv->rsv_node;
345 	ext2_fsblk_t start = rsv->rsv_start;
346 
347 	struct rb_node ** p = &root->rb_node;
348 	struct rb_node * parent = NULL;
349 	struct ext2_reserve_window_node *this;
350 
351 	while (*p)
352 	{
353 		parent = *p;
354 		this = rb_entry(parent, struct ext2_reserve_window_node, rsv_node);
355 
356 		if (start < this->rsv_start)
357 			p = &(*p)->rb_left;
358 		else if (start > this->rsv_end)
359 			p = &(*p)->rb_right;
360 		else {
361 			rsv_window_dump(root, 1);
362 			BUG();
363 		}
364 	}
365 
366 	rb_link_node(node, parent, p);
367 	rb_insert_color(node, root);
368 }
369 
370 /**
371  * rsv_window_remove() -- unlink a window from the reservation rb tree
372  * @sb:			super block
373  * @rsv:		reservation window to remove
374  *
375  * Mark the block reservation window as not allocated, and unlink it
376  * from the filesystem reservation window rb tree. Must be called with
377  * rsv_lock held.
378  */
379 static void rsv_window_remove(struct super_block *sb,
380 			      struct ext2_reserve_window_node *rsv)
381 {
382 	rsv->rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
383 	rsv->rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
384 	rsv->rsv_alloc_hit = 0;
385 	rb_erase(&rsv->rsv_node, &EXT2_SB(sb)->s_rsv_window_root);
386 }
387 
388 /*
389  * rsv_is_empty() -- Check if the reservation window is allocated.
390  * @rsv:		given reservation window to check
391  *
392  * returns 1 if the end block is EXT2_RESERVE_WINDOW_NOT_ALLOCATED.
393  */
394 static inline int rsv_is_empty(struct ext2_reserve_window *rsv)
395 {
396 	/* a valid reservation end block could not be 0 */
397 	return (rsv->_rsv_end == EXT2_RESERVE_WINDOW_NOT_ALLOCATED);
398 }
399 
400 /**
401  * ext2_init_block_alloc_info()
402  * @inode:		file inode structure
403  *
404  * Allocate and initialize the  reservation window structure, and
405  * link the window to the ext2 inode structure at last
406  *
407  * The reservation window structure is only dynamically allocated
408  * and linked to ext2 inode the first time the open file
409  * needs a new block. So, before every ext2_new_block(s) call, for
410  * regular files, we should check whether the reservation window
411  * structure exists or not. In the latter case, this function is called.
412  * Fail to do so will result in block reservation being turned off for that
413  * open file.
414  *
415  * This function is called from ext2_get_blocks_handle(), also called
416  * when setting the reservation window size through ioctl before the file
417  * is open for write (needs block allocation).
418  *
419  * Needs truncate_mutex protection prior to calling this function.
420  */
421 void ext2_init_block_alloc_info(struct inode *inode)
422 {
423 	struct ext2_inode_info *ei = EXT2_I(inode);
424 	struct ext2_block_alloc_info *block_i = ei->i_block_alloc_info;
425 	struct super_block *sb = inode->i_sb;
426 
427 	block_i = kmalloc(sizeof(*block_i), GFP_NOFS);
428 	if (block_i) {
429 		struct ext2_reserve_window_node *rsv = &block_i->rsv_window_node;
430 
431 		rsv->rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
432 		rsv->rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
433 
434 	 	/*
435 		 * if filesystem is mounted with NORESERVATION, the goal
436 		 * reservation window size is set to zero to indicate
437 		 * block reservation is off
438 		 */
439 		if (!test_opt(sb, RESERVATION))
440 			rsv->rsv_goal_size = 0;
441 		else
442 			rsv->rsv_goal_size = EXT2_DEFAULT_RESERVE_BLOCKS;
443 		rsv->rsv_alloc_hit = 0;
444 		block_i->last_alloc_logical_block = 0;
445 		block_i->last_alloc_physical_block = 0;
446 	}
447 	ei->i_block_alloc_info = block_i;
448 }
449 
450 /**
451  * ext2_discard_reservation()
452  * @inode:		inode
453  *
454  * Discard(free) block reservation window on last file close, or truncate
455  * or at last iput().
456  *
457  * It is being called in three cases:
458  * 	ext2_release_file(): last writer closes the file
459  * 	ext2_clear_inode(): last iput(), when nobody links to this file.
460  * 	ext2_truncate(): when the block indirect map is about to change.
461  */
462 void ext2_discard_reservation(struct inode *inode)
463 {
464 	struct ext2_inode_info *ei = EXT2_I(inode);
465 	struct ext2_block_alloc_info *block_i = ei->i_block_alloc_info;
466 	struct ext2_reserve_window_node *rsv;
467 	spinlock_t *rsv_lock = &EXT2_SB(inode->i_sb)->s_rsv_window_lock;
468 
469 	if (!block_i)
470 		return;
471 
472 	rsv = &block_i->rsv_window_node;
473 	if (!rsv_is_empty(&rsv->rsv_window)) {
474 		spin_lock(rsv_lock);
475 		if (!rsv_is_empty(&rsv->rsv_window))
476 			rsv_window_remove(inode->i_sb, rsv);
477 		spin_unlock(rsv_lock);
478 	}
479 }
480 
481 /**
482  * ext2_free_blocks_sb() -- Free given blocks and update quota and i_blocks
483  * @inode:		inode
484  * @block:		start physcial block to free
485  * @count:		number of blocks to free
486  */
487 void ext2_free_blocks (struct inode * inode, unsigned long block,
488 		       unsigned long count)
489 {
490 	struct buffer_head *bitmap_bh = NULL;
491 	struct buffer_head * bh2;
492 	unsigned long block_group;
493 	unsigned long bit;
494 	unsigned long i;
495 	unsigned long overflow;
496 	struct super_block * sb = inode->i_sb;
497 	struct ext2_sb_info * sbi = EXT2_SB(sb);
498 	struct ext2_group_desc * desc;
499 	struct ext2_super_block * es = sbi->s_es;
500 	unsigned freed = 0, group_freed;
501 
502 	if (block < le32_to_cpu(es->s_first_data_block) ||
503 	    block + count < block ||
504 	    block + count > le32_to_cpu(es->s_blocks_count)) {
505 		ext2_error (sb, "ext2_free_blocks",
506 			    "Freeing blocks not in datazone - "
507 			    "block = %lu, count = %lu", block, count);
508 		goto error_return;
509 	}
510 
511 	ext2_debug ("freeing block(s) %lu-%lu\n", block, block + count - 1);
512 
513 do_more:
514 	overflow = 0;
515 	block_group = (block - le32_to_cpu(es->s_first_data_block)) /
516 		      EXT2_BLOCKS_PER_GROUP(sb);
517 	bit = (block - le32_to_cpu(es->s_first_data_block)) %
518 		      EXT2_BLOCKS_PER_GROUP(sb);
519 	/*
520 	 * Check to see if we are freeing blocks across a group
521 	 * boundary.
522 	 */
523 	if (bit + count > EXT2_BLOCKS_PER_GROUP(sb)) {
524 		overflow = bit + count - EXT2_BLOCKS_PER_GROUP(sb);
525 		count -= overflow;
526 	}
527 	brelse(bitmap_bh);
528 	bitmap_bh = read_block_bitmap(sb, block_group);
529 	if (!bitmap_bh)
530 		goto error_return;
531 
532 	desc = ext2_get_group_desc (sb, block_group, &bh2);
533 	if (!desc)
534 		goto error_return;
535 
536 	if (in_range (le32_to_cpu(desc->bg_block_bitmap), block, count) ||
537 	    in_range (le32_to_cpu(desc->bg_inode_bitmap), block, count) ||
538 	    in_range (block, le32_to_cpu(desc->bg_inode_table),
539 		      sbi->s_itb_per_group) ||
540 	    in_range (block + count - 1, le32_to_cpu(desc->bg_inode_table),
541 		      sbi->s_itb_per_group)) {
542 		ext2_error (sb, "ext2_free_blocks",
543 			    "Freeing blocks in system zones - "
544 			    "Block = %lu, count = %lu",
545 			    block, count);
546 		goto error_return;
547 	}
548 
549 	for (i = 0, group_freed = 0; i < count; i++) {
550 		if (!ext2_clear_bit_atomic(sb_bgl_lock(sbi, block_group),
551 						bit + i, bitmap_bh->b_data)) {
552 			ext2_error(sb, __func__,
553 				"bit already cleared for block %lu", block + i);
554 		} else {
555 			group_freed++;
556 		}
557 	}
558 
559 	mark_buffer_dirty(bitmap_bh);
560 	if (sb->s_flags & MS_SYNCHRONOUS)
561 		sync_dirty_buffer(bitmap_bh);
562 
563 	group_adjust_blocks(sb, block_group, desc, bh2, group_freed);
564 	freed += group_freed;
565 
566 	if (overflow) {
567 		block += count;
568 		count = overflow;
569 		goto do_more;
570 	}
571 error_return:
572 	brelse(bitmap_bh);
573 	release_blocks(sb, freed);
574 	dquot_free_block_nodirty(inode, freed);
575 }
576 
577 /**
578  * bitmap_search_next_usable_block()
579  * @start:		the starting block (group relative) of the search
580  * @bh:			bufferhead contains the block group bitmap
581  * @maxblocks:		the ending block (group relative) of the reservation
582  *
583  * The bitmap search --- search forward through the actual bitmap on disk until
584  * we find a bit free.
585  */
586 static ext2_grpblk_t
587 bitmap_search_next_usable_block(ext2_grpblk_t start, struct buffer_head *bh,
588 					ext2_grpblk_t maxblocks)
589 {
590 	ext2_grpblk_t next;
591 
592 	next = ext2_find_next_zero_bit(bh->b_data, maxblocks, start);
593 	if (next >= maxblocks)
594 		return -1;
595 	return next;
596 }
597 
598 /**
599  * find_next_usable_block()
600  * @start:		the starting block (group relative) to find next
601  * 			allocatable block in bitmap.
602  * @bh:			bufferhead contains the block group bitmap
603  * @maxblocks:		the ending block (group relative) for the search
604  *
605  * Find an allocatable block in a bitmap.  We perform the "most
606  * appropriate allocation" algorithm of looking for a free block near
607  * the initial goal; then for a free byte somewhere in the bitmap;
608  * then for any free bit in the bitmap.
609  */
610 static ext2_grpblk_t
611 find_next_usable_block(int start, struct buffer_head *bh, int maxblocks)
612 {
613 	ext2_grpblk_t here, next;
614 	char *p, *r;
615 
616 	if (start > 0) {
617 		/*
618 		 * The goal was occupied; search forward for a free
619 		 * block within the next XX blocks.
620 		 *
621 		 * end_goal is more or less random, but it has to be
622 		 * less than EXT2_BLOCKS_PER_GROUP. Aligning up to the
623 		 * next 64-bit boundary is simple..
624 		 */
625 		ext2_grpblk_t end_goal = (start + 63) & ~63;
626 		if (end_goal > maxblocks)
627 			end_goal = maxblocks;
628 		here = ext2_find_next_zero_bit(bh->b_data, end_goal, start);
629 		if (here < end_goal)
630 			return here;
631 		ext2_debug("Bit not found near goal\n");
632 	}
633 
634 	here = start;
635 	if (here < 0)
636 		here = 0;
637 
638 	p = ((char *)bh->b_data) + (here >> 3);
639 	r = memscan(p, 0, ((maxblocks + 7) >> 3) - (here >> 3));
640 	next = (r - ((char *)bh->b_data)) << 3;
641 
642 	if (next < maxblocks && next >= here)
643 		return next;
644 
645 	here = bitmap_search_next_usable_block(here, bh, maxblocks);
646 	return here;
647 }
648 
649 /**
650  * ext2_try_to_allocate()
651  * @sb:			superblock
652  * @group:		given allocation block group
653  * @bitmap_bh:		bufferhead holds the block bitmap
654  * @grp_goal:		given target block within the group
655  * @count:		target number of blocks to allocate
656  * @my_rsv:		reservation window
657  *
658  * Attempt to allocate blocks within a give range. Set the range of allocation
659  * first, then find the first free bit(s) from the bitmap (within the range),
660  * and at last, allocate the blocks by claiming the found free bit as allocated.
661  *
662  * To set the range of this allocation:
663  * 	if there is a reservation window, only try to allocate block(s)
664  * 	from the file's own reservation window;
665  * 	Otherwise, the allocation range starts from the give goal block,
666  * 	ends at the block group's last block.
667  *
668  * If we failed to allocate the desired block then we may end up crossing to a
669  * new bitmap.
670  */
671 static int
672 ext2_try_to_allocate(struct super_block *sb, int group,
673 			struct buffer_head *bitmap_bh, ext2_grpblk_t grp_goal,
674 			unsigned long *count,
675 			struct ext2_reserve_window *my_rsv)
676 {
677 	ext2_fsblk_t group_first_block;
678        	ext2_grpblk_t start, end;
679 	unsigned long num = 0;
680 
681 	/* we do allocation within the reservation window if we have a window */
682 	if (my_rsv) {
683 		group_first_block = ext2_group_first_block_no(sb, group);
684 		if (my_rsv->_rsv_start >= group_first_block)
685 			start = my_rsv->_rsv_start - group_first_block;
686 		else
687 			/* reservation window cross group boundary */
688 			start = 0;
689 		end = my_rsv->_rsv_end - group_first_block + 1;
690 		if (end > EXT2_BLOCKS_PER_GROUP(sb))
691 			/* reservation window crosses group boundary */
692 			end = EXT2_BLOCKS_PER_GROUP(sb);
693 		if ((start <= grp_goal) && (grp_goal < end))
694 			start = grp_goal;
695 		else
696 			grp_goal = -1;
697 	} else {
698 		if (grp_goal > 0)
699 			start = grp_goal;
700 		else
701 			start = 0;
702 		end = EXT2_BLOCKS_PER_GROUP(sb);
703 	}
704 
705 	BUG_ON(start > EXT2_BLOCKS_PER_GROUP(sb));
706 
707 repeat:
708 	if (grp_goal < 0) {
709 		grp_goal = find_next_usable_block(start, bitmap_bh, end);
710 		if (grp_goal < 0)
711 			goto fail_access;
712 		if (!my_rsv) {
713 			int i;
714 
715 			for (i = 0; i < 7 && grp_goal > start &&
716 					!ext2_test_bit(grp_goal - 1,
717 					     		bitmap_bh->b_data);
718 			     		i++, grp_goal--)
719 				;
720 		}
721 	}
722 	start = grp_goal;
723 
724 	if (ext2_set_bit_atomic(sb_bgl_lock(EXT2_SB(sb), group), grp_goal,
725 			       				bitmap_bh->b_data)) {
726 		/*
727 		 * The block was allocated by another thread, or it was
728 		 * allocated and then freed by another thread
729 		 */
730 		start++;
731 		grp_goal++;
732 		if (start >= end)
733 			goto fail_access;
734 		goto repeat;
735 	}
736 	num++;
737 	grp_goal++;
738 	while (num < *count && grp_goal < end
739 		&& !ext2_set_bit_atomic(sb_bgl_lock(EXT2_SB(sb), group),
740 					grp_goal, bitmap_bh->b_data)) {
741 		num++;
742 		grp_goal++;
743 	}
744 	*count = num;
745 	return grp_goal - num;
746 fail_access:
747 	*count = num;
748 	return -1;
749 }
750 
751 /**
752  * 	find_next_reservable_window():
753  *		find a reservable space within the given range.
754  *		It does not allocate the reservation window for now:
755  *		alloc_new_reservation() will do the work later.
756  *
757  * 	@search_head: the head of the searching list;
758  *		This is not necessarily the list head of the whole filesystem
759  *
760  *		We have both head and start_block to assist the search
761  *		for the reservable space. The list starts from head,
762  *		but we will shift to the place where start_block is,
763  *		then start from there, when looking for a reservable space.
764  *
765  * 	@size: the target new reservation window size
766  *
767  * 	@group_first_block: the first block we consider to start
768  *			the real search from
769  *
770  * 	@last_block:
771  *		the maximum block number that our goal reservable space
772  *		could start from. This is normally the last block in this
773  *		group. The search will end when we found the start of next
774  *		possible reservable space is out of this boundary.
775  *		This could handle the cross boundary reservation window
776  *		request.
777  *
778  * 	basically we search from the given range, rather than the whole
779  * 	reservation double linked list, (start_block, last_block)
780  * 	to find a free region that is of my size and has not
781  * 	been reserved.
782  *
783  */
784 static int find_next_reservable_window(
785 				struct ext2_reserve_window_node *search_head,
786 				struct ext2_reserve_window_node *my_rsv,
787 				struct super_block * sb,
788 				ext2_fsblk_t start_block,
789 				ext2_fsblk_t last_block)
790 {
791 	struct rb_node *next;
792 	struct ext2_reserve_window_node *rsv, *prev;
793 	ext2_fsblk_t cur;
794 	int size = my_rsv->rsv_goal_size;
795 
796 	/* TODO: make the start of the reservation window byte-aligned */
797 	/* cur = *start_block & ~7;*/
798 	cur = start_block;
799 	rsv = search_head;
800 	if (!rsv)
801 		return -1;
802 
803 	while (1) {
804 		if (cur <= rsv->rsv_end)
805 			cur = rsv->rsv_end + 1;
806 
807 		/* TODO?
808 		 * in the case we could not find a reservable space
809 		 * that is what is expected, during the re-search, we could
810 		 * remember what's the largest reservable space we could have
811 		 * and return that one.
812 		 *
813 		 * For now it will fail if we could not find the reservable
814 		 * space with expected-size (or more)...
815 		 */
816 		if (cur > last_block)
817 			return -1;		/* fail */
818 
819 		prev = rsv;
820 		next = rb_next(&rsv->rsv_node);
821 		rsv = rb_entry(next,struct ext2_reserve_window_node,rsv_node);
822 
823 		/*
824 		 * Reached the last reservation, we can just append to the
825 		 * previous one.
826 		 */
827 		if (!next)
828 			break;
829 
830 		if (cur + size <= rsv->rsv_start) {
831 			/*
832 			 * Found a reserveable space big enough.  We could
833 			 * have a reservation across the group boundary here
834 		 	 */
835 			break;
836 		}
837 	}
838 	/*
839 	 * we come here either :
840 	 * when we reach the end of the whole list,
841 	 * and there is empty reservable space after last entry in the list.
842 	 * append it to the end of the list.
843 	 *
844 	 * or we found one reservable space in the middle of the list,
845 	 * return the reservation window that we could append to.
846 	 * succeed.
847 	 */
848 
849 	if ((prev != my_rsv) && (!rsv_is_empty(&my_rsv->rsv_window)))
850 		rsv_window_remove(sb, my_rsv);
851 
852 	/*
853 	 * Let's book the whole available window for now.  We will check the
854 	 * disk bitmap later and then, if there are free blocks then we adjust
855 	 * the window size if it's larger than requested.
856 	 * Otherwise, we will remove this node from the tree next time
857 	 * call find_next_reservable_window.
858 	 */
859 	my_rsv->rsv_start = cur;
860 	my_rsv->rsv_end = cur + size - 1;
861 	my_rsv->rsv_alloc_hit = 0;
862 
863 	if (prev != my_rsv)
864 		ext2_rsv_window_add(sb, my_rsv);
865 
866 	return 0;
867 }
868 
869 /**
870  * 	alloc_new_reservation()--allocate a new reservation window
871  *
872  *		To make a new reservation, we search part of the filesystem
873  *		reservation list (the list that inside the group). We try to
874  *		allocate a new reservation window near the allocation goal,
875  *		or the beginning of the group, if there is no goal.
876  *
877  *		We first find a reservable space after the goal, then from
878  *		there, we check the bitmap for the first free block after
879  *		it. If there is no free block until the end of group, then the
880  *		whole group is full, we failed. Otherwise, check if the free
881  *		block is inside the expected reservable space, if so, we
882  *		succeed.
883  *		If the first free block is outside the reservable space, then
884  *		start from the first free block, we search for next available
885  *		space, and go on.
886  *
887  *	on succeed, a new reservation will be found and inserted into the list
888  *	It contains at least one free block, and it does not overlap with other
889  *	reservation windows.
890  *
891  *	failed: we failed to find a reservation window in this group
892  *
893  *	@rsv: the reservation
894  *
895  *	@grp_goal: The goal (group-relative).  It is where the search for a
896  *		free reservable space should start from.
897  *		if we have a goal(goal >0 ), then start from there,
898  *		no goal(goal = -1), we start from the first block
899  *		of the group.
900  *
901  *	@sb: the super block
902  *	@group: the group we are trying to allocate in
903  *	@bitmap_bh: the block group block bitmap
904  *
905  */
906 static int alloc_new_reservation(struct ext2_reserve_window_node *my_rsv,
907 		ext2_grpblk_t grp_goal, struct super_block *sb,
908 		unsigned int group, struct buffer_head *bitmap_bh)
909 {
910 	struct ext2_reserve_window_node *search_head;
911 	ext2_fsblk_t group_first_block, group_end_block, start_block;
912 	ext2_grpblk_t first_free_block;
913 	struct rb_root *fs_rsv_root = &EXT2_SB(sb)->s_rsv_window_root;
914 	unsigned long size;
915 	int ret;
916 	spinlock_t *rsv_lock = &EXT2_SB(sb)->s_rsv_window_lock;
917 
918 	group_first_block = ext2_group_first_block_no(sb, group);
919 	group_end_block = group_first_block + (EXT2_BLOCKS_PER_GROUP(sb) - 1);
920 
921 	if (grp_goal < 0)
922 		start_block = group_first_block;
923 	else
924 		start_block = grp_goal + group_first_block;
925 
926 	size = my_rsv->rsv_goal_size;
927 
928 	if (!rsv_is_empty(&my_rsv->rsv_window)) {
929 		/*
930 		 * if the old reservation is cross group boundary
931 		 * and if the goal is inside the old reservation window,
932 		 * we will come here when we just failed to allocate from
933 		 * the first part of the window. We still have another part
934 		 * that belongs to the next group. In this case, there is no
935 		 * point to discard our window and try to allocate a new one
936 		 * in this group(which will fail). we should
937 		 * keep the reservation window, just simply move on.
938 		 *
939 		 * Maybe we could shift the start block of the reservation
940 		 * window to the first block of next group.
941 		 */
942 
943 		if ((my_rsv->rsv_start <= group_end_block) &&
944 				(my_rsv->rsv_end > group_end_block) &&
945 				(start_block >= my_rsv->rsv_start))
946 			return -1;
947 
948 		if ((my_rsv->rsv_alloc_hit >
949 		     (my_rsv->rsv_end - my_rsv->rsv_start + 1) / 2)) {
950 			/*
951 			 * if the previously allocation hit ratio is
952 			 * greater than 1/2, then we double the size of
953 			 * the reservation window the next time,
954 			 * otherwise we keep the same size window
955 			 */
956 			size = size * 2;
957 			if (size > EXT2_MAX_RESERVE_BLOCKS)
958 				size = EXT2_MAX_RESERVE_BLOCKS;
959 			my_rsv->rsv_goal_size= size;
960 		}
961 	}
962 
963 	spin_lock(rsv_lock);
964 	/*
965 	 * shift the search start to the window near the goal block
966 	 */
967 	search_head = search_reserve_window(fs_rsv_root, start_block);
968 
969 	/*
970 	 * find_next_reservable_window() simply finds a reservable window
971 	 * inside the given range(start_block, group_end_block).
972 	 *
973 	 * To make sure the reservation window has a free bit inside it, we
974 	 * need to check the bitmap after we found a reservable window.
975 	 */
976 retry:
977 	ret = find_next_reservable_window(search_head, my_rsv, sb,
978 						start_block, group_end_block);
979 
980 	if (ret == -1) {
981 		if (!rsv_is_empty(&my_rsv->rsv_window))
982 			rsv_window_remove(sb, my_rsv);
983 		spin_unlock(rsv_lock);
984 		return -1;
985 	}
986 
987 	/*
988 	 * On success, find_next_reservable_window() returns the
989 	 * reservation window where there is a reservable space after it.
990 	 * Before we reserve this reservable space, we need
991 	 * to make sure there is at least a free block inside this region.
992 	 *
993 	 * Search the first free bit on the block bitmap.  Search starts from
994 	 * the start block of the reservable space we just found.
995 	 */
996 	spin_unlock(rsv_lock);
997 	first_free_block = bitmap_search_next_usable_block(
998 			my_rsv->rsv_start - group_first_block,
999 			bitmap_bh, group_end_block - group_first_block + 1);
1000 
1001 	if (first_free_block < 0) {
1002 		/*
1003 		 * no free block left on the bitmap, no point
1004 		 * to reserve the space. return failed.
1005 		 */
1006 		spin_lock(rsv_lock);
1007 		if (!rsv_is_empty(&my_rsv->rsv_window))
1008 			rsv_window_remove(sb, my_rsv);
1009 		spin_unlock(rsv_lock);
1010 		return -1;		/* failed */
1011 	}
1012 
1013 	start_block = first_free_block + group_first_block;
1014 	/*
1015 	 * check if the first free block is within the
1016 	 * free space we just reserved
1017 	 */
1018 	if (start_block >= my_rsv->rsv_start && start_block <= my_rsv->rsv_end)
1019 		return 0;		/* success */
1020 	/*
1021 	 * if the first free bit we found is out of the reservable space
1022 	 * continue search for next reservable space,
1023 	 * start from where the free block is,
1024 	 * we also shift the list head to where we stopped last time
1025 	 */
1026 	search_head = my_rsv;
1027 	spin_lock(rsv_lock);
1028 	goto retry;
1029 }
1030 
1031 /**
1032  * try_to_extend_reservation()
1033  * @my_rsv:		given reservation window
1034  * @sb:			super block
1035  * @size:		the delta to extend
1036  *
1037  * Attempt to expand the reservation window large enough to have
1038  * required number of free blocks
1039  *
1040  * Since ext2_try_to_allocate() will always allocate blocks within
1041  * the reservation window range, if the window size is too small,
1042  * multiple blocks allocation has to stop at the end of the reservation
1043  * window. To make this more efficient, given the total number of
1044  * blocks needed and the current size of the window, we try to
1045  * expand the reservation window size if necessary on a best-effort
1046  * basis before ext2_new_blocks() tries to allocate blocks.
1047  */
1048 static void try_to_extend_reservation(struct ext2_reserve_window_node *my_rsv,
1049 			struct super_block *sb, int size)
1050 {
1051 	struct ext2_reserve_window_node *next_rsv;
1052 	struct rb_node *next;
1053 	spinlock_t *rsv_lock = &EXT2_SB(sb)->s_rsv_window_lock;
1054 
1055 	if (!spin_trylock(rsv_lock))
1056 		return;
1057 
1058 	next = rb_next(&my_rsv->rsv_node);
1059 
1060 	if (!next)
1061 		my_rsv->rsv_end += size;
1062 	else {
1063 		next_rsv = rb_entry(next, struct ext2_reserve_window_node, rsv_node);
1064 
1065 		if ((next_rsv->rsv_start - my_rsv->rsv_end - 1) >= size)
1066 			my_rsv->rsv_end += size;
1067 		else
1068 			my_rsv->rsv_end = next_rsv->rsv_start - 1;
1069 	}
1070 	spin_unlock(rsv_lock);
1071 }
1072 
1073 /**
1074  * ext2_try_to_allocate_with_rsv()
1075  * @sb:			superblock
1076  * @group:		given allocation block group
1077  * @bitmap_bh:		bufferhead holds the block bitmap
1078  * @grp_goal:		given target block within the group
1079  * @count:		target number of blocks to allocate
1080  * @my_rsv:		reservation window
1081  *
1082  * This is the main function used to allocate a new block and its reservation
1083  * window.
1084  *
1085  * Each time when a new block allocation is need, first try to allocate from
1086  * its own reservation.  If it does not have a reservation window, instead of
1087  * looking for a free bit on bitmap first, then look up the reservation list to
1088  * see if it is inside somebody else's reservation window, we try to allocate a
1089  * reservation window for it starting from the goal first. Then do the block
1090  * allocation within the reservation window.
1091  *
1092  * This will avoid keeping on searching the reservation list again and
1093  * again when somebody is looking for a free block (without
1094  * reservation), and there are lots of free blocks, but they are all
1095  * being reserved.
1096  *
1097  * We use a red-black tree for the per-filesystem reservation list.
1098  */
1099 static ext2_grpblk_t
1100 ext2_try_to_allocate_with_rsv(struct super_block *sb, unsigned int group,
1101 			struct buffer_head *bitmap_bh, ext2_grpblk_t grp_goal,
1102 			struct ext2_reserve_window_node * my_rsv,
1103 			unsigned long *count)
1104 {
1105 	ext2_fsblk_t group_first_block, group_last_block;
1106 	ext2_grpblk_t ret = 0;
1107 	unsigned long num = *count;
1108 
1109 	/*
1110 	 * we don't deal with reservation when
1111 	 * filesystem is mounted without reservation
1112 	 * or the file is not a regular file
1113 	 * or last attempt to allocate a block with reservation turned on failed
1114 	 */
1115 	if (my_rsv == NULL) {
1116 		return ext2_try_to_allocate(sb, group, bitmap_bh,
1117 						grp_goal, count, NULL);
1118 	}
1119 	/*
1120 	 * grp_goal is a group relative block number (if there is a goal)
1121 	 * 0 <= grp_goal < EXT2_BLOCKS_PER_GROUP(sb)
1122 	 * first block is a filesystem wide block number
1123 	 * first block is the block number of the first block in this group
1124 	 */
1125 	group_first_block = ext2_group_first_block_no(sb, group);
1126 	group_last_block = group_first_block + (EXT2_BLOCKS_PER_GROUP(sb) - 1);
1127 
1128 	/*
1129 	 * Basically we will allocate a new block from inode's reservation
1130 	 * window.
1131 	 *
1132 	 * We need to allocate a new reservation window, if:
1133 	 * a) inode does not have a reservation window; or
1134 	 * b) last attempt to allocate a block from existing reservation
1135 	 *    failed; or
1136 	 * c) we come here with a goal and with a reservation window
1137 	 *
1138 	 * We do not need to allocate a new reservation window if we come here
1139 	 * at the beginning with a goal and the goal is inside the window, or
1140 	 * we don't have a goal but already have a reservation window.
1141 	 * then we could go to allocate from the reservation window directly.
1142 	 */
1143 	while (1) {
1144 		if (rsv_is_empty(&my_rsv->rsv_window) || (ret < 0) ||
1145 			!goal_in_my_reservation(&my_rsv->rsv_window,
1146 						grp_goal, group, sb)) {
1147 			if (my_rsv->rsv_goal_size < *count)
1148 				my_rsv->rsv_goal_size = *count;
1149 			ret = alloc_new_reservation(my_rsv, grp_goal, sb,
1150 							group, bitmap_bh);
1151 			if (ret < 0)
1152 				break;			/* failed */
1153 
1154 			if (!goal_in_my_reservation(&my_rsv->rsv_window,
1155 							grp_goal, group, sb))
1156 				grp_goal = -1;
1157 		} else if (grp_goal >= 0) {
1158 			int curr = my_rsv->rsv_end -
1159 					(grp_goal + group_first_block) + 1;
1160 
1161 			if (curr < *count)
1162 				try_to_extend_reservation(my_rsv, sb,
1163 							*count - curr);
1164 		}
1165 
1166 		if ((my_rsv->rsv_start > group_last_block) ||
1167 				(my_rsv->rsv_end < group_first_block)) {
1168 			rsv_window_dump(&EXT2_SB(sb)->s_rsv_window_root, 1);
1169 			BUG();
1170 		}
1171 		ret = ext2_try_to_allocate(sb, group, bitmap_bh, grp_goal,
1172 					   &num, &my_rsv->rsv_window);
1173 		if (ret >= 0) {
1174 			my_rsv->rsv_alloc_hit += num;
1175 			*count = num;
1176 			break;				/* succeed */
1177 		}
1178 		num = *count;
1179 	}
1180 	return ret;
1181 }
1182 
1183 /**
1184  * ext2_has_free_blocks()
1185  * @sbi:		in-core super block structure.
1186  *
1187  * Check if filesystem has at least 1 free block available for allocation.
1188  */
1189 static int ext2_has_free_blocks(struct ext2_sb_info *sbi)
1190 {
1191 	ext2_fsblk_t free_blocks, root_blocks;
1192 
1193 	free_blocks = percpu_counter_read_positive(&sbi->s_freeblocks_counter);
1194 	root_blocks = le32_to_cpu(sbi->s_es->s_r_blocks_count);
1195 	if (free_blocks < root_blocks + 1 && !capable(CAP_SYS_RESOURCE) &&
1196 		sbi->s_resuid != current_fsuid() &&
1197 		(sbi->s_resgid == 0 || !in_group_p (sbi->s_resgid))) {
1198 		return 0;
1199 	}
1200 	return 1;
1201 }
1202 
1203 /*
1204  * ext2_new_blocks() -- core block(s) allocation function
1205  * @inode:		file inode
1206  * @goal:		given target block(filesystem wide)
1207  * @count:		target number of blocks to allocate
1208  * @errp:		error code
1209  *
1210  * ext2_new_blocks uses a goal block to assist allocation.  If the goal is
1211  * free, or there is a free block within 32 blocks of the goal, that block
1212  * is allocated.  Otherwise a forward search is made for a free block; within
1213  * each block group the search first looks for an entire free byte in the block
1214  * bitmap, and then for any free bit if that fails.
1215  * This function also updates quota and i_blocks field.
1216  */
1217 ext2_fsblk_t ext2_new_blocks(struct inode *inode, ext2_fsblk_t goal,
1218 		    unsigned long *count, int *errp)
1219 {
1220 	struct buffer_head *bitmap_bh = NULL;
1221 	struct buffer_head *gdp_bh;
1222 	int group_no;
1223 	int goal_group;
1224 	ext2_grpblk_t grp_target_blk;	/* blockgroup relative goal block */
1225 	ext2_grpblk_t grp_alloc_blk;	/* blockgroup-relative allocated block*/
1226 	ext2_fsblk_t ret_block;		/* filesyetem-wide allocated block */
1227 	int bgi;			/* blockgroup iteration index */
1228 	int performed_allocation = 0;
1229 	ext2_grpblk_t free_blocks;	/* number of free blocks in a group */
1230 	struct super_block *sb;
1231 	struct ext2_group_desc *gdp;
1232 	struct ext2_super_block *es;
1233 	struct ext2_sb_info *sbi;
1234 	struct ext2_reserve_window_node *my_rsv = NULL;
1235 	struct ext2_block_alloc_info *block_i;
1236 	unsigned short windowsz = 0;
1237 	unsigned long ngroups;
1238 	unsigned long num = *count;
1239 	int ret;
1240 
1241 	*errp = -ENOSPC;
1242 	sb = inode->i_sb;
1243 	if (!sb) {
1244 		printk("ext2_new_blocks: nonexistent device");
1245 		return 0;
1246 	}
1247 
1248 	/*
1249 	 * Check quota for allocation of this block.
1250 	 */
1251 	ret = dquot_alloc_block(inode, num);
1252 	if (ret) {
1253 		*errp = ret;
1254 		return 0;
1255 	}
1256 
1257 	sbi = EXT2_SB(sb);
1258 	es = EXT2_SB(sb)->s_es;
1259 	ext2_debug("goal=%lu.\n", goal);
1260 	/*
1261 	 * Allocate a block from reservation only when
1262 	 * filesystem is mounted with reservation(default,-o reservation), and
1263 	 * it's a regular file, and
1264 	 * the desired window size is greater than 0 (One could use ioctl
1265 	 * command EXT2_IOC_SETRSVSZ to set the window size to 0 to turn off
1266 	 * reservation on that particular file)
1267 	 */
1268 	block_i = EXT2_I(inode)->i_block_alloc_info;
1269 	if (block_i) {
1270 		windowsz = block_i->rsv_window_node.rsv_goal_size;
1271 		if (windowsz > 0)
1272 			my_rsv = &block_i->rsv_window_node;
1273 	}
1274 
1275 	if (!ext2_has_free_blocks(sbi)) {
1276 		*errp = -ENOSPC;
1277 		goto out;
1278 	}
1279 
1280 	/*
1281 	 * First, test whether the goal block is free.
1282 	 */
1283 	if (goal < le32_to_cpu(es->s_first_data_block) ||
1284 	    goal >= le32_to_cpu(es->s_blocks_count))
1285 		goal = le32_to_cpu(es->s_first_data_block);
1286 	group_no = (goal - le32_to_cpu(es->s_first_data_block)) /
1287 			EXT2_BLOCKS_PER_GROUP(sb);
1288 	goal_group = group_no;
1289 retry_alloc:
1290 	gdp = ext2_get_group_desc(sb, group_no, &gdp_bh);
1291 	if (!gdp)
1292 		goto io_error;
1293 
1294 	free_blocks = le16_to_cpu(gdp->bg_free_blocks_count);
1295 	/*
1296 	 * if there is not enough free blocks to make a new resevation
1297 	 * turn off reservation for this allocation
1298 	 */
1299 	if (my_rsv && (free_blocks < windowsz)
1300 		&& (free_blocks > 0)
1301 		&& (rsv_is_empty(&my_rsv->rsv_window)))
1302 		my_rsv = NULL;
1303 
1304 	if (free_blocks > 0) {
1305 		grp_target_blk = ((goal - le32_to_cpu(es->s_first_data_block)) %
1306 				EXT2_BLOCKS_PER_GROUP(sb));
1307 		bitmap_bh = read_block_bitmap(sb, group_no);
1308 		if (!bitmap_bh)
1309 			goto io_error;
1310 		grp_alloc_blk = ext2_try_to_allocate_with_rsv(sb, group_no,
1311 					bitmap_bh, grp_target_blk,
1312 					my_rsv, &num);
1313 		if (grp_alloc_blk >= 0)
1314 			goto allocated;
1315 	}
1316 
1317 	ngroups = EXT2_SB(sb)->s_groups_count;
1318 	smp_rmb();
1319 
1320 	/*
1321 	 * Now search the rest of the groups.  We assume that
1322 	 * group_no and gdp correctly point to the last group visited.
1323 	 */
1324 	for (bgi = 0; bgi < ngroups; bgi++) {
1325 		group_no++;
1326 		if (group_no >= ngroups)
1327 			group_no = 0;
1328 		gdp = ext2_get_group_desc(sb, group_no, &gdp_bh);
1329 		if (!gdp)
1330 			goto io_error;
1331 
1332 		free_blocks = le16_to_cpu(gdp->bg_free_blocks_count);
1333 		/*
1334 		 * skip this group (and avoid loading bitmap) if there
1335 		 * are no free blocks
1336 		 */
1337 		if (!free_blocks)
1338 			continue;
1339 		/*
1340 		 * skip this group if the number of
1341 		 * free blocks is less than half of the reservation
1342 		 * window size.
1343 		 */
1344 		if (my_rsv && (free_blocks <= (windowsz/2)))
1345 			continue;
1346 
1347 		brelse(bitmap_bh);
1348 		bitmap_bh = read_block_bitmap(sb, group_no);
1349 		if (!bitmap_bh)
1350 			goto io_error;
1351 		/*
1352 		 * try to allocate block(s) from this group, without a goal(-1).
1353 		 */
1354 		grp_alloc_blk = ext2_try_to_allocate_with_rsv(sb, group_no,
1355 					bitmap_bh, -1, my_rsv, &num);
1356 		if (grp_alloc_blk >= 0)
1357 			goto allocated;
1358 	}
1359 	/*
1360 	 * We may end up a bogus earlier ENOSPC error due to
1361 	 * filesystem is "full" of reservations, but
1362 	 * there maybe indeed free blocks available on disk
1363 	 * In this case, we just forget about the reservations
1364 	 * just do block allocation as without reservations.
1365 	 */
1366 	if (my_rsv) {
1367 		my_rsv = NULL;
1368 		windowsz = 0;
1369 		group_no = goal_group;
1370 		goto retry_alloc;
1371 	}
1372 	/* No space left on the device */
1373 	*errp = -ENOSPC;
1374 	goto out;
1375 
1376 allocated:
1377 
1378 	ext2_debug("using block group %d(%d)\n",
1379 			group_no, gdp->bg_free_blocks_count);
1380 
1381 	ret_block = grp_alloc_blk + ext2_group_first_block_no(sb, group_no);
1382 
1383 	if (in_range(le32_to_cpu(gdp->bg_block_bitmap), ret_block, num) ||
1384 	    in_range(le32_to_cpu(gdp->bg_inode_bitmap), ret_block, num) ||
1385 	    in_range(ret_block, le32_to_cpu(gdp->bg_inode_table),
1386 		      EXT2_SB(sb)->s_itb_per_group) ||
1387 	    in_range(ret_block + num - 1, le32_to_cpu(gdp->bg_inode_table),
1388 		      EXT2_SB(sb)->s_itb_per_group)) {
1389 		ext2_error(sb, "ext2_new_blocks",
1390 			    "Allocating block in system zone - "
1391 			    "blocks from "E2FSBLK", length %lu",
1392 			    ret_block, num);
1393 		/*
1394 		 * ext2_try_to_allocate marked the blocks we allocated as in
1395 		 * use.  So we may want to selectively mark some of the blocks
1396 		 * as free
1397 		 */
1398 		goto retry_alloc;
1399 	}
1400 
1401 	performed_allocation = 1;
1402 
1403 	if (ret_block + num - 1 >= le32_to_cpu(es->s_blocks_count)) {
1404 		ext2_error(sb, "ext2_new_blocks",
1405 			    "block("E2FSBLK") >= blocks count(%d) - "
1406 			    "block_group = %d, es == %p ", ret_block,
1407 			le32_to_cpu(es->s_blocks_count), group_no, es);
1408 		goto out;
1409 	}
1410 
1411 	group_adjust_blocks(sb, group_no, gdp, gdp_bh, -num);
1412 	percpu_counter_sub(&sbi->s_freeblocks_counter, num);
1413 
1414 	mark_buffer_dirty(bitmap_bh);
1415 	if (sb->s_flags & MS_SYNCHRONOUS)
1416 		sync_dirty_buffer(bitmap_bh);
1417 
1418 	*errp = 0;
1419 	brelse(bitmap_bh);
1420 	dquot_free_block_nodirty(inode, *count-num);
1421 	mark_inode_dirty(inode);
1422 	*count = num;
1423 	return ret_block;
1424 
1425 io_error:
1426 	*errp = -EIO;
1427 out:
1428 	/*
1429 	 * Undo the block allocation
1430 	 */
1431 	if (!performed_allocation) {
1432 		dquot_free_block_nodirty(inode, *count);
1433 		mark_inode_dirty(inode);
1434 	}
1435 	brelse(bitmap_bh);
1436 	return 0;
1437 }
1438 
1439 ext2_fsblk_t ext2_new_block(struct inode *inode, unsigned long goal, int *errp)
1440 {
1441 	unsigned long count = 1;
1442 
1443 	return ext2_new_blocks(inode, goal, &count, errp);
1444 }
1445 
1446 #ifdef EXT2FS_DEBUG
1447 
1448 static const int nibblemap[] = {4, 3, 3, 2, 3, 2, 2, 1, 3, 2, 2, 1, 2, 1, 1, 0};
1449 
1450 unsigned long ext2_count_free (struct buffer_head * map, unsigned int numchars)
1451 {
1452 	unsigned int i;
1453 	unsigned long sum = 0;
1454 
1455 	if (!map)
1456 		return (0);
1457 	for (i = 0; i < numchars; i++)
1458 		sum += nibblemap[map->b_data[i] & 0xf] +
1459 			nibblemap[(map->b_data[i] >> 4) & 0xf];
1460 	return (sum);
1461 }
1462 
1463 #endif  /*  EXT2FS_DEBUG  */
1464 
1465 unsigned long ext2_count_free_blocks (struct super_block * sb)
1466 {
1467 	struct ext2_group_desc * desc;
1468 	unsigned long desc_count = 0;
1469 	int i;
1470 #ifdef EXT2FS_DEBUG
1471 	unsigned long bitmap_count, x;
1472 	struct ext2_super_block *es;
1473 
1474 	es = EXT2_SB(sb)->s_es;
1475 	desc_count = 0;
1476 	bitmap_count = 0;
1477 	desc = NULL;
1478 	for (i = 0; i < EXT2_SB(sb)->s_groups_count; i++) {
1479 		struct buffer_head *bitmap_bh;
1480 		desc = ext2_get_group_desc (sb, i, NULL);
1481 		if (!desc)
1482 			continue;
1483 		desc_count += le16_to_cpu(desc->bg_free_blocks_count);
1484 		bitmap_bh = read_block_bitmap(sb, i);
1485 		if (!bitmap_bh)
1486 			continue;
1487 
1488 		x = ext2_count_free(bitmap_bh, sb->s_blocksize);
1489 		printk ("group %d: stored = %d, counted = %lu\n",
1490 			i, le16_to_cpu(desc->bg_free_blocks_count), x);
1491 		bitmap_count += x;
1492 		brelse(bitmap_bh);
1493 	}
1494 	printk("ext2_count_free_blocks: stored = %lu, computed = %lu, %lu\n",
1495 		(long)le32_to_cpu(es->s_free_blocks_count),
1496 		desc_count, bitmap_count);
1497 	return bitmap_count;
1498 #else
1499         for (i = 0; i < EXT2_SB(sb)->s_groups_count; i++) {
1500                 desc = ext2_get_group_desc (sb, i, NULL);
1501                 if (!desc)
1502                         continue;
1503                 desc_count += le16_to_cpu(desc->bg_free_blocks_count);
1504 	}
1505 	return desc_count;
1506 #endif
1507 }
1508 
1509 static inline int test_root(int a, int b)
1510 {
1511 	int num = b;
1512 
1513 	while (a > num)
1514 		num *= b;
1515 	return num == a;
1516 }
1517 
1518 static int ext2_group_sparse(int group)
1519 {
1520 	if (group <= 1)
1521 		return 1;
1522 	return (test_root(group, 3) || test_root(group, 5) ||
1523 		test_root(group, 7));
1524 }
1525 
1526 /**
1527  *	ext2_bg_has_super - number of blocks used by the superblock in group
1528  *	@sb: superblock for filesystem
1529  *	@group: group number to check
1530  *
1531  *	Return the number of blocks used by the superblock (primary or backup)
1532  *	in this group.  Currently this will be only 0 or 1.
1533  */
1534 int ext2_bg_has_super(struct super_block *sb, int group)
1535 {
1536 	if (EXT2_HAS_RO_COMPAT_FEATURE(sb,EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER)&&
1537 	    !ext2_group_sparse(group))
1538 		return 0;
1539 	return 1;
1540 }
1541 
1542 /**
1543  *	ext2_bg_num_gdb - number of blocks used by the group table in group
1544  *	@sb: superblock for filesystem
1545  *	@group: group number to check
1546  *
1547  *	Return the number of blocks used by the group descriptor table
1548  *	(primary or backup) in this group.  In the future there may be a
1549  *	different number of descriptor blocks in each group.
1550  */
1551 unsigned long ext2_bg_num_gdb(struct super_block *sb, int group)
1552 {
1553 	return ext2_bg_has_super(sb, group) ? EXT2_SB(sb)->s_gdb_count : 0;
1554 }
1555 
1556