blob: 92fd0338a6ebe3df16739b65feec4e4ef034206f [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/ext3/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
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/time.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080015#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/fs.h>
17#include <linux/jbd.h>
18#include <linux/ext3_fs.h>
19#include <linux/ext3_jbd.h>
20#include <linux/quotaops.h>
21#include <linux/buffer_head.h>
22
23/*
24 * balloc.c contains the blocks allocation and deallocation routines
25 */
26
27/*
28 * The free blocks are managed by bitmaps. A file system contains several
29 * blocks groups. Each group contains 1 bitmap block for blocks, 1 bitmap
30 * block for inodes, N blocks for the inode table and data blocks.
31 *
32 * The file system contains group descriptors which are located after the
33 * super block. Each descriptor contains the number of the bitmap block and
34 * the free blocks count in the block. The descriptors are loaded in memory
Aneesh Kumar K.Ve6274322007-02-20 13:57:58 -080035 * when a file system is mounted (see ext3_fill_super).
Linus Torvalds1da177e2005-04-16 15:20:36 -070036 */
37
38
39#define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
40
Mingming Cao36faadc142006-09-27 01:49:32 -070041/**
42 * ext3_get_group_desc() -- load group descriptor from disk
Dave Kleikampe9ad5622006-09-27 01:49:35 -070043 * @sb: super block
Mingming Cao36faadc142006-09-27 01:49:32 -070044 * @block_group: given block group
45 * @bh: pointer to the buffer head to store the block
46 * group descriptor
47 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070048struct ext3_group_desc * ext3_get_group_desc(struct super_block * sb,
49 unsigned int block_group,
50 struct buffer_head ** bh)
51{
52 unsigned long group_desc;
53 unsigned long offset;
54 struct ext3_group_desc * desc;
55 struct ext3_sb_info *sbi = EXT3_SB(sb);
56
57 if (block_group >= sbi->s_groups_count) {
58 ext3_error (sb, "ext3_get_group_desc",
59 "block_group >= groups_count - "
60 "block_group = %d, groups_count = %lu",
61 block_group, sbi->s_groups_count);
62
63 return NULL;
64 }
65 smp_rmb();
66
67 group_desc = block_group >> EXT3_DESC_PER_BLOCK_BITS(sb);
68 offset = block_group & (EXT3_DESC_PER_BLOCK(sb) - 1);
69 if (!sbi->s_group_desc[group_desc]) {
70 ext3_error (sb, "ext3_get_group_desc",
71 "Group descriptor not loaded - "
72 "block_group = %d, group_desc = %lu, desc = %lu",
73 block_group, group_desc, offset);
74 return NULL;
75 }
76
77 desc = (struct ext3_group_desc *) sbi->s_group_desc[group_desc]->b_data;
78 if (bh)
79 *bh = sbi->s_group_desc[group_desc];
80 return desc + offset;
81}
82
Aneesh Kumar K.Vf762e902008-02-06 01:40:09 -080083static int ext3_valid_block_bitmap(struct super_block *sb,
84 struct ext3_group_desc *desc,
85 unsigned int block_group,
86 struct buffer_head *bh)
87{
88 ext3_grpblk_t offset;
89 ext3_grpblk_t next_zero_bit;
90 ext3_fsblk_t bitmap_blk;
91 ext3_fsblk_t group_first_block;
92
93 group_first_block = ext3_group_first_block_no(sb, block_group);
94
95 /* check whether block bitmap block number is set */
96 bitmap_blk = le32_to_cpu(desc->bg_block_bitmap);
97 offset = bitmap_blk - group_first_block;
98 if (!ext3_test_bit(offset, bh->b_data))
99 /* bad block bitmap */
100 goto err_out;
101
102 /* check whether the inode bitmap block number is set */
103 bitmap_blk = le32_to_cpu(desc->bg_inode_bitmap);
104 offset = bitmap_blk - group_first_block;
105 if (!ext3_test_bit(offset, bh->b_data))
106 /* bad block bitmap */
107 goto err_out;
108
109 /* check whether the inode table block number is set */
110 bitmap_blk = le32_to_cpu(desc->bg_inode_table);
111 offset = bitmap_blk - group_first_block;
112 next_zero_bit = ext3_find_next_zero_bit(bh->b_data,
113 offset + EXT3_SB(sb)->s_itb_per_group,
114 offset);
115 if (next_zero_bit >= offset + EXT3_SB(sb)->s_itb_per_group)
116 /* good bitmap for inode tables */
117 return 1;
118
119err_out:
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700120 ext3_error(sb, __func__,
Aneesh Kumar K.Vf762e902008-02-06 01:40:09 -0800121 "Invalid block bitmap - "
122 "block_group = %d, block = %lu",
123 block_group, bitmap_blk);
124 return 0;
125}
126
Mingming Cao36faadc142006-09-27 01:49:32 -0700127/**
128 * read_block_bitmap()
129 * @sb: super block
130 * @block_group: given block group
131 *
Aneesh Kumar K.Vf762e902008-02-06 01:40:09 -0800132 * Read the bitmap for a given block_group,and validate the
133 * bits for block/inode/inode tables are set in the bitmaps
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 *
135 * Return buffer_head on success or NULL in case of failure.
136 */
137static struct buffer_head *
138read_block_bitmap(struct super_block *sb, unsigned int block_group)
139{
140 struct ext3_group_desc * desc;
141 struct buffer_head * bh = NULL;
Aneesh Kumar K.Vf762e902008-02-06 01:40:09 -0800142 ext3_fsblk_t bitmap_blk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143
Aneesh Kumar K.Vf762e902008-02-06 01:40:09 -0800144 desc = ext3_get_group_desc(sb, block_group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 if (!desc)
Aneesh Kumar K.Vf762e902008-02-06 01:40:09 -0800146 return NULL;
147 bitmap_blk = le32_to_cpu(desc->bg_block_bitmap);
148 bh = sb_getblk(sb, bitmap_blk);
149 if (unlikely(!bh)) {
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700150 ext3_error(sb, __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151 "Cannot read block bitmap - "
152 "block_group = %d, block_bitmap = %u",
153 block_group, le32_to_cpu(desc->bg_block_bitmap));
Aneesh Kumar K.Vf762e902008-02-06 01:40:09 -0800154 return NULL;
155 }
156 if (likely(bh_uptodate_or_lock(bh)))
157 return bh;
158
159 if (bh_submit_read(bh) < 0) {
160 brelse(bh);
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700161 ext3_error(sb, __func__,
Aneesh Kumar K.Vf762e902008-02-06 01:40:09 -0800162 "Cannot read block bitmap - "
163 "block_group = %d, block_bitmap = %u",
164 block_group, le32_to_cpu(desc->bg_block_bitmap));
165 return NULL;
166 }
Aneesh Kumar K.V2588ef82008-04-28 02:16:14 -0700167 ext3_valid_block_bitmap(sb, desc, block_group, bh);
168 /*
169 * file system mounted not to panic on error, continue with corrupt
170 * bitmap
171 */
Linus Torvalds0b832a42007-11-13 08:07:31 -0800172 return bh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173}
174/*
175 * The reservation window structure operations
176 * --------------------------------------------
177 * Operations include:
178 * dump, find, add, remove, is_empty, find_next_reservable_window, etc.
179 *
Mingming Cao36faadc142006-09-27 01:49:32 -0700180 * We use a red-black tree to represent per-filesystem reservation
181 * windows.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 *
Mingming Cao36faadc142006-09-27 01:49:32 -0700183 */
184
185/**
186 * __rsv_window_dump() -- Dump the filesystem block allocation reservation map
187 * @rb_root: root of per-filesystem reservation rb tree
188 * @verbose: verbose mode
189 * @fn: function which wishes to dump the reservation map
190 *
191 * If verbose is turned on, it will print the whole block reservation
192 * windows(start, end). Otherwise, it will only print out the "bad" windows,
193 * those windows that overlap with their immediate neighbors.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 */
Mingming Cao321fb9e2006-09-27 01:49:32 -0700195#if 1
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196static void __rsv_window_dump(struct rb_root *root, int verbose,
197 const char *fn)
198{
199 struct rb_node *n;
200 struct ext3_reserve_window_node *rsv, *prev;
201 int bad;
202
203restart:
204 n = rb_first(root);
205 bad = 0;
206 prev = NULL;
207
208 printk("Block Allocation Reservation Windows Map (%s):\n", fn);
209 while (n) {
Hugh Dickinsc56d2562006-12-06 20:41:27 -0800210 rsv = rb_entry(n, struct ext3_reserve_window_node, rsv_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211 if (verbose)
212 printk("reservation window 0x%p "
Mingming Cao321fb9e2006-09-27 01:49:32 -0700213 "start: %lu, end: %lu\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 rsv, rsv->rsv_start, rsv->rsv_end);
215 if (rsv->rsv_start && rsv->rsv_start >= rsv->rsv_end) {
216 printk("Bad reservation %p (start >= end)\n",
217 rsv);
218 bad = 1;
219 }
220 if (prev && prev->rsv_end >= rsv->rsv_start) {
221 printk("Bad reservation %p (prev->end >= start)\n",
222 rsv);
223 bad = 1;
224 }
225 if (bad) {
226 if (!verbose) {
227 printk("Restarting reservation walk in verbose mode\n");
228 verbose = 1;
229 goto restart;
230 }
231 }
232 n = rb_next(n);
233 prev = rsv;
234 }
235 printk("Window map complete.\n");
Julia Lawall269b2612008-04-28 02:16:09 -0700236 BUG_ON(bad);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237}
238#define rsv_window_dump(root, verbose) \
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700239 __rsv_window_dump((root), (verbose), __func__)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240#else
241#define rsv_window_dump(root, verbose) do {} while (0)
242#endif
243
Mingming Cao36faadc142006-09-27 01:49:32 -0700244/**
245 * goal_in_my_reservation()
246 * @rsv: inode's reservation window
247 * @grp_goal: given goal block relative to the allocation block group
248 * @group: the current allocation block group
249 * @sb: filesystem super block
250 *
251 * Test if the given goal block (group relative) is within the file's
252 * own block reservation window range.
253 *
254 * If the reservation window is outside the goal allocation group, return 0;
255 * grp_goal (given goal block) could be -1, which means no specific
256 * goal block. In this case, always return 1.
257 * If the goal block is within the reservation window, return 1;
258 * otherwise, return 0;
259 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260static int
Mingming Cao1c2bf372006-06-25 05:48:06 -0700261goal_in_my_reservation(struct ext3_reserve_window *rsv, ext3_grpblk_t grp_goal,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 unsigned int group, struct super_block * sb)
263{
Mingming Cao1c2bf372006-06-25 05:48:06 -0700264 ext3_fsblk_t group_first_block, group_last_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265
Mingming Cao43d23f92006-06-25 05:48:07 -0700266 group_first_block = ext3_group_first_block_no(sb, group);
Eric Sandeen32c2d2b2006-09-27 01:49:36 -0700267 group_last_block = group_first_block + (EXT3_BLOCKS_PER_GROUP(sb) - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268
269 if ((rsv->_rsv_start > group_last_block) ||
270 (rsv->_rsv_end < group_first_block))
271 return 0;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700272 if ((grp_goal >= 0) && ((grp_goal + group_first_block < rsv->_rsv_start)
273 || (grp_goal + group_first_block > rsv->_rsv_end)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 return 0;
275 return 1;
276}
277
Mingming Cao36faadc142006-09-27 01:49:32 -0700278/**
279 * search_reserve_window()
280 * @rb_root: root of reservation tree
281 * @goal: target allocation block
282 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283 * Find the reserved window which includes the goal, or the previous one
284 * if the goal is not in any window.
285 * Returns NULL if there are no windows or if all windows start after the goal.
286 */
287static struct ext3_reserve_window_node *
Mingming Cao1c2bf372006-06-25 05:48:06 -0700288search_reserve_window(struct rb_root *root, ext3_fsblk_t goal)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289{
290 struct rb_node *n = root->rb_node;
291 struct ext3_reserve_window_node *rsv;
292
293 if (!n)
294 return NULL;
295
296 do {
297 rsv = rb_entry(n, struct ext3_reserve_window_node, rsv_node);
298
299 if (goal < rsv->rsv_start)
300 n = n->rb_left;
301 else if (goal > rsv->rsv_end)
302 n = n->rb_right;
303 else
304 return rsv;
305 } while (n);
306 /*
307 * We've fallen off the end of the tree: the goal wasn't inside
308 * any particular node. OK, the previous node must be to one
309 * side of the interval containing the goal. If it's the RHS,
310 * we need to back up one.
311 */
312 if (rsv->rsv_start > goal) {
313 n = rb_prev(&rsv->rsv_node);
314 rsv = rb_entry(n, struct ext3_reserve_window_node, rsv_node);
315 }
316 return rsv;
317}
318
Mingming Cao36faadc142006-09-27 01:49:32 -0700319/**
320 * ext3_rsv_window_add() -- Insert a window to the block reservation rb tree.
321 * @sb: super block
322 * @rsv: reservation window to add
323 *
324 * Must be called with rsv_lock hold.
325 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326void ext3_rsv_window_add(struct super_block *sb,
327 struct ext3_reserve_window_node *rsv)
328{
329 struct rb_root *root = &EXT3_SB(sb)->s_rsv_window_root;
330 struct rb_node *node = &rsv->rsv_node;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700331 ext3_fsblk_t start = rsv->rsv_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332
333 struct rb_node ** p = &root->rb_node;
334 struct rb_node * parent = NULL;
335 struct ext3_reserve_window_node *this;
336
337 while (*p)
338 {
339 parent = *p;
340 this = rb_entry(parent, struct ext3_reserve_window_node, rsv_node);
341
342 if (start < this->rsv_start)
343 p = &(*p)->rb_left;
344 else if (start > this->rsv_end)
345 p = &(*p)->rb_right;
Mingming Cao321fb9e2006-09-27 01:49:32 -0700346 else {
347 rsv_window_dump(root, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 BUG();
Mingming Cao321fb9e2006-09-27 01:49:32 -0700349 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 }
351
352 rb_link_node(node, parent, p);
353 rb_insert_color(node, root);
354}
355
Mingming Cao36faadc142006-09-27 01:49:32 -0700356/**
357 * ext3_rsv_window_remove() -- unlink a window from the reservation rb tree
358 * @sb: super block
359 * @rsv: reservation window to remove
360 *
361 * Mark the block reservation window as not allocated, and unlink it
362 * from the filesystem reservation window rb tree. Must be called with
363 * rsv_lock hold.
364 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365static void rsv_window_remove(struct super_block *sb,
366 struct ext3_reserve_window_node *rsv)
367{
368 rsv->rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
369 rsv->rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
370 rsv->rsv_alloc_hit = 0;
371 rb_erase(&rsv->rsv_node, &EXT3_SB(sb)->s_rsv_window_root);
372}
373
Mingming Cao36faadc142006-09-27 01:49:32 -0700374/*
375 * rsv_is_empty() -- Check if the reservation window is allocated.
376 * @rsv: given reservation window to check
377 *
378 * returns 1 if the end block is EXT3_RESERVE_WINDOW_NOT_ALLOCATED.
379 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380static inline int rsv_is_empty(struct ext3_reserve_window *rsv)
381{
382 /* a valid reservation end block could not be 0 */
Mingming Cao36faadc142006-09-27 01:49:32 -0700383 return rsv->_rsv_end == EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384}
Mingming Cao36faadc142006-09-27 01:49:32 -0700385
386/**
387 * ext3_init_block_alloc_info()
388 * @inode: file inode structure
389 *
390 * Allocate and initialize the reservation window structure, and
391 * link the window to the ext3 inode structure at last
392 *
393 * The reservation window structure is only dynamically allocated
394 * and linked to ext3 inode the first time the open file
395 * needs a new block. So, before every ext3_new_block(s) call, for
396 * regular files, we should check whether the reservation window
397 * structure exists or not. In the latter case, this function is called.
398 * Fail to do so will result in block reservation being turned off for that
399 * open file.
400 *
401 * This function is called from ext3_get_blocks_handle(), also called
402 * when setting the reservation window size through ioctl before the file
403 * is open for write (needs block allocation).
404 *
405 * Needs truncate_mutex protection prior to call this function.
406 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407void ext3_init_block_alloc_info(struct inode *inode)
408{
409 struct ext3_inode_info *ei = EXT3_I(inode);
410 struct ext3_block_alloc_info *block_i = ei->i_block_alloc_info;
411 struct super_block *sb = inode->i_sb;
412
413 block_i = kmalloc(sizeof(*block_i), GFP_NOFS);
414 if (block_i) {
415 struct ext3_reserve_window_node *rsv = &block_i->rsv_window_node;
416
417 rsv->rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
418 rsv->rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
419
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700420 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 * if filesystem is mounted with NORESERVATION, the goal
422 * reservation window size is set to zero to indicate
423 * block reservation is off
424 */
425 if (!test_opt(sb, RESERVATION))
426 rsv->rsv_goal_size = 0;
427 else
428 rsv->rsv_goal_size = EXT3_DEFAULT_RESERVE_BLOCKS;
429 rsv->rsv_alloc_hit = 0;
430 block_i->last_alloc_logical_block = 0;
431 block_i->last_alloc_physical_block = 0;
432 }
433 ei->i_block_alloc_info = block_i;
434}
435
Mingming Cao36faadc142006-09-27 01:49:32 -0700436/**
437 * ext3_discard_reservation()
438 * @inode: inode
439 *
440 * Discard(free) block reservation window on last file close, or truncate
441 * or at last iput().
442 *
443 * It is being called in three cases:
444 * ext3_release_file(): last writer close the file
445 * ext3_clear_inode(): last iput(), when nobody link to this file.
446 * ext3_truncate(): when the block indirect map is about to change.
447 *
448 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449void ext3_discard_reservation(struct inode *inode)
450{
451 struct ext3_inode_info *ei = EXT3_I(inode);
452 struct ext3_block_alloc_info *block_i = ei->i_block_alloc_info;
453 struct ext3_reserve_window_node *rsv;
454 spinlock_t *rsv_lock = &EXT3_SB(inode->i_sb)->s_rsv_window_lock;
455
456 if (!block_i)
457 return;
458
459 rsv = &block_i->rsv_window_node;
460 if (!rsv_is_empty(&rsv->rsv_window)) {
461 spin_lock(rsv_lock);
462 if (!rsv_is_empty(&rsv->rsv_window))
463 rsv_window_remove(inode->i_sb, rsv);
464 spin_unlock(rsv_lock);
465 }
466}
467
Mingming Cao36faadc142006-09-27 01:49:32 -0700468/**
469 * ext3_free_blocks_sb() -- Free given blocks and update quota
470 * @handle: handle to this transaction
471 * @sb: super block
472 * @block: start physcial block to free
473 * @count: number of blocks to free
474 * @pdquot_freed_blocks: pointer to quota
475 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476void ext3_free_blocks_sb(handle_t *handle, struct super_block *sb,
Mingming Cao1c2bf372006-06-25 05:48:06 -0700477 ext3_fsblk_t block, unsigned long count,
478 unsigned long *pdquot_freed_blocks)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479{
480 struct buffer_head *bitmap_bh = NULL;
481 struct buffer_head *gd_bh;
482 unsigned long block_group;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700483 ext3_grpblk_t bit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 unsigned long i;
485 unsigned long overflow;
486 struct ext3_group_desc * desc;
487 struct ext3_super_block * es;
488 struct ext3_sb_info *sbi;
489 int err = 0, ret;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700490 ext3_grpblk_t group_freed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491
492 *pdquot_freed_blocks = 0;
493 sbi = EXT3_SB(sb);
494 es = sbi->s_es;
495 if (block < le32_to_cpu(es->s_first_data_block) ||
496 block + count < block ||
497 block + count > le32_to_cpu(es->s_blocks_count)) {
498 ext3_error (sb, "ext3_free_blocks",
499 "Freeing blocks not in datazone - "
Mingming Cao1c2bf372006-06-25 05:48:06 -0700500 "block = "E3FSBLK", count = %lu", block, count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 goto error_return;
502 }
503
504 ext3_debug ("freeing block(s) %lu-%lu\n", block, block + count - 1);
505
506do_more:
507 overflow = 0;
508 block_group = (block - le32_to_cpu(es->s_first_data_block)) /
509 EXT3_BLOCKS_PER_GROUP(sb);
510 bit = (block - le32_to_cpu(es->s_first_data_block)) %
511 EXT3_BLOCKS_PER_GROUP(sb);
512 /*
513 * Check to see if we are freeing blocks across a group
514 * boundary.
515 */
516 if (bit + count > EXT3_BLOCKS_PER_GROUP(sb)) {
517 overflow = bit + count - EXT3_BLOCKS_PER_GROUP(sb);
518 count -= overflow;
519 }
520 brelse(bitmap_bh);
521 bitmap_bh = read_block_bitmap(sb, block_group);
522 if (!bitmap_bh)
523 goto error_return;
524 desc = ext3_get_group_desc (sb, block_group, &gd_bh);
525 if (!desc)
526 goto error_return;
527
528 if (in_range (le32_to_cpu(desc->bg_block_bitmap), block, count) ||
529 in_range (le32_to_cpu(desc->bg_inode_bitmap), block, count) ||
530 in_range (block, le32_to_cpu(desc->bg_inode_table),
531 sbi->s_itb_per_group) ||
532 in_range (block + count - 1, le32_to_cpu(desc->bg_inode_table),
Aneesh Kumar K.Vfeda58d2008-02-06 01:40:13 -0800533 sbi->s_itb_per_group)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 ext3_error (sb, "ext3_free_blocks",
535 "Freeing blocks in system zones - "
Mingming Cao1c2bf372006-06-25 05:48:06 -0700536 "Block = "E3FSBLK", count = %lu",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 block, count);
Aneesh Kumar K.Vfeda58d2008-02-06 01:40:13 -0800538 goto error_return;
539 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540
541 /*
542 * We are about to start releasing blocks in the bitmap,
543 * so we need undo access.
544 */
545 /* @@@ check errors */
546 BUFFER_TRACE(bitmap_bh, "getting undo access");
547 err = ext3_journal_get_undo_access(handle, bitmap_bh);
548 if (err)
549 goto error_return;
550
551 /*
552 * We are about to modify some metadata. Call the journal APIs
553 * to unshare ->b_data if a currently-committing transaction is
554 * using it
555 */
556 BUFFER_TRACE(gd_bh, "get_write_access");
557 err = ext3_journal_get_write_access(handle, gd_bh);
558 if (err)
559 goto error_return;
560
561 jbd_lock_bh_state(bitmap_bh);
562
563 for (i = 0, group_freed = 0; i < count; i++) {
564 /*
565 * An HJ special. This is expensive...
566 */
567#ifdef CONFIG_JBD_DEBUG
568 jbd_unlock_bh_state(bitmap_bh);
569 {
570 struct buffer_head *debug_bh;
571 debug_bh = sb_find_get_block(sb, block + i);
572 if (debug_bh) {
573 BUFFER_TRACE(debug_bh, "Deleted!");
574 if (!bh2jh(bitmap_bh)->b_committed_data)
575 BUFFER_TRACE(debug_bh,
576 "No commited data in bitmap");
577 BUFFER_TRACE2(debug_bh, bitmap_bh, "bitmap");
578 __brelse(debug_bh);
579 }
580 }
581 jbd_lock_bh_state(bitmap_bh);
582#endif
583 if (need_resched()) {
584 jbd_unlock_bh_state(bitmap_bh);
585 cond_resched();
586 jbd_lock_bh_state(bitmap_bh);
587 }
588 /* @@@ This prevents newly-allocated data from being
589 * freed and then reallocated within the same
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700590 * transaction.
591 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 * Ideally we would want to allow that to happen, but to
593 * do so requires making journal_forget() capable of
594 * revoking the queued write of a data block, which
595 * implies blocking on the journal lock. *forget()
596 * cannot block due to truncate races.
597 *
598 * Eventually we can fix this by making journal_forget()
599 * return a status indicating whether or not it was able
600 * to revoke the buffer. On successful revoke, it is
601 * safe not to set the allocation bit in the committed
602 * bitmap, because we know that there is no outstanding
603 * activity on the buffer any more and so it is safe to
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700604 * reallocate it.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605 */
606 BUFFER_TRACE(bitmap_bh, "set in b_committed_data");
607 J_ASSERT_BH(bitmap_bh,
608 bh2jh(bitmap_bh)->b_committed_data != NULL);
609 ext3_set_bit_atomic(sb_bgl_lock(sbi, block_group), bit + i,
610 bh2jh(bitmap_bh)->b_committed_data);
611
612 /*
613 * We clear the bit in the bitmap after setting the committed
614 * data bit, because this is the reverse order to that which
615 * the allocator uses.
616 */
617 BUFFER_TRACE(bitmap_bh, "clear bit");
618 if (!ext3_clear_bit_atomic(sb_bgl_lock(sbi, block_group),
619 bit + i, bitmap_bh->b_data)) {
620 jbd_unlock_bh_state(bitmap_bh);
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700621 ext3_error(sb, __func__,
Mingming Cao1c2bf372006-06-25 05:48:06 -0700622 "bit already cleared for block "E3FSBLK,
623 block + i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 jbd_lock_bh_state(bitmap_bh);
625 BUFFER_TRACE(bitmap_bh, "bit already cleared");
626 } else {
627 group_freed++;
628 }
629 }
630 jbd_unlock_bh_state(bitmap_bh);
631
632 spin_lock(sb_bgl_lock(sbi, block_group));
Marcin Slusarz50e8a282008-02-08 04:20:13 -0800633 le16_add_cpu(&desc->bg_free_blocks_count, group_freed);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 spin_unlock(sb_bgl_lock(sbi, block_group));
Peter Zijlstraaa0dff22007-10-16 23:25:42 -0700635 percpu_counter_add(&sbi->s_freeblocks_counter, count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636
637 /* We dirtied the bitmap block */
638 BUFFER_TRACE(bitmap_bh, "dirtied bitmap block");
639 err = ext3_journal_dirty_metadata(handle, bitmap_bh);
640
641 /* And the group descriptor block */
642 BUFFER_TRACE(gd_bh, "dirtied group descriptor block");
643 ret = ext3_journal_dirty_metadata(handle, gd_bh);
644 if (!err) err = ret;
645 *pdquot_freed_blocks += group_freed;
646
647 if (overflow && !err) {
648 block += count;
649 count = overflow;
650 goto do_more;
651 }
652 sb->s_dirt = 1;
653error_return:
654 brelse(bitmap_bh);
655 ext3_std_error(sb, err);
656 return;
657}
658
Mingming Cao36faadc142006-09-27 01:49:32 -0700659/**
660 * ext3_free_blocks() -- Free given blocks and update quota
661 * @handle: handle for this transaction
662 * @inode: inode
663 * @block: start physical block to free
664 * @count: number of blocks to count
665 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666void ext3_free_blocks(handle_t *handle, struct inode *inode,
Mingming Cao1c2bf372006-06-25 05:48:06 -0700667 ext3_fsblk_t block, unsigned long count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668{
669 struct super_block * sb;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700670 unsigned long dquot_freed_blocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
672 sb = inode->i_sb;
673 if (!sb) {
674 printk ("ext3_free_blocks: nonexistent device");
675 return;
676 }
677 ext3_free_blocks_sb(handle, sb, block, count, &dquot_freed_blocks);
678 if (dquot_freed_blocks)
679 DQUOT_FREE_BLOCK(inode, dquot_freed_blocks);
680 return;
681}
682
Mingming Cao36faadc142006-09-27 01:49:32 -0700683/**
684 * ext3_test_allocatable()
685 * @nr: given allocation block group
686 * @bh: bufferhead contains the bitmap of the given block group
687 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 * For ext3 allocations, we must not reuse any blocks which are
689 * allocated in the bitmap buffer's "last committed data" copy. This
690 * prevents deletes from freeing up the page for reuse until we have
691 * committed the delete transaction.
692 *
693 * If we didn't do this, then deleting something and reallocating it as
694 * data would allow the old block to be overwritten before the
695 * transaction committed (because we force data to disk before commit).
696 * This would lead to corruption if we crashed between overwriting the
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700697 * data and committing the delete.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 *
699 * @@@ We may want to make this allocation behaviour conditional on
700 * data-writes at some point, and disable it for metadata allocations or
701 * sync-data inodes.
702 */
Mingming Cao1c2bf372006-06-25 05:48:06 -0700703static int ext3_test_allocatable(ext3_grpblk_t nr, struct buffer_head *bh)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704{
705 int ret;
706 struct journal_head *jh = bh2jh(bh);
707
708 if (ext3_test_bit(nr, bh->b_data))
709 return 0;
710
711 jbd_lock_bh_state(bh);
712 if (!jh->b_committed_data)
713 ret = 1;
714 else
715 ret = !ext3_test_bit(nr, jh->b_committed_data);
716 jbd_unlock_bh_state(bh);
717 return ret;
718}
719
Mingming Cao36faadc142006-09-27 01:49:32 -0700720/**
721 * bitmap_search_next_usable_block()
722 * @start: the starting block (group relative) of the search
723 * @bh: bufferhead contains the block group bitmap
724 * @maxblocks: the ending block (group relative) of the reservation
725 *
726 * The bitmap search --- search forward alternately through the actual
727 * bitmap on disk and the last-committed copy in journal, until we find a
728 * bit free in both bitmaps.
729 */
Mingming Cao1c2bf372006-06-25 05:48:06 -0700730static ext3_grpblk_t
731bitmap_search_next_usable_block(ext3_grpblk_t start, struct buffer_head *bh,
732 ext3_grpblk_t maxblocks)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733{
Mingming Cao1c2bf372006-06-25 05:48:06 -0700734 ext3_grpblk_t next;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 struct journal_head *jh = bh2jh(bh);
736
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737 while (start < maxblocks) {
738 next = ext3_find_next_zero_bit(bh->b_data, maxblocks, start);
739 if (next >= maxblocks)
740 return -1;
741 if (ext3_test_allocatable(next, bh))
742 return next;
743 jbd_lock_bh_state(bh);
744 if (jh->b_committed_data)
745 start = ext3_find_next_zero_bit(jh->b_committed_data,
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700746 maxblocks, next);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 jbd_unlock_bh_state(bh);
748 }
749 return -1;
750}
751
Mingming Cao36faadc142006-09-27 01:49:32 -0700752/**
753 * find_next_usable_block()
754 * @start: the starting block (group relative) to find next
755 * allocatable block in bitmap.
756 * @bh: bufferhead contains the block group bitmap
757 * @maxblocks: the ending block (group relative) for the search
758 *
759 * Find an allocatable block in a bitmap. We honor both the bitmap and
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760 * its last-committed copy (if that exists), and perform the "most
761 * appropriate allocation" algorithm of looking for a free block near
762 * the initial goal; then for a free byte somewhere in the bitmap; then
763 * for any free bit in the bitmap.
764 */
Mingming Cao1c2bf372006-06-25 05:48:06 -0700765static ext3_grpblk_t
766find_next_usable_block(ext3_grpblk_t start, struct buffer_head *bh,
767 ext3_grpblk_t maxblocks)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768{
Mingming Cao1c2bf372006-06-25 05:48:06 -0700769 ext3_grpblk_t here, next;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 char *p, *r;
771
772 if (start > 0) {
773 /*
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700774 * The goal was occupied; search forward for a free
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 * block within the next XX blocks.
776 *
777 * end_goal is more or less random, but it has to be
778 * less than EXT3_BLOCKS_PER_GROUP. Aligning up to the
779 * next 64-bit boundary is simple..
780 */
Mingming Cao1c2bf372006-06-25 05:48:06 -0700781 ext3_grpblk_t end_goal = (start + 63) & ~63;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 if (end_goal > maxblocks)
783 end_goal = maxblocks;
784 here = ext3_find_next_zero_bit(bh->b_data, end_goal, start);
785 if (here < end_goal && ext3_test_allocatable(here, bh))
786 return here;
787 ext3_debug("Bit not found near goal\n");
788 }
789
790 here = start;
791 if (here < 0)
792 here = 0;
793
794 p = ((char *)bh->b_data) + (here >> 3);
Hugh Dickins7d1c5202006-12-06 20:41:30 -0800795 r = memscan(p, 0, ((maxblocks + 7) >> 3) - (here >> 3));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 next = (r - ((char *)bh->b_data)) << 3;
797
798 if (next < maxblocks && next >= start && ext3_test_allocatable(next, bh))
799 return next;
800
801 /*
802 * The bitmap search --- search forward alternately through the actual
803 * bitmap and the last-committed copy until we find a bit free in
804 * both
805 */
806 here = bitmap_search_next_usable_block(here, bh, maxblocks);
807 return here;
808}
809
Mingming Cao36faadc142006-09-27 01:49:32 -0700810/**
811 * claim_block()
812 * @block: the free block (group relative) to allocate
813 * @bh: the bufferhead containts the block group bitmap
814 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 * We think we can allocate this block in this bitmap. Try to set the bit.
816 * If that succeeds then check that nobody has allocated and then freed the
817 * block since we saw that is was not marked in b_committed_data. If it _was_
818 * allocated and freed then clear the bit in the bitmap again and return
819 * zero (failure).
820 */
821static inline int
Mingming Cao1c2bf372006-06-25 05:48:06 -0700822claim_block(spinlock_t *lock, ext3_grpblk_t block, struct buffer_head *bh)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823{
824 struct journal_head *jh = bh2jh(bh);
825 int ret;
826
827 if (ext3_set_bit_atomic(lock, block, bh->b_data))
828 return 0;
829 jbd_lock_bh_state(bh);
830 if (jh->b_committed_data && ext3_test_bit(block,jh->b_committed_data)) {
831 ext3_clear_bit_atomic(lock, block, bh->b_data);
832 ret = 0;
833 } else {
834 ret = 1;
835 }
836 jbd_unlock_bh_state(bh);
837 return ret;
838}
839
Mingming Cao36faadc142006-09-27 01:49:32 -0700840/**
841 * ext3_try_to_allocate()
842 * @sb: superblock
843 * @handle: handle to this transaction
844 * @group: given allocation block group
845 * @bitmap_bh: bufferhead holds the block bitmap
846 * @grp_goal: given target block within the group
847 * @count: target number of blocks to allocate
848 * @my_rsv: reservation window
849 *
850 * Attempt to allocate blocks within a give range. Set the range of allocation
851 * first, then find the first free bit(s) from the bitmap (within the range),
852 * and at last, allocate the blocks by claiming the found free bit as allocated.
853 *
854 * To set the range of this allocation:
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700855 * if there is a reservation window, only try to allocate block(s) from the
Mingming Cao36faadc142006-09-27 01:49:32 -0700856 * file's own reservation window;
857 * Otherwise, the allocation range starts from the give goal block, ends at
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700858 * the block group's last block.
Mingming Cao36faadc142006-09-27 01:49:32 -0700859 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 * If we failed to allocate the desired block then we may end up crossing to a
861 * new bitmap. In that case we must release write access to the old one via
862 * ext3_journal_release_buffer(), else we'll run out of credits.
863 */
Mingming Cao1c2bf372006-06-25 05:48:06 -0700864static ext3_grpblk_t
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865ext3_try_to_allocate(struct super_block *sb, handle_t *handle, int group,
Mingming Cao1c2bf372006-06-25 05:48:06 -0700866 struct buffer_head *bitmap_bh, ext3_grpblk_t grp_goal,
Mingming Caob54e41e2006-03-26 01:37:57 -0800867 unsigned long *count, struct ext3_reserve_window *my_rsv)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868{
Mingming Cao1c2bf372006-06-25 05:48:06 -0700869 ext3_fsblk_t group_first_block;
870 ext3_grpblk_t start, end;
Mingming Caob54e41e2006-03-26 01:37:57 -0800871 unsigned long num = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872
873 /* we do allocation within the reservation window if we have a window */
874 if (my_rsv) {
Mingming Cao43d23f92006-06-25 05:48:07 -0700875 group_first_block = ext3_group_first_block_no(sb, group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876 if (my_rsv->_rsv_start >= group_first_block)
877 start = my_rsv->_rsv_start - group_first_block;
878 else
879 /* reservation window cross group boundary */
880 start = 0;
881 end = my_rsv->_rsv_end - group_first_block + 1;
882 if (end > EXT3_BLOCKS_PER_GROUP(sb))
883 /* reservation window crosses group boundary */
884 end = EXT3_BLOCKS_PER_GROUP(sb);
Mingming Cao1c2bf372006-06-25 05:48:06 -0700885 if ((start <= grp_goal) && (grp_goal < end))
886 start = grp_goal;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 else
Mingming Cao1c2bf372006-06-25 05:48:06 -0700888 grp_goal = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 } else {
Mingming Cao1c2bf372006-06-25 05:48:06 -0700890 if (grp_goal > 0)
891 start = grp_goal;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892 else
893 start = 0;
894 end = EXT3_BLOCKS_PER_GROUP(sb);
895 }
896
897 BUG_ON(start > EXT3_BLOCKS_PER_GROUP(sb));
898
899repeat:
Mingming Cao1c2bf372006-06-25 05:48:06 -0700900 if (grp_goal < 0 || !ext3_test_allocatable(grp_goal, bitmap_bh)) {
901 grp_goal = find_next_usable_block(start, bitmap_bh, end);
902 if (grp_goal < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 goto fail_access;
904 if (!my_rsv) {
905 int i;
906
Mingming Cao1c2bf372006-06-25 05:48:06 -0700907 for (i = 0; i < 7 && grp_goal > start &&
908 ext3_test_allocatable(grp_goal - 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909 bitmap_bh);
Mingming Cao1c2bf372006-06-25 05:48:06 -0700910 i++, grp_goal--)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 ;
912 }
913 }
Mingming Cao1c2bf372006-06-25 05:48:06 -0700914 start = grp_goal;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915
Mingming Cao36faadc142006-09-27 01:49:32 -0700916 if (!claim_block(sb_bgl_lock(EXT3_SB(sb), group),
917 grp_goal, bitmap_bh)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 /*
919 * The block was allocated by another thread, or it was
920 * allocated and then freed by another thread
921 */
922 start++;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700923 grp_goal++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 if (start >= end)
925 goto fail_access;
926 goto repeat;
927 }
Mingming Caob54e41e2006-03-26 01:37:57 -0800928 num++;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700929 grp_goal++;
930 while (num < *count && grp_goal < end
931 && ext3_test_allocatable(grp_goal, bitmap_bh)
Mingming Cao36faadc142006-09-27 01:49:32 -0700932 && claim_block(sb_bgl_lock(EXT3_SB(sb), group),
933 grp_goal, bitmap_bh)) {
Mingming Caob54e41e2006-03-26 01:37:57 -0800934 num++;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700935 grp_goal++;
Mingming Caob54e41e2006-03-26 01:37:57 -0800936 }
937 *count = num;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700938 return grp_goal - num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939fail_access:
Mingming Caob54e41e2006-03-26 01:37:57 -0800940 *count = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 return -1;
942}
943
944/**
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700945 * find_next_reservable_window():
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 * find a reservable space within the given range.
947 * It does not allocate the reservation window for now:
948 * alloc_new_reservation() will do the work later.
949 *
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700950 * @search_head: the head of the searching list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951 * This is not necessarily the list head of the whole filesystem
952 *
953 * We have both head and start_block to assist the search
954 * for the reservable space. The list starts from head,
955 * but we will shift to the place where start_block is,
956 * then start from there, when looking for a reservable space.
957 *
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700958 * @size: the target new reservation window size
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 *
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700960 * @group_first_block: the first block we consider to start
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961 * the real search from
962 *
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700963 * @last_block:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 * the maximum block number that our goal reservable space
965 * could start from. This is normally the last block in this
966 * group. The search will end when we found the start of next
967 * possible reservable space is out of this boundary.
968 * This could handle the cross boundary reservation window
969 * request.
970 *
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700971 * basically we search from the given range, rather than the whole
972 * reservation double linked list, (start_block, last_block)
973 * to find a free region that is of my size and has not
974 * been reserved.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976 */
Mingming Cao21fe3472005-06-28 20:45:16 -0700977static int find_next_reservable_window(
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 struct ext3_reserve_window_node *search_head,
Mingming Cao21fe3472005-06-28 20:45:16 -0700979 struct ext3_reserve_window_node *my_rsv,
Mingming Cao1c2bf372006-06-25 05:48:06 -0700980 struct super_block * sb,
981 ext3_fsblk_t start_block,
982 ext3_fsblk_t last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983{
984 struct rb_node *next;
985 struct ext3_reserve_window_node *rsv, *prev;
Mingming Cao1c2bf372006-06-25 05:48:06 -0700986 ext3_fsblk_t cur;
Mingming Cao21fe3472005-06-28 20:45:16 -0700987 int size = my_rsv->rsv_goal_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988
989 /* TODO: make the start of the reservation window byte-aligned */
990 /* cur = *start_block & ~7;*/
Mingming Cao21fe3472005-06-28 20:45:16 -0700991 cur = start_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 rsv = search_head;
993 if (!rsv)
Mingming Cao21fe3472005-06-28 20:45:16 -0700994 return -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995
996 while (1) {
997 if (cur <= rsv->rsv_end)
998 cur = rsv->rsv_end + 1;
999
1000 /* TODO?
1001 * in the case we could not find a reservable space
1002 * that is what is expected, during the re-search, we could
1003 * remember what's the largest reservable space we could have
1004 * and return that one.
1005 *
1006 * For now it will fail if we could not find the reservable
1007 * space with expected-size (or more)...
1008 */
1009 if (cur > last_block)
Mingming Cao21fe3472005-06-28 20:45:16 -07001010 return -1; /* fail */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011
1012 prev = rsv;
1013 next = rb_next(&rsv->rsv_node);
Hugh Dickinsc56d2562006-12-06 20:41:27 -08001014 rsv = rb_entry(next,struct ext3_reserve_window_node,rsv_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015
1016 /*
1017 * Reached the last reservation, we can just append to the
1018 * previous one.
1019 */
1020 if (!next)
1021 break;
1022
1023 if (cur + size <= rsv->rsv_start) {
1024 /*
1025 * Found a reserveable space big enough. We could
1026 * have a reservation across the group boundary here
Dave Kleikampe9ad5622006-09-27 01:49:35 -07001027 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 break;
1029 }
1030 }
1031 /*
1032 * we come here either :
1033 * when we reach the end of the whole list,
1034 * and there is empty reservable space after last entry in the list.
1035 * append it to the end of the list.
1036 *
1037 * or we found one reservable space in the middle of the list,
1038 * return the reservation window that we could append to.
1039 * succeed.
1040 */
Mingming Cao21fe3472005-06-28 20:45:16 -07001041
1042 if ((prev != my_rsv) && (!rsv_is_empty(&my_rsv->rsv_window)))
1043 rsv_window_remove(sb, my_rsv);
1044
1045 /*
1046 * Let's book the whole avaliable window for now. We will check the
1047 * disk bitmap later and then, if there are free blocks then we adjust
1048 * the window size if it's larger than requested.
1049 * Otherwise, we will remove this node from the tree next time
1050 * call find_next_reservable_window.
1051 */
1052 my_rsv->rsv_start = cur;
1053 my_rsv->rsv_end = cur + size - 1;
1054 my_rsv->rsv_alloc_hit = 0;
1055
1056 if (prev != my_rsv)
1057 ext3_rsv_window_add(sb, my_rsv);
1058
1059 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060}
1061
1062/**
Dave Kleikampe9ad5622006-09-27 01:49:35 -07001063 * alloc_new_reservation()--allocate a new reservation window
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064 *
1065 * To make a new reservation, we search part of the filesystem
1066 * reservation list (the list that inside the group). We try to
1067 * allocate a new reservation window near the allocation goal,
1068 * or the beginning of the group, if there is no goal.
1069 *
1070 * We first find a reservable space after the goal, then from
1071 * there, we check the bitmap for the first free block after
1072 * it. If there is no free block until the end of group, then the
1073 * whole group is full, we failed. Otherwise, check if the free
1074 * block is inside the expected reservable space, if so, we
1075 * succeed.
1076 * If the first free block is outside the reservable space, then
1077 * start from the first free block, we search for next available
1078 * space, and go on.
1079 *
1080 * on succeed, a new reservation will be found and inserted into the list
1081 * It contains at least one free block, and it does not overlap with other
1082 * reservation windows.
1083 *
1084 * failed: we failed to find a reservation window in this group
1085 *
1086 * @rsv: the reservation
1087 *
Mingming Cao1c2bf372006-06-25 05:48:06 -07001088 * @grp_goal: The goal (group-relative). It is where the search for a
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 * free reservable space should start from.
Mingming Cao1c2bf372006-06-25 05:48:06 -07001090 * if we have a grp_goal(grp_goal >0 ), then start from there,
1091 * no grp_goal(grp_goal = -1), we start from the first block
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092 * of the group.
1093 *
1094 * @sb: the super block
1095 * @group: the group we are trying to allocate in
1096 * @bitmap_bh: the block group block bitmap
Mingming Cao21fe3472005-06-28 20:45:16 -07001097 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 */
1099static int alloc_new_reservation(struct ext3_reserve_window_node *my_rsv,
Mingming Cao1c2bf372006-06-25 05:48:06 -07001100 ext3_grpblk_t grp_goal, struct super_block *sb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 unsigned int group, struct buffer_head *bitmap_bh)
1102{
1103 struct ext3_reserve_window_node *search_head;
Mingming Cao1c2bf372006-06-25 05:48:06 -07001104 ext3_fsblk_t group_first_block, group_end_block, start_block;
1105 ext3_grpblk_t first_free_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 struct rb_root *fs_rsv_root = &EXT3_SB(sb)->s_rsv_window_root;
1107 unsigned long size;
Mingming Cao21fe3472005-06-28 20:45:16 -07001108 int ret;
1109 spinlock_t *rsv_lock = &EXT3_SB(sb)->s_rsv_window_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110
Mingming Cao43d23f92006-06-25 05:48:07 -07001111 group_first_block = ext3_group_first_block_no(sb, group);
Eric Sandeen32c2d2b2006-09-27 01:49:36 -07001112 group_end_block = group_first_block + (EXT3_BLOCKS_PER_GROUP(sb) - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113
Mingming Cao1c2bf372006-06-25 05:48:06 -07001114 if (grp_goal < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 start_block = group_first_block;
1116 else
Mingming Cao1c2bf372006-06-25 05:48:06 -07001117 start_block = grp_goal + group_first_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118
1119 size = my_rsv->rsv_goal_size;
Mingming Cao21fe3472005-06-28 20:45:16 -07001120
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 if (!rsv_is_empty(&my_rsv->rsv_window)) {
1122 /*
1123 * if the old reservation is cross group boundary
1124 * and if the goal is inside the old reservation window,
1125 * we will come here when we just failed to allocate from
1126 * the first part of the window. We still have another part
1127 * that belongs to the next group. In this case, there is no
1128 * point to discard our window and try to allocate a new one
1129 * in this group(which will fail). we should
1130 * keep the reservation window, just simply move on.
1131 *
1132 * Maybe we could shift the start block of the reservation
1133 * window to the first block of next group.
1134 */
1135
1136 if ((my_rsv->rsv_start <= group_end_block) &&
1137 (my_rsv->rsv_end > group_end_block) &&
1138 (start_block >= my_rsv->rsv_start))
1139 return -1;
1140
1141 if ((my_rsv->rsv_alloc_hit >
1142 (my_rsv->rsv_end - my_rsv->rsv_start + 1) / 2)) {
1143 /*
Mingming Cao36faadc142006-09-27 01:49:32 -07001144 * if the previously allocation hit ratio is
1145 * greater than 1/2, then we double the size of
1146 * the reservation window the next time,
1147 * otherwise we keep the same size window
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148 */
1149 size = size * 2;
1150 if (size > EXT3_MAX_RESERVE_BLOCKS)
1151 size = EXT3_MAX_RESERVE_BLOCKS;
1152 my_rsv->rsv_goal_size= size;
1153 }
1154 }
Mingming Cao21fe3472005-06-28 20:45:16 -07001155
1156 spin_lock(rsv_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 /*
1158 * shift the search start to the window near the goal block
1159 */
1160 search_head = search_reserve_window(fs_rsv_root, start_block);
1161
1162 /*
1163 * find_next_reservable_window() simply finds a reservable window
1164 * inside the given range(start_block, group_end_block).
1165 *
1166 * To make sure the reservation window has a free bit inside it, we
1167 * need to check the bitmap after we found a reservable window.
1168 */
1169retry:
Mingming Cao21fe3472005-06-28 20:45:16 -07001170 ret = find_next_reservable_window(search_head, my_rsv, sb,
1171 start_block, group_end_block);
1172
1173 if (ret == -1) {
1174 if (!rsv_is_empty(&my_rsv->rsv_window))
1175 rsv_window_remove(sb, my_rsv);
1176 spin_unlock(rsv_lock);
1177 return -1;
1178 }
1179
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 /*
1181 * On success, find_next_reservable_window() returns the
1182 * reservation window where there is a reservable space after it.
1183 * Before we reserve this reservable space, we need
1184 * to make sure there is at least a free block inside this region.
1185 *
1186 * searching the first free bit on the block bitmap and copy of
1187 * last committed bitmap alternatively, until we found a allocatable
1188 * block. Search start from the start block of the reservable space
1189 * we just found.
1190 */
Mingming Cao21fe3472005-06-28 20:45:16 -07001191 spin_unlock(rsv_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 first_free_block = bitmap_search_next_usable_block(
Mingming Cao21fe3472005-06-28 20:45:16 -07001193 my_rsv->rsv_start - group_first_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194 bitmap_bh, group_end_block - group_first_block + 1);
1195
1196 if (first_free_block < 0) {
1197 /*
1198 * no free block left on the bitmap, no point
1199 * to reserve the space. return failed.
1200 */
Mingming Cao21fe3472005-06-28 20:45:16 -07001201 spin_lock(rsv_lock);
1202 if (!rsv_is_empty(&my_rsv->rsv_window))
1203 rsv_window_remove(sb, my_rsv);
1204 spin_unlock(rsv_lock);
1205 return -1; /* failed */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206 }
Mingming Cao21fe3472005-06-28 20:45:16 -07001207
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208 start_block = first_free_block + group_first_block;
1209 /*
1210 * check if the first free block is within the
Mingming Cao21fe3472005-06-28 20:45:16 -07001211 * free space we just reserved
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212 */
Hugh Dickinsff50dc52006-12-06 20:41:25 -08001213 if (start_block >= my_rsv->rsv_start && start_block <= my_rsv->rsv_end)
Mingming Cao21fe3472005-06-28 20:45:16 -07001214 return 0; /* success */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215 /*
1216 * if the first free bit we found is out of the reservable space
Mingming Cao21fe3472005-06-28 20:45:16 -07001217 * continue search for next reservable space,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 * start from where the free block is,
1219 * we also shift the list head to where we stopped last time
1220 */
Mingming Cao21fe3472005-06-28 20:45:16 -07001221 search_head = my_rsv;
1222 spin_lock(rsv_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223 goto retry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224}
1225
Mingming Cao36faadc142006-09-27 01:49:32 -07001226/**
1227 * try_to_extend_reservation()
1228 * @my_rsv: given reservation window
1229 * @sb: super block
1230 * @size: the delta to extend
1231 *
1232 * Attempt to expand the reservation window large enough to have
1233 * required number of free blocks
1234 *
1235 * Since ext3_try_to_allocate() will always allocate blocks within
1236 * the reservation window range, if the window size is too small,
1237 * multiple blocks allocation has to stop at the end of the reservation
1238 * window. To make this more efficient, given the total number of
1239 * blocks needed and the current size of the window, we try to
1240 * expand the reservation window size if necessary on a best-effort
1241 * basis before ext3_new_blocks() tries to allocate blocks,
1242 */
Mingming Caod48589b2006-03-26 01:37:59 -08001243static void try_to_extend_reservation(struct ext3_reserve_window_node *my_rsv,
1244 struct super_block *sb, int size)
1245{
1246 struct ext3_reserve_window_node *next_rsv;
1247 struct rb_node *next;
1248 spinlock_t *rsv_lock = &EXT3_SB(sb)->s_rsv_window_lock;
1249
1250 if (!spin_trylock(rsv_lock))
1251 return;
1252
1253 next = rb_next(&my_rsv->rsv_node);
1254
1255 if (!next)
1256 my_rsv->rsv_end += size;
1257 else {
Hugh Dickinsc56d2562006-12-06 20:41:27 -08001258 next_rsv = rb_entry(next, struct ext3_reserve_window_node, rsv_node);
Mingming Caod48589b2006-03-26 01:37:59 -08001259
1260 if ((next_rsv->rsv_start - my_rsv->rsv_end - 1) >= size)
1261 my_rsv->rsv_end += size;
1262 else
1263 my_rsv->rsv_end = next_rsv->rsv_start - 1;
1264 }
1265 spin_unlock(rsv_lock);
1266}
1267
Mingming Cao36faadc142006-09-27 01:49:32 -07001268/**
1269 * ext3_try_to_allocate_with_rsv()
1270 * @sb: superblock
1271 * @handle: handle to this transaction
1272 * @group: given allocation block group
1273 * @bitmap_bh: bufferhead holds the block bitmap
1274 * @grp_goal: given target block within the group
1275 * @count: target number of blocks to allocate
1276 * @my_rsv: reservation window
1277 * @errp: pointer to store the error code
1278 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 * This is the main function used to allocate a new block and its reservation
1280 * window.
1281 *
1282 * Each time when a new block allocation is need, first try to allocate from
1283 * its own reservation. If it does not have a reservation window, instead of
1284 * looking for a free bit on bitmap first, then look up the reservation list to
1285 * see if it is inside somebody else's reservation window, we try to allocate a
1286 * reservation window for it starting from the goal first. Then do the block
1287 * allocation within the reservation window.
1288 *
1289 * This will avoid keeping on searching the reservation list again and
Glauber de Oliveira Costa5b116872005-10-30 15:02:48 -08001290 * again when somebody is looking for a free block (without
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 * reservation), and there are lots of free blocks, but they are all
1292 * being reserved.
1293 *
Mingming Cao36faadc142006-09-27 01:49:32 -07001294 * We use a red-black tree for the per-filesystem reservation list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295 *
1296 */
Mingming Cao1c2bf372006-06-25 05:48:06 -07001297static ext3_grpblk_t
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298ext3_try_to_allocate_with_rsv(struct super_block *sb, handle_t *handle,
1299 unsigned int group, struct buffer_head *bitmap_bh,
Mingming Cao1c2bf372006-06-25 05:48:06 -07001300 ext3_grpblk_t grp_goal,
1301 struct ext3_reserve_window_node * my_rsv,
Mingming Caob54e41e2006-03-26 01:37:57 -08001302 unsigned long *count, int *errp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303{
Eric Sandeen32c2d2b2006-09-27 01:49:36 -07001304 ext3_fsblk_t group_first_block, group_last_block;
Mingming Cao1c2bf372006-06-25 05:48:06 -07001305 ext3_grpblk_t ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306 int fatal;
Mingming Caob54e41e2006-03-26 01:37:57 -08001307 unsigned long num = *count;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308
1309 *errp = 0;
1310
1311 /*
1312 * Make sure we use undo access for the bitmap, because it is critical
1313 * that we do the frozen_data COW on bitmap buffers in all cases even
1314 * if the buffer is in BJ_Forget state in the committing transaction.
1315 */
1316 BUFFER_TRACE(bitmap_bh, "get undo access for new block");
1317 fatal = ext3_journal_get_undo_access(handle, bitmap_bh);
1318 if (fatal) {
1319 *errp = fatal;
1320 return -1;
1321 }
1322
1323 /*
1324 * we don't deal with reservation when
1325 * filesystem is mounted without reservation
1326 * or the file is not a regular file
1327 * or last attempt to allocate a block with reservation turned on failed
1328 */
1329 if (my_rsv == NULL ) {
Mingming Caob54e41e2006-03-26 01:37:57 -08001330 ret = ext3_try_to_allocate(sb, handle, group, bitmap_bh,
Mingming Cao1c2bf372006-06-25 05:48:06 -07001331 grp_goal, count, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 goto out;
1333 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334 /*
Mingming Cao1c2bf372006-06-25 05:48:06 -07001335 * grp_goal is a group relative block number (if there is a goal)
Hugh Dickins16502422006-12-06 20:41:25 -08001336 * 0 <= grp_goal < EXT3_BLOCKS_PER_GROUP(sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 * first block is a filesystem wide block number
1338 * first block is the block number of the first block in this group
1339 */
Mingming Cao43d23f92006-06-25 05:48:07 -07001340 group_first_block = ext3_group_first_block_no(sb, group);
Eric Sandeen32c2d2b2006-09-27 01:49:36 -07001341 group_last_block = group_first_block + (EXT3_BLOCKS_PER_GROUP(sb) - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342
1343 /*
1344 * Basically we will allocate a new block from inode's reservation
1345 * window.
1346 *
1347 * We need to allocate a new reservation window, if:
1348 * a) inode does not have a reservation window; or
1349 * b) last attempt to allocate a block from existing reservation
1350 * failed; or
1351 * c) we come here with a goal and with a reservation window
1352 *
1353 * We do not need to allocate a new reservation window if we come here
1354 * at the beginning with a goal and the goal is inside the window, or
1355 * we don't have a goal but already have a reservation window.
1356 * then we could go to allocate from the reservation window directly.
1357 */
1358 while (1) {
Mingming Cao21fe3472005-06-28 20:45:16 -07001359 if (rsv_is_empty(&my_rsv->rsv_window) || (ret < 0) ||
Mingming Cao36faadc142006-09-27 01:49:32 -07001360 !goal_in_my_reservation(&my_rsv->rsv_window,
1361 grp_goal, group, sb)) {
Mingming Caod48589b2006-03-26 01:37:59 -08001362 if (my_rsv->rsv_goal_size < *count)
1363 my_rsv->rsv_goal_size = *count;
Mingming Cao1c2bf372006-06-25 05:48:06 -07001364 ret = alloc_new_reservation(my_rsv, grp_goal, sb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 group, bitmap_bh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366 if (ret < 0)
1367 break; /* failed */
1368
Mingming Cao36faadc142006-09-27 01:49:32 -07001369 if (!goal_in_my_reservation(&my_rsv->rsv_window,
1370 grp_goal, group, sb))
Mingming Cao1c2bf372006-06-25 05:48:06 -07001371 grp_goal = -1;
Hugh Dickins16502422006-12-06 20:41:25 -08001372 } else if (grp_goal >= 0) {
Mingming Cao2bd94bd2006-12-06 20:38:18 -08001373 int curr = my_rsv->rsv_end -
1374 (grp_goal + group_first_block) + 1;
1375
1376 if (curr < *count)
1377 try_to_extend_reservation(my_rsv, sb,
1378 *count - curr);
1379 }
Mingming Caod48589b2006-03-26 01:37:59 -08001380
Eric Sandeen32c2d2b2006-09-27 01:49:36 -07001381 if ((my_rsv->rsv_start > group_last_block) ||
1382 (my_rsv->rsv_end < group_first_block)) {
Mingming Cao321fb9e2006-09-27 01:49:32 -07001383 rsv_window_dump(&EXT3_SB(sb)->s_rsv_window_root, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384 BUG();
Mingming Cao321fb9e2006-09-27 01:49:32 -07001385 }
Mingming Cao36faadc142006-09-27 01:49:32 -07001386 ret = ext3_try_to_allocate(sb, handle, group, bitmap_bh,
1387 grp_goal, &num, &my_rsv->rsv_window);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 if (ret >= 0) {
Mingming Caob54e41e2006-03-26 01:37:57 -08001389 my_rsv->rsv_alloc_hit += num;
1390 *count = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391 break; /* succeed */
1392 }
Mingming Caob54e41e2006-03-26 01:37:57 -08001393 num = *count;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 }
1395out:
1396 if (ret >= 0) {
1397 BUFFER_TRACE(bitmap_bh, "journal_dirty_metadata for "
1398 "bitmap block");
1399 fatal = ext3_journal_dirty_metadata(handle, bitmap_bh);
1400 if (fatal) {
1401 *errp = fatal;
1402 return -1;
1403 }
1404 return ret;
1405 }
1406
1407 BUFFER_TRACE(bitmap_bh, "journal_release_buffer");
1408 ext3_journal_release_buffer(handle, bitmap_bh);
1409 return ret;
1410}
1411
Mingming Cao36faadc142006-09-27 01:49:32 -07001412/**
1413 * ext3_has_free_blocks()
1414 * @sbi: in-core super block structure.
1415 *
1416 * Check if filesystem has at least 1 free block available for allocation.
1417 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418static int ext3_has_free_blocks(struct ext3_sb_info *sbi)
1419{
Mingming Cao1c2bf372006-06-25 05:48:06 -07001420 ext3_fsblk_t free_blocks, root_blocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421
1422 free_blocks = percpu_counter_read_positive(&sbi->s_freeblocks_counter);
1423 root_blocks = le32_to_cpu(sbi->s_es->s_r_blocks_count);
1424 if (free_blocks < root_blocks + 1 && !capable(CAP_SYS_RESOURCE) &&
1425 sbi->s_resuid != current->fsuid &&
1426 (sbi->s_resgid == 0 || !in_group_p (sbi->s_resgid))) {
1427 return 0;
1428 }
1429 return 1;
1430}
1431
Mingming Cao36faadc142006-09-27 01:49:32 -07001432/**
1433 * ext3_should_retry_alloc()
1434 * @sb: super block
1435 * @retries number of attemps has been made
1436 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437 * ext3_should_retry_alloc() is called when ENOSPC is returned, and if
1438 * it is profitable to retry the operation, this function will wait
1439 * for the current or commiting transaction to complete, and then
1440 * return TRUE.
Mingming Cao36faadc142006-09-27 01:49:32 -07001441 *
1442 * if the total number of retries exceed three times, return FALSE.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 */
1444int ext3_should_retry_alloc(struct super_block *sb, int *retries)
1445{
1446 if (!ext3_has_free_blocks(EXT3_SB(sb)) || (*retries)++ > 3)
1447 return 0;
1448
1449 jbd_debug(1, "%s: retrying operation after ENOSPC\n", sb->s_id);
1450
1451 return journal_force_commit_nested(EXT3_SB(sb)->s_journal);
1452}
1453
Mingming Cao36faadc142006-09-27 01:49:32 -07001454/**
1455 * ext3_new_blocks() -- core block(s) allocation function
1456 * @handle: handle to this transaction
1457 * @inode: file inode
1458 * @goal: given target block(filesystem wide)
1459 * @count: target number of blocks to allocate
1460 * @errp: error code
1461 *
1462 * ext3_new_blocks uses a goal block to assist allocation. It tries to
1463 * allocate block(s) from the block group contains the goal block first. If that
1464 * fails, it will try to allocate block(s) from other block groups without
1465 * any specific goal block.
1466 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467 */
Mingming Cao1c2bf372006-06-25 05:48:06 -07001468ext3_fsblk_t ext3_new_blocks(handle_t *handle, struct inode *inode,
1469 ext3_fsblk_t goal, unsigned long *count, int *errp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470{
1471 struct buffer_head *bitmap_bh = NULL;
1472 struct buffer_head *gdp_bh;
1473 int group_no;
1474 int goal_group;
Mingming Cao1c2bf372006-06-25 05:48:06 -07001475 ext3_grpblk_t grp_target_blk; /* blockgroup relative goal block */
1476 ext3_grpblk_t grp_alloc_blk; /* blockgroup-relative allocated block*/
1477 ext3_fsblk_t ret_block; /* filesyetem-wide allocated block */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 int bgi; /* blockgroup iteration index */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 int fatal = 0, err;
1480 int performed_allocation = 0;
Mingming Cao1c2bf372006-06-25 05:48:06 -07001481 ext3_grpblk_t free_blocks; /* number of free blocks in a group */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482 struct super_block *sb;
1483 struct ext3_group_desc *gdp;
1484 struct ext3_super_block *es;
1485 struct ext3_sb_info *sbi;
1486 struct ext3_reserve_window_node *my_rsv = NULL;
1487 struct ext3_block_alloc_info *block_i;
1488 unsigned short windowsz = 0;
1489#ifdef EXT3FS_DEBUG
1490 static int goal_hits, goal_attempts;
1491#endif
1492 unsigned long ngroups;
Mingming Caob54e41e2006-03-26 01:37:57 -08001493 unsigned long num = *count;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494
1495 *errp = -ENOSPC;
1496 sb = inode->i_sb;
1497 if (!sb) {
1498 printk("ext3_new_block: nonexistent device");
1499 return 0;
1500 }
1501
1502 /*
1503 * Check quota for allocation of this block.
1504 */
Mingming Caofaa56972006-03-26 01:37:58 -08001505 if (DQUOT_ALLOC_BLOCK(inode, num)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506 *errp = -EDQUOT;
1507 return 0;
1508 }
1509
1510 sbi = EXT3_SB(sb);
1511 es = EXT3_SB(sb)->s_es;
1512 ext3_debug("goal=%lu.\n", goal);
1513 /*
1514 * Allocate a block from reservation only when
1515 * filesystem is mounted with reservation(default,-o reservation), and
1516 * it's a regular file, and
1517 * the desired window size is greater than 0 (One could use ioctl
1518 * command EXT3_IOC_SETRSVSZ to set the window size to 0 to turn off
1519 * reservation on that particular file)
1520 */
1521 block_i = EXT3_I(inode)->i_block_alloc_info;
1522 if (block_i && ((windowsz = block_i->rsv_window_node.rsv_goal_size) > 0))
1523 my_rsv = &block_i->rsv_window_node;
1524
1525 if (!ext3_has_free_blocks(sbi)) {
1526 *errp = -ENOSPC;
1527 goto out;
1528 }
1529
1530 /*
1531 * First, test whether the goal block is free.
1532 */
1533 if (goal < le32_to_cpu(es->s_first_data_block) ||
1534 goal >= le32_to_cpu(es->s_blocks_count))
1535 goal = le32_to_cpu(es->s_first_data_block);
1536 group_no = (goal - le32_to_cpu(es->s_first_data_block)) /
1537 EXT3_BLOCKS_PER_GROUP(sb);
Mingming Cao08fb306f2006-08-27 01:23:44 -07001538 goal_group = group_no;
1539retry_alloc:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 gdp = ext3_get_group_desc(sb, group_no, &gdp_bh);
1541 if (!gdp)
1542 goto io_error;
1543
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 free_blocks = le16_to_cpu(gdp->bg_free_blocks_count);
1545 /*
1546 * if there is not enough free blocks to make a new resevation
1547 * turn off reservation for this allocation
1548 */
1549 if (my_rsv && (free_blocks < windowsz)
1550 && (rsv_is_empty(&my_rsv->rsv_window)))
1551 my_rsv = NULL;
1552
1553 if (free_blocks > 0) {
Mingming Cao1c2bf372006-06-25 05:48:06 -07001554 grp_target_blk = ((goal - le32_to_cpu(es->s_first_data_block)) %
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 EXT3_BLOCKS_PER_GROUP(sb));
1556 bitmap_bh = read_block_bitmap(sb, group_no);
1557 if (!bitmap_bh)
1558 goto io_error;
Mingming Cao1c2bf372006-06-25 05:48:06 -07001559 grp_alloc_blk = ext3_try_to_allocate_with_rsv(sb, handle,
1560 group_no, bitmap_bh, grp_target_blk,
1561 my_rsv, &num, &fatal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562 if (fatal)
1563 goto out;
Mingming Cao1c2bf372006-06-25 05:48:06 -07001564 if (grp_alloc_blk >= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 goto allocated;
1566 }
1567
1568 ngroups = EXT3_SB(sb)->s_groups_count;
1569 smp_rmb();
1570
1571 /*
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001572 * Now search the rest of the groups. We assume that
Akinobu Mita144704e2008-02-06 01:40:15 -08001573 * group_no and gdp correctly point to the last group visited.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574 */
1575 for (bgi = 0; bgi < ngroups; bgi++) {
1576 group_no++;
1577 if (group_no >= ngroups)
1578 group_no = 0;
1579 gdp = ext3_get_group_desc(sb, group_no, &gdp_bh);
Hugh Dickins2823b552006-12-06 20:41:28 -08001580 if (!gdp)
1581 goto io_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 free_blocks = le16_to_cpu(gdp->bg_free_blocks_count);
1583 /*
1584 * skip this group if the number of
1585 * free blocks is less than half of the reservation
1586 * window size.
1587 */
1588 if (free_blocks <= (windowsz/2))
1589 continue;
1590
1591 brelse(bitmap_bh);
1592 bitmap_bh = read_block_bitmap(sb, group_no);
1593 if (!bitmap_bh)
1594 goto io_error;
Mingming Cao1c2bf372006-06-25 05:48:06 -07001595 /*
1596 * try to allocate block(s) from this group, without a goal(-1).
1597 */
1598 grp_alloc_blk = ext3_try_to_allocate_with_rsv(sb, handle,
1599 group_no, bitmap_bh, -1, my_rsv,
1600 &num, &fatal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601 if (fatal)
1602 goto out;
Mingming Cao1c2bf372006-06-25 05:48:06 -07001603 if (grp_alloc_blk >= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604 goto allocated;
1605 }
1606 /*
1607 * We may end up a bogus ealier ENOSPC error due to
1608 * filesystem is "full" of reservations, but
1609 * there maybe indeed free blocks avaliable on disk
1610 * In this case, we just forget about the reservations
1611 * just do block allocation as without reservations.
1612 */
1613 if (my_rsv) {
1614 my_rsv = NULL;
Hugh Dickinsef503672006-12-06 20:41:23 -08001615 windowsz = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616 group_no = goal_group;
Mingming Cao08fb306f2006-08-27 01:23:44 -07001617 goto retry_alloc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618 }
1619 /* No space left on the device */
1620 *errp = -ENOSPC;
1621 goto out;
1622
1623allocated:
1624
1625 ext3_debug("using block group %d(%d)\n",
1626 group_no, gdp->bg_free_blocks_count);
1627
1628 BUFFER_TRACE(gdp_bh, "get_write_access");
1629 fatal = ext3_journal_get_write_access(handle, gdp_bh);
1630 if (fatal)
1631 goto out;
1632
Mingming Cao43d23f92006-06-25 05:48:07 -07001633 ret_block = grp_alloc_blk + ext3_group_first_block_no(sb, group_no);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634
Mingming Cao1c2bf372006-06-25 05:48:06 -07001635 if (in_range(le32_to_cpu(gdp->bg_block_bitmap), ret_block, num) ||
1636 in_range(le32_to_cpu(gdp->bg_inode_bitmap), ret_block, num) ||
1637 in_range(ret_block, le32_to_cpu(gdp->bg_inode_table),
Mingming Caofaa56972006-03-26 01:37:58 -08001638 EXT3_SB(sb)->s_itb_per_group) ||
Mingming Cao1c2bf372006-06-25 05:48:06 -07001639 in_range(ret_block + num - 1, le32_to_cpu(gdp->bg_inode_table),
Aneesh Kumar K.Vfeda58d2008-02-06 01:40:13 -08001640 EXT3_SB(sb)->s_itb_per_group)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641 ext3_error(sb, "ext3_new_block",
1642 "Allocating block in system zone - "
Mingming Cao1c2bf372006-06-25 05:48:06 -07001643 "blocks from "E3FSBLK", length %lu",
1644 ret_block, num);
Aneesh Kumar K.V2588ef82008-04-28 02:16:14 -07001645 /*
1646 * claim_block() marked the blocks we allocated as in use. So we
1647 * may want to selectively mark some of the blocks as free.
1648 */
1649 goto retry_alloc;
Aneesh Kumar K.Vfeda58d2008-02-06 01:40:13 -08001650 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651
1652 performed_allocation = 1;
1653
1654#ifdef CONFIG_JBD_DEBUG
1655 {
1656 struct buffer_head *debug_bh;
1657
1658 /* Record bitmap buffer state in the newly allocated block */
Mingming Cao1c2bf372006-06-25 05:48:06 -07001659 debug_bh = sb_find_get_block(sb, ret_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660 if (debug_bh) {
1661 BUFFER_TRACE(debug_bh, "state when allocated");
1662 BUFFER_TRACE2(debug_bh, bitmap_bh, "bitmap state");
1663 brelse(debug_bh);
1664 }
1665 }
1666 jbd_lock_bh_state(bitmap_bh);
1667 spin_lock(sb_bgl_lock(sbi, group_no));
1668 if (buffer_jbd(bitmap_bh) && bh2jh(bitmap_bh)->b_committed_data) {
Mingming Caofaa56972006-03-26 01:37:58 -08001669 int i;
1670
1671 for (i = 0; i < num; i++) {
Mingming Cao1c2bf372006-06-25 05:48:06 -07001672 if (ext3_test_bit(grp_alloc_blk+i,
Mingming Caofaa56972006-03-26 01:37:58 -08001673 bh2jh(bitmap_bh)->b_committed_data)) {
1674 printk("%s: block was unexpectedly set in "
Harvey Harrisone05b6b52008-04-28 02:16:15 -07001675 "b_committed_data\n", __func__);
Mingming Caofaa56972006-03-26 01:37:58 -08001676 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677 }
1678 }
Mingming Cao1c2bf372006-06-25 05:48:06 -07001679 ext3_debug("found bit %d\n", grp_alloc_blk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 spin_unlock(sb_bgl_lock(sbi, group_no));
1681 jbd_unlock_bh_state(bitmap_bh);
1682#endif
1683
Mingming Caofaa56972006-03-26 01:37:58 -08001684 if (ret_block + num - 1 >= le32_to_cpu(es->s_blocks_count)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 ext3_error(sb, "ext3_new_block",
Mingming Cao1c2bf372006-06-25 05:48:06 -07001686 "block("E3FSBLK") >= blocks count(%d) - "
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 "block_group = %d, es == %p ", ret_block,
1688 le32_to_cpu(es->s_blocks_count), group_no, es);
1689 goto out;
1690 }
1691
1692 /*
1693 * It is up to the caller to add the new buffer to a journal
1694 * list of some description. We don't know in advance whether
1695 * the caller wants to use it as metadata or data.
1696 */
Mingming Cao1c2bf372006-06-25 05:48:06 -07001697 ext3_debug("allocating block %lu. Goal hits %d of %d.\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698 ret_block, goal_hits, goal_attempts);
1699
1700 spin_lock(sb_bgl_lock(sbi, group_no));
Marcin Slusarz50e8a282008-02-08 04:20:13 -08001701 le16_add_cpu(&gdp->bg_free_blocks_count, -num);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702 spin_unlock(sb_bgl_lock(sbi, group_no));
Peter Zijlstra3cb4f9f2007-10-16 23:25:42 -07001703 percpu_counter_sub(&sbi->s_freeblocks_counter, num);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704
1705 BUFFER_TRACE(gdp_bh, "journal_dirty_metadata for group descriptor");
1706 err = ext3_journal_dirty_metadata(handle, gdp_bh);
1707 if (!fatal)
1708 fatal = err;
1709
1710 sb->s_dirt = 1;
1711 if (fatal)
1712 goto out;
1713
1714 *errp = 0;
1715 brelse(bitmap_bh);
Mingming Caofaa56972006-03-26 01:37:58 -08001716 DQUOT_FREE_BLOCK(inode, *count-num);
Mingming Caob54e41e2006-03-26 01:37:57 -08001717 *count = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718 return ret_block;
1719
1720io_error:
1721 *errp = -EIO;
1722out:
1723 if (fatal) {
1724 *errp = fatal;
1725 ext3_std_error(sb, fatal);
1726 }
1727 /*
1728 * Undo the block allocation
1729 */
1730 if (!performed_allocation)
Mingming Caofaa56972006-03-26 01:37:58 -08001731 DQUOT_FREE_BLOCK(inode, *count);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732 brelse(bitmap_bh);
1733 return 0;
1734}
1735
Mingming Cao1c2bf372006-06-25 05:48:06 -07001736ext3_fsblk_t ext3_new_block(handle_t *handle, struct inode *inode,
1737 ext3_fsblk_t goal, int *errp)
Mingming Caob54e41e2006-03-26 01:37:57 -08001738{
1739 unsigned long count = 1;
1740
1741 return ext3_new_blocks(handle, inode, goal, &count, errp);
1742}
1743
Mingming Cao36faadc142006-09-27 01:49:32 -07001744/**
1745 * ext3_count_free_blocks() -- count filesystem free blocks
1746 * @sb: superblock
1747 *
1748 * Adds up the number of free blocks from each block group.
1749 */
Mingming Cao43d23f92006-06-25 05:48:07 -07001750ext3_fsblk_t ext3_count_free_blocks(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751{
Mingming Cao43d23f92006-06-25 05:48:07 -07001752 ext3_fsblk_t desc_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 struct ext3_group_desc *gdp;
1754 int i;
Glauber de Oliveira Costa8bdac5d2005-09-22 21:44:26 -07001755 unsigned long ngroups = EXT3_SB(sb)->s_groups_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756#ifdef EXT3FS_DEBUG
1757 struct ext3_super_block *es;
Mingming Cao43d23f92006-06-25 05:48:07 -07001758 ext3_fsblk_t bitmap_count;
1759 unsigned long x;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760 struct buffer_head *bitmap_bh = NULL;
1761
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762 es = EXT3_SB(sb)->s_es;
1763 desc_count = 0;
1764 bitmap_count = 0;
1765 gdp = NULL;
Glauber de Oliveira Costa8bdac5d2005-09-22 21:44:26 -07001766
Glauber de Oliveira Costa5b116872005-10-30 15:02:48 -08001767 smp_rmb();
Glauber de Oliveira Costa8bdac5d2005-09-22 21:44:26 -07001768 for (i = 0; i < ngroups; i++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769 gdp = ext3_get_group_desc(sb, i, NULL);
1770 if (!gdp)
1771 continue;
1772 desc_count += le16_to_cpu(gdp->bg_free_blocks_count);
1773 brelse(bitmap_bh);
1774 bitmap_bh = read_block_bitmap(sb, i);
1775 if (bitmap_bh == NULL)
1776 continue;
1777
1778 x = ext3_count_free(bitmap_bh, sb->s_blocksize);
1779 printk("group %d: stored = %d, counted = %lu\n",
1780 i, le16_to_cpu(gdp->bg_free_blocks_count), x);
1781 bitmap_count += x;
1782 }
1783 brelse(bitmap_bh);
Mingming Cao43d23f92006-06-25 05:48:07 -07001784 printk("ext3_count_free_blocks: stored = "E3FSBLK
1785 ", computed = "E3FSBLK", "E3FSBLK"\n",
1786 le32_to_cpu(es->s_free_blocks_count),
1787 desc_count, bitmap_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788 return bitmap_count;
1789#else
1790 desc_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791 smp_rmb();
1792 for (i = 0; i < ngroups; i++) {
1793 gdp = ext3_get_group_desc(sb, i, NULL);
1794 if (!gdp)
1795 continue;
1796 desc_count += le16_to_cpu(gdp->bg_free_blocks_count);
1797 }
1798
1799 return desc_count;
1800#endif
1801}
1802
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803static inline int test_root(int a, int b)
1804{
1805 int num = b;
1806
1807 while (a > num)
1808 num *= b;
1809 return num == a;
1810}
1811
1812static int ext3_group_sparse(int group)
1813{
1814 if (group <= 1)
1815 return 1;
1816 if (!(group & 1))
1817 return 0;
1818 return (test_root(group, 7) || test_root(group, 5) ||
1819 test_root(group, 3));
1820}
1821
1822/**
1823 * ext3_bg_has_super - number of blocks used by the superblock in group
1824 * @sb: superblock for filesystem
1825 * @group: group number to check
1826 *
1827 * Return the number of blocks used by the superblock (primary or backup)
1828 * in this group. Currently this will be only 0 or 1.
1829 */
1830int ext3_bg_has_super(struct super_block *sb, int group)
1831{
Glauber de Oliveira Costab5a7c4f2006-03-24 03:18:37 -08001832 if (EXT3_HAS_RO_COMPAT_FEATURE(sb,
1833 EXT3_FEATURE_RO_COMPAT_SPARSE_SUPER) &&
1834 !ext3_group_sparse(group))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001835 return 0;
1836 return 1;
1837}
1838
Glauber de Oliveira Costab5a7c4f2006-03-24 03:18:37 -08001839static unsigned long ext3_bg_num_gdb_meta(struct super_block *sb, int group)
1840{
1841 unsigned long metagroup = group / EXT3_DESC_PER_BLOCK(sb);
1842 unsigned long first = metagroup * EXT3_DESC_PER_BLOCK(sb);
1843 unsigned long last = first + EXT3_DESC_PER_BLOCK(sb) - 1;
1844
1845 if (group == first || group == first + 1 || group == last)
1846 return 1;
1847 return 0;
1848}
1849
1850static unsigned long ext3_bg_num_gdb_nometa(struct super_block *sb, int group)
1851{
Akinobu Mita859cb932008-02-06 01:40:17 -08001852 return ext3_bg_has_super(sb, group) ? EXT3_SB(sb)->s_gdb_count : 0;
Glauber de Oliveira Costab5a7c4f2006-03-24 03:18:37 -08001853}
1854
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855/**
1856 * ext3_bg_num_gdb - number of blocks used by the group table in group
1857 * @sb: superblock for filesystem
1858 * @group: group number to check
1859 *
1860 * Return the number of blocks used by the group descriptor table
1861 * (primary or backup) in this group. In the future there may be a
1862 * different number of descriptor blocks in each group.
1863 */
1864unsigned long ext3_bg_num_gdb(struct super_block *sb, int group)
1865{
Glauber de Oliveira Costab5a7c4f2006-03-24 03:18:37 -08001866 unsigned long first_meta_bg =
1867 le32_to_cpu(EXT3_SB(sb)->s_es->s_first_meta_bg);
1868 unsigned long metagroup = group / EXT3_DESC_PER_BLOCK(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869
Glauber de Oliveira Costab5a7c4f2006-03-24 03:18:37 -08001870 if (!EXT3_HAS_INCOMPAT_FEATURE(sb,EXT3_FEATURE_INCOMPAT_META_BG) ||
1871 metagroup < first_meta_bg)
1872 return ext3_bg_num_gdb_nometa(sb,group);
1873
1874 return ext3_bg_num_gdb_meta(sb,group);
1875
1876}