]>
Commit | Line | Data |
---|---|---|
00e5a55c BS |
1 | From: Mingming Cao <cmm@us.ibm.com> |
2 | Subject: Ext4 update | |
3 | References: fate#303783 | |
4 | ||
5 | Bring ext4 codebase to the state of 2.6.28-rc8. It has lots of bugfixes, some | |
6 | of them really important ones (data corruption, easily triggerable kernel | |
7 | oopses with delayed allocation, ...). | |
8 | ||
9 | Signed-off-by: Jan Kara <jack@suse.cz> | |
10 | ||
11 | diff -rup b/fs/ext4//acl.h a/fs/ext4///acl.h | |
12 | --- b/fs/ext4/acl.h 2009-02-11 14:37:58.000000000 +0100 | |
13 | +++ a/fs/ext4/acl.h 2009-02-10 21:40:14.000000000 +0100 | |
14 | @@ -51,18 +51,18 @@ static inline int ext4_acl_count(size_t | |
15 | } | |
16 | } | |
17 | ||
18 | -#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL | |
19 | +#if defined(CONFIG_EXT4_FS_POSIX_ACL) || defined(CONFIG_EXT4DEV_FS_POSIX_ACL) | |
20 | ||
21 | /* Value for inode->u.ext4_i.i_acl and inode->u.ext4_i.i_default_acl | |
22 | if the ACL has not been cached */ | |
23 | #define EXT4_ACL_NOT_CACHED ((void *)-1) | |
24 | ||
25 | /* acl.c */ | |
26 | -extern int ext4_permission (struct inode *, int); | |
27 | -extern int ext4_acl_chmod (struct inode *); | |
28 | -extern int ext4_init_acl (handle_t *, struct inode *, struct inode *); | |
29 | +extern int ext4_permission(struct inode *, int); | |
30 | +extern int ext4_acl_chmod(struct inode *); | |
31 | +extern int ext4_init_acl(handle_t *, struct inode *, struct inode *); | |
32 | ||
33 | -#else /* CONFIG_EXT4DEV_FS_POSIX_ACL */ | |
34 | +#else /* CONFIG_EXT4_FS_POSIX_ACL */ | |
35 | #include <linux/sched.h> | |
36 | #define ext4_permission NULL | |
37 | ||
38 | @@ -77,5 +77,5 @@ ext4_init_acl(handle_t *handle, struct i | |
39 | { | |
40 | return 0; | |
41 | } | |
42 | -#endif /* CONFIG_EXT4DEV_FS_POSIX_ACL */ | |
43 | +#endif /* CONFIG_EXT4_FS_POSIX_ACL */ | |
44 | ||
45 | diff -rup b/fs/ext4//balloc.c a/fs/ext4///balloc.c | |
46 | --- b/fs/ext4/balloc.c 2009-02-11 14:37:58.000000000 +0100 | |
47 | +++ a/fs/ext4/balloc.c 2009-02-10 21:40:11.000000000 +0100 | |
48 | @@ -20,6 +20,7 @@ | |
49 | #include "ext4.h" | |
50 | #include "ext4_jbd2.h" | |
51 | #include "group.h" | |
52 | +#include "mballoc.h" | |
53 | ||
54 | /* | |
55 | * balloc.c contains the blocks allocation and deallocation routines | |
56 | @@ -83,6 +84,7 @@ static int ext4_group_used_meta_blocks(s | |
57 | } | |
58 | return used_blocks; | |
59 | } | |
60 | + | |
61 | /* Initializes an uninitialized block bitmap if given, and returns the | |
62 | * number of blocks free in the group. */ | |
63 | unsigned ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh, | |
64 | @@ -99,10 +101,10 @@ unsigned ext4_init_block_bitmap(struct s | |
65 | * essentially implementing a per-group read-only flag. */ | |
66 | if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) { | |
67 | ext4_error(sb, __func__, | |
68 | - "Checksum bad for group %lu\n", block_group); | |
69 | - gdp->bg_free_blocks_count = 0; | |
70 | - gdp->bg_free_inodes_count = 0; | |
71 | - gdp->bg_itable_unused = 0; | |
72 | + "Checksum bad for group %u\n", block_group); | |
73 | + ext4_free_blks_set(sb, gdp, 0); | |
74 | + ext4_free_inodes_set(sb, gdp, 0); | |
75 | + ext4_itable_unused_set(sb, gdp, 0); | |
76 | memset(bh->b_data, 0xff, sb->s_blocksize); | |
77 | return 0; | |
78 | } | |
79 | @@ -132,7 +134,7 @@ unsigned ext4_init_block_bitmap(struct s | |
80 | */ | |
81 | group_blocks = ext4_blocks_count(sbi->s_es) - | |
82 | le32_to_cpu(sbi->s_es->s_first_data_block) - | |
83 | - (EXT4_BLOCKS_PER_GROUP(sb) * (sbi->s_groups_count -1)); | |
84 | + (EXT4_BLOCKS_PER_GROUP(sb) * (sbi->s_groups_count - 1)); | |
85 | } else { | |
86 | group_blocks = EXT4_BLOCKS_PER_GROUP(sb); | |
87 | } | |
88 | @@ -200,20 +202,20 @@ unsigned ext4_init_block_bitmap(struct s | |
89 | * @bh: pointer to the buffer head to store the block | |
90 | * group descriptor | |
91 | */ | |
92 | -struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb, | |
93 | +struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb, | |
94 | ext4_group_t block_group, | |
95 | - struct buffer_head ** bh) | |
96 | + struct buffer_head **bh) | |
97 | { | |
98 | - unsigned long group_desc; | |
99 | - unsigned long offset; | |
100 | - struct ext4_group_desc * desc; | |
101 | + unsigned int group_desc; | |
102 | + unsigned int offset; | |
103 | + struct ext4_group_desc *desc; | |
104 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
105 | ||
106 | if (block_group >= sbi->s_groups_count) { | |
107 | - ext4_error (sb, "ext4_get_group_desc", | |
108 | - "block_group >= groups_count - " | |
109 | - "block_group = %lu, groups_count = %lu", | |
110 | - block_group, sbi->s_groups_count); | |
111 | + ext4_error(sb, "ext4_get_group_desc", | |
112 | + "block_group >= groups_count - " | |
113 | + "block_group = %u, groups_count = %u", | |
114 | + block_group, sbi->s_groups_count); | |
115 | ||
116 | return NULL; | |
117 | } | |
118 | @@ -222,10 +224,10 @@ struct ext4_group_desc * ext4_get_group_ | |
119 | group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb); | |
120 | offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1); | |
121 | if (!sbi->s_group_desc[group_desc]) { | |
122 | - ext4_error (sb, "ext4_get_group_desc", | |
123 | - "Group descriptor not loaded - " | |
124 | - "block_group = %lu, group_desc = %lu, desc = %lu", | |
125 | - block_group, group_desc, offset); | |
126 | + ext4_error(sb, "ext4_get_group_desc", | |
127 | + "Group descriptor not loaded - " | |
128 | + "block_group = %u, group_desc = %u, desc = %u", | |
129 | + block_group, group_desc, offset); | |
130 | return NULL; | |
131 | } | |
132 | ||
133 | @@ -302,8 +304,8 @@ err_out: | |
134 | struct buffer_head * | |
135 | ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group) | |
136 | { | |
137 | - struct ext4_group_desc * desc; | |
138 | - struct buffer_head * bh = NULL; | |
139 | + struct ext4_group_desc *desc; | |
140 | + struct buffer_head *bh = NULL; | |
141 | ext4_fsblk_t bitmap_blk; | |
142 | ||
143 | desc = ext4_get_group_desc(sb, block_group, NULL); | |
144 | @@ -314,27 +316,50 @@ ext4_read_block_bitmap(struct super_bloc | |
145 | if (unlikely(!bh)) { | |
146 | ext4_error(sb, __func__, | |
147 | "Cannot read block bitmap - " | |
148 | - "block_group = %lu, block_bitmap = %llu", | |
149 | + "block_group = %u, block_bitmap = %llu", | |
150 | block_group, bitmap_blk); | |
151 | return NULL; | |
152 | } | |
153 | - if (bh_uptodate_or_lock(bh)) | |
154 | + | |
155 | + if (bitmap_uptodate(bh)) | |
156 | return bh; | |
157 | ||
158 | + lock_buffer(bh); | |
159 | + if (bitmap_uptodate(bh)) { | |
160 | + unlock_buffer(bh); | |
161 | + return bh; | |
162 | + } | |
163 | spin_lock(sb_bgl_lock(EXT4_SB(sb), block_group)); | |
164 | if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { | |
165 | ext4_init_block_bitmap(sb, bh, block_group, desc); | |
166 | + set_bitmap_uptodate(bh); | |
167 | set_buffer_uptodate(bh); | |
168 | - unlock_buffer(bh); | |
169 | spin_unlock(sb_bgl_lock(EXT4_SB(sb), block_group)); | |
170 | + unlock_buffer(bh); | |
171 | return bh; | |
172 | } | |
173 | spin_unlock(sb_bgl_lock(EXT4_SB(sb), block_group)); | |
174 | + if (buffer_uptodate(bh)) { | |
175 | + /* | |
176 | + * if not uninit if bh is uptodate, | |
177 | + * bitmap is also uptodate | |
178 | + */ | |
179 | + set_bitmap_uptodate(bh); | |
180 | + unlock_buffer(bh); | |
181 | + return bh; | |
182 | + } | |
183 | + /* | |
184 | + * submit the buffer_head for read. We can | |
185 | + * safely mark the bitmap as uptodate now. | |
186 | + * We do it here so the bitmap uptodate bit | |
187 | + * get set with buffer lock held. | |
188 | + */ | |
189 | + set_bitmap_uptodate(bh); | |
190 | if (bh_submit_read(bh) < 0) { | |
191 | put_bh(bh); | |
192 | ext4_error(sb, __func__, | |
193 | "Cannot read block bitmap - " | |
194 | - "block_group = %lu, block_bitmap = %llu", | |
195 | + "block_group = %u, block_bitmap = %llu", | |
196 | block_group, bitmap_blk); | |
197 | return NULL; | |
198 | } | |
199 | @@ -345,356 +370,50 @@ ext4_read_block_bitmap(struct super_bloc | |
200 | */ | |
201 | return bh; | |
202 | } | |
203 | -/* | |
204 | - * The reservation window structure operations | |
205 | - * -------------------------------------------- | |
206 | - * Operations include: | |
207 | - * dump, find, add, remove, is_empty, find_next_reservable_window, etc. | |
208 | - * | |
209 | - * We use a red-black tree to represent per-filesystem reservation | |
210 | - * windows. | |
211 | - * | |
212 | - */ | |
213 | - | |
214 | -/** | |
215 | - * __rsv_window_dump() -- Dump the filesystem block allocation reservation map | |
216 | - * @rb_root: root of per-filesystem reservation rb tree | |
217 | - * @verbose: verbose mode | |
218 | - * @fn: function which wishes to dump the reservation map | |
219 | - * | |
220 | - * If verbose is turned on, it will print the whole block reservation | |
221 | - * windows(start, end). Otherwise, it will only print out the "bad" windows, | |
222 | - * those windows that overlap with their immediate neighbors. | |
223 | - */ | |
224 | -#if 1 | |
225 | -static void __rsv_window_dump(struct rb_root *root, int verbose, | |
226 | - const char *fn) | |
227 | -{ | |
228 | - struct rb_node *n; | |
229 | - struct ext4_reserve_window_node *rsv, *prev; | |
230 | - int bad; | |
231 | - | |
232 | -restart: | |
233 | - n = rb_first(root); | |
234 | - bad = 0; | |
235 | - prev = NULL; | |
236 | - | |
237 | - printk("Block Allocation Reservation Windows Map (%s):\n", fn); | |
238 | - while (n) { | |
239 | - rsv = rb_entry(n, struct ext4_reserve_window_node, rsv_node); | |
240 | - if (verbose) | |
241 | - printk("reservation window 0x%p " | |
242 | - "start: %llu, end: %llu\n", | |
243 | - rsv, rsv->rsv_start, rsv->rsv_end); | |
244 | - if (rsv->rsv_start && rsv->rsv_start >= rsv->rsv_end) { | |
245 | - printk("Bad reservation %p (start >= end)\n", | |
246 | - rsv); | |
247 | - bad = 1; | |
248 | - } | |
249 | - if (prev && prev->rsv_end >= rsv->rsv_start) { | |
250 | - printk("Bad reservation %p (prev->end >= start)\n", | |
251 | - rsv); | |
252 | - bad = 1; | |
253 | - } | |
254 | - if (bad) { | |
255 | - if (!verbose) { | |
256 | - printk("Restarting reservation walk in verbose mode\n"); | |
257 | - verbose = 1; | |
258 | - goto restart; | |
259 | - } | |
260 | - } | |
261 | - n = rb_next(n); | |
262 | - prev = rsv; | |
263 | - } | |
264 | - printk("Window map complete.\n"); | |
265 | - BUG_ON(bad); | |
266 | -} | |
267 | -#define rsv_window_dump(root, verbose) \ | |
268 | - __rsv_window_dump((root), (verbose), __func__) | |
269 | -#else | |
270 | -#define rsv_window_dump(root, verbose) do {} while (0) | |
271 | -#endif | |
272 | - | |
273 | -/** | |
274 | - * goal_in_my_reservation() | |
275 | - * @rsv: inode's reservation window | |
276 | - * @grp_goal: given goal block relative to the allocation block group | |
277 | - * @group: the current allocation block group | |
278 | - * @sb: filesystem super block | |
279 | - * | |
280 | - * Test if the given goal block (group relative) is within the file's | |
281 | - * own block reservation window range. | |
282 | - * | |
283 | - * If the reservation window is outside the goal allocation group, return 0; | |
284 | - * grp_goal (given goal block) could be -1, which means no specific | |
285 | - * goal block. In this case, always return 1. | |
286 | - * If the goal block is within the reservation window, return 1; | |
287 | - * otherwise, return 0; | |
288 | - */ | |
289 | -static int | |
290 | -goal_in_my_reservation(struct ext4_reserve_window *rsv, ext4_grpblk_t grp_goal, | |
291 | - ext4_group_t group, struct super_block *sb) | |
292 | -{ | |
293 | - ext4_fsblk_t group_first_block, group_last_block; | |
294 | - | |
295 | - group_first_block = ext4_group_first_block_no(sb, group); | |
296 | - group_last_block = group_first_block + (EXT4_BLOCKS_PER_GROUP(sb) - 1); | |
297 | - | |
298 | - if ((rsv->_rsv_start > group_last_block) || | |
299 | - (rsv->_rsv_end < group_first_block)) | |
300 | - return 0; | |
301 | - if ((grp_goal >= 0) && ((grp_goal + group_first_block < rsv->_rsv_start) | |
302 | - || (grp_goal + group_first_block > rsv->_rsv_end))) | |
303 | - return 0; | |
304 | - return 1; | |
305 | -} | |
306 | ||
307 | /** | |
308 | - * search_reserve_window() | |
309 | - * @rb_root: root of reservation tree | |
310 | - * @goal: target allocation block | |
311 | - * | |
312 | - * Find the reserved window which includes the goal, or the previous one | |
313 | - * if the goal is not in any window. | |
314 | - * Returns NULL if there are no windows or if all windows start after the goal. | |
315 | - */ | |
316 | -static struct ext4_reserve_window_node * | |
317 | -search_reserve_window(struct rb_root *root, ext4_fsblk_t goal) | |
318 | -{ | |
319 | - struct rb_node *n = root->rb_node; | |
320 | - struct ext4_reserve_window_node *rsv; | |
321 | - | |
322 | - if (!n) | |
323 | - return NULL; | |
324 | - | |
325 | - do { | |
326 | - rsv = rb_entry(n, struct ext4_reserve_window_node, rsv_node); | |
327 | - | |
328 | - if (goal < rsv->rsv_start) | |
329 | - n = n->rb_left; | |
330 | - else if (goal > rsv->rsv_end) | |
331 | - n = n->rb_right; | |
332 | - else | |
333 | - return rsv; | |
334 | - } while (n); | |
335 | - /* | |
336 | - * We've fallen off the end of the tree: the goal wasn't inside | |
337 | - * any particular node. OK, the previous node must be to one | |
338 | - * side of the interval containing the goal. If it's the RHS, | |
339 | - * we need to back up one. | |
340 | - */ | |
341 | - if (rsv->rsv_start > goal) { | |
342 | - n = rb_prev(&rsv->rsv_node); | |
343 | - rsv = rb_entry(n, struct ext4_reserve_window_node, rsv_node); | |
344 | - } | |
345 | - return rsv; | |
346 | -} | |
347 | - | |
348 | -/** | |
349 | - * ext4_rsv_window_add() -- Insert a window to the block reservation rb tree. | |
350 | - * @sb: super block | |
351 | - * @rsv: reservation window to add | |
352 | - * | |
353 | - * Must be called with rsv_lock hold. | |
354 | - */ | |
355 | -void ext4_rsv_window_add(struct super_block *sb, | |
356 | - struct ext4_reserve_window_node *rsv) | |
357 | -{ | |
358 | - struct rb_root *root = &EXT4_SB(sb)->s_rsv_window_root; | |
359 | - struct rb_node *node = &rsv->rsv_node; | |
360 | - ext4_fsblk_t start = rsv->rsv_start; | |
361 | - | |
362 | - struct rb_node ** p = &root->rb_node; | |
363 | - struct rb_node * parent = NULL; | |
364 | - struct ext4_reserve_window_node *this; | |
365 | - | |
366 | - while (*p) | |
367 | - { | |
368 | - parent = *p; | |
369 | - this = rb_entry(parent, struct ext4_reserve_window_node, rsv_node); | |
370 | - | |
371 | - if (start < this->rsv_start) | |
372 | - p = &(*p)->rb_left; | |
373 | - else if (start > this->rsv_end) | |
374 | - p = &(*p)->rb_right; | |
375 | - else { | |
376 | - rsv_window_dump(root, 1); | |
377 | - BUG(); | |
378 | - } | |
379 | - } | |
380 | - | |
381 | - rb_link_node(node, parent, p); | |
382 | - rb_insert_color(node, root); | |
383 | -} | |
384 | - | |
385 | -/** | |
386 | - * ext4_rsv_window_remove() -- unlink a window from the reservation rb tree | |
387 | - * @sb: super block | |
388 | - * @rsv: reservation window to remove | |
389 | - * | |
390 | - * Mark the block reservation window as not allocated, and unlink it | |
391 | - * from the filesystem reservation window rb tree. Must be called with | |
392 | - * rsv_lock hold. | |
393 | - */ | |
394 | -static void rsv_window_remove(struct super_block *sb, | |
395 | - struct ext4_reserve_window_node *rsv) | |
396 | -{ | |
397 | - rsv->rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED; | |
398 | - rsv->rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED; | |
399 | - rsv->rsv_alloc_hit = 0; | |
400 | - rb_erase(&rsv->rsv_node, &EXT4_SB(sb)->s_rsv_window_root); | |
401 | -} | |
402 | - | |
403 | -/* | |
404 | - * rsv_is_empty() -- Check if the reservation window is allocated. | |
405 | - * @rsv: given reservation window to check | |
406 | - * | |
407 | - * returns 1 if the end block is EXT4_RESERVE_WINDOW_NOT_ALLOCATED. | |
408 | - */ | |
409 | -static inline int rsv_is_empty(struct ext4_reserve_window *rsv) | |
410 | -{ | |
411 | - /* a valid reservation end block could not be 0 */ | |
412 | - return rsv->_rsv_end == EXT4_RESERVE_WINDOW_NOT_ALLOCATED; | |
413 | -} | |
414 | - | |
415 | -/** | |
416 | - * ext4_init_block_alloc_info() | |
417 | - * @inode: file inode structure | |
418 | - * | |
419 | - * Allocate and initialize the reservation window structure, and | |
420 | - * link the window to the ext4 inode structure at last | |
421 | - * | |
422 | - * The reservation window structure is only dynamically allocated | |
423 | - * and linked to ext4 inode the first time the open file | |
424 | - * needs a new block. So, before every ext4_new_block(s) call, for | |
425 | - * regular files, we should check whether the reservation window | |
426 | - * structure exists or not. In the latter case, this function is called. | |
427 | - * Fail to do so will result in block reservation being turned off for that | |
428 | - * open file. | |
429 | - * | |
430 | - * This function is called from ext4_get_blocks_handle(), also called | |
431 | - * when setting the reservation window size through ioctl before the file | |
432 | - * is open for write (needs block allocation). | |
433 | - * | |
434 | - * Needs down_write(i_data_sem) protection prior to call this function. | |
435 | - */ | |
436 | -void ext4_init_block_alloc_info(struct inode *inode) | |
437 | -{ | |
438 | - struct ext4_inode_info *ei = EXT4_I(inode); | |
439 | - struct ext4_block_alloc_info *block_i = ei->i_block_alloc_info; | |
440 | - struct super_block *sb = inode->i_sb; | |
441 | - | |
442 | - block_i = kmalloc(sizeof(*block_i), GFP_NOFS); | |
443 | - if (block_i) { | |
444 | - struct ext4_reserve_window_node *rsv = &block_i->rsv_window_node; | |
445 | - | |
446 | - rsv->rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED; | |
447 | - rsv->rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED; | |
448 | - | |
449 | - /* | |
450 | - * if filesystem is mounted with NORESERVATION, the goal | |
451 | - * reservation window size is set to zero to indicate | |
452 | - * block reservation is off | |
453 | - */ | |
454 | - if (!test_opt(sb, RESERVATION)) | |
455 | - rsv->rsv_goal_size = 0; | |
456 | - else | |
457 | - rsv->rsv_goal_size = EXT4_DEFAULT_RESERVE_BLOCKS; | |
458 | - rsv->rsv_alloc_hit = 0; | |
459 | - block_i->last_alloc_logical_block = 0; | |
460 | - block_i->last_alloc_physical_block = 0; | |
461 | - } | |
462 | - ei->i_block_alloc_info = block_i; | |
463 | -} | |
464 | - | |
465 | -/** | |
466 | - * ext4_discard_reservation() | |
467 | - * @inode: inode | |
468 | - * | |
469 | - * Discard(free) block reservation window on last file close, or truncate | |
470 | - * or at last iput(). | |
471 | - * | |
472 | - * It is being called in three cases: | |
473 | - * ext4_release_file(): last writer close the file | |
474 | - * ext4_clear_inode(): last iput(), when nobody link to this file. | |
475 | - * ext4_truncate(): when the block indirect map is about to change. | |
476 | - * | |
477 | - */ | |
478 | -void ext4_discard_reservation(struct inode *inode) | |
479 | -{ | |
480 | - struct ext4_inode_info *ei = EXT4_I(inode); | |
481 | - struct ext4_block_alloc_info *block_i = ei->i_block_alloc_info; | |
482 | - struct ext4_reserve_window_node *rsv; | |
483 | - spinlock_t *rsv_lock = &EXT4_SB(inode->i_sb)->s_rsv_window_lock; | |
484 | - | |
485 | - ext4_mb_discard_inode_preallocations(inode); | |
486 | - | |
487 | - if (!block_i) | |
488 | - return; | |
489 | - | |
490 | - rsv = &block_i->rsv_window_node; | |
491 | - if (!rsv_is_empty(&rsv->rsv_window)) { | |
492 | - spin_lock(rsv_lock); | |
493 | - if (!rsv_is_empty(&rsv->rsv_window)) | |
494 | - rsv_window_remove(inode->i_sb, rsv); | |
495 | - spin_unlock(rsv_lock); | |
496 | - } | |
497 | -} | |
498 | - | |
499 | -/** | |
500 | - * ext4_free_blocks_sb() -- Free given blocks and update quota | |
501 | + * ext4_add_groupblocks() -- Add given blocks to an existing group | |
502 | * @handle: handle to this transaction | |
503 | * @sb: super block | |
504 | - * @block: start physcial block to free | |
505 | + * @block: start physcial block to add to the block group | |
506 | * @count: number of blocks to free | |
507 | - * @pdquot_freed_blocks: pointer to quota | |
508 | + * | |
509 | + * This marks the blocks as free in the bitmap. We ask the | |
510 | + * mballoc to reload the buddy after this by setting group | |
511 | + * EXT4_GROUP_INFO_NEED_INIT_BIT flag | |
512 | */ | |
513 | -void ext4_free_blocks_sb(handle_t *handle, struct super_block *sb, | |
514 | - ext4_fsblk_t block, unsigned long count, | |
515 | - unsigned long *pdquot_freed_blocks) | |
516 | +void ext4_add_groupblocks(handle_t *handle, struct super_block *sb, | |
517 | + ext4_fsblk_t block, unsigned long count) | |
518 | { | |
519 | struct buffer_head *bitmap_bh = NULL; | |
520 | struct buffer_head *gd_bh; | |
521 | ext4_group_t block_group; | |
522 | ext4_grpblk_t bit; | |
523 | - unsigned long i; | |
524 | - unsigned long overflow; | |
525 | - struct ext4_group_desc * desc; | |
526 | - struct ext4_super_block * es; | |
527 | + unsigned int i; | |
528 | + struct ext4_group_desc *desc; | |
529 | + struct ext4_super_block *es; | |
530 | struct ext4_sb_info *sbi; | |
531 | - int err = 0, ret; | |
532 | - ext4_grpblk_t group_freed; | |
533 | + int err = 0, ret, blk_free_count; | |
534 | + ext4_grpblk_t blocks_freed; | |
535 | + struct ext4_group_info *grp; | |
536 | ||
537 | - *pdquot_freed_blocks = 0; | |
538 | sbi = EXT4_SB(sb); | |
539 | es = sbi->s_es; | |
540 | - if (block < le32_to_cpu(es->s_first_data_block) || | |
541 | - block + count < block || | |
542 | - block + count > ext4_blocks_count(es)) { | |
543 | - ext4_error (sb, "ext4_free_blocks", | |
544 | - "Freeing blocks not in datazone - " | |
545 | - "block = %llu, count = %lu", block, count); | |
546 | - goto error_return; | |
547 | - } | |
548 | - | |
549 | - ext4_debug ("freeing block(s) %llu-%llu\n", block, block + count - 1); | |
550 | + ext4_debug("Adding block(s) %llu-%llu\n", block, block + count - 1); | |
551 | ||
552 | -do_more: | |
553 | - overflow = 0; | |
554 | ext4_get_group_no_and_offset(sb, block, &block_group, &bit); | |
555 | + grp = ext4_get_group_info(sb, block_group); | |
556 | /* | |
557 | * Check to see if we are freeing blocks across a group | |
558 | * boundary. | |
559 | */ | |
560 | if (bit + count > EXT4_BLOCKS_PER_GROUP(sb)) { | |
561 | - overflow = bit + count - EXT4_BLOCKS_PER_GROUP(sb); | |
562 | - count -= overflow; | |
563 | + goto error_return; | |
564 | } | |
565 | - brelse(bitmap_bh); | |
566 | bitmap_bh = ext4_read_block_bitmap(sb, block_group); | |
567 | if (!bitmap_bh) | |
568 | goto error_return; | |
569 | - desc = ext4_get_group_desc (sb, block_group, &gd_bh); | |
570 | + desc = ext4_get_group_desc(sb, block_group, &gd_bh); | |
571 | if (!desc) | |
572 | goto error_return; | |
573 | ||
574 | @@ -703,18 +422,17 @@ do_more: | |
575 | in_range(block, ext4_inode_table(sb, desc), sbi->s_itb_per_group) || | |
576 | in_range(block + count - 1, ext4_inode_table(sb, desc), | |
577 | sbi->s_itb_per_group)) { | |
578 | - ext4_error (sb, "ext4_free_blocks", | |
579 | - "Freeing blocks in system zones - " | |
580 | - "Block = %llu, count = %lu", | |
581 | - block, count); | |
582 | + ext4_error(sb, __func__, | |
583 | + "Adding blocks in system zones - " | |
584 | + "Block = %llu, count = %lu", | |
585 | + block, count); | |
586 | goto error_return; | |
587 | } | |
588 | ||
589 | /* | |
590 | - * We are about to start releasing blocks in the bitmap, | |
591 | + * We are about to add blocks to the bitmap, | |
592 | * so we need undo access. | |
593 | */ | |
594 | - /* @@@ check errors */ | |
595 | BUFFER_TRACE(bitmap_bh, "getting undo access"); | |
596 | err = ext4_journal_get_undo_access(handle, bitmap_bh); | |
597 | if (err) | |
598 | @@ -729,90 +447,43 @@ do_more: | |
599 | err = ext4_journal_get_write_access(handle, gd_bh); | |
600 | if (err) | |
601 | goto error_return; | |
602 | - | |
603 | - jbd_lock_bh_state(bitmap_bh); | |
604 | - | |
605 | - for (i = 0, group_freed = 0; i < count; i++) { | |
606 | - /* | |
607 | - * An HJ special. This is expensive... | |
608 | - */ | |
609 | -#ifdef CONFIG_JBD2_DEBUG | |
610 | - jbd_unlock_bh_state(bitmap_bh); | |
611 | - { | |
612 | - struct buffer_head *debug_bh; | |
613 | - debug_bh = sb_find_get_block(sb, block + i); | |
614 | - if (debug_bh) { | |
615 | - BUFFER_TRACE(debug_bh, "Deleted!"); | |
616 | - if (!bh2jh(bitmap_bh)->b_committed_data) | |
617 | - BUFFER_TRACE(debug_bh, | |
618 | - "No commited data in bitmap"); | |
619 | - BUFFER_TRACE2(debug_bh, bitmap_bh, "bitmap"); | |
620 | - __brelse(debug_bh); | |
621 | - } | |
622 | - } | |
623 | - jbd_lock_bh_state(bitmap_bh); | |
624 | -#endif | |
625 | - if (need_resched()) { | |
626 | - jbd_unlock_bh_state(bitmap_bh); | |
627 | - cond_resched(); | |
628 | - jbd_lock_bh_state(bitmap_bh); | |
629 | - } | |
630 | - /* @@@ This prevents newly-allocated data from being | |
631 | - * freed and then reallocated within the same | |
632 | - * transaction. | |
633 | - * | |
634 | - * Ideally we would want to allow that to happen, but to | |
635 | - * do so requires making jbd2_journal_forget() capable of | |
636 | - * revoking the queued write of a data block, which | |
637 | - * implies blocking on the journal lock. *forget() | |
638 | - * cannot block due to truncate races. | |
639 | - * | |
640 | - * Eventually we can fix this by making jbd2_journal_forget() | |
641 | - * return a status indicating whether or not it was able | |
642 | - * to revoke the buffer. On successful revoke, it is | |
643 | - * safe not to set the allocation bit in the committed | |
644 | - * bitmap, because we know that there is no outstanding | |
645 | - * activity on the buffer any more and so it is safe to | |
646 | - * reallocate it. | |
647 | - */ | |
648 | - BUFFER_TRACE(bitmap_bh, "set in b_committed_data"); | |
649 | - J_ASSERT_BH(bitmap_bh, | |
650 | - bh2jh(bitmap_bh)->b_committed_data != NULL); | |
651 | - ext4_set_bit_atomic(sb_bgl_lock(sbi, block_group), bit + i, | |
652 | - bh2jh(bitmap_bh)->b_committed_data); | |
653 | - | |
654 | - /* | |
655 | - * We clear the bit in the bitmap after setting the committed | |
656 | - * data bit, because this is the reverse order to that which | |
657 | - * the allocator uses. | |
658 | - */ | |
659 | + /* | |
660 | + * make sure we don't allow a parallel init on other groups in the | |
661 | + * same buddy cache | |
662 | + */ | |
663 | + down_write(&grp->alloc_sem); | |
664 | + for (i = 0, blocks_freed = 0; i < count; i++) { | |
665 | BUFFER_TRACE(bitmap_bh, "clear bit"); | |
666 | if (!ext4_clear_bit_atomic(sb_bgl_lock(sbi, block_group), | |
667 | bit + i, bitmap_bh->b_data)) { | |
668 | - jbd_unlock_bh_state(bitmap_bh); | |
669 | ext4_error(sb, __func__, | |
670 | "bit already cleared for block %llu", | |
671 | (ext4_fsblk_t)(block + i)); | |
672 | - jbd_lock_bh_state(bitmap_bh); | |
673 | BUFFER_TRACE(bitmap_bh, "bit already cleared"); | |
674 | } else { | |
675 | - group_freed++; | |
676 | + blocks_freed++; | |
677 | } | |
678 | } | |
679 | - jbd_unlock_bh_state(bitmap_bh); | |
680 | - | |
681 | spin_lock(sb_bgl_lock(sbi, block_group)); | |
682 | - le16_add_cpu(&desc->bg_free_blocks_count, group_freed); | |
683 | + blk_free_count = blocks_freed + ext4_free_blks_count(sb, desc); | |
684 | + ext4_free_blks_set(sb, desc, blk_free_count); | |
685 | desc->bg_checksum = ext4_group_desc_csum(sbi, block_group, desc); | |
686 | spin_unlock(sb_bgl_lock(sbi, block_group)); | |
687 | - percpu_counter_add(&sbi->s_freeblocks_counter, count); | |
688 | + percpu_counter_add(&sbi->s_freeblocks_counter, blocks_freed); | |
689 | ||
690 | if (sbi->s_log_groups_per_flex) { | |
691 | ext4_group_t flex_group = ext4_flex_group(sbi, block_group); | |
692 | spin_lock(sb_bgl_lock(sbi, flex_group)); | |
693 | - sbi->s_flex_groups[flex_group].free_blocks += count; | |
694 | + sbi->s_flex_groups[flex_group].free_blocks += blocks_freed; | |
695 | spin_unlock(sb_bgl_lock(sbi, flex_group)); | |
696 | } | |
697 | + /* | |
698 | + * request to reload the buddy with the | |
699 | + * new bitmap information | |
700 | + */ | |
701 | + set_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &(grp->bb_state)); | |
702 | + ext4_mb_update_group_info(grp, blocks_freed); | |
703 | + up_write(&grp->alloc_sem); | |
704 | ||
705 | /* We dirtied the bitmap block */ | |
706 | BUFFER_TRACE(bitmap_bh, "dirtied bitmap block"); | |
707 | @@ -821,15 +492,10 @@ do_more: | |
708 | /* And the group descriptor block */ | |
709 | BUFFER_TRACE(gd_bh, "dirtied group descriptor block"); | |
710 | ret = ext4_journal_dirty_metadata(handle, gd_bh); | |
711 | - if (!err) err = ret; | |
712 | - *pdquot_freed_blocks += group_freed; | |
713 | - | |
714 | - if (overflow && !err) { | |
715 | - block += count; | |
716 | - count = overflow; | |
717 | - goto do_more; | |
718 | - } | |
719 | + if (!err) | |
720 | + err = ret; | |
721 | sb->s_dirt = 1; | |
722 | + | |
723 | error_return: | |
724 | brelse(bitmap_bh); | |
725 | ext4_std_error(sb, err); | |
726 | @@ -848,792 +514,86 @@ void ext4_free_blocks(handle_t *handle, | |
727 | ext4_fsblk_t block, unsigned long count, | |
728 | int metadata) | |
729 | { | |
730 | - struct super_block * sb; | |
731 | + struct super_block *sb; | |
732 | unsigned long dquot_freed_blocks; | |
733 | ||
734 | /* this isn't the right place to decide whether block is metadata | |
735 | * inode.c/extents.c knows better, but for safety ... */ | |
736 | - if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode) || | |
737 | - ext4_should_journal_data(inode)) | |
738 | + if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) | |
739 | + metadata = 1; | |
740 | + | |
741 | + /* We need to make sure we don't reuse | |
742 | + * block released untill the transaction commit. | |
743 | + * writeback mode have weak data consistency so | |
744 | + * don't force data as metadata when freeing block | |
745 | + * for writeback mode. | |
746 | + */ | |
747 | + if (metadata == 0 && !ext4_should_writeback_data(inode)) | |
748 | metadata = 1; | |
749 | ||
750 | sb = inode->i_sb; | |
751 | ||
752 | - if (!test_opt(sb, MBALLOC) || !EXT4_SB(sb)->s_group_info) | |
753 | - ext4_free_blocks_sb(handle, sb, block, count, | |
754 | - &dquot_freed_blocks); | |
755 | - else | |
756 | - ext4_mb_free_blocks(handle, inode, block, count, | |
757 | - metadata, &dquot_freed_blocks); | |
758 | + ext4_mb_free_blocks(handle, inode, block, count, | |
759 | + metadata, &dquot_freed_blocks); | |
760 | if (dquot_freed_blocks) | |
761 | DQUOT_FREE_BLOCK(inode, dquot_freed_blocks); | |
762 | return; | |
763 | } | |
764 | ||
765 | /** | |
766 | - * ext4_test_allocatable() | |
767 | - * @nr: given allocation block group | |
768 | - * @bh: bufferhead contains the bitmap of the given block group | |
769 | - * | |
770 | - * For ext4 allocations, we must not reuse any blocks which are | |
771 | - * allocated in the bitmap buffer's "last committed data" copy. This | |
772 | - * prevents deletes from freeing up the page for reuse until we have | |
773 | - * committed the delete transaction. | |
774 | - * | |
775 | - * If we didn't do this, then deleting something and reallocating it as | |
776 | - * data would allow the old block to be overwritten before the | |
777 | - * transaction committed (because we force data to disk before commit). | |
778 | - * This would lead to corruption if we crashed between overwriting the | |
779 | - * data and committing the delete. | |
780 | - * | |
781 | - * @@@ We may want to make this allocation behaviour conditional on | |
782 | - * data-writes at some point, and disable it for metadata allocations or | |
783 | - * sync-data inodes. | |
784 | - */ | |
785 | -static int ext4_test_allocatable(ext4_grpblk_t nr, struct buffer_head *bh) | |
786 | -{ | |
787 | - int ret; | |
788 | - struct journal_head *jh = bh2jh(bh); | |
789 | - | |
790 | - if (ext4_test_bit(nr, bh->b_data)) | |
791 | - return 0; | |
792 | - | |
793 | - jbd_lock_bh_state(bh); | |
794 | - if (!jh->b_committed_data) | |
795 | - ret = 1; | |
796 | - else | |
797 | - ret = !ext4_test_bit(nr, jh->b_committed_data); | |
798 | - jbd_unlock_bh_state(bh); | |
799 | - return ret; | |
800 | -} | |
801 | - | |
802 | -/** | |
803 | - * bitmap_search_next_usable_block() | |
804 | - * @start: the starting block (group relative) of the search | |
805 | - * @bh: bufferhead contains the block group bitmap | |
806 | - * @maxblocks: the ending block (group relative) of the reservation | |
807 | - * | |
808 | - * The bitmap search --- search forward alternately through the actual | |
809 | - * bitmap on disk and the last-committed copy in journal, until we find a | |
810 | - * bit free in both bitmaps. | |
811 | - */ | |
812 | -static ext4_grpblk_t | |
813 | -bitmap_search_next_usable_block(ext4_grpblk_t start, struct buffer_head *bh, | |
814 | - ext4_grpblk_t maxblocks) | |
815 | -{ | |
816 | - ext4_grpblk_t next; | |
817 | - struct journal_head *jh = bh2jh(bh); | |
818 | - | |
819 | - while (start < maxblocks) { | |
820 | - next = ext4_find_next_zero_bit(bh->b_data, maxblocks, start); | |
821 | - if (next >= maxblocks) | |
822 | - return -1; | |
823 | - if (ext4_test_allocatable(next, bh)) | |
824 | - return next; | |
825 | - jbd_lock_bh_state(bh); | |
826 | - if (jh->b_committed_data) | |
827 | - start = ext4_find_next_zero_bit(jh->b_committed_data, | |
828 | - maxblocks, next); | |
829 | - jbd_unlock_bh_state(bh); | |
830 | - } | |
831 | - return -1; | |
832 | -} | |
833 | - | |
834 | -/** | |
835 | - * find_next_usable_block() | |
836 | - * @start: the starting block (group relative) to find next | |
837 | - * allocatable block in bitmap. | |
838 | - * @bh: bufferhead contains the block group bitmap | |
839 | - * @maxblocks: the ending block (group relative) for the search | |
840 | - * | |
841 | - * Find an allocatable block in a bitmap. We honor both the bitmap and | |
842 | - * its last-committed copy (if that exists), and perform the "most | |
843 | - * appropriate allocation" algorithm of looking for a free block near | |
844 | - * the initial goal; then for a free byte somewhere in the bitmap; then | |
845 | - * for any free bit in the bitmap. | |
846 | - */ | |
847 | -static ext4_grpblk_t | |
848 | -find_next_usable_block(ext4_grpblk_t start, struct buffer_head *bh, | |
849 | - ext4_grpblk_t maxblocks) | |
850 | -{ | |
851 | - ext4_grpblk_t here, next; | |
852 | - char *p, *r; | |
853 | - | |
854 | - if (start > 0) { | |
855 | - /* | |
856 | - * The goal was occupied; search forward for a free | |
857 | - * block within the next XX blocks. | |
858 | - * | |
859 | - * end_goal is more or less random, but it has to be | |
860 | - * less than EXT4_BLOCKS_PER_GROUP. Aligning up to the | |
861 | - * next 64-bit boundary is simple.. | |
862 | - */ | |
863 | - ext4_grpblk_t end_goal = (start + 63) & ~63; | |
864 | - if (end_goal > maxblocks) | |
865 | - end_goal = maxblocks; | |
866 | - here = ext4_find_next_zero_bit(bh->b_data, end_goal, start); | |
867 | - if (here < end_goal && ext4_test_allocatable(here, bh)) | |
868 | - return here; | |
869 | - ext4_debug("Bit not found near goal\n"); | |
870 | - } | |
871 | - | |
872 | - here = start; | |
873 | - if (here < 0) | |
874 | - here = 0; | |
875 | - | |
876 | - p = ((char *)bh->b_data) + (here >> 3); | |
877 | - r = memscan(p, 0, ((maxblocks + 7) >> 3) - (here >> 3)); | |
878 | - next = (r - ((char *)bh->b_data)) << 3; | |
879 | - | |
880 | - if (next < maxblocks && next >= start && ext4_test_allocatable(next, bh)) | |
881 | - return next; | |
882 | - | |
883 | - /* | |
884 | - * The bitmap search --- search forward alternately through the actual | |
885 | - * bitmap and the last-committed copy until we find a bit free in | |
886 | - * both | |
887 | - */ | |
888 | - here = bitmap_search_next_usable_block(here, bh, maxblocks); | |
889 | - return here; | |
890 | -} | |
891 | - | |
892 | -/** | |
893 | - * claim_block() | |
894 | - * @block: the free block (group relative) to allocate | |
895 | - * @bh: the bufferhead containts the block group bitmap | |
896 | - * | |
897 | - * We think we can allocate this block in this bitmap. Try to set the bit. | |
898 | - * If that succeeds then check that nobody has allocated and then freed the | |
899 | - * block since we saw that is was not marked in b_committed_data. If it _was_ | |
900 | - * allocated and freed then clear the bit in the bitmap again and return | |
901 | - * zero (failure). | |
902 | - */ | |
903 | -static inline int | |
904 | -claim_block(spinlock_t *lock, ext4_grpblk_t block, struct buffer_head *bh) | |
905 | -{ | |
906 | - struct journal_head *jh = bh2jh(bh); | |
907 | - int ret; | |
908 | - | |
909 | - if (ext4_set_bit_atomic(lock, block, bh->b_data)) | |
910 | - return 0; | |
911 | - jbd_lock_bh_state(bh); | |
912 | - if (jh->b_committed_data && ext4_test_bit(block,jh->b_committed_data)) { | |
913 | - ext4_clear_bit_atomic(lock, block, bh->b_data); | |
914 | - ret = 0; | |
915 | - } else { | |
916 | - ret = 1; | |
917 | - } | |
918 | - jbd_unlock_bh_state(bh); | |
919 | - return ret; | |
920 | -} | |
921 | - | |
922 | -/** | |
923 | - * ext4_try_to_allocate() | |
924 | - * @sb: superblock | |
925 | - * @handle: handle to this transaction | |
926 | - * @group: given allocation block group | |
927 | - * @bitmap_bh: bufferhead holds the block bitmap | |
928 | - * @grp_goal: given target block within the group | |
929 | - * @count: target number of blocks to allocate | |
930 | - * @my_rsv: reservation window | |
931 | - * | |
932 | - * Attempt to allocate blocks within a give range. Set the range of allocation | |
933 | - * first, then find the first free bit(s) from the bitmap (within the range), | |
934 | - * and at last, allocate the blocks by claiming the found free bit as allocated. | |
935 | - * | |
936 | - * To set the range of this allocation: | |
937 | - * if there is a reservation window, only try to allocate block(s) from the | |
938 | - * file's own reservation window; | |
939 | - * Otherwise, the allocation range starts from the give goal block, ends at | |
940 | - * the block group's last block. | |
941 | - * | |
942 | - * If we failed to allocate the desired block then we may end up crossing to a | |
943 | - * new bitmap. In that case we must release write access to the old one via | |
944 | - * ext4_journal_release_buffer(), else we'll run out of credits. | |
945 | - */ | |
946 | -static ext4_grpblk_t | |
947 | -ext4_try_to_allocate(struct super_block *sb, handle_t *handle, | |
948 | - ext4_group_t group, struct buffer_head *bitmap_bh, | |
949 | - ext4_grpblk_t grp_goal, unsigned long *count, | |
950 | - struct ext4_reserve_window *my_rsv) | |
951 | -{ | |
952 | - ext4_fsblk_t group_first_block; | |
953 | - ext4_grpblk_t start, end; | |
954 | - unsigned long num = 0; | |
955 | - | |
956 | - /* we do allocation within the reservation window if we have a window */ | |
957 | - if (my_rsv) { | |
958 | - group_first_block = ext4_group_first_block_no(sb, group); | |
959 | - if (my_rsv->_rsv_start >= group_first_block) | |
960 | - start = my_rsv->_rsv_start - group_first_block; | |
961 | - else | |
962 | - /* reservation window cross group boundary */ | |
963 | - start = 0; | |
964 | - end = my_rsv->_rsv_end - group_first_block + 1; | |
965 | - if (end > EXT4_BLOCKS_PER_GROUP(sb)) | |
966 | - /* reservation window crosses group boundary */ | |
967 | - end = EXT4_BLOCKS_PER_GROUP(sb); | |
968 | - if ((start <= grp_goal) && (grp_goal < end)) | |
969 | - start = grp_goal; | |
970 | - else | |
971 | - grp_goal = -1; | |
972 | - } else { | |
973 | - if (grp_goal > 0) | |
974 | - start = grp_goal; | |
975 | - else | |
976 | - start = 0; | |
977 | - end = EXT4_BLOCKS_PER_GROUP(sb); | |
978 | - } | |
979 | - | |
980 | - BUG_ON(start > EXT4_BLOCKS_PER_GROUP(sb)); | |
981 | - | |
982 | -repeat: | |
983 | - if (grp_goal < 0 || !ext4_test_allocatable(grp_goal, bitmap_bh)) { | |
984 | - grp_goal = find_next_usable_block(start, bitmap_bh, end); | |
985 | - if (grp_goal < 0) | |
986 | - goto fail_access; | |
987 | - if (!my_rsv) { | |
988 | - int i; | |
989 | - | |
990 | - for (i = 0; i < 7 && grp_goal > start && | |
991 | - ext4_test_allocatable(grp_goal - 1, | |
992 | - bitmap_bh); | |
993 | - i++, grp_goal--) | |
994 | - ; | |
995 | - } | |
996 | - } | |
997 | - start = grp_goal; | |
998 | - | |
999 | - if (!claim_block(sb_bgl_lock(EXT4_SB(sb), group), | |
1000 | - grp_goal, bitmap_bh)) { | |
1001 | - /* | |
1002 | - * The block was allocated by another thread, or it was | |
1003 | - * allocated and then freed by another thread | |
1004 | - */ | |
1005 | - start++; | |
1006 | - grp_goal++; | |
1007 | - if (start >= end) | |
1008 | - goto fail_access; | |
1009 | - goto repeat; | |
1010 | - } | |
1011 | - num++; | |
1012 | - grp_goal++; | |
1013 | - while (num < *count && grp_goal < end | |
1014 | - && ext4_test_allocatable(grp_goal, bitmap_bh) | |
1015 | - && claim_block(sb_bgl_lock(EXT4_SB(sb), group), | |
1016 | - grp_goal, bitmap_bh)) { | |
1017 | - num++; | |
1018 | - grp_goal++; | |
1019 | - } | |
1020 | - *count = num; | |
1021 | - return grp_goal - num; | |
1022 | -fail_access: | |
1023 | - *count = num; | |
1024 | - return -1; | |
1025 | -} | |
1026 | - | |
1027 | -/** | |
1028 | - * find_next_reservable_window(): | |
1029 | - * find a reservable space within the given range. | |
1030 | - * It does not allocate the reservation window for now: | |
1031 | - * alloc_new_reservation() will do the work later. | |
1032 | - * | |
1033 | - * @search_head: the head of the searching list; | |
1034 | - * This is not necessarily the list head of the whole filesystem | |
1035 | - * | |
1036 | - * We have both head and start_block to assist the search | |
1037 | - * for the reservable space. The list starts from head, | |
1038 | - * but we will shift to the place where start_block is, | |
1039 | - * then start from there, when looking for a reservable space. | |
1040 | - * | |
1041 | - * @size: the target new reservation window size | |
1042 | - * | |
1043 | - * @group_first_block: the first block we consider to start | |
1044 | - * the real search from | |
1045 | - * | |
1046 | - * @last_block: | |
1047 | - * the maximum block number that our goal reservable space | |
1048 | - * could start from. This is normally the last block in this | |
1049 | - * group. The search will end when we found the start of next | |
1050 | - * possible reservable space is out of this boundary. | |
1051 | - * This could handle the cross boundary reservation window | |
1052 | - * request. | |
1053 | - * | |
1054 | - * basically we search from the given range, rather than the whole | |
1055 | - * reservation double linked list, (start_block, last_block) | |
1056 | - * to find a free region that is of my size and has not | |
1057 | - * been reserved. | |
1058 | - * | |
1059 | - */ | |
1060 | -static int find_next_reservable_window( | |
1061 | - struct ext4_reserve_window_node *search_head, | |
1062 | - struct ext4_reserve_window_node *my_rsv, | |
1063 | - struct super_block * sb, | |
1064 | - ext4_fsblk_t start_block, | |
1065 | - ext4_fsblk_t last_block) | |
1066 | -{ | |
1067 | - struct rb_node *next; | |
1068 | - struct ext4_reserve_window_node *rsv, *prev; | |
1069 | - ext4_fsblk_t cur; | |
1070 | - int size = my_rsv->rsv_goal_size; | |
1071 | - | |
1072 | - /* TODO: make the start of the reservation window byte-aligned */ | |
1073 | - /* cur = *start_block & ~7;*/ | |
1074 | - cur = start_block; | |
1075 | - rsv = search_head; | |
1076 | - if (!rsv) | |
1077 | - return -1; | |
1078 | - | |
1079 | - while (1) { | |
1080 | - if (cur <= rsv->rsv_end) | |
1081 | - cur = rsv->rsv_end + 1; | |
1082 | - | |
1083 | - /* TODO? | |
1084 | - * in the case we could not find a reservable space | |
1085 | - * that is what is expected, during the re-search, we could | |
1086 | - * remember what's the largest reservable space we could have | |
1087 | - * and return that one. | |
1088 | - * | |
1089 | - * For now it will fail if we could not find the reservable | |
1090 | - * space with expected-size (or more)... | |
1091 | - */ | |
1092 | - if (cur > last_block) | |
1093 | - return -1; /* fail */ | |
1094 | - | |
1095 | - prev = rsv; | |
1096 | - next = rb_next(&rsv->rsv_node); | |
1097 | - rsv = rb_entry(next,struct ext4_reserve_window_node,rsv_node); | |
1098 | - | |
1099 | - /* | |
1100 | - * Reached the last reservation, we can just append to the | |
1101 | - * previous one. | |
1102 | - */ | |
1103 | - if (!next) | |
1104 | - break; | |
1105 | - | |
1106 | - if (cur + size <= rsv->rsv_start) { | |
1107 | - /* | |
1108 | - * Found a reserveable space big enough. We could | |
1109 | - * have a reservation across the group boundary here | |
1110 | - */ | |
1111 | - break; | |
1112 | - } | |
1113 | - } | |
1114 | - /* | |
1115 | - * we come here either : | |
1116 | - * when we reach the end of the whole list, | |
1117 | - * and there is empty reservable space after last entry in the list. | |
1118 | - * append it to the end of the list. | |
1119 | - * | |
1120 | - * or we found one reservable space in the middle of the list, | |
1121 | - * return the reservation window that we could append to. | |
1122 | - * succeed. | |
1123 | - */ | |
1124 | - | |
1125 | - if ((prev != my_rsv) && (!rsv_is_empty(&my_rsv->rsv_window))) | |
1126 | - rsv_window_remove(sb, my_rsv); | |
1127 | - | |
1128 | - /* | |
1129 | - * Let's book the whole avaliable window for now. We will check the | |
1130 | - * disk bitmap later and then, if there are free blocks then we adjust | |
1131 | - * the window size if it's larger than requested. | |
1132 | - * Otherwise, we will remove this node from the tree next time | |
1133 | - * call find_next_reservable_window. | |
1134 | - */ | |
1135 | - my_rsv->rsv_start = cur; | |
1136 | - my_rsv->rsv_end = cur + size - 1; | |
1137 | - my_rsv->rsv_alloc_hit = 0; | |
1138 | - | |
1139 | - if (prev != my_rsv) | |
1140 | - ext4_rsv_window_add(sb, my_rsv); | |
1141 | - | |
1142 | - return 0; | |
1143 | -} | |
1144 | - | |
1145 | -/** | |
1146 | - * alloc_new_reservation()--allocate a new reservation window | |
1147 | - * | |
1148 | - * To make a new reservation, we search part of the filesystem | |
1149 | - * reservation list (the list that inside the group). We try to | |
1150 | - * allocate a new reservation window near the allocation goal, | |
1151 | - * or the beginning of the group, if there is no goal. | |
1152 | - * | |
1153 | - * We first find a reservable space after the goal, then from | |
1154 | - * there, we check the bitmap for the first free block after | |
1155 | - * it. If there is no free block until the end of group, then the | |
1156 | - * whole group is full, we failed. Otherwise, check if the free | |
1157 | - * block is inside the expected reservable space, if so, we | |
1158 | - * succeed. | |
1159 | - * If the first free block is outside the reservable space, then | |
1160 | - * start from the first free block, we search for next available | |
1161 | - * space, and go on. | |
1162 | - * | |
1163 | - * on succeed, a new reservation will be found and inserted into the list | |
1164 | - * It contains at least one free block, and it does not overlap with other | |
1165 | - * reservation windows. | |
1166 | - * | |
1167 | - * failed: we failed to find a reservation window in this group | |
1168 | - * | |
1169 | - * @rsv: the reservation | |
1170 | - * | |
1171 | - * @grp_goal: The goal (group-relative). It is where the search for a | |
1172 | - * free reservable space should start from. | |
1173 | - * if we have a grp_goal(grp_goal >0 ), then start from there, | |
1174 | - * no grp_goal(grp_goal = -1), we start from the first block | |
1175 | - * of the group. | |
1176 | - * | |
1177 | - * @sb: the super block | |
1178 | - * @group: the group we are trying to allocate in | |
1179 | - * @bitmap_bh: the block group block bitmap | |
1180 | + * ext4_has_free_blocks() | |
1181 | + * @sbi: in-core super block structure. | |
1182 | + * @nblocks: number of needed blocks | |
1183 | * | |
1184 | + * Check if filesystem has nblocks free & available for allocation. | |
1185 | + * On success return 1, return 0 on failure. | |
1186 | */ | |
1187 | -static int alloc_new_reservation(struct ext4_reserve_window_node *my_rsv, | |
1188 | - ext4_grpblk_t grp_goal, struct super_block *sb, | |
1189 | - ext4_group_t group, struct buffer_head *bitmap_bh) | |
1190 | +int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks) | |
1191 | { | |
1192 | - struct ext4_reserve_window_node *search_head; | |
1193 | - ext4_fsblk_t group_first_block, group_end_block, start_block; | |
1194 | - ext4_grpblk_t first_free_block; | |
1195 | - struct rb_root *fs_rsv_root = &EXT4_SB(sb)->s_rsv_window_root; | |
1196 | - unsigned long size; | |
1197 | - int ret; | |
1198 | - spinlock_t *rsv_lock = &EXT4_SB(sb)->s_rsv_window_lock; | |
1199 | - | |
1200 | - group_first_block = ext4_group_first_block_no(sb, group); | |
1201 | - group_end_block = group_first_block + (EXT4_BLOCKS_PER_GROUP(sb) - 1); | |
1202 | - | |
1203 | - if (grp_goal < 0) | |
1204 | - start_block = group_first_block; | |
1205 | - else | |
1206 | - start_block = grp_goal + group_first_block; | |
1207 | + s64 free_blocks, dirty_blocks, root_blocks; | |
1208 | + struct percpu_counter *fbc = &sbi->s_freeblocks_counter; | |
1209 | + struct percpu_counter *dbc = &sbi->s_dirtyblocks_counter; | |
1210 | ||
1211 | - size = my_rsv->rsv_goal_size; | |
1212 | + free_blocks = percpu_counter_read_positive(fbc); | |
1213 | + dirty_blocks = percpu_counter_read_positive(dbc); | |
1214 | + root_blocks = ext4_r_blocks_count(sbi->s_es); | |
1215 | ||
1216 | - if (!rsv_is_empty(&my_rsv->rsv_window)) { | |
1217 | - /* | |
1218 | - * if the old reservation is cross group boundary | |
1219 | - * and if the goal is inside the old reservation window, | |
1220 | - * we will come here when we just failed to allocate from | |
1221 | - * the first part of the window. We still have another part | |
1222 | - * that belongs to the next group. In this case, there is no | |
1223 | - * point to discard our window and try to allocate a new one | |
1224 | - * in this group(which will fail). we should | |
1225 | - * keep the reservation window, just simply move on. | |
1226 | - * | |
1227 | - * Maybe we could shift the start block of the reservation | |
1228 | - * window to the first block of next group. | |
1229 | - */ | |
1230 | - | |
1231 | - if ((my_rsv->rsv_start <= group_end_block) && | |
1232 | - (my_rsv->rsv_end > group_end_block) && | |
1233 | - (start_block >= my_rsv->rsv_start)) | |
1234 | - return -1; | |
1235 | - | |
1236 | - if ((my_rsv->rsv_alloc_hit > | |
1237 | - (my_rsv->rsv_end - my_rsv->rsv_start + 1) / 2)) { | |
1238 | - /* | |
1239 | - * if the previously allocation hit ratio is | |
1240 | - * greater than 1/2, then we double the size of | |
1241 | - * the reservation window the next time, | |
1242 | - * otherwise we keep the same size window | |
1243 | - */ | |
1244 | - size = size * 2; | |
1245 | - if (size > EXT4_MAX_RESERVE_BLOCKS) | |
1246 | - size = EXT4_MAX_RESERVE_BLOCKS; | |
1247 | - my_rsv->rsv_goal_size= size; | |
1248 | + if (free_blocks - (nblocks + root_blocks + dirty_blocks) < | |
1249 | + EXT4_FREEBLOCKS_WATERMARK) { | |
1250 | + free_blocks = percpu_counter_sum_positive(fbc); | |
1251 | + dirty_blocks = percpu_counter_sum_positive(dbc); | |
1252 | + if (dirty_blocks < 0) { | |
1253 | + printk(KERN_CRIT "Dirty block accounting " | |
1254 | + "went wrong %lld\n", | |
1255 | + (long long)dirty_blocks); | |
1256 | } | |
1257 | } | |
1258 | - | |
1259 | - spin_lock(rsv_lock); | |
1260 | - /* | |
1261 | - * shift the search start to the window near the goal block | |
1262 | - */ | |
1263 | - search_head = search_reserve_window(fs_rsv_root, start_block); | |
1264 | - | |
1265 | - /* | |
1266 | - * find_next_reservable_window() simply finds a reservable window | |
1267 | - * inside the given range(start_block, group_end_block). | |
1268 | - * | |
1269 | - * To make sure the reservation window has a free bit inside it, we | |
1270 | - * need to check the bitmap after we found a reservable window. | |
1271 | - */ | |
1272 | -retry: | |
1273 | - ret = find_next_reservable_window(search_head, my_rsv, sb, | |
1274 | - start_block, group_end_block); | |
1275 | - | |
1276 | - if (ret == -1) { | |
1277 | - if (!rsv_is_empty(&my_rsv->rsv_window)) | |
1278 | - rsv_window_remove(sb, my_rsv); | |
1279 | - spin_unlock(rsv_lock); | |
1280 | - return -1; | |
1281 | - } | |
1282 | - | |
1283 | - /* | |
1284 | - * On success, find_next_reservable_window() returns the | |
1285 | - * reservation window where there is a reservable space after it. | |
1286 | - * Before we reserve this reservable space, we need | |
1287 | - * to make sure there is at least a free block inside this region. | |
1288 | - * | |
1289 | - * searching the first free bit on the block bitmap and copy of | |
1290 | - * last committed bitmap alternatively, until we found a allocatable | |
1291 | - * block. Search start from the start block of the reservable space | |
1292 | - * we just found. | |
1293 | - */ | |
1294 | - spin_unlock(rsv_lock); | |
1295 | - first_free_block = bitmap_search_next_usable_block( | |
1296 | - my_rsv->rsv_start - group_first_block, | |
1297 | - bitmap_bh, group_end_block - group_first_block + 1); | |
1298 | - | |
1299 | - if (first_free_block < 0) { | |
1300 | - /* | |
1301 | - * no free block left on the bitmap, no point | |
1302 | - * to reserve the space. return failed. | |
1303 | - */ | |
1304 | - spin_lock(rsv_lock); | |
1305 | - if (!rsv_is_empty(&my_rsv->rsv_window)) | |
1306 | - rsv_window_remove(sb, my_rsv); | |
1307 | - spin_unlock(rsv_lock); | |
1308 | - return -1; /* failed */ | |
1309 | - } | |
1310 | - | |
1311 | - start_block = first_free_block + group_first_block; | |
1312 | - /* | |
1313 | - * check if the first free block is within the | |
1314 | - * free space we just reserved | |
1315 | - */ | |
1316 | - if (start_block >= my_rsv->rsv_start && start_block <= my_rsv->rsv_end) | |
1317 | - return 0; /* success */ | |
1318 | - /* | |
1319 | - * if the first free bit we found is out of the reservable space | |
1320 | - * continue search for next reservable space, | |
1321 | - * start from where the free block is, | |
1322 | - * we also shift the list head to where we stopped last time | |
1323 | - */ | |
1324 | - search_head = my_rsv; | |
1325 | - spin_lock(rsv_lock); | |
1326 | - goto retry; | |
1327 | -} | |
1328 | - | |
1329 | -/** | |
1330 | - * try_to_extend_reservation() | |
1331 | - * @my_rsv: given reservation window | |
1332 | - * @sb: super block | |
1333 | - * @size: the delta to extend | |
1334 | - * | |
1335 | - * Attempt to expand the reservation window large enough to have | |
1336 | - * required number of free blocks | |
1337 | - * | |
1338 | - * Since ext4_try_to_allocate() will always allocate blocks within | |
1339 | - * the reservation window range, if the window size is too small, | |
1340 | - * multiple blocks allocation has to stop at the end of the reservation | |
1341 | - * window. To make this more efficient, given the total number of | |
1342 | - * blocks needed and the current size of the window, we try to | |
1343 | - * expand the reservation window size if necessary on a best-effort | |
1344 | - * basis before ext4_new_blocks() tries to allocate blocks, | |
1345 | - */ | |
1346 | -static void try_to_extend_reservation(struct ext4_reserve_window_node *my_rsv, | |
1347 | - struct super_block *sb, int size) | |
1348 | -{ | |
1349 | - struct ext4_reserve_window_node *next_rsv; | |
1350 | - struct rb_node *next; | |
1351 | - spinlock_t *rsv_lock = &EXT4_SB(sb)->s_rsv_window_lock; | |
1352 | - | |
1353 | - if (!spin_trylock(rsv_lock)) | |
1354 | - return; | |
1355 | - | |
1356 | - next = rb_next(&my_rsv->rsv_node); | |
1357 | - | |
1358 | - if (!next) | |
1359 | - my_rsv->rsv_end += size; | |
1360 | - else { | |
1361 | - next_rsv = rb_entry(next, struct ext4_reserve_window_node, rsv_node); | |
1362 | - | |
1363 | - if ((next_rsv->rsv_start - my_rsv->rsv_end - 1) >= size) | |
1364 | - my_rsv->rsv_end += size; | |
1365 | - else | |
1366 | - my_rsv->rsv_end = next_rsv->rsv_start - 1; | |
1367 | - } | |
1368 | - spin_unlock(rsv_lock); | |
1369 | -} | |
1370 | - | |
1371 | -/** | |
1372 | - * ext4_try_to_allocate_with_rsv() | |
1373 | - * @sb: superblock | |
1374 | - * @handle: handle to this transaction | |
1375 | - * @group: given allocation block group | |
1376 | - * @bitmap_bh: bufferhead holds the block bitmap | |
1377 | - * @grp_goal: given target block within the group | |
1378 | - * @count: target number of blocks to allocate | |
1379 | - * @my_rsv: reservation window | |
1380 | - * @errp: pointer to store the error code | |
1381 | - * | |
1382 | - * This is the main function used to allocate a new block and its reservation | |
1383 | - * window. | |
1384 | - * | |
1385 | - * Each time when a new block allocation is need, first try to allocate from | |
1386 | - * its own reservation. If it does not have a reservation window, instead of | |
1387 | - * looking for a free bit on bitmap first, then look up the reservation list to | |
1388 | - * see if it is inside somebody else's reservation window, we try to allocate a | |
1389 | - * reservation window for it starting from the goal first. Then do the block | |
1390 | - * allocation within the reservation window. | |
1391 | - * | |
1392 | - * This will avoid keeping on searching the reservation list again and | |
1393 | - * again when somebody is looking for a free block (without | |
1394 | - * reservation), and there are lots of free blocks, but they are all | |
1395 | - * being reserved. | |
1396 | - * | |
1397 | - * We use a red-black tree for the per-filesystem reservation list. | |
1398 | - * | |
1399 | - */ | |
1400 | -static ext4_grpblk_t | |
1401 | -ext4_try_to_allocate_with_rsv(struct super_block *sb, handle_t *handle, | |
1402 | - ext4_group_t group, struct buffer_head *bitmap_bh, | |
1403 | - ext4_grpblk_t grp_goal, | |
1404 | - struct ext4_reserve_window_node * my_rsv, | |
1405 | - unsigned long *count, int *errp) | |
1406 | -{ | |
1407 | - ext4_fsblk_t group_first_block, group_last_block; | |
1408 | - ext4_grpblk_t ret = 0; | |
1409 | - int fatal; | |
1410 | - unsigned long num = *count; | |
1411 | - | |
1412 | - *errp = 0; | |
1413 | - | |
1414 | - /* | |
1415 | - * Make sure we use undo access for the bitmap, because it is critical | |
1416 | - * that we do the frozen_data COW on bitmap buffers in all cases even | |
1417 | - * if the buffer is in BJ_Forget state in the committing transaction. | |
1418 | + /* Check whether we have space after | |
1419 | + * accounting for current dirty blocks & root reserved blocks. | |
1420 | */ | |
1421 | - BUFFER_TRACE(bitmap_bh, "get undo access for new block"); | |
1422 | - fatal = ext4_journal_get_undo_access(handle, bitmap_bh); | |
1423 | - if (fatal) { | |
1424 | - *errp = fatal; | |
1425 | - return -1; | |
1426 | - } | |
1427 | - | |
1428 | - /* | |
1429 | - * we don't deal with reservation when | |
1430 | - * filesystem is mounted without reservation | |
1431 | - * or the file is not a regular file | |
1432 | - * or last attempt to allocate a block with reservation turned on failed | |
1433 | - */ | |
1434 | - if (my_rsv == NULL ) { | |
1435 | - ret = ext4_try_to_allocate(sb, handle, group, bitmap_bh, | |
1436 | - grp_goal, count, NULL); | |
1437 | - goto out; | |
1438 | - } | |
1439 | - /* | |
1440 | - * grp_goal is a group relative block number (if there is a goal) | |
1441 | - * 0 <= grp_goal < EXT4_BLOCKS_PER_GROUP(sb) | |
1442 | - * first block is a filesystem wide block number | |
1443 | - * first block is the block number of the first block in this group | |
1444 | - */ | |
1445 | - group_first_block = ext4_group_first_block_no(sb, group); | |
1446 | - group_last_block = group_first_block + (EXT4_BLOCKS_PER_GROUP(sb) - 1); | |
1447 | - | |
1448 | - /* | |
1449 | - * Basically we will allocate a new block from inode's reservation | |
1450 | - * window. | |
1451 | - * | |
1452 | - * We need to allocate a new reservation window, if: | |
1453 | - * a) inode does not have a reservation window; or | |
1454 | - * b) last attempt to allocate a block from existing reservation | |
1455 | - * failed; or | |
1456 | - * c) we come here with a goal and with a reservation window | |
1457 | - * | |
1458 | - * We do not need to allocate a new reservation window if we come here | |
1459 | - * at the beginning with a goal and the goal is inside the window, or | |
1460 | - * we don't have a goal but already have a reservation window. | |
1461 | - * then we could go to allocate from the reservation window directly. | |
1462 | - */ | |
1463 | - while (1) { | |
1464 | - if (rsv_is_empty(&my_rsv->rsv_window) || (ret < 0) || | |
1465 | - !goal_in_my_reservation(&my_rsv->rsv_window, | |
1466 | - grp_goal, group, sb)) { | |
1467 | - if (my_rsv->rsv_goal_size < *count) | |
1468 | - my_rsv->rsv_goal_size = *count; | |
1469 | - ret = alloc_new_reservation(my_rsv, grp_goal, sb, | |
1470 | - group, bitmap_bh); | |
1471 | - if (ret < 0) | |
1472 | - break; /* failed */ | |
1473 | - | |
1474 | - if (!goal_in_my_reservation(&my_rsv->rsv_window, | |
1475 | - grp_goal, group, sb)) | |
1476 | - grp_goal = -1; | |
1477 | - } else if (grp_goal >= 0) { | |
1478 | - int curr = my_rsv->rsv_end - | |
1479 | - (grp_goal + group_first_block) + 1; | |
1480 | - | |
1481 | - if (curr < *count) | |
1482 | - try_to_extend_reservation(my_rsv, sb, | |
1483 | - *count - curr); | |
1484 | - } | |
1485 | + if (free_blocks >= ((root_blocks + nblocks) + dirty_blocks)) | |
1486 | + return 1; | |
1487 | ||
1488 | - if ((my_rsv->rsv_start > group_last_block) || | |
1489 | - (my_rsv->rsv_end < group_first_block)) { | |
1490 | - rsv_window_dump(&EXT4_SB(sb)->s_rsv_window_root, 1); | |
1491 | - BUG(); | |
1492 | - } | |
1493 | - ret = ext4_try_to_allocate(sb, handle, group, bitmap_bh, | |
1494 | - grp_goal, &num, &my_rsv->rsv_window); | |
1495 | - if (ret >= 0) { | |
1496 | - my_rsv->rsv_alloc_hit += num; | |
1497 | - *count = num; | |
1498 | - break; /* succeed */ | |
1499 | - } | |
1500 | - num = *count; | |
1501 | - } | |
1502 | -out: | |
1503 | - if (ret >= 0) { | |
1504 | - BUFFER_TRACE(bitmap_bh, "journal_dirty_metadata for " | |
1505 | - "bitmap block"); | |
1506 | - fatal = ext4_journal_dirty_metadata(handle, bitmap_bh); | |
1507 | - if (fatal) { | |
1508 | - *errp = fatal; | |
1509 | - return -1; | |
1510 | - } | |
1511 | - return ret; | |
1512 | + /* Hm, nope. Are (enough) root reserved blocks available? */ | |
1513 | + if (sbi->s_resuid == current_fsuid() || | |
1514 | + ((sbi->s_resgid != 0) && in_group_p(sbi->s_resgid)) || | |
1515 | + capable(CAP_SYS_RESOURCE)) { | |
1516 | + if (free_blocks >= (nblocks + dirty_blocks)) | |
1517 | + return 1; | |
1518 | } | |
1519 | ||
1520 | - BUFFER_TRACE(bitmap_bh, "journal_release_buffer"); | |
1521 | - ext4_journal_release_buffer(handle, bitmap_bh); | |
1522 | - return ret; | |
1523 | + return 0; | |
1524 | } | |
1525 | ||
1526 | -/** | |
1527 | - * ext4_has_free_blocks() | |
1528 | - * @sbi: in-core super block structure. | |
1529 | - * @nblocks: number of neeed blocks | |
1530 | - * | |
1531 | - * Check if filesystem has free blocks available for allocation. | |
1532 | - * Return the number of blocks avaible for allocation for this request | |
1533 | - * On success, return nblocks | |
1534 | - */ | |
1535 | -ext4_fsblk_t ext4_has_free_blocks(struct ext4_sb_info *sbi, | |
1536 | - ext4_fsblk_t nblocks) | |
1537 | +int ext4_claim_free_blocks(struct ext4_sb_info *sbi, | |
1538 | + s64 nblocks) | |
1539 | { | |
1540 | - ext4_fsblk_t free_blocks; | |
1541 | - ext4_fsblk_t root_blocks = 0; | |
1542 | - | |
1543 | - free_blocks = percpu_counter_read_positive(&sbi->s_freeblocks_counter); | |
1544 | - | |
1545 | - if (!capable(CAP_SYS_RESOURCE) && | |
1546 | - sbi->s_resuid != current->fsuid && | |
1547 | - (sbi->s_resgid == 0 || !in_group_p(sbi->s_resgid))) | |
1548 | - root_blocks = ext4_r_blocks_count(sbi->s_es); | |
1549 | -#ifdef CONFIG_SMP | |
1550 | - if (free_blocks - root_blocks < FBC_BATCH) | |
1551 | - free_blocks = | |
1552 | - percpu_counter_sum(&sbi->s_freeblocks_counter); | |
1553 | -#endif | |
1554 | - if (free_blocks <= root_blocks) | |
1555 | - /* we don't have free space */ | |
1556 | + if (ext4_has_free_blocks(sbi, nblocks)) { | |
1557 | + percpu_counter_add(&sbi->s_dirtyblocks_counter, nblocks); | |
1558 | return 0; | |
1559 | - if (free_blocks - root_blocks < nblocks) | |
1560 | - return free_blocks - root_blocks; | |
1561 | - return nblocks; | |
1562 | - } | |
1563 | - | |
1564 | + } else | |
1565 | + return -ENOSPC; | |
1566 | +} | |
1567 | ||
1568 | /** | |
1569 | * ext4_should_retry_alloc() | |
1570 | @@ -1657,402 +617,45 @@ int ext4_should_retry_alloc(struct super | |
1571 | return jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal); | |
1572 | } | |
1573 | ||
1574 | -/** | |
1575 | - * ext4_old_new_blocks() -- core block bitmap based block allocation function | |
1576 | - * | |
1577 | - * @handle: handle to this transaction | |
1578 | - * @inode: file inode | |
1579 | - * @goal: given target block(filesystem wide) | |
1580 | - * @count: target number of blocks to allocate | |
1581 | - * @errp: error code | |
1582 | - * | |
1583 | - * ext4_old_new_blocks uses a goal block to assist allocation and look up | |
1584 | - * the block bitmap directly to do block allocation. It tries to | |
1585 | - * allocate block(s) from the block group contains the goal block first. If | |
1586 | - * that fails, it will try to allocate block(s) from other block groups | |
1587 | - * without any specific goal block. | |
1588 | - * | |
1589 | - * This function is called when -o nomballoc mount option is enabled | |
1590 | - * | |
1591 | - */ | |
1592 | -ext4_fsblk_t ext4_old_new_blocks(handle_t *handle, struct inode *inode, | |
1593 | - ext4_fsblk_t goal, unsigned long *count, int *errp) | |
1594 | -{ | |
1595 | - struct buffer_head *bitmap_bh = NULL; | |
1596 | - struct buffer_head *gdp_bh; | |
1597 | - ext4_group_t group_no; | |
1598 | - ext4_group_t goal_group; | |
1599 | - ext4_grpblk_t grp_target_blk; /* blockgroup relative goal block */ | |
1600 | - ext4_grpblk_t grp_alloc_blk; /* blockgroup-relative allocated block*/ | |
1601 | - ext4_fsblk_t ret_block; /* filesyetem-wide allocated block */ | |
1602 | - ext4_group_t bgi; /* blockgroup iteration index */ | |
1603 | - int fatal = 0, err; | |
1604 | - int performed_allocation = 0; | |
1605 | - ext4_grpblk_t free_blocks; /* number of free blocks in a group */ | |
1606 | - struct super_block *sb; | |
1607 | - struct ext4_group_desc *gdp; | |
1608 | - struct ext4_super_block *es; | |
1609 | - struct ext4_sb_info *sbi; | |
1610 | - struct ext4_reserve_window_node *my_rsv = NULL; | |
1611 | - struct ext4_block_alloc_info *block_i; | |
1612 | - unsigned short windowsz = 0; | |
1613 | - ext4_group_t ngroups; | |
1614 | - unsigned long num = *count; | |
1615 | - | |
1616 | - sb = inode->i_sb; | |
1617 | - if (!sb) { | |
1618 | - *errp = -ENODEV; | |
1619 | - printk("ext4_new_block: nonexistent device"); | |
1620 | - return 0; | |
1621 | - } | |
1622 | - | |
1623 | - sbi = EXT4_SB(sb); | |
1624 | - if (!EXT4_I(inode)->i_delalloc_reserved_flag) { | |
1625 | - /* | |
1626 | - * With delalloc we already reserved the blocks | |
1627 | - */ | |
1628 | - *count = ext4_has_free_blocks(sbi, *count); | |
1629 | - } | |
1630 | - if (*count == 0) { | |
1631 | - *errp = -ENOSPC; | |
1632 | - return 0; /*return with ENOSPC error */ | |
1633 | - } | |
1634 | - num = *count; | |
1635 | - | |
1636 | - /* | |
1637 | - * Check quota for allocation of this block. | |
1638 | - */ | |
1639 | - if (DQUOT_ALLOC_BLOCK(inode, num)) { | |
1640 | - *errp = -EDQUOT; | |
1641 | - return 0; | |
1642 | - } | |
1643 | - | |
1644 | - sbi = EXT4_SB(sb); | |
1645 | - es = EXT4_SB(sb)->s_es; | |
1646 | - ext4_debug("goal=%llu.\n", goal); | |
1647 | - /* | |
1648 | - * Allocate a block from reservation only when | |
1649 | - * filesystem is mounted with reservation(default,-o reservation), and | |
1650 | - * it's a regular file, and | |
1651 | - * the desired window size is greater than 0 (One could use ioctl | |
1652 | - * command EXT4_IOC_SETRSVSZ to set the window size to 0 to turn off | |
1653 | - * reservation on that particular file) | |
1654 | - */ | |
1655 | - block_i = EXT4_I(inode)->i_block_alloc_info; | |
1656 | - if (block_i && ((windowsz = block_i->rsv_window_node.rsv_goal_size) > 0)) | |
1657 | - my_rsv = &block_i->rsv_window_node; | |
1658 | - | |
1659 | - /* | |
1660 | - * First, test whether the goal block is free. | |
1661 | - */ | |
1662 | - if (goal < le32_to_cpu(es->s_first_data_block) || | |
1663 | - goal >= ext4_blocks_count(es)) | |
1664 | - goal = le32_to_cpu(es->s_first_data_block); | |
1665 | - ext4_get_group_no_and_offset(sb, goal, &group_no, &grp_target_blk); | |
1666 | - goal_group = group_no; | |
1667 | -retry_alloc: | |
1668 | - gdp = ext4_get_group_desc(sb, group_no, &gdp_bh); | |
1669 | - if (!gdp) | |
1670 | - goto io_error; | |
1671 | - | |
1672 | - free_blocks = le16_to_cpu(gdp->bg_free_blocks_count); | |
1673 | - /* | |
1674 | - * if there is not enough free blocks to make a new resevation | |
1675 | - * turn off reservation for this allocation | |
1676 | - */ | |
1677 | - if (my_rsv && (free_blocks < windowsz) | |
1678 | - && (rsv_is_empty(&my_rsv->rsv_window))) | |
1679 | - my_rsv = NULL; | |
1680 | - | |
1681 | - if (free_blocks > 0) { | |
1682 | - bitmap_bh = ext4_read_block_bitmap(sb, group_no); | |
1683 | - if (!bitmap_bh) | |
1684 | - goto io_error; | |
1685 | - grp_alloc_blk = ext4_try_to_allocate_with_rsv(sb, handle, | |
1686 | - group_no, bitmap_bh, grp_target_blk, | |
1687 | - my_rsv, &num, &fatal); | |
1688 | - if (fatal) | |
1689 | - goto out; | |
1690 | - if (grp_alloc_blk >= 0) | |
1691 | - goto allocated; | |
1692 | - } | |
1693 | - | |
1694 | - ngroups = EXT4_SB(sb)->s_groups_count; | |
1695 | - smp_rmb(); | |
1696 | - | |
1697 | - /* | |
1698 | - * Now search the rest of the groups. We assume that | |
1699 | - * group_no and gdp correctly point to the last group visited. | |
1700 | - */ | |
1701 | - for (bgi = 0; bgi < ngroups; bgi++) { | |
1702 | - group_no++; | |
1703 | - if (group_no >= ngroups) | |
1704 | - group_no = 0; | |
1705 | - gdp = ext4_get_group_desc(sb, group_no, &gdp_bh); | |
1706 | - if (!gdp) | |
1707 | - goto io_error; | |
1708 | - free_blocks = le16_to_cpu(gdp->bg_free_blocks_count); | |
1709 | - /* | |
1710 | - * skip this group if the number of | |
1711 | - * free blocks is less than half of the reservation | |
1712 | - * window size. | |
1713 | - */ | |
1714 | - if (free_blocks <= (windowsz/2)) | |
1715 | - continue; | |
1716 | - | |
1717 | - brelse(bitmap_bh); | |
1718 | - bitmap_bh = ext4_read_block_bitmap(sb, group_no); | |
1719 | - if (!bitmap_bh) | |
1720 | - goto io_error; | |
1721 | - /* | |
1722 | - * try to allocate block(s) from this group, without a goal(-1). | |
1723 | - */ | |
1724 | - grp_alloc_blk = ext4_try_to_allocate_with_rsv(sb, handle, | |
1725 | - group_no, bitmap_bh, -1, my_rsv, | |
1726 | - &num, &fatal); | |
1727 | - if (fatal) | |
1728 | - goto out; | |
1729 | - if (grp_alloc_blk >= 0) | |
1730 | - goto allocated; | |
1731 | - } | |
1732 | - /* | |
1733 | - * We may end up a bogus ealier ENOSPC error due to | |
1734 | - * filesystem is "full" of reservations, but | |
1735 | - * there maybe indeed free blocks avaliable on disk | |
1736 | - * In this case, we just forget about the reservations | |
1737 | - * just do block allocation as without reservations. | |
1738 | - */ | |
1739 | - if (my_rsv) { | |
1740 | - my_rsv = NULL; | |
1741 | - windowsz = 0; | |
1742 | - group_no = goal_group; | |
1743 | - goto retry_alloc; | |
1744 | - } | |
1745 | - /* No space left on the device */ | |
1746 | - *errp = -ENOSPC; | |
1747 | - goto out; | |
1748 | - | |
1749 | -allocated: | |
1750 | - | |
1751 | - ext4_debug("using block group %lu(%d)\n", | |
1752 | - group_no, gdp->bg_free_blocks_count); | |
1753 | - | |
1754 | - BUFFER_TRACE(gdp_bh, "get_write_access"); | |
1755 | - fatal = ext4_journal_get_write_access(handle, gdp_bh); | |
1756 | - if (fatal) | |
1757 | - goto out; | |
1758 | - | |
1759 | - ret_block = grp_alloc_blk + ext4_group_first_block_no(sb, group_no); | |
1760 | - | |
1761 | - if (in_range(ext4_block_bitmap(sb, gdp), ret_block, num) || | |
1762 | - in_range(ext4_inode_bitmap(sb, gdp), ret_block, num) || | |
1763 | - in_range(ret_block, ext4_inode_table(sb, gdp), | |
1764 | - EXT4_SB(sb)->s_itb_per_group) || | |
1765 | - in_range(ret_block + num - 1, ext4_inode_table(sb, gdp), | |
1766 | - EXT4_SB(sb)->s_itb_per_group)) { | |
1767 | - ext4_error(sb, "ext4_new_block", | |
1768 | - "Allocating block in system zone - " | |
1769 | - "blocks from %llu, length %lu", | |
1770 | - ret_block, num); | |
1771 | - /* | |
1772 | - * claim_block marked the blocks we allocated | |
1773 | - * as in use. So we may want to selectively | |
1774 | - * mark some of the blocks as free | |
1775 | - */ | |
1776 | - goto retry_alloc; | |
1777 | - } | |
1778 | - | |
1779 | - performed_allocation = 1; | |
1780 | - | |
1781 | -#ifdef CONFIG_JBD2_DEBUG | |
1782 | - { | |
1783 | - struct buffer_head *debug_bh; | |
1784 | - | |
1785 | - /* Record bitmap buffer state in the newly allocated block */ | |
1786 | - debug_bh = sb_find_get_block(sb, ret_block); | |
1787 | - if (debug_bh) { | |
1788 | - BUFFER_TRACE(debug_bh, "state when allocated"); | |
1789 | - BUFFER_TRACE2(debug_bh, bitmap_bh, "bitmap state"); | |
1790 | - brelse(debug_bh); | |
1791 | - } | |
1792 | - } | |
1793 | - jbd_lock_bh_state(bitmap_bh); | |
1794 | - spin_lock(sb_bgl_lock(sbi, group_no)); | |
1795 | - if (buffer_jbd(bitmap_bh) && bh2jh(bitmap_bh)->b_committed_data) { | |
1796 | - int i; | |
1797 | - | |
1798 | - for (i = 0; i < num; i++) { | |
1799 | - if (ext4_test_bit(grp_alloc_blk+i, | |
1800 | - bh2jh(bitmap_bh)->b_committed_data)) { | |
1801 | - printk("%s: block was unexpectedly set in " | |
1802 | - "b_committed_data\n", __func__); | |
1803 | - } | |
1804 | - } | |
1805 | - } | |
1806 | - ext4_debug("found bit %d\n", grp_alloc_blk); | |
1807 | - spin_unlock(sb_bgl_lock(sbi, group_no)); | |
1808 | - jbd_unlock_bh_state(bitmap_bh); | |
1809 | -#endif | |
1810 | - | |
1811 | - if (ret_block + num - 1 >= ext4_blocks_count(es)) { | |
1812 | - ext4_error(sb, "ext4_new_block", | |
1813 | - "block(%llu) >= blocks count(%llu) - " | |
1814 | - "block_group = %lu, es == %p ", ret_block, | |
1815 | - ext4_blocks_count(es), group_no, es); | |
1816 | - goto out; | |
1817 | - } | |
1818 | - | |
1819 | - /* | |
1820 | - * It is up to the caller to add the new buffer to a journal | |
1821 | - * list of some description. We don't know in advance whether | |
1822 | - * the caller wants to use it as metadata or data. | |
1823 | - */ | |
1824 | - spin_lock(sb_bgl_lock(sbi, group_no)); | |
1825 | - if (gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) | |
1826 | - gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT); | |
1827 | - le16_add_cpu(&gdp->bg_free_blocks_count, -num); | |
1828 | - gdp->bg_checksum = ext4_group_desc_csum(sbi, group_no, gdp); | |
1829 | - spin_unlock(sb_bgl_lock(sbi, group_no)); | |
1830 | - if (!EXT4_I(inode)->i_delalloc_reserved_flag) | |
1831 | - percpu_counter_sub(&sbi->s_freeblocks_counter, num); | |
1832 | - | |
1833 | - if (sbi->s_log_groups_per_flex) { | |
1834 | - ext4_group_t flex_group = ext4_flex_group(sbi, group_no); | |
1835 | - spin_lock(sb_bgl_lock(sbi, flex_group)); | |
1836 | - sbi->s_flex_groups[flex_group].free_blocks -= num; | |
1837 | - spin_unlock(sb_bgl_lock(sbi, flex_group)); | |
1838 | - } | |
1839 | - | |
1840 | - BUFFER_TRACE(gdp_bh, "journal_dirty_metadata for group descriptor"); | |
1841 | - err = ext4_journal_dirty_metadata(handle, gdp_bh); | |
1842 | - if (!fatal) | |
1843 | - fatal = err; | |
1844 | - | |
1845 | - sb->s_dirt = 1; | |
1846 | - if (fatal) | |
1847 | - goto out; | |
1848 | - | |
1849 | - *errp = 0; | |
1850 | - brelse(bitmap_bh); | |
1851 | - DQUOT_FREE_BLOCK(inode, *count-num); | |
1852 | - *count = num; | |
1853 | - return ret_block; | |
1854 | - | |
1855 | -io_error: | |
1856 | - *errp = -EIO; | |
1857 | -out: | |
1858 | - if (fatal) { | |
1859 | - *errp = fatal; | |
1860 | - ext4_std_error(sb, fatal); | |
1861 | - } | |
1862 | - /* | |
1863 | - * Undo the block allocation | |
1864 | - */ | |
1865 | - if (!performed_allocation) | |
1866 | - DQUOT_FREE_BLOCK(inode, *count); | |
1867 | - brelse(bitmap_bh); | |
1868 | - return 0; | |
1869 | -} | |
1870 | - | |
1871 | -#define EXT4_META_BLOCK 0x1 | |
1872 | - | |
1873 | -static ext4_fsblk_t do_blk_alloc(handle_t *handle, struct inode *inode, | |
1874 | - ext4_lblk_t iblock, ext4_fsblk_t goal, | |
1875 | - unsigned long *count, int *errp, int flags) | |
1876 | -{ | |
1877 | - struct ext4_allocation_request ar; | |
1878 | - ext4_fsblk_t ret; | |
1879 | - | |
1880 | - if (!test_opt(inode->i_sb, MBALLOC)) { | |
1881 | - return ext4_old_new_blocks(handle, inode, goal, count, errp); | |
1882 | - } | |
1883 | - | |
1884 | - memset(&ar, 0, sizeof(ar)); | |
1885 | - /* Fill with neighbour allocated blocks */ | |
1886 | - | |
1887 | - ar.inode = inode; | |
1888 | - ar.goal = goal; | |
1889 | - ar.len = *count; | |
1890 | - ar.logical = iblock; | |
1891 | - | |
1892 | - if (S_ISREG(inode->i_mode) && !(flags & EXT4_META_BLOCK)) | |
1893 | - /* enable in-core preallocation for data block allocation */ | |
1894 | - ar.flags = EXT4_MB_HINT_DATA; | |
1895 | - else | |
1896 | - /* disable in-core preallocation for non-regular files */ | |
1897 | - ar.flags = 0; | |
1898 | - | |
1899 | - ret = ext4_mb_new_blocks(handle, &ar, errp); | |
1900 | - *count = ar.len; | |
1901 | - return ret; | |
1902 | -} | |
1903 | - | |
1904 | /* | |
1905 | * ext4_new_meta_blocks() -- allocate block for meta data (indexing) blocks | |
1906 | * | |
1907 | * @handle: handle to this transaction | |
1908 | * @inode: file inode | |
1909 | * @goal: given target block(filesystem wide) | |
1910 | - * @count: total number of blocks need | |
1911 | + * @count: pointer to total number of blocks needed | |
1912 | * @errp: error code | |
1913 | * | |
1914 | - * Return 1st allocated block numberon success, *count stores total account | |
1915 | + * Return 1st allocated block number on success, *count stores total account | |
1916 | * error stores in errp pointer | |
1917 | */ | |
1918 | ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode, | |
1919 | ext4_fsblk_t goal, unsigned long *count, int *errp) | |
1920 | { | |
1921 | + struct ext4_allocation_request ar; | |
1922 | ext4_fsblk_t ret; | |
1923 | - ret = do_blk_alloc(handle, inode, 0, goal, | |
1924 | - count, errp, EXT4_META_BLOCK); | |
1925 | + | |
1926 | + memset(&ar, 0, sizeof(ar)); | |
1927 | + /* Fill with neighbour allocated blocks */ | |
1928 | + ar.inode = inode; | |
1929 | + ar.goal = goal; | |
1930 | + ar.len = count ? *count : 1; | |
1931 | + | |
1932 | + ret = ext4_mb_new_blocks(handle, &ar, errp); | |
1933 | + if (count) | |
1934 | + *count = ar.len; | |
1935 | + | |
1936 | /* | |
1937 | * Account for the allocated meta blocks | |
1938 | */ | |
1939 | - if (!(*errp)) { | |
1940 | + if (!(*errp) && EXT4_I(inode)->i_delalloc_reserved_flag) { | |
1941 | spin_lock(&EXT4_I(inode)->i_block_reservation_lock); | |
1942 | - EXT4_I(inode)->i_allocated_meta_blocks += *count; | |
1943 | + EXT4_I(inode)->i_allocated_meta_blocks += ar.len; | |
1944 | spin_unlock(&EXT4_I(inode)->i_block_reservation_lock); | |
1945 | } | |
1946 | return ret; | |
1947 | } | |
1948 | ||
1949 | -/* | |
1950 | - * ext4_new_meta_block() -- allocate block for meta data (indexing) blocks | |
1951 | - * | |
1952 | - * @handle: handle to this transaction | |
1953 | - * @inode: file inode | |
1954 | - * @goal: given target block(filesystem wide) | |
1955 | - * @errp: error code | |
1956 | - * | |
1957 | - * Return allocated block number on success | |
1958 | - */ | |
1959 | -ext4_fsblk_t ext4_new_meta_block(handle_t *handle, struct inode *inode, | |
1960 | - ext4_fsblk_t goal, int *errp) | |
1961 | -{ | |
1962 | - unsigned long count = 1; | |
1963 | - return ext4_new_meta_blocks(handle, inode, goal, &count, errp); | |
1964 | -} | |
1965 | - | |
1966 | -/* | |
1967 | - * ext4_new_blocks() -- allocate data blocks | |
1968 | - * | |
1969 | - * @handle: handle to this transaction | |
1970 | - * @inode: file inode | |
1971 | - * @goal: given target block(filesystem wide) | |
1972 | - * @count: total number of blocks need | |
1973 | - * @errp: error code | |
1974 | - * | |
1975 | - * Return 1st allocated block numberon success, *count stores total account | |
1976 | - * error stores in errp pointer | |
1977 | - */ | |
1978 | - | |
1979 | -ext4_fsblk_t ext4_new_blocks(handle_t *handle, struct inode *inode, | |
1980 | - ext4_lblk_t iblock, ext4_fsblk_t goal, | |
1981 | - unsigned long *count, int *errp) | |
1982 | -{ | |
1983 | - return do_blk_alloc(handle, inode, iblock, goal, count, errp, 0); | |
1984 | -} | |
1985 | - | |
1986 | /** | |
1987 | * ext4_count_free_blocks() -- count filesystem free blocks | |
1988 | * @sb: superblock | |
1989 | @@ -2068,7 +671,7 @@ ext4_fsblk_t ext4_count_free_blocks(stru | |
1990 | #ifdef EXT4FS_DEBUG | |
1991 | struct ext4_super_block *es; | |
1992 | ext4_fsblk_t bitmap_count; | |
1993 | - unsigned long x; | |
1994 | + unsigned int x; | |
1995 | struct buffer_head *bitmap_bh = NULL; | |
1996 | ||
1997 | es = EXT4_SB(sb)->s_es; | |
1998 | @@ -2088,15 +691,14 @@ ext4_fsblk_t ext4_count_free_blocks(stru | |
1999 | continue; | |
2000 | ||
2001 | x = ext4_count_free(bitmap_bh, sb->s_blocksize); | |
2002 | - printk(KERN_DEBUG "group %lu: stored = %d, counted = %lu\n", | |
2003 | + printk(KERN_DEBUG "group %lu: stored = %d, counted = %u\n", | |
2004 | i, le16_to_cpu(gdp->bg_free_blocks_count), x); | |
2005 | bitmap_count += x; | |
2006 | } | |
2007 | brelse(bitmap_bh); | |
2008 | - printk("ext4_count_free_blocks: stored = %llu" | |
2009 | - ", computed = %llu, %llu\n", | |
2010 | - ext4_free_blocks_count(es), | |
2011 | - desc_count, bitmap_count); | |
2012 | + printk(KERN_DEBUG "ext4_count_free_blocks: stored = %llu" | |
2013 | + ", computed = %llu, %llu\n", ext4_free_blocks_count(es), | |
2014 | + desc_count, bitmap_count); | |
2015 | return bitmap_count; | |
2016 | #else | |
2017 | desc_count = 0; | |
2018 | @@ -2105,7 +707,7 @@ ext4_fsblk_t ext4_count_free_blocks(stru | |
2019 | gdp = ext4_get_group_desc(sb, i, NULL); | |
2020 | if (!gdp) | |
2021 | continue; | |
2022 | - desc_count += le16_to_cpu(gdp->bg_free_blocks_count); | |
2023 | + desc_count += ext4_free_blks_count(sb, gdp); | |
2024 | } | |
2025 | ||
2026 | return desc_count; | |
2027 | @@ -2183,8 +785,9 @@ unsigned long ext4_bg_num_gdb(struct sup | |
2028 | ||
2029 | if (!EXT4_HAS_INCOMPAT_FEATURE(sb,EXT4_FEATURE_INCOMPAT_META_BG) || | |
2030 | metagroup < first_meta_bg) | |
2031 | - return ext4_bg_num_gdb_nometa(sb,group); | |
2032 | + return ext4_bg_num_gdb_nometa(sb, group); | |
2033 | ||
2034 | return ext4_bg_num_gdb_meta(sb,group); | |
2035 | ||
2036 | } | |
2037 | + | |
2038 | diff -rup b/fs/ext4//bitmap.c a/fs/ext4///bitmap.c | |
2039 | --- b/fs/ext4/bitmap.c 2009-02-11 14:37:58.000000000 +0100 | |
2040 | +++ a/fs/ext4/bitmap.c 2009-02-10 21:40:11.000000000 +0100 | |
2041 | @@ -15,17 +15,16 @@ | |
2042 | ||
2043 | static const int nibblemap[] = {4, 3, 3, 2, 3, 2, 2, 1, 3, 2, 2, 1, 2, 1, 1, 0}; | |
2044 | ||
2045 | -unsigned long ext4_count_free (struct buffer_head * map, unsigned int numchars) | |
2046 | +unsigned int ext4_count_free(struct buffer_head *map, unsigned int numchars) | |
2047 | { | |
2048 | - unsigned int i; | |
2049 | - unsigned long sum = 0; | |
2050 | + unsigned int i, sum = 0; | |
2051 | ||
2052 | if (!map) | |
2053 | - return (0); | |
2054 | + return 0; | |
2055 | for (i = 0; i < numchars; i++) | |
2056 | sum += nibblemap[map->b_data[i] & 0xf] + | |
2057 | nibblemap[(map->b_data[i] >> 4) & 0xf]; | |
2058 | - return (sum); | |
2059 | + return sum; | |
2060 | } | |
2061 | ||
2062 | #endif /* EXT4FS_DEBUG */ | |
2063 | diff -rup b/fs/ext4//dir.c a/fs/ext4///dir.c | |
2064 | --- b/fs/ext4/dir.c 2009-02-11 14:37:58.000000000 +0100 | |
2065 | +++ a/fs/ext4/dir.c 2009-02-10 21:40:11.000000000 +0100 | |
2066 | @@ -33,10 +33,10 @@ static unsigned char ext4_filetype_table | |
2067 | }; | |
2068 | ||
2069 | static int ext4_readdir(struct file *, void *, filldir_t); | |
2070 | -static int ext4_dx_readdir(struct file * filp, | |
2071 | - void * dirent, filldir_t filldir); | |
2072 | -static int ext4_release_dir (struct inode * inode, | |
2073 | - struct file * filp); | |
2074 | +static int ext4_dx_readdir(struct file *filp, | |
2075 | + void *dirent, filldir_t filldir); | |
2076 | +static int ext4_release_dir(struct inode *inode, | |
2077 | + struct file *filp); | |
2078 | ||
2079 | const struct file_operations ext4_dir_operations = { | |
2080 | .llseek = generic_file_llseek, | |
2081 | @@ -61,12 +61,12 @@ static unsigned char get_dtype(struct su | |
2082 | } | |
2083 | ||
2084 | ||
2085 | -int ext4_check_dir_entry (const char * function, struct inode * dir, | |
2086 | - struct ext4_dir_entry_2 * de, | |
2087 | - struct buffer_head * bh, | |
2088 | - unsigned long offset) | |
2089 | +int ext4_check_dir_entry(const char *function, struct inode *dir, | |
2090 | + struct ext4_dir_entry_2 *de, | |
2091 | + struct buffer_head *bh, | |
2092 | + unsigned int offset) | |
2093 | { | |
2094 | - const char * error_msg = NULL; | |
2095 | + const char *error_msg = NULL; | |
2096 | const int rlen = ext4_rec_len_from_disk(de->rec_len); | |
2097 | ||
2098 | if (rlen < EXT4_DIR_REC_LEN(1)) | |
2099 | @@ -82,20 +82,20 @@ int ext4_check_dir_entry (const char * f | |
2100 | error_msg = "inode out of bounds"; | |
2101 | ||
2102 | if (error_msg != NULL) | |
2103 | - ext4_error (dir->i_sb, function, | |
2104 | + ext4_error(dir->i_sb, function, | |
2105 | "bad entry in directory #%lu: %s - " | |
2106 | - "offset=%lu, inode=%lu, rec_len=%d, name_len=%d", | |
2107 | + "offset=%u, inode=%u, rec_len=%d, name_len=%d", | |
2108 | dir->i_ino, error_msg, offset, | |
2109 | - (unsigned long) le32_to_cpu(de->inode), | |
2110 | + le32_to_cpu(de->inode), | |
2111 | rlen, de->name_len); | |
2112 | return error_msg == NULL ? 1 : 0; | |
2113 | } | |
2114 | ||
2115 | -static int ext4_readdir(struct file * filp, | |
2116 | - void * dirent, filldir_t filldir) | |
2117 | +static int ext4_readdir(struct file *filp, | |
2118 | + void *dirent, filldir_t filldir) | |
2119 | { | |
2120 | int error = 0; | |
2121 | - unsigned long offset; | |
2122 | + unsigned int offset; | |
2123 | int i, stored; | |
2124 | struct ext4_dir_entry_2 *de; | |
2125 | struct super_block *sb; | |
2126 | @@ -192,14 +192,14 @@ revalidate: | |
2127 | while (!error && filp->f_pos < inode->i_size | |
2128 | && offset < sb->s_blocksize) { | |
2129 | de = (struct ext4_dir_entry_2 *) (bh->b_data + offset); | |
2130 | - if (!ext4_check_dir_entry ("ext4_readdir", inode, de, | |
2131 | - bh, offset)) { | |
2132 | + if (!ext4_check_dir_entry("ext4_readdir", inode, de, | |
2133 | + bh, offset)) { | |
2134 | /* | |
2135 | * On error, skip the f_pos to the next block | |
2136 | */ | |
2137 | filp->f_pos = (filp->f_pos | | |
2138 | (sb->s_blocksize - 1)) + 1; | |
2139 | - brelse (bh); | |
2140 | + brelse(bh); | |
2141 | ret = stored; | |
2142 | goto out; | |
2143 | } | |
2144 | @@ -223,12 +223,12 @@ revalidate: | |
2145 | break; | |
2146 | if (version != filp->f_version) | |
2147 | goto revalidate; | |
2148 | - stored ++; | |
2149 | + stored++; | |
2150 | } | |
2151 | filp->f_pos += ext4_rec_len_from_disk(de->rec_len); | |
2152 | } | |
2153 | offset = 0; | |
2154 | - brelse (bh); | |
2155 | + brelse(bh); | |
2156 | } | |
2157 | out: | |
2158 | return ret; | |
2159 | @@ -295,9 +295,9 @@ static void free_rb_tree_fname(struct rb | |
2160 | parent = rb_parent(n); | |
2161 | fname = rb_entry(n, struct fname, rb_hash); | |
2162 | while (fname) { | |
2163 | - struct fname * old = fname; | |
2164 | + struct fname *old = fname; | |
2165 | fname = fname->next; | |
2166 | - kfree (old); | |
2167 | + kfree(old); | |
2168 | } | |
2169 | if (!parent) | |
2170 | root->rb_node = NULL; | |
2171 | @@ -336,7 +336,7 @@ int ext4_htree_store_dirent(struct file | |
2172 | struct ext4_dir_entry_2 *dirent) | |
2173 | { | |
2174 | struct rb_node **p, *parent = NULL; | |
2175 | - struct fname * fname, *new_fn; | |
2176 | + struct fname *fname, *new_fn; | |
2177 | struct dir_private_info *info; | |
2178 | int len; | |
2179 | ||
2180 | @@ -393,19 +393,20 @@ int ext4_htree_store_dirent(struct file | |
2181 | * for all entres on the fname linked list. (Normally there is only | |
2182 | * one entry on the linked list, unless there are 62 bit hash collisions.) | |
2183 | */ | |
2184 | -static int call_filldir(struct file * filp, void * dirent, | |
2185 | +static int call_filldir(struct file *filp, void *dirent, | |
2186 | filldir_t filldir, struct fname *fname) | |
2187 | { | |
2188 | struct dir_private_info *info = filp->private_data; | |
2189 | loff_t curr_pos; | |
2190 | struct inode *inode = filp->f_path.dentry->d_inode; | |
2191 | - struct super_block * sb; | |
2192 | + struct super_block *sb; | |
2193 | int error; | |
2194 | ||
2195 | sb = inode->i_sb; | |
2196 | ||
2197 | if (!fname) { | |
2198 | - printk("call_filldir: called with null fname?!?\n"); | |
2199 | + printk(KERN_ERR "ext4: call_filldir: called with " | |
2200 | + "null fname?!?\n"); | |
2201 | return 0; | |
2202 | } | |
2203 | curr_pos = hash2pos(fname->hash, fname->minor_hash); | |
2204 | @@ -424,8 +425,8 @@ static int call_filldir(struct file * fi | |
2205 | return 0; | |
2206 | } | |
2207 | ||
2208 | -static int ext4_dx_readdir(struct file * filp, | |
2209 | - void * dirent, filldir_t filldir) | |
2210 | +static int ext4_dx_readdir(struct file *filp, | |
2211 | + void *dirent, filldir_t filldir) | |
2212 | { | |
2213 | struct dir_private_info *info = filp->private_data; | |
2214 | struct inode *inode = filp->f_path.dentry->d_inode; | |
2215 | @@ -512,7 +513,7 @@ finished: | |
2216 | return 0; | |
2217 | } | |
2218 | ||
2219 | -static int ext4_release_dir (struct inode * inode, struct file * filp) | |
2220 | +static int ext4_release_dir(struct inode *inode, struct file *filp) | |
2221 | { | |
2222 | if (filp->private_data) | |
2223 | ext4_htree_free_dir_info(filp->private_data); | |
2224 | diff -rup b/fs/ext4//ext4_extents.h a/fs/ext4///ext4_extents.h | |
2225 | --- b/fs/ext4/ext4_extents.h 2009-02-11 14:37:58.000000000 +0100 | |
2226 | +++ a/fs/ext4/ext4_extents.h 2009-02-10 21:40:14.000000000 +0100 | |
2227 | @@ -181,11 +181,6 @@ static inline unsigned short ext_depth(s | |
2228 | return le16_to_cpu(ext_inode_hdr(inode)->eh_depth); | |
2229 | } | |
2230 | ||
2231 | -static inline void ext4_ext_tree_changed(struct inode *inode) | |
2232 | -{ | |
2233 | - EXT4_I(inode)->i_ext_generation++; | |
2234 | -} | |
2235 | - | |
2236 | static inline void | |
2237 | ext4_ext_invalidate_cache(struct inode *inode) | |
2238 | { | |
2239 | diff -rup b/fs/ext4//ext4.h a/fs/ext4///ext4.h | |
2240 | --- b/fs/ext4/ext4.h 2009-02-11 14:37:58.000000000 +0100 | |
2241 | +++ a/fs/ext4/ext4.h 2009-02-10 21:40:14.000000000 +0100 | |
2242 | @@ -19,6 +19,7 @@ | |
2243 | #include <linux/types.h> | |
2244 | #include <linux/blkdev.h> | |
2245 | #include <linux/magic.h> | |
2246 | +#include <linux/jbd2.h> | |
2247 | #include "ext4_i.h" | |
2248 | ||
2249 | /* | |
2250 | @@ -44,9 +45,9 @@ | |
2251 | #ifdef EXT4FS_DEBUG | |
2252 | #define ext4_debug(f, a...) \ | |
2253 | do { \ | |
2254 | - printk (KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:", \ | |
2255 | + printk(KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:", \ | |
2256 | __FILE__, __LINE__, __func__); \ | |
2257 | - printk (KERN_DEBUG f, ## a); \ | |
2258 | + printk(KERN_DEBUG f, ## a); \ | |
2259 | } while (0) | |
2260 | #else | |
2261 | #define ext4_debug(f, a...) do {} while (0) | |
2262 | @@ -94,9 +95,9 @@ struct ext4_allocation_request { | |
2263 | /* phys. block for ^^^ */ | |
2264 | ext4_fsblk_t pright; | |
2265 | /* how many blocks we want to allocate */ | |
2266 | - unsigned long len; | |
2267 | + unsigned int len; | |
2268 | /* flags. see above EXT4_MB_HINT_* */ | |
2269 | - unsigned long flags; | |
2270 | + unsigned int flags; | |
2271 | }; | |
2272 | ||
2273 | /* | |
2274 | @@ -128,7 +129,7 @@ struct ext4_allocation_request { | |
2275 | #else | |
2276 | # define EXT4_BLOCK_SIZE(s) (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size) | |
2277 | #endif | |
2278 | -#define EXT4_ADDR_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / sizeof (__u32)) | |
2279 | +#define EXT4_ADDR_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / sizeof(__u32)) | |
2280 | #ifdef __KERNEL__ | |
2281 | # define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits) | |
2282 | #else | |
2283 | @@ -156,12 +157,12 @@ struct ext4_group_desc | |
2284 | __le32 bg_block_bitmap_lo; /* Blocks bitmap block */ | |
2285 | __le32 bg_inode_bitmap_lo; /* Inodes bitmap block */ | |
2286 | __le32 bg_inode_table_lo; /* Inodes table block */ | |
2287 | - __le16 bg_free_blocks_count; /* Free blocks count */ | |
2288 | - __le16 bg_free_inodes_count; /* Free inodes count */ | |
2289 | - __le16 bg_used_dirs_count; /* Directories count */ | |
2290 | + __le16 bg_free_blocks_count_lo;/* Free blocks count */ | |
2291 | + __le16 bg_free_inodes_count_lo;/* Free inodes count */ | |
2292 | + __le16 bg_used_dirs_count_lo; /* Directories count */ | |
2293 | __le16 bg_flags; /* EXT4_BG_flags (INODE_UNINIT, etc) */ | |
2294 | __u32 bg_reserved[2]; /* Likely block/inode bitmap checksum */ | |
2295 | - __le16 bg_itable_unused; /* Unused inodes count */ | |
2296 | + __le16 bg_itable_unused_lo; /* Unused inodes count */ | |
2297 | __le16 bg_checksum; /* crc16(sb_uuid+group+desc) */ | |
2298 | __le32 bg_block_bitmap_hi; /* Blocks bitmap block MSB */ | |
2299 | __le32 bg_inode_bitmap_hi; /* Inodes bitmap block MSB */ | |
2300 | @@ -169,7 +170,7 @@ struct ext4_group_desc | |
2301 | __le16 bg_free_blocks_count_hi;/* Free blocks count MSB */ | |
2302 | __le16 bg_free_inodes_count_hi;/* Free inodes count MSB */ | |
2303 | __le16 bg_used_dirs_count_hi; /* Directories count MSB */ | |
2304 | - __le16 bg_itable_unused_hi; /* Unused inodes count MSB */ | |
2305 | + __le16 bg_itable_unused_hi; /* Unused inodes count MSB */ | |
2306 | __u32 bg_reserved2[3]; | |
2307 | }; | |
2308 | ||
2309 | @@ -245,7 +246,7 @@ struct flex_groups { | |
2310 | #define EXT4_RESERVED_FL 0x80000000 /* reserved for ext4 lib */ | |
2311 | ||
2312 | #define EXT4_FL_USER_VISIBLE 0x000BDFFF /* User visible flags */ | |
2313 | -#define EXT4_FL_USER_MODIFIABLE 0x000380FF /* User modifiable flags */ | |
2314 | +#define EXT4_FL_USER_MODIFIABLE 0x000B80FF /* User modifiable flags */ | |
2315 | ||
2316 | /* | |
2317 | * Inode dynamic state flags | |
2318 | @@ -511,7 +512,6 @@ do { \ | |
2319 | /* | |
2320 | * Mount flags | |
2321 | */ | |
2322 | -#define EXT4_MOUNT_CHECK 0x00001 /* Do mount-time checks */ | |
2323 | #define EXT4_MOUNT_OLDALLOC 0x00002 /* Don't use the new Orlov allocator */ | |
2324 | #define EXT4_MOUNT_GRPID 0x00004 /* Create files with directory's group */ | |
2325 | #define EXT4_MOUNT_DEBUG 0x00008 /* Some debugging messages */ | |
2326 | @@ -539,7 +539,6 @@ do { \ | |
2327 | #define EXT4_MOUNT_JOURNAL_CHECKSUM 0x800000 /* Journal checksums */ | |
2328 | #define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */ | |
2329 | #define EXT4_MOUNT_I_VERSION 0x2000000 /* i_version support */ | |
2330 | -#define EXT4_MOUNT_MBALLOC 0x4000000 /* Buddy allocation support */ | |
2331 | #define EXT4_MOUNT_DELALLOC 0x8000000 /* Delalloc support */ | |
2332 | /* Compatibility, for having both ext2_fs.h and ext4_fs.h included at once */ | |
2333 | #ifndef _LINUX_EXT2_FS_H | |
2334 | @@ -668,7 +667,7 @@ struct ext4_super_block { | |
2335 | }; | |
2336 | ||
2337 | #ifdef __KERNEL__ | |
2338 | -static inline struct ext4_sb_info * EXT4_SB(struct super_block *sb) | |
2339 | +static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb) | |
2340 | { | |
2341 | return sb->s_fs_info; | |
2342 | } | |
2343 | @@ -726,11 +725,11 @@ static inline int ext4_valid_inum(struct | |
2344 | */ | |
2345 | ||
2346 | #define EXT4_HAS_COMPAT_FEATURE(sb,mask) \ | |
2347 | - ( EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask) ) | |
2348 | + ((EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask)) != 0) | |
2349 | #define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask) \ | |
2350 | - ( EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask) ) | |
2351 | + ((EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask)) != 0) | |
2352 | #define EXT4_HAS_INCOMPAT_FEATURE(sb,mask) \ | |
2353 | - ( EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask) ) | |
2354 | + ((EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask)) != 0) | |
2355 | #define EXT4_SET_COMPAT_FEATURE(sb,mask) \ | |
2356 | EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask) | |
2357 | #define EXT4_SET_RO_COMPAT_FEATURE(sb,mask) \ | |
2358 | @@ -790,6 +789,8 @@ static inline int ext4_valid_inum(struct | |
2359 | #define EXT4_DEF_RESUID 0 | |
2360 | #define EXT4_DEF_RESGID 0 | |
2361 | ||
2362 | +#define EXT4_DEF_INODE_READAHEAD_BLKS 32 | |
2363 | + | |
2364 | /* | |
2365 | * Default mount options | |
2366 | */ | |
2367 | @@ -889,6 +890,9 @@ static inline __le16 ext4_rec_len_to_dis | |
2368 | #define DX_HASH_LEGACY 0 | |
2369 | #define DX_HASH_HALF_MD4 1 | |
2370 | #define DX_HASH_TEA 2 | |
2371 | +#define DX_HASH_LEGACY_UNSIGNED 3 | |
2372 | +#define DX_HASH_HALF_MD4_UNSIGNED 4 | |
2373 | +#define DX_HASH_TEA_UNSIGNED 5 | |
2374 | ||
2375 | #ifdef __KERNEL__ | |
2376 | ||
2377 | @@ -953,7 +957,25 @@ ext4_group_first_block_no(struct super_b | |
2378 | #define ERR_BAD_DX_DIR -75000 | |
2379 | ||
2380 | void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr, | |
2381 | - unsigned long *blockgrpp, ext4_grpblk_t *offsetp); | |
2382 | + ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp); | |
2383 | + | |
2384 | +extern struct proc_dir_entry *ext4_proc_root; | |
2385 | + | |
2386 | +#ifdef CONFIG_PROC_FS | |
2387 | +extern const struct file_operations ext4_ui_proc_fops; | |
2388 | + | |
2389 | +#define EXT4_PROC_HANDLER(name, var) \ | |
2390 | +do { \ | |
2391 | + proc = proc_create_data(name, mode, sbi->s_proc, \ | |
2392 | + &ext4_ui_proc_fops, &sbi->s_##var); \ | |
2393 | + if (proc == NULL) { \ | |
2394 | + printk(KERN_ERR "EXT4-fs: can't create %s\n", name); \ | |
2395 | + goto err_out; \ | |
2396 | + } \ | |
2397 | +} while (0) | |
2398 | +#else | |
2399 | +#define EXT4_PROC_HANDLER(name, var) | |
2400 | +#endif | |
2401 | ||
2402 | /* | |
2403 | * Function prototypes | |
2404 | @@ -967,6 +989,9 @@ void ext4_get_group_no_and_offset(struct | |
2405 | # define ATTRIB_NORET __attribute__((noreturn)) | |
2406 | # define NORET_AND noreturn, | |
2407 | ||
2408 | +/* bitmap.c */ | |
2409 | +extern unsigned int ext4_count_free(struct buffer_head *, unsigned); | |
2410 | + | |
2411 | /* balloc.c */ | |
2412 | extern unsigned int ext4_block_group(struct super_block *sb, | |
2413 | ext4_fsblk_t blocknr); | |
2414 | @@ -975,55 +1000,44 @@ extern ext4_grpblk_t ext4_block_group_of | |
2415 | extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group); | |
2416 | extern unsigned long ext4_bg_num_gdb(struct super_block *sb, | |
2417 | ext4_group_t group); | |
2418 | -extern ext4_fsblk_t ext4_new_meta_block(handle_t *handle, struct inode *inode, | |
2419 | - ext4_fsblk_t goal, int *errp); | |
2420 | extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode, | |
2421 | ext4_fsblk_t goal, unsigned long *count, int *errp); | |
2422 | -extern ext4_fsblk_t ext4_new_blocks(handle_t *handle, struct inode *inode, | |
2423 | - ext4_lblk_t iblock, ext4_fsblk_t goal, | |
2424 | - unsigned long *count, int *errp); | |
2425 | -extern ext4_fsblk_t ext4_old_new_blocks(handle_t *handle, struct inode *inode, | |
2426 | - ext4_fsblk_t goal, unsigned long *count, int *errp); | |
2427 | -extern ext4_fsblk_t ext4_has_free_blocks(struct ext4_sb_info *sbi, | |
2428 | - ext4_fsblk_t nblocks); | |
2429 | -extern void ext4_free_blocks (handle_t *handle, struct inode *inode, | |
2430 | +extern int ext4_claim_free_blocks(struct ext4_sb_info *sbi, s64 nblocks); | |
2431 | +extern int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks); | |
2432 | +extern void ext4_free_blocks(handle_t *handle, struct inode *inode, | |
2433 | ext4_fsblk_t block, unsigned long count, int metadata); | |
2434 | -extern void ext4_free_blocks_sb (handle_t *handle, struct super_block *sb, | |
2435 | - ext4_fsblk_t block, unsigned long count, | |
2436 | - unsigned long *pdquot_freed_blocks); | |
2437 | -extern ext4_fsblk_t ext4_count_free_blocks (struct super_block *); | |
2438 | -extern void ext4_check_blocks_bitmap (struct super_block *); | |
2439 | +extern void ext4_add_groupblocks(handle_t *handle, struct super_block *sb, | |
2440 | + ext4_fsblk_t block, unsigned long count); | |
2441 | +extern ext4_fsblk_t ext4_count_free_blocks(struct super_block *); | |
2442 | +extern void ext4_check_blocks_bitmap(struct super_block *); | |
2443 | extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb, | |
2444 | ext4_group_t block_group, | |
2445 | struct buffer_head ** bh); | |
2446 | extern int ext4_should_retry_alloc(struct super_block *sb, int *retries); | |
2447 | -extern void ext4_init_block_alloc_info(struct inode *); | |
2448 | -extern void ext4_rsv_window_add(struct super_block *sb, struct ext4_reserve_window_node *rsv); | |
2449 | ||
2450 | /* dir.c */ | |
2451 | extern int ext4_check_dir_entry(const char *, struct inode *, | |
2452 | struct ext4_dir_entry_2 *, | |
2453 | - struct buffer_head *, unsigned long); | |
2454 | + struct buffer_head *, unsigned int); | |
2455 | extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash, | |
2456 | __u32 minor_hash, | |
2457 | struct ext4_dir_entry_2 *dirent); | |
2458 | extern void ext4_htree_free_dir_info(struct dir_private_info *p); | |
2459 | ||
2460 | /* fsync.c */ | |
2461 | -extern int ext4_sync_file (struct file *, struct dentry *, int); | |
2462 | +extern int ext4_sync_file(struct file *, struct dentry *, int); | |
2463 | ||
2464 | /* hash.c */ | |
2465 | extern int ext4fs_dirhash(const char *name, int len, struct | |
2466 | dx_hash_info *hinfo); | |
2467 | ||
2468 | /* ialloc.c */ | |
2469 | -extern struct inode * ext4_new_inode (handle_t *, struct inode *, int); | |
2470 | -extern void ext4_free_inode (handle_t *, struct inode *); | |
2471 | -extern struct inode * ext4_orphan_get (struct super_block *, unsigned long); | |
2472 | -extern unsigned long ext4_count_free_inodes (struct super_block *); | |
2473 | -extern unsigned long ext4_count_dirs (struct super_block *); | |
2474 | -extern void ext4_check_inodes_bitmap (struct super_block *); | |
2475 | -extern unsigned long ext4_count_free (struct buffer_head *, unsigned); | |
2476 | +extern struct inode * ext4_new_inode(handle_t *, struct inode *, int); | |
2477 | +extern void ext4_free_inode(handle_t *, struct inode *); | |
2478 | +extern struct inode * ext4_orphan_get(struct super_block *, unsigned long); | |
2479 | +extern unsigned long ext4_count_free_inodes(struct super_block *); | |
2480 | +extern unsigned long ext4_count_dirs(struct super_block *); | |
2481 | +extern void ext4_check_inodes_bitmap(struct super_block *); | |
2482 | ||
2483 | /* mballoc.c */ | |
2484 | extern long ext4_mb_stats; | |
2485 | @@ -1033,17 +1047,18 @@ extern int ext4_mb_release(struct super_ | |
2486 | extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *, | |
2487 | struct ext4_allocation_request *, int *); | |
2488 | extern int ext4_mb_reserve_blocks(struct super_block *, int); | |
2489 | -extern void ext4_mb_discard_inode_preallocations(struct inode *); | |
2490 | +extern void ext4_discard_preallocations(struct inode *); | |
2491 | extern int __init init_ext4_mballoc(void); | |
2492 | extern void exit_ext4_mballoc(void); | |
2493 | extern void ext4_mb_free_blocks(handle_t *, struct inode *, | |
2494 | unsigned long, unsigned long, int, unsigned long *); | |
2495 | -extern int ext4_mb_add_more_groupinfo(struct super_block *sb, | |
2496 | +extern int ext4_mb_add_groupinfo(struct super_block *sb, | |
2497 | ext4_group_t i, struct ext4_group_desc *desc); | |
2498 | extern void ext4_mb_update_group_info(struct ext4_group_info *grp, | |
2499 | ext4_grpblk_t add); | |
2500 | - | |
2501 | - | |
2502 | +extern int ext4_mb_get_buddy_cache_lock(struct super_block *, ext4_group_t); | |
2503 | +extern void ext4_mb_put_buddy_cache_lock(struct super_block *, | |
2504 | + ext4_group_t, int); | |
2505 | /* inode.c */ | |
2506 | int ext4_forget(handle_t *handle, int is_metadata, struct inode *inode, | |
2507 | struct buffer_head *bh, ext4_fsblk_t blocknr); | |
2508 | @@ -1051,24 +1066,19 @@ struct buffer_head *ext4_getblk(handle_t | |
2509 | ext4_lblk_t, int, int *); | |
2510 | struct buffer_head *ext4_bread(handle_t *, struct inode *, | |
2511 | ext4_lblk_t, int, int *); | |
2512 | -int ext4_get_blocks_handle(handle_t *handle, struct inode *inode, | |
2513 | - ext4_lblk_t iblock, unsigned long maxblocks, | |
2514 | - struct buffer_head *bh_result, | |
2515 | - int create, int extend_disksize); | |
2516 | ||
2517 | extern struct inode *ext4_iget(struct super_block *, unsigned long); | |
2518 | -extern int ext4_write_inode (struct inode *, int); | |
2519 | -extern int ext4_setattr (struct dentry *, struct iattr *); | |
2520 | +extern int ext4_write_inode(struct inode *, int); | |
2521 | +extern int ext4_setattr(struct dentry *, struct iattr *); | |
2522 | extern int ext4_getattr(struct vfsmount *mnt, struct dentry *dentry, | |
2523 | struct kstat *stat); | |
2524 | -extern void ext4_delete_inode (struct inode *); | |
2525 | -extern int ext4_sync_inode (handle_t *, struct inode *); | |
2526 | -extern void ext4_discard_reservation (struct inode *); | |
2527 | +extern void ext4_delete_inode(struct inode *); | |
2528 | +extern int ext4_sync_inode(handle_t *, struct inode *); | |
2529 | extern void ext4_dirty_inode(struct inode *); | |
2530 | extern int ext4_change_inode_journal_flag(struct inode *, int); | |
2531 | extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *); | |
2532 | extern int ext4_can_truncate(struct inode *inode); | |
2533 | -extern void ext4_truncate (struct inode *); | |
2534 | +extern void ext4_truncate(struct inode *); | |
2535 | extern void ext4_set_inode_flags(struct inode *); | |
2536 | extern void ext4_get_inode_flags(struct ext4_inode_info *); | |
2537 | extern void ext4_set_aops(struct inode *inode); | |
2538 | @@ -1081,7 +1091,7 @@ extern int ext4_page_mkwrite(struct vm_a | |
2539 | ||
2540 | /* ioctl.c */ | |
2541 | extern long ext4_ioctl(struct file *, unsigned int, unsigned long); | |
2542 | -extern long ext4_compat_ioctl (struct file *, unsigned int, unsigned long); | |
2543 | +extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long); | |
2544 | ||
2545 | /* migrate.c */ | |
2546 | extern int ext4_ext_migrate(struct inode *); | |
2547 | @@ -1099,14 +1109,17 @@ extern int ext4_group_extend(struct supe | |
2548 | ext4_fsblk_t n_blocks_count); | |
2549 | ||
2550 | /* super.c */ | |
2551 | -extern void ext4_error (struct super_block *, const char *, const char *, ...) | |
2552 | +extern void ext4_error(struct super_block *, const char *, const char *, ...) | |
2553 | __attribute__ ((format (printf, 3, 4))); | |
2554 | -extern void __ext4_std_error (struct super_block *, const char *, int); | |
2555 | -extern void ext4_abort (struct super_block *, const char *, const char *, ...) | |
2556 | +extern void __ext4_std_error(struct super_block *, const char *, int); | |
2557 | +extern void ext4_abort(struct super_block *, const char *, const char *, ...) | |
2558 | __attribute__ ((format (printf, 3, 4))); | |
2559 | -extern void ext4_warning (struct super_block *, const char *, const char *, ...) | |
2560 | +extern void ext4_warning(struct super_block *, const char *, const char *, ...) | |
2561 | __attribute__ ((format (printf, 3, 4))); | |
2562 | -extern void ext4_update_dynamic_rev (struct super_block *sb); | |
2563 | +extern void ext4_grp_locked_error(struct super_block *, ext4_group_t, | |
2564 | + const char *, const char *, ...) | |
2565 | + __attribute__ ((format (printf, 4, 5))); | |
2566 | +extern void ext4_update_dynamic_rev(struct super_block *sb); | |
2567 | extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb, | |
2568 | __u32 compat); | |
2569 | extern int ext4_update_rocompat_feature(handle_t *handle, | |
2570 | @@ -1119,12 +1132,28 @@ extern ext4_fsblk_t ext4_inode_bitmap(st | |
2571 | struct ext4_group_desc *bg); | |
2572 | extern ext4_fsblk_t ext4_inode_table(struct super_block *sb, | |
2573 | struct ext4_group_desc *bg); | |
2574 | +extern __u32 ext4_free_blks_count(struct super_block *sb, | |
2575 | + struct ext4_group_desc *bg); | |
2576 | +extern __u32 ext4_free_inodes_count(struct super_block *sb, | |
2577 | + struct ext4_group_desc *bg); | |
2578 | +extern __u32 ext4_used_dirs_count(struct super_block *sb, | |
2579 | + struct ext4_group_desc *bg); | |
2580 | +extern __u32 ext4_itable_unused_count(struct super_block *sb, | |
2581 | + struct ext4_group_desc *bg); | |
2582 | extern void ext4_block_bitmap_set(struct super_block *sb, | |
2583 | struct ext4_group_desc *bg, ext4_fsblk_t blk); | |
2584 | extern void ext4_inode_bitmap_set(struct super_block *sb, | |
2585 | struct ext4_group_desc *bg, ext4_fsblk_t blk); | |
2586 | extern void ext4_inode_table_set(struct super_block *sb, | |
2587 | struct ext4_group_desc *bg, ext4_fsblk_t blk); | |
2588 | +extern void ext4_free_blks_set(struct super_block *sb, | |
2589 | + struct ext4_group_desc *bg, __u32 count); | |
2590 | +extern void ext4_free_inodes_set(struct super_block *sb, | |
2591 | + struct ext4_group_desc *bg, __u32 count); | |
2592 | +extern void ext4_used_dirs_set(struct super_block *sb, | |
2593 | + struct ext4_group_desc *bg, __u32 count); | |
2594 | +extern void ext4_itable_unused_set(struct super_block *sb, | |
2595 | + struct ext4_group_desc *bg, __u32 count); | |
2596 | ||
2597 | static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es) | |
2598 | { | |
2599 | @@ -1179,7 +1208,7 @@ static inline void ext4_isize_set(struct | |
2600 | ||
2601 | static inline | |
2602 | struct ext4_group_info *ext4_get_group_info(struct super_block *sb, | |
2603 | - ext4_group_t group) | |
2604 | + ext4_group_t group) | |
2605 | { | |
2606 | struct ext4_group_info ***grp_info; | |
2607 | long indexv, indexh; | |
2608 | @@ -1207,6 +1236,72 @@ do { \ | |
2609 | __ext4_std_error((sb), __func__, (errno)); \ | |
2610 | } while (0) | |
2611 | ||
2612 | +#ifdef CONFIG_SMP | |
2613 | +/* Each CPU can accumulate FBC_BATCH blocks in their local | |
2614 | + * counters. So we need to make sure we have free blocks more | |
2615 | + * than FBC_BATCH * nr_cpu_ids. Also add a window of 4 times. | |
2616 | + */ | |
2617 | +#define EXT4_FREEBLOCKS_WATERMARK (4 * (FBC_BATCH * nr_cpu_ids)) | |
2618 | +#else | |
2619 | +#define EXT4_FREEBLOCKS_WATERMARK 0 | |
2620 | +#endif | |
2621 | + | |
2622 | +static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize) | |
2623 | +{ | |
2624 | + /* | |
2625 | + * XXX: replace with spinlock if seen contended -bzzz | |
2626 | + */ | |
2627 | + down_write(&EXT4_I(inode)->i_data_sem); | |
2628 | + if (newsize > EXT4_I(inode)->i_disksize) | |
2629 | + EXT4_I(inode)->i_disksize = newsize; | |
2630 | + up_write(&EXT4_I(inode)->i_data_sem); | |
2631 | + return ; | |
2632 | +} | |
2633 | + | |
2634 | +struct ext4_group_info { | |
2635 | + unsigned long bb_state; | |
2636 | + struct rb_root bb_free_root; | |
2637 | + unsigned short bb_first_free; | |
2638 | + unsigned short bb_free; | |
2639 | + unsigned short bb_fragments; | |
2640 | + struct list_head bb_prealloc_list; | |
2641 | +#ifdef DOUBLE_CHECK | |
2642 | + void *bb_bitmap; | |
2643 | +#endif | |
2644 | + struct rw_semaphore alloc_sem; | |
2645 | + unsigned short bb_counters[]; | |
2646 | +}; | |
2647 | + | |
2648 | +#define EXT4_GROUP_INFO_NEED_INIT_BIT 0 | |
2649 | +#define EXT4_GROUP_INFO_LOCKED_BIT 1 | |
2650 | + | |
2651 | +#define EXT4_MB_GRP_NEED_INIT(grp) \ | |
2652 | + (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state))) | |
2653 | + | |
2654 | +static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group) | |
2655 | +{ | |
2656 | + struct ext4_group_info *grinfo = ext4_get_group_info(sb, group); | |
2657 | + | |
2658 | + bit_spin_lock(EXT4_GROUP_INFO_LOCKED_BIT, &(grinfo->bb_state)); | |
2659 | +} | |
2660 | + | |
2661 | +static inline void ext4_unlock_group(struct super_block *sb, | |
2662 | + ext4_group_t group) | |
2663 | +{ | |
2664 | + struct ext4_group_info *grinfo = ext4_get_group_info(sb, group); | |
2665 | + | |
2666 | + bit_spin_unlock(EXT4_GROUP_INFO_LOCKED_BIT, &(grinfo->bb_state)); | |
2667 | +} | |
2668 | + | |
2669 | +static inline int ext4_is_group_locked(struct super_block *sb, | |
2670 | + ext4_group_t group) | |
2671 | +{ | |
2672 | + struct ext4_group_info *grinfo = ext4_get_group_info(sb, group); | |
2673 | + | |
2674 | + return bit_spin_is_locked(EXT4_GROUP_INFO_LOCKED_BIT, | |
2675 | + &(grinfo->bb_state)); | |
2676 | +} | |
2677 | + | |
2678 | /* | |
2679 | * Inodes and files operations | |
2680 | */ | |
2681 | @@ -1232,18 +1327,37 @@ extern int ext4_ext_writepage_trans_bloc | |
2682 | extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks, | |
2683 | int chunk); | |
2684 | extern int ext4_ext_get_blocks(handle_t *handle, struct inode *inode, | |
2685 | - ext4_lblk_t iblock, | |
2686 | - unsigned long max_blocks, struct buffer_head *bh_result, | |
2687 | - int create, int extend_disksize); | |
2688 | + ext4_lblk_t iblock, unsigned int max_blocks, | |
2689 | + struct buffer_head *bh_result, | |
2690 | + int create, int extend_disksize); | |
2691 | extern void ext4_ext_truncate(struct inode *); | |
2692 | extern void ext4_ext_init(struct super_block *); | |
2693 | extern void ext4_ext_release(struct super_block *); | |
2694 | extern long ext4_fallocate(struct inode *inode, int mode, loff_t offset, | |
2695 | loff_t len); | |
2696 | extern int ext4_get_blocks_wrap(handle_t *handle, struct inode *inode, | |
2697 | - sector_t block, unsigned long max_blocks, | |
2698 | + sector_t block, unsigned int max_blocks, | |
2699 | struct buffer_head *bh, int create, | |
2700 | int extend_disksize, int flag); | |
2701 | + | |
2702 | +#define BH_JBDPrivateStart (BH_Unshadow+1) | |
2703 | +/* | |
2704 | + * Add new method to test wether block and inode bitmaps are properly | |
2705 | + * initialized. With uninit_bg reading the block from disk is not enough | |
2706 | + * to mark the bitmap uptodate. We need to also zero-out the bitmap | |
2707 | + */ | |
2708 | +#define BH_BITMAP_UPTODATE BH_JBDPrivateStart | |
2709 | + | |
2710 | +static inline int bitmap_uptodate(struct buffer_head *bh) | |
2711 | +{ | |
2712 | + return (buffer_uptodate(bh) && | |
2713 | + test_bit(BH_BITMAP_UPTODATE, &(bh)->b_state)); | |
2714 | +} | |
2715 | +static inline void set_bitmap_uptodate(struct buffer_head *bh) | |
2716 | +{ | |
2717 | + set_bit(BH_BITMAP_UPTODATE, &(bh)->b_state); | |
2718 | +} | |
2719 | + | |
2720 | #endif /* __KERNEL__ */ | |
2721 | ||
2722 | #endif /* _EXT4_H */ | |
2723 | diff -rup b/fs/ext4//ext4_i.h a/fs/ext4///ext4_i.h | |
2724 | --- b/fs/ext4/ext4_i.h 2009-02-11 14:37:58.000000000 +0100 | |
2725 | +++ a/fs/ext4/ext4_i.h 2009-02-10 21:40:14.000000000 +0100 | |
2726 | @@ -31,39 +31,7 @@ typedef unsigned long long ext4_fsblk_t; | |
2727 | typedef __u32 ext4_lblk_t; | |
2728 | ||
2729 | /* data type for block group number */ | |
2730 | -typedef unsigned long ext4_group_t; | |
2731 | - | |
2732 | -struct ext4_reserve_window { | |
2733 | - ext4_fsblk_t _rsv_start; /* First byte reserved */ | |
2734 | - ext4_fsblk_t _rsv_end; /* Last byte reserved or 0 */ | |
2735 | -}; | |
2736 | - | |
2737 | -struct ext4_reserve_window_node { | |
2738 | - struct rb_node rsv_node; | |
2739 | - __u32 rsv_goal_size; | |
2740 | - __u32 rsv_alloc_hit; | |
2741 | - struct ext4_reserve_window rsv_window; | |
2742 | -}; | |
2743 | - | |
2744 | -struct ext4_block_alloc_info { | |
2745 | - /* information about reservation window */ | |
2746 | - struct ext4_reserve_window_node rsv_window_node; | |
2747 | - /* | |
2748 | - * was i_next_alloc_block in ext4_inode_info | |
2749 | - * is the logical (file-relative) number of the | |
2750 | - * most-recently-allocated block in this file. | |
2751 | - * We use this for detecting linearly ascending allocation requests. | |
2752 | - */ | |
2753 | - ext4_lblk_t last_alloc_logical_block; | |
2754 | - /* | |
2755 | - * Was i_next_alloc_goal in ext4_inode_info | |
2756 | - * is the *physical* companion to i_next_alloc_block. | |
2757 | - * it the physical block number of the block which was most-recentl | |
2758 | - * allocated to this file. This give us the goal (target) for the next | |
2759 | - * allocation when we detect linearly ascending requests. | |
2760 | - */ | |
2761 | - ext4_fsblk_t last_alloc_physical_block; | |
2762 | -}; | |
2763 | +typedef unsigned int ext4_group_t; | |
2764 | ||
2765 | #define rsv_start rsv_window._rsv_start | |
2766 | #define rsv_end rsv_window._rsv_end | |
2767 | @@ -97,11 +65,8 @@ struct ext4_inode_info { | |
2768 | ext4_group_t i_block_group; | |
2769 | __u32 i_state; /* Dynamic state flags for ext4 */ | |
2770 | ||
2771 | - /* block reservation info */ | |
2772 | - struct ext4_block_alloc_info *i_block_alloc_info; | |
2773 | - | |
2774 | ext4_lblk_t i_dir_start_lookup; | |
2775 | -#ifdef CONFIG_EXT4DEV_FS_XATTR | |
2776 | +#if defined(CONFIG_EXT4_FS_XATTR) || defined(CONFIG_EXT4DEV_FS_XATTR) | |
2777 | /* | |
2778 | * Extended attributes can be read independently of the main file | |
2779 | * data. Taking i_mutex even when reading would cause contention | |
2780 | @@ -111,7 +76,7 @@ struct ext4_inode_info { | |
2781 | */ | |
2782 | struct rw_semaphore xattr_sem; | |
2783 | #endif | |
2784 | -#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL | |
2785 | +#if defined(CONFIG_EXT4_FS_POSIX_ACL) || defined(CONFIG_EXT4DEV_FS_POSIX_ACL) | |
2786 | struct posix_acl *i_acl; | |
2787 | struct posix_acl *i_default_acl; | |
2788 | #endif | |
2789 | @@ -135,9 +100,6 @@ struct ext4_inode_info { | |
2790 | */ | |
2791 | loff_t i_disksize; | |
2792 | ||
2793 | - /* on-disk additional length */ | |
2794 | - __u16 i_extra_isize; | |
2795 | - | |
2796 | /* | |
2797 | * i_data_sem is for serialising ext4_truncate() against | |
2798 | * ext4_getblock(). In the 2.4 ext2 design, great chunks of inode's | |
2799 | @@ -152,7 +114,6 @@ struct ext4_inode_info { | |
2800 | struct inode vfs_inode; | |
2801 | struct jbd2_inode jinode; | |
2802 | ||
2803 | - unsigned long i_ext_generation; | |
2804 | struct ext4_ext_cache i_cached_extent; | |
2805 | /* | |
2806 | * File creation time. Its function is same as that of | |
2807 | @@ -165,10 +126,14 @@ struct ext4_inode_info { | |
2808 | spinlock_t i_prealloc_lock; | |
2809 | ||
2810 | /* allocation reservation info for delalloc */ | |
2811 | - unsigned long i_reserved_data_blocks; | |
2812 | - unsigned long i_reserved_meta_blocks; | |
2813 | - unsigned long i_allocated_meta_blocks; | |
2814 | + unsigned int i_reserved_data_blocks; | |
2815 | + unsigned int i_reserved_meta_blocks; | |
2816 | + unsigned int i_allocated_meta_blocks; | |
2817 | unsigned short i_delalloc_reserved_flag; | |
2818 | + | |
2819 | + /* on-disk additional length */ | |
2820 | + __u16 i_extra_isize; | |
2821 | + | |
2822 | spinlock_t i_block_reservation_lock; | |
2823 | }; | |
2824 | ||
2825 | diff -rup b/fs/ext4//ext4_sb.h a/fs/ext4///ext4_sb.h | |
2826 | --- b/fs/ext4/ext4_sb.h 2009-02-11 14:37:58.000000000 +0100 | |
2827 | +++ a/fs/ext4/ext4_sb.h 2009-02-10 21:40:14.000000000 +0100 | |
2828 | @@ -40,8 +40,8 @@ struct ext4_sb_info { | |
2829 | unsigned long s_blocks_last; /* Last seen block count */ | |
2830 | loff_t s_bitmap_maxbytes; /* max bytes for bitmap files */ | |
2831 | struct buffer_head * s_sbh; /* Buffer containing the super block */ | |
2832 | - struct ext4_super_block * s_es; /* Pointer to the super block in the buffer */ | |
2833 | - struct buffer_head ** s_group_desc; | |
2834 | + struct ext4_super_block *s_es; /* Pointer to the super block in the buffer */ | |
2835 | + struct buffer_head **s_group_desc; | |
2836 | unsigned long s_mount_opt; | |
2837 | ext4_fsblk_t s_sb_block; | |
2838 | uid_t s_resuid; | |
2839 | @@ -52,23 +52,26 @@ struct ext4_sb_info { | |
2840 | int s_desc_per_block_bits; | |
2841 | int s_inode_size; | |
2842 | int s_first_ino; | |
2843 | + unsigned int s_inode_readahead_blks; | |
2844 | spinlock_t s_next_gen_lock; | |
2845 | u32 s_next_generation; | |
2846 | u32 s_hash_seed[4]; | |
2847 | int s_def_hash_version; | |
2848 | + int s_hash_unsigned; /* 3 if hash should be signed, 0 if not */ | |
2849 | struct percpu_counter s_freeblocks_counter; | |
2850 | struct percpu_counter s_freeinodes_counter; | |
2851 | struct percpu_counter s_dirs_counter; | |
2852 | + struct percpu_counter s_dirtyblocks_counter; | |
2853 | struct blockgroup_lock s_blockgroup_lock; | |
2854 | + struct proc_dir_entry *s_proc; | |
2855 | ||
2856 | /* root of the per fs reservation window tree */ | |
2857 | spinlock_t s_rsv_window_lock; | |
2858 | struct rb_root s_rsv_window_root; | |
2859 | - struct ext4_reserve_window_node s_rsv_window_head; | |
2860 | ||
2861 | /* Journaling */ | |
2862 | - struct inode * s_journal_inode; | |
2863 | - struct journal_s * s_journal; | |
2864 | + struct inode *s_journal_inode; | |
2865 | + struct journal_s *s_journal; | |
2866 | struct list_head s_orphan; | |
2867 | unsigned long s_commit_interval; | |
2868 | struct block_device *journal_bdev; | |
2869 | @@ -106,12 +109,12 @@ struct ext4_sb_info { | |
2870 | ||
2871 | /* tunables */ | |
2872 | unsigned long s_stripe; | |
2873 | - unsigned long s_mb_stream_request; | |
2874 | - unsigned long s_mb_max_to_scan; | |
2875 | - unsigned long s_mb_min_to_scan; | |
2876 | - unsigned long s_mb_stats; | |
2877 | - unsigned long s_mb_order2_reqs; | |
2878 | - unsigned long s_mb_group_prealloc; | |
2879 | + unsigned int s_mb_stream_request; | |
2880 | + unsigned int s_mb_max_to_scan; | |
2881 | + unsigned int s_mb_min_to_scan; | |
2882 | + unsigned int s_mb_stats; | |
2883 | + unsigned int s_mb_order2_reqs; | |
2884 | + unsigned int s_mb_group_prealloc; | |
2885 | /* where last allocation was done - for stream allocation */ | |
2886 | unsigned long s_mb_last_group; | |
2887 | unsigned long s_mb_last_start; | |
2888 | @@ -121,7 +124,6 @@ struct ext4_sb_info { | |
2889 | int s_mb_history_cur; | |
2890 | int s_mb_history_max; | |
2891 | int s_mb_history_num; | |
2892 | - struct proc_dir_entry *s_mb_proc; | |
2893 | spinlock_t s_mb_history_lock; | |
2894 | int s_mb_history_filter; | |
2895 | ||
2896 | diff -rup b/fs/ext4//extents.c a/fs/ext4///extents.c | |
2897 | --- b/fs/ext4/extents.c 2009-02-11 14:37:58.000000000 +0100 | |
2898 | +++ a/fs/ext4/extents.c 2009-02-10 21:40:11.000000000 +0100 | |
2899 | @@ -190,7 +190,7 @@ ext4_ext_new_meta_block(handle_t *handle | |
2900 | ext4_fsblk_t goal, newblock; | |
2901 | ||
2902 | goal = ext4_ext_find_goal(inode, path, le32_to_cpu(ex->ee_block)); | |
2903 | - newblock = ext4_new_meta_block(handle, inode, goal, err); | |
2904 | + newblock = ext4_new_meta_blocks(handle, inode, goal, NULL, err); | |
2905 | return newblock; | |
2906 | } | |
2907 | ||
2908 | @@ -383,8 +383,8 @@ static void ext4_ext_show_leaf(struct in | |
2909 | ext_debug("\n"); | |
2910 | } | |
2911 | #else | |
2912 | -#define ext4_ext_show_path(inode,path) | |
2913 | -#define ext4_ext_show_leaf(inode,path) | |
2914 | +#define ext4_ext_show_path(inode, path) | |
2915 | +#define ext4_ext_show_leaf(inode, path) | |
2916 | #endif | |
2917 | ||
2918 | void ext4_ext_drop_refs(struct ext4_ext_path *path) | |
2919 | @@ -440,9 +440,10 @@ ext4_ext_binsearch_idx(struct inode *ino | |
2920 | for (k = 0; k < le16_to_cpu(eh->eh_entries); k++, ix++) { | |
2921 | if (k != 0 && | |
2922 | le32_to_cpu(ix->ei_block) <= le32_to_cpu(ix[-1].ei_block)) { | |
2923 | - printk("k=%d, ix=0x%p, first=0x%p\n", k, | |
2924 | - ix, EXT_FIRST_INDEX(eh)); | |
2925 | - printk("%u <= %u\n", | |
2926 | + printk(KERN_DEBUG "k=%d, ix=0x%p, " | |
2927 | + "first=0x%p\n", k, | |
2928 | + ix, EXT_FIRST_INDEX(eh)); | |
2929 | + printk(KERN_DEBUG "%u <= %u\n", | |
2930 | le32_to_cpu(ix->ei_block), | |
2931 | le32_to_cpu(ix[-1].ei_block)); | |
2932 | } | |
2933 | @@ -1158,15 +1159,13 @@ ext4_ext_search_right(struct inode *inod | |
2934 | while (--depth >= 0) { | |
2935 | ix = path[depth].p_idx; | |
2936 | if (ix != EXT_LAST_INDEX(path[depth].p_hdr)) | |
2937 | - break; | |
2938 | + goto got_index; | |
2939 | } | |
2940 | ||
2941 | - if (depth < 0) { | |
2942 | - /* we've gone up to the root and | |
2943 | - * found no index to the right */ | |
2944 | - return 0; | |
2945 | - } | |
2946 | + /* we've gone up to the root and found no index to the right */ | |
2947 | + return 0; | |
2948 | ||
2949 | +got_index: | |
2950 | /* we've found index to the right, let's | |
2951 | * follow it and find the closest allocated | |
2952 | * block to the right */ | |
2953 | @@ -1199,7 +1198,6 @@ ext4_ext_search_right(struct inode *inod | |
2954 | *phys = ext_pblock(ex); | |
2955 | put_bh(bh); | |
2956 | return 0; | |
2957 | - | |
2958 | } | |
2959 | ||
2960 | /* | |
2961 | @@ -1475,7 +1473,7 @@ int ext4_ext_insert_extent(handle_t *han | |
2962 | struct ext4_ext_path *path, | |
2963 | struct ext4_extent *newext) | |
2964 | { | |
2965 | - struct ext4_extent_header * eh; | |
2966 | + struct ext4_extent_header *eh; | |
2967 | struct ext4_extent *ex, *fex; | |
2968 | struct ext4_extent *nearex; /* nearest extent */ | |
2969 | struct ext4_ext_path *npath = NULL; | |
2970 | @@ -1620,7 +1618,6 @@ cleanup: | |
2971 | ext4_ext_drop_refs(npath); | |
2972 | kfree(npath); | |
2973 | } | |
2974 | - ext4_ext_tree_changed(inode); | |
2975 | ext4_ext_invalidate_cache(inode); | |
2976 | return err; | |
2977 | } | |
2978 | @@ -2124,7 +2121,6 @@ static int ext4_ext_remove_space(struct | |
2979 | } | |
2980 | } | |
2981 | out: | |
2982 | - ext4_ext_tree_changed(inode); | |
2983 | ext4_ext_drop_refs(path); | |
2984 | kfree(path); | |
2985 | ext4_journal_stop(handle); | |
2986 | @@ -2142,7 +2138,7 @@ void ext4_ext_init(struct super_block *s | |
2987 | */ | |
2988 | ||
2989 | if (test_opt(sb, EXTENTS)) { | |
2990 | - printk("EXT4-fs: file extents enabled"); | |
2991 | + printk(KERN_INFO "EXT4-fs: file extents enabled"); | |
2992 | #ifdef AGGRESSIVE_TEST | |
2993 | printk(", aggressive tests"); | |
2994 | #endif | |
2995 | @@ -2271,7 +2267,7 @@ static int ext4_ext_convert_to_initializ | |
2996 | struct inode *inode, | |
2997 | struct ext4_ext_path *path, | |
2998 | ext4_lblk_t iblock, | |
2999 | - unsigned long max_blocks) | |
3000 | + unsigned int max_blocks) | |
3001 | { | |
3002 | struct ext4_extent *ex, newex, orig_ex; | |
3003 | struct ext4_extent *ex1 = NULL; | |
3004 | @@ -2569,26 +2565,26 @@ fix_extent_len: | |
3005 | */ | |
3006 | int ext4_ext_get_blocks(handle_t *handle, struct inode *inode, | |
3007 | ext4_lblk_t iblock, | |
3008 | - unsigned long max_blocks, struct buffer_head *bh_result, | |
3009 | + unsigned int max_blocks, struct buffer_head *bh_result, | |
3010 | int create, int extend_disksize) | |
3011 | { | |
3012 | struct ext4_ext_path *path = NULL; | |
3013 | struct ext4_extent_header *eh; | |
3014 | struct ext4_extent newex, *ex; | |
3015 | - ext4_fsblk_t goal, newblock; | |
3016 | - int err = 0, depth, ret; | |
3017 | - unsigned long allocated = 0; | |
3018 | + ext4_fsblk_t newblock; | |
3019 | + int err = 0, depth, ret, cache_type; | |
3020 | + unsigned int allocated = 0; | |
3021 | struct ext4_allocation_request ar; | |
3022 | loff_t disksize; | |
3023 | ||
3024 | __clear_bit(BH_New, &bh_result->b_state); | |
3025 | - ext_debug("blocks %u/%lu requested for inode %u\n", | |
3026 | + ext_debug("blocks %u/%u requested for inode %u\n", | |
3027 | iblock, max_blocks, inode->i_ino); | |
3028 | ||
3029 | /* check in cache */ | |
3030 | - goal = ext4_ext_in_cache(inode, iblock, &newex); | |
3031 | - if (goal) { | |
3032 | - if (goal == EXT4_EXT_CACHE_GAP) { | |
3033 | + cache_type = ext4_ext_in_cache(inode, iblock, &newex); | |
3034 | + if (cache_type) { | |
3035 | + if (cache_type == EXT4_EXT_CACHE_GAP) { | |
3036 | if (!create) { | |
3037 | /* | |
3038 | * block isn't allocated yet and | |
3039 | @@ -2597,7 +2593,7 @@ int ext4_ext_get_blocks(handle_t *handle | |
3040 | goto out2; | |
3041 | } | |
3042 | /* we should allocate requested block */ | |
3043 | - } else if (goal == EXT4_EXT_CACHE_EXTENT) { | |
3044 | + } else if (cache_type == EXT4_EXT_CACHE_EXTENT) { | |
3045 | /* block is already allocated */ | |
3046 | newblock = iblock | |
3047 | - le32_to_cpu(newex.ee_block) | |
3048 | @@ -2696,11 +2692,8 @@ int ext4_ext_get_blocks(handle_t *handle | |
3049 | goto out2; | |
3050 | } | |
3051 | /* | |
3052 | - * Okay, we need to do block allocation. Lazily initialize the block | |
3053 | - * allocation info here if necessary. | |
3054 | + * Okay, we need to do block allocation. | |
3055 | */ | |
3056 | - if (S_ISREG(inode->i_mode) && (!EXT4_I(inode)->i_block_alloc_info)) | |
3057 | - ext4_init_block_alloc_info(inode); | |
3058 | ||
3059 | /* find neighbour allocated blocks */ | |
3060 | ar.lleft = iblock; | |
3061 | @@ -2748,7 +2741,7 @@ int ext4_ext_get_blocks(handle_t *handle | |
3062 | if (!newblock) | |
3063 | goto out2; | |
3064 | ext_debug("allocate new block: goal %llu, found %llu/%lu\n", | |
3065 | - goal, newblock, allocated); | |
3066 | + ar.goal, newblock, allocated); | |
3067 | ||
3068 | /* try to insert new extent into found leaf and return */ | |
3069 | ext4_ext_store_pblock(&newex, newblock); | |
3070 | @@ -2760,7 +2753,7 @@ int ext4_ext_get_blocks(handle_t *handle | |
3071 | /* free data blocks we just allocated */ | |
3072 | /* not a good idea to call discard here directly, | |
3073 | * but otherwise we'd need to call it every free() */ | |
3074 | - ext4_mb_discard_inode_preallocations(inode); | |
3075 | + ext4_discard_preallocations(inode); | |
3076 | ext4_free_blocks(handle, inode, ext_pblock(&newex), | |
3077 | ext4_ext_get_actual_len(&newex), 0); | |
3078 | goto out2; | |
3079 | @@ -2824,7 +2817,7 @@ void ext4_ext_truncate(struct inode *ino | |
3080 | down_write(&EXT4_I(inode)->i_data_sem); | |
3081 | ext4_ext_invalidate_cache(inode); | |
3082 | ||
3083 | - ext4_discard_reservation(inode); | |
3084 | + ext4_discard_preallocations(inode); | |
3085 | ||
3086 | /* | |
3087 | * TODO: optimization is possible here. | |
3088 | @@ -2877,10 +2870,11 @@ static void ext4_falloc_update_inode(str | |
3089 | * Update only when preallocation was requested beyond | |
3090 | * the file size. | |
3091 | */ | |
3092 | - if (!(mode & FALLOC_FL_KEEP_SIZE) && | |
3093 | - new_size > i_size_read(inode)) { | |
3094 | - i_size_write(inode, new_size); | |
3095 | - EXT4_I(inode)->i_disksize = new_size; | |
3096 | + if (!(mode & FALLOC_FL_KEEP_SIZE)) { | |
3097 | + if (new_size > i_size_read(inode)) | |
3098 | + i_size_write(inode, new_size); | |
3099 | + if (new_size > EXT4_I(inode)->i_disksize) | |
3100 | + ext4_update_i_disksize(inode, new_size); | |
3101 | } | |
3102 | ||
3103 | } | |
3104 | @@ -2897,7 +2891,7 @@ long ext4_fallocate(struct inode *inode, | |
3105 | handle_t *handle; | |
3106 | ext4_lblk_t block; | |
3107 | loff_t new_size; | |
3108 | - unsigned long max_blocks; | |
3109 | + unsigned int max_blocks; | |
3110 | int ret = 0; | |
3111 | int ret2 = 0; | |
3112 | int retries = 0; | |
3113 | diff -rup b/fs/ext4//file.c a/fs/ext4///file.c | |
3114 | --- b/fs/ext4/file.c 2009-02-11 14:37:58.000000000 +0100 | |
3115 | +++ a/fs/ext4/file.c 2009-02-10 21:40:11.000000000 +0100 | |
3116 | @@ -31,14 +31,14 @@ | |
3117 | * from ext4_file_open: open gets called at every open, but release | |
3118 | * gets called only when /all/ the files are closed. | |
3119 | */ | |
3120 | -static int ext4_release_file (struct inode * inode, struct file * filp) | |
3121 | +static int ext4_release_file(struct inode *inode, struct file *filp) | |
3122 | { | |
3123 | /* if we are the last writer on the inode, drop the block reservation */ | |
3124 | if ((filp->f_mode & FMODE_WRITE) && | |
3125 | (atomic_read(&inode->i_writecount) == 1)) | |
3126 | { | |
3127 | down_write(&EXT4_I(inode)->i_data_sem); | |
3128 | - ext4_discard_reservation(inode); | |
3129 | + ext4_discard_preallocations(inode); | |
3130 | up_write(&EXT4_I(inode)->i_data_sem); | |
3131 | } | |
3132 | if (is_dx(inode) && filp->private_data) | |
3133 | @@ -162,7 +162,7 @@ const struct inode_operations ext4_file_ | |
3134 | .truncate = ext4_truncate, | |
3135 | .setattr = ext4_setattr, | |
3136 | .getattr = ext4_getattr, | |
3137 | -#ifdef CONFIG_EXT4DEV_FS_XATTR | |
3138 | +#if defined(CONFIG_EXT4_FS_XATTR) || defined(CONFIG_EXT4DEV_FS_XATTR) | |
3139 | .setxattr = generic_setxattr, | |
3140 | .getxattr = generic_getxattr, | |
3141 | .listxattr = ext4_listxattr, | |
3142 | diff -rup b/fs/ext4//fsync.c a/fs/ext4///fsync.c | |
3143 | --- b/fs/ext4/fsync.c 2009-02-11 14:37:58.000000000 +0100 | |
3144 | +++ a/fs/ext4/fsync.c 2009-02-10 21:40:11.000000000 +0100 | |
3145 | @@ -28,6 +28,7 @@ | |
3146 | #include <linux/writeback.h> | |
3147 | #include <linux/jbd2.h> | |
3148 | #include <linux/blkdev.h> | |
3149 | +#include <linux/marker.h> | |
3150 | #include "ext4.h" | |
3151 | #include "ext4_jbd2.h" | |
3152 | ||
3153 | @@ -43,7 +44,7 @@ | |
3154 | * inode to disk. | |
3155 | */ | |
3156 | ||
3157 | -int ext4_sync_file(struct file * file, struct dentry *dentry, int datasync) | |
3158 | +int ext4_sync_file(struct file *file, struct dentry *dentry, int datasync) | |
3159 | { | |
3160 | struct inode *inode = dentry->d_inode; | |
3161 | journal_t *journal = EXT4_SB(inode->i_sb)->s_journal; | |
3162 | @@ -51,6 +52,10 @@ int ext4_sync_file(struct file * file, s | |
3163 | ||
3164 | J_ASSERT(ext4_journal_current_handle() == NULL); | |
3165 | ||
3166 | + trace_mark(ext4_sync_file, "dev %s datasync %d ino %ld parent %ld", | |
3167 | + inode->i_sb->s_id, datasync, inode->i_ino, | |
3168 | + dentry->d_parent->d_inode->i_ino); | |
3169 | + | |
3170 | /* | |
3171 | * data=writeback: | |
3172 | * The caller's filemap_fdatawrite()/wait will sync the data. | |
3173 | diff -rup b/fs/ext4//hash.c a/fs/ext4///hash.c | |
3174 | --- b/fs/ext4/hash.c 2009-02-11 14:37:58.000000000 +0100 | |
3175 | +++ a/fs/ext4/hash.c 2009-02-10 21:40:11.000000000 +0100 | |
3176 | @@ -27,7 +27,7 @@ static void TEA_transform(__u32 buf[4], | |
3177 | sum += DELTA; | |
3178 | b0 += ((b1 << 4)+a) ^ (b1+sum) ^ ((b1 >> 5)+b); | |
3179 | b1 += ((b0 << 4)+c) ^ (b0+sum) ^ ((b0 >> 5)+d); | |
3180 | - } while(--n); | |
3181 | + } while (--n); | |
3182 | ||
3183 | buf[0] += b0; | |
3184 | buf[1] += b1; | |
3185 | @@ -35,23 +35,43 @@ static void TEA_transform(__u32 buf[4], | |
3186 | ||
3187 | ||
3188 | /* The old legacy hash */ | |
3189 | -static __u32 dx_hack_hash (const char *name, int len) | |
3190 | +static __u32 dx_hack_hash_unsigned(const char *name, int len) | |
3191 | { | |
3192 | - __u32 hash0 = 0x12a3fe2d, hash1 = 0x37abe8f9; | |
3193 | + __u32 hash, hash0 = 0x12a3fe2d, hash1 = 0x37abe8f9; | |
3194 | + const unsigned char *ucp = (const unsigned char *) name; | |
3195 | + | |
3196 | + while (len--) { | |
3197 | + hash = hash1 + (hash0 ^ (((int) *ucp++) * 7152373)); | |
3198 | + | |
3199 | + if (hash & 0x80000000) | |
3200 | + hash -= 0x7fffffff; | |
3201 | + hash1 = hash0; | |
3202 | + hash0 = hash; | |
3203 | + } | |
3204 | + return hash0 << 1; | |
3205 | +} | |
3206 | + | |
3207 | +static __u32 dx_hack_hash_signed(const char *name, int len) | |
3208 | +{ | |
3209 | + __u32 hash, hash0 = 0x12a3fe2d, hash1 = 0x37abe8f9; | |
3210 | + const signed char *scp = (const signed char *) name; | |
3211 | + | |
3212 | while (len--) { | |
3213 | - __u32 hash = hash1 + (hash0 ^ (*name++ * 7152373)); | |
3214 | + hash = hash1 + (hash0 ^ (((int) *scp++) * 7152373)); | |
3215 | ||
3216 | - if (hash & 0x80000000) hash -= 0x7fffffff; | |
3217 | + if (hash & 0x80000000) | |
3218 | + hash -= 0x7fffffff; | |
3219 | hash1 = hash0; | |
3220 | hash0 = hash; | |
3221 | } | |
3222 | - return (hash0 << 1); | |
3223 | + return hash0 << 1; | |
3224 | } | |
3225 | ||
3226 | -static void str2hashbuf(const char *msg, int len, __u32 *buf, int num) | |
3227 | +static void str2hashbuf_signed(const char *msg, int len, __u32 *buf, int num) | |
3228 | { | |
3229 | __u32 pad, val; | |
3230 | int i; | |
3231 | + const signed char *scp = (const signed char *) msg; | |
3232 | ||
3233 | pad = (__u32)len | ((__u32)len << 8); | |
3234 | pad |= pad << 16; | |
3235 | @@ -59,10 +79,38 @@ static void str2hashbuf(const char *msg, | |
3236 | val = pad; | |
3237 | if (len > num*4) | |
3238 | len = num * 4; | |
3239 | - for (i=0; i < len; i++) { | |
3240 | + for (i = 0; i < len; i++) { | |
3241 | if ((i % 4) == 0) | |
3242 | val = pad; | |
3243 | - val = msg[i] + (val << 8); | |
3244 | + val = ((int) scp[i]) + (val << 8); | |
3245 | + if ((i % 4) == 3) { | |
3246 | + *buf++ = val; | |
3247 | + val = pad; | |
3248 | + num--; | |
3249 | + } | |
3250 | + } | |
3251 | + if (--num >= 0) | |
3252 | + *buf++ = val; | |
3253 | + while (--num >= 0) | |
3254 | + *buf++ = pad; | |
3255 | +} | |
3256 | + | |
3257 | +static void str2hashbuf_unsigned(const char *msg, int len, __u32 *buf, int num) | |
3258 | +{ | |
3259 | + __u32 pad, val; | |
3260 | + int i; | |
3261 | + const unsigned char *ucp = (const unsigned char *) msg; | |
3262 | + | |
3263 | + pad = (__u32)len | ((__u32)len << 8); | |
3264 | + pad |= pad << 16; | |
3265 | + | |
3266 | + val = pad; | |
3267 | + if (len > num*4) | |
3268 | + len = num * 4; | |
3269 | + for (i = 0; i < len; i++) { | |
3270 | + if ((i % 4) == 0) | |
3271 | + val = pad; | |
3272 | + val = ((int) ucp[i]) + (val << 8); | |
3273 | if ((i % 4) == 3) { | |
3274 | *buf++ = val; | |
3275 | val = pad; | |
3276 | @@ -95,6 +143,8 @@ int ext4fs_dirhash(const char *name, int | |
3277 | const char *p; | |
3278 | int i; | |
3279 | __u32 in[8], buf[4]; | |
3280 | + void (*str2hashbuf)(const char *, int, __u32 *, int) = | |
3281 | + str2hashbuf_signed; | |
3282 | ||
3283 | /* Initialize the default seed for the hash checksum functions */ | |
3284 | buf[0] = 0x67452301; | |
3285 | @@ -104,7 +154,7 @@ int ext4fs_dirhash(const char *name, int | |
3286 | ||
3287 | /* Check to see if the seed is all zero's */ | |
3288 | if (hinfo->seed) { | |
3289 | - for (i=0; i < 4; i++) { | |
3290 | + for (i = 0; i < 4; i++) { | |
3291 | if (hinfo->seed[i]) | |
3292 | break; | |
3293 | } | |
3294 | @@ -113,13 +163,18 @@ int ext4fs_dirhash(const char *name, int | |
3295 | } | |
3296 | ||
3297 | switch (hinfo->hash_version) { | |
3298 | + case DX_HASH_LEGACY_UNSIGNED: | |
3299 | + hash = dx_hack_hash_unsigned(name, len); | |
3300 | + break; | |
3301 | case DX_HASH_LEGACY: | |
3302 | - hash = dx_hack_hash(name, len); | |
3303 | + hash = dx_hack_hash_signed(name, len); | |
3304 | break; | |
3305 | + case DX_HASH_HALF_MD4_UNSIGNED: | |
3306 | + str2hashbuf = str2hashbuf_unsigned; | |
3307 | case DX_HASH_HALF_MD4: | |
3308 | p = name; | |
3309 | while (len > 0) { | |
3310 | - str2hashbuf(p, len, in, 8); | |
3311 | + (*str2hashbuf)(p, len, in, 8); | |
3312 | half_md4_transform(buf, in); | |
3313 | len -= 32; | |
3314 | p += 32; | |
3315 | @@ -127,10 +182,12 @@ int ext4fs_dirhash(const char *name, int | |
3316 | minor_hash = buf[2]; | |
3317 | hash = buf[1]; | |
3318 | break; | |
3319 | + case DX_HASH_TEA_UNSIGNED: | |
3320 | + str2hashbuf = str2hashbuf_unsigned; | |
3321 | case DX_HASH_TEA: | |
3322 | p = name; | |
3323 | while (len > 0) { | |
3324 | - str2hashbuf(p, len, in, 4); | |
3325 | + (*str2hashbuf)(p, len, in, 4); | |
3326 | TEA_transform(buf, in); | |
3327 | len -= 16; | |
3328 | p += 16; | |
3329 | diff -rup b/fs/ext4//ialloc.c a/fs/ext4///ialloc.c | |
3330 | --- b/fs/ext4/ialloc.c 2009-02-11 14:37:58.000000000 +0100 | |
3331 | +++ a/fs/ext4/ialloc.c 2009-02-10 21:40:11.000000000 +0100 | |
3332 | @@ -74,17 +74,17 @@ unsigned ext4_init_inode_bitmap(struct s | |
3333 | /* If checksum is bad mark all blocks and inodes use to prevent | |
3334 | * allocation, essentially implementing a per-group read-only flag. */ | |
3335 | if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) { | |
3336 | - ext4_error(sb, __func__, "Checksum bad for group %lu\n", | |
3337 | + ext4_error(sb, __func__, "Checksum bad for group %u\n", | |
3338 | block_group); | |
3339 | - gdp->bg_free_blocks_count = 0; | |
3340 | - gdp->bg_free_inodes_count = 0; | |
3341 | - gdp->bg_itable_unused = 0; | |
3342 | + ext4_free_blks_set(sb, gdp, 0); | |
3343 | + ext4_free_inodes_set(sb, gdp, 0); | |
3344 | + ext4_itable_unused_set(sb, gdp, 0); | |
3345 | memset(bh->b_data, 0xff, sb->s_blocksize); | |
3346 | return 0; | |
3347 | } | |
3348 | ||
3349 | memset(bh->b_data, 0, (EXT4_INODES_PER_GROUP(sb) + 7) / 8); | |
3350 | - mark_bitmap_end(EXT4_INODES_PER_GROUP(sb), EXT4_BLOCKS_PER_GROUP(sb), | |
3351 | + mark_bitmap_end(EXT4_INODES_PER_GROUP(sb), sb->s_blocksize * 8, | |
3352 | bh->b_data); | |
3353 | ||
3354 | return EXT4_INODES_PER_GROUP(sb); | |
3355 | @@ -111,27 +111,49 @@ ext4_read_inode_bitmap(struct super_bloc | |
3356 | if (unlikely(!bh)) { | |
3357 | ext4_error(sb, __func__, | |
3358 | "Cannot read inode bitmap - " | |
3359 | - "block_group = %lu, inode_bitmap = %llu", | |
3360 | + "block_group = %u, inode_bitmap = %llu", | |
3361 | block_group, bitmap_blk); | |
3362 | return NULL; | |
3363 | } | |
3364 | - if (bh_uptodate_or_lock(bh)) | |
3365 | + if (bitmap_uptodate(bh)) | |
3366 | return bh; | |
3367 | ||
3368 | + lock_buffer(bh); | |
3369 | + if (bitmap_uptodate(bh)) { | |
3370 | + unlock_buffer(bh); | |
3371 | + return bh; | |
3372 | + } | |
3373 | spin_lock(sb_bgl_lock(EXT4_SB(sb), block_group)); | |
3374 | if (desc->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)) { | |
3375 | ext4_init_inode_bitmap(sb, bh, block_group, desc); | |
3376 | + set_bitmap_uptodate(bh); | |
3377 | set_buffer_uptodate(bh); | |
3378 | - unlock_buffer(bh); | |
3379 | spin_unlock(sb_bgl_lock(EXT4_SB(sb), block_group)); | |
3380 | + unlock_buffer(bh); | |
3381 | return bh; | |
3382 | } | |
3383 | spin_unlock(sb_bgl_lock(EXT4_SB(sb), block_group)); | |
3384 | + if (buffer_uptodate(bh)) { | |
3385 | + /* | |
3386 | + * if not uninit if bh is uptodate, | |
3387 | + * bitmap is also uptodate | |
3388 | + */ | |
3389 | + set_bitmap_uptodate(bh); | |
3390 | + unlock_buffer(bh); | |
3391 | + return bh; | |
3392 | + } | |
3393 | + /* | |
3394 | + * submit the buffer_head for read. We can | |
3395 | + * safely mark the bitmap as uptodate now. | |
3396 | + * We do it here so the bitmap uptodate bit | |
3397 | + * get set with buffer lock held. | |
3398 | + */ | |
3399 | + set_bitmap_uptodate(bh); | |
3400 | if (bh_submit_read(bh) < 0) { | |
3401 | put_bh(bh); | |
3402 | ext4_error(sb, __func__, | |
3403 | "Cannot read inode bitmap - " | |
3404 | - "block_group = %lu, inode_bitmap = %llu", | |
3405 | + "block_group = %u, inode_bitmap = %llu", | |
3406 | block_group, bitmap_blk); | |
3407 | return NULL; | |
3408 | } | |
3409 | @@ -154,39 +176,40 @@ ext4_read_inode_bitmap(struct super_bloc | |
3410 | * though), and then we'd have two inodes sharing the | |
3411 | * same inode number and space on the harddisk. | |
3412 | */ | |
3413 | -void ext4_free_inode (handle_t *handle, struct inode * inode) | |
3414 | +void ext4_free_inode(handle_t *handle, struct inode *inode) | |
3415 | { | |
3416 | - struct super_block * sb = inode->i_sb; | |
3417 | + struct super_block *sb = inode->i_sb; | |
3418 | int is_directory; | |
3419 | unsigned long ino; | |
3420 | struct buffer_head *bitmap_bh = NULL; | |
3421 | struct buffer_head *bh2; | |
3422 | ext4_group_t block_group; | |
3423 | unsigned long bit; | |
3424 | - struct ext4_group_desc * gdp; | |
3425 | - struct ext4_super_block * es; | |
3426 | + struct ext4_group_desc *gdp; | |
3427 | + struct ext4_super_block *es; | |
3428 | struct ext4_sb_info *sbi; | |
3429 | - int fatal = 0, err; | |
3430 | + int fatal = 0, err, count; | |
3431 | ext4_group_t flex_group; | |
3432 | ||
3433 | if (atomic_read(&inode->i_count) > 1) { | |
3434 | - printk ("ext4_free_inode: inode has count=%d\n", | |
3435 | - atomic_read(&inode->i_count)); | |
3436 | + printk(KERN_ERR "ext4_free_inode: inode has count=%d\n", | |
3437 | + atomic_read(&inode->i_count)); | |
3438 | return; | |
3439 | } | |
3440 | if (inode->i_nlink) { | |
3441 | - printk ("ext4_free_inode: inode has nlink=%d\n", | |
3442 | - inode->i_nlink); | |
3443 | + printk(KERN_ERR "ext4_free_inode: inode has nlink=%d\n", | |
3444 | + inode->i_nlink); | |
3445 | return; | |
3446 | } | |
3447 | if (!sb) { | |
3448 | - printk("ext4_free_inode: inode on nonexistent device\n"); | |
3449 | + printk(KERN_ERR "ext4_free_inode: inode on " | |
3450 | + "nonexistent device\n"); | |
3451 | return; | |
3452 | } | |
3453 | sbi = EXT4_SB(sb); | |
3454 | ||
3455 | ino = inode->i_ino; | |
3456 | - ext4_debug ("freeing inode %lu\n", ino); | |
3457 | + ext4_debug("freeing inode %lu\n", ino); | |
3458 | ||
3459 | /* | |
3460 | * Note: we must free any quota before locking the superblock, | |
3461 | @@ -200,12 +223,12 @@ void ext4_free_inode (handle_t *handle, | |
3462 | is_directory = S_ISDIR(inode->i_mode); | |
3463 | ||
3464 | /* Do this BEFORE marking the inode not in use or returning an error */ | |
3465 | - clear_inode (inode); | |
3466 | + clear_inode(inode); | |
3467 | ||
3468 | es = EXT4_SB(sb)->s_es; | |
3469 | if (ino < EXT4_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) { | |
3470 | - ext4_error (sb, "ext4_free_inode", | |
3471 | - "reserved or nonexistent inode %lu", ino); | |
3472 | + ext4_error(sb, "ext4_free_inode", | |
3473 | + "reserved or nonexistent inode %lu", ino); | |
3474 | goto error_return; | |
3475 | } | |
3476 | block_group = (ino - 1) / EXT4_INODES_PER_GROUP(sb); | |
3477 | @@ -222,10 +245,10 @@ void ext4_free_inode (handle_t *handle, | |
3478 | /* Ok, now we can actually update the inode bitmaps.. */ | |
3479 | if (!ext4_clear_bit_atomic(sb_bgl_lock(sbi, block_group), | |
3480 | bit, bitmap_bh->b_data)) | |
3481 | - ext4_error (sb, "ext4_free_inode", | |
3482 | - "bit already cleared for inode %lu", ino); | |
3483 | + ext4_error(sb, "ext4_free_inode", | |
3484 | + "bit already cleared for inode %lu", ino); | |
3485 | else { | |
3486 | - gdp = ext4_get_group_desc (sb, block_group, &bh2); | |
3487 | + gdp = ext4_get_group_desc(sb, block_group, &bh2); | |
3488 | ||
3489 | BUFFER_TRACE(bh2, "get_write_access"); | |
3490 | fatal = ext4_journal_get_write_access(handle, bh2); | |
3491 | @@ -233,9 +256,12 @@ void ext4_free_inode (handle_t *handle, | |
3492 | ||
3493 | if (gdp) { | |
3494 | spin_lock(sb_bgl_lock(sbi, block_group)); | |
3495 | - le16_add_cpu(&gdp->bg_free_inodes_count, 1); | |
3496 | - if (is_directory) | |
3497 | - le16_add_cpu(&gdp->bg_used_dirs_count, -1); | |
3498 | + count = ext4_free_inodes_count(sb, gdp) + 1; | |
3499 | + ext4_free_inodes_set(sb, gdp, count); | |
3500 | + if (is_directory) { | |
3501 | + count = ext4_used_dirs_count(sb, gdp) - 1; | |
3502 | + ext4_used_dirs_set(sb, gdp, count); | |
3503 | + } | |
3504 | gdp->bg_checksum = ext4_group_desc_csum(sbi, | |
3505 | block_group, gdp); | |
3506 | spin_unlock(sb_bgl_lock(sbi, block_group)); | |
3507 | @@ -287,14 +313,14 @@ static int find_group_dir(struct super_b | |
3508 | avefreei = freei / ngroups; | |
3509 | ||
3510 | for (group = 0; group < ngroups; group++) { | |
3511 | - desc = ext4_get_group_desc (sb, group, NULL); | |
3512 | - if (!desc || !desc->bg_free_inodes_count) | |
3513 | + desc = ext4_get_group_desc(sb, group, NULL); | |
3514 | + if (!desc || !ext4_free_inodes_count(sb, desc)) | |
3515 | continue; | |
3516 | - if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei) | |
3517 | + if (ext4_free_inodes_count(sb, desc) < avefreei) | |
3518 | continue; | |
3519 | if (!best_desc || | |
3520 | - (le16_to_cpu(desc->bg_free_blocks_count) > | |
3521 | - le16_to_cpu(best_desc->bg_free_blocks_count))) { | |
3522 | + (ext4_free_blks_count(sb, desc) > | |
3523 | + ext4_free_blks_count(sb, best_desc))) { | |
3524 | *best_group = group; | |
3525 | best_desc = desc; | |
3526 | ret = 0; | |
3527 | @@ -366,7 +392,7 @@ found_flexbg: | |
3528 | for (i = best_flex * flex_size; i < ngroups && | |
3529 | i < (best_flex + 1) * flex_size; i++) { | |
3530 | desc = ext4_get_group_desc(sb, i, &bh); | |
3531 | - if (le16_to_cpu(desc->bg_free_inodes_count)) { | |
3532 | + if (ext4_free_inodes_count(sb, desc)) { | |
3533 | *best_group = i; | |
3534 | goto out; | |
3535 | } | |
3536 | @@ -440,17 +466,17 @@ static int find_group_orlov(struct super | |
3537 | for (i = 0; i < ngroups; i++) { | |
3538 | grp = (parent_group + i) % ngroups; | |
3539 | desc = ext4_get_group_desc(sb, grp, NULL); | |
3540 | - if (!desc || !desc->bg_free_inodes_count) | |
3541 | + if (!desc || !ext4_free_inodes_count(sb, desc)) | |
3542 | continue; | |
3543 | - if (le16_to_cpu(desc->bg_used_dirs_count) >= best_ndir) | |
3544 | + if (ext4_used_dirs_count(sb, desc) >= best_ndir) | |
3545 | continue; | |
3546 | - if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei) | |
3547 | + if (ext4_free_inodes_count(sb, desc) < avefreei) | |
3548 | continue; | |
3549 | - if (le16_to_cpu(desc->bg_free_blocks_count) < avefreeb) | |
3550 | + if (ext4_free_blks_count(sb, desc) < avefreeb) | |
3551 | continue; | |
3552 | *group = grp; | |
3553 | ret = 0; | |
3554 | - best_ndir = le16_to_cpu(desc->bg_used_dirs_count); | |
3555 | + best_ndir = ext4_used_dirs_count(sb, desc); | |
3556 | } | |
3557 | if (ret == 0) | |
3558 | return ret; | |
3559 | @@ -476,13 +502,13 @@ static int find_group_orlov(struct super | |
3560 | for (i = 0; i < ngroups; i++) { | |
3561 | *group = (parent_group + i) % ngroups; | |
3562 | desc = ext4_get_group_desc(sb, *group, NULL); | |
3563 | - if (!desc || !desc->bg_free_inodes_count) | |
3564 | + if (!desc || !ext4_free_inodes_count(sb, desc)) | |
3565 | continue; | |
3566 | - if (le16_to_cpu(desc->bg_used_dirs_count) >= max_dirs) | |
3567 | + if (ext4_used_dirs_count(sb, desc) >= max_dirs) | |
3568 | continue; | |
3569 | - if (le16_to_cpu(desc->bg_free_inodes_count) < min_inodes) | |
3570 | + if (ext4_free_inodes_count(sb, desc) < min_inodes) | |
3571 | continue; | |
3572 | - if (le16_to_cpu(desc->bg_free_blocks_count) < min_blocks) | |
3573 | + if (ext4_free_blks_count(sb, desc) < min_blocks) | |
3574 | continue; | |
3575 | return 0; | |
3576 | } | |
3577 | @@ -491,8 +517,8 @@ fallback: | |
3578 | for (i = 0; i < ngroups; i++) { | |
3579 | *group = (parent_group + i) % ngroups; | |
3580 | desc = ext4_get_group_desc(sb, *group, NULL); | |
3581 | - if (desc && desc->bg_free_inodes_count && | |
3582 | - le16_to_cpu(desc->bg_free_inodes_count) >= avefreei) | |
3583 | + if (desc && ext4_free_inodes_count(sb, desc) && | |
3584 | + ext4_free_inodes_count(sb, desc) >= avefreei) | |
3585 | return 0; | |
3586 | } | |
3587 | ||
3588 | @@ -521,8 +547,8 @@ static int find_group_other(struct super | |
3589 | */ | |
3590 | *group = parent_group; | |
3591 | desc = ext4_get_group_desc(sb, *group, NULL); | |
3592 | - if (desc && le16_to_cpu(desc->bg_free_inodes_count) && | |
3593 | - le16_to_cpu(desc->bg_free_blocks_count)) | |
3594 | + if (desc && ext4_free_inodes_count(sb, desc) && | |
3595 | + ext4_free_blks_count(sb, desc)) | |
3596 | return 0; | |
3597 | ||
3598 | /* | |
3599 | @@ -545,8 +571,8 @@ static int find_group_other(struct super | |
3600 | if (*group >= ngroups) | |
3601 | *group -= ngroups; | |
3602 | desc = ext4_get_group_desc(sb, *group, NULL); | |
3603 | - if (desc && le16_to_cpu(desc->bg_free_inodes_count) && | |
3604 | - le16_to_cpu(desc->bg_free_blocks_count)) | |
3605 | + if (desc && ext4_free_inodes_count(sb, desc) && | |
3606 | + ext4_free_blks_count(sb, desc)) | |
3607 | return 0; | |
3608 | } | |
3609 | ||
3610 | @@ -559,7 +585,7 @@ static int find_group_other(struct super | |
3611 | if (++*group >= ngroups) | |
3612 | *group = 0; | |
3613 | desc = ext4_get_group_desc(sb, *group, NULL); | |
3614 | - if (desc && le16_to_cpu(desc->bg_free_inodes_count)) | |
3615 | + if (desc && ext4_free_inodes_count(sb, desc)) | |
3616 | return 0; | |
3617 | } | |
3618 | ||
3619 | @@ -567,6 +593,79 @@ static int find_group_other(struct super | |
3620 | } | |
3621 | ||
3622 | /* | |
3623 | + * claim the inode from the inode bitmap. If the group | |
3624 | + * is uninit we need to take the groups's sb_bgl_lock | |
3625 | + * and clear the uninit flag. The inode bitmap update | |
3626 | + * and group desc uninit flag clear should be done | |
3627 | + * after holding sb_bgl_lock so that ext4_read_inode_bitmap | |
3628 | + * doesn't race with the ext4_claim_inode | |
3629 | + */ | |
3630 | +static int ext4_claim_inode(struct super_block *sb, | |
3631 | + struct buffer_head *inode_bitmap_bh, | |
3632 | + unsigned long ino, ext4_group_t group, int mode) | |
3633 | +{ | |
3634 | + int free = 0, retval = 0, count; | |
3635 | + struct ext4_sb_info *sbi = EXT4_SB(sb); | |
3636 | + struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); | |
3637 | + | |
3638 | + spin_lock(sb_bgl_lock(sbi, group)); | |
3639 | + if (ext4_set_bit(ino, inode_bitmap_bh->b_data)) { | |
3640 | + /* not a free inode */ | |
3641 | + retval = 1; | |
3642 | + goto err_ret; | |
3643 | + } | |
3644 | + ino++; | |
3645 | + if ((group == 0 && ino < EXT4_FIRST_INO(sb)) || | |
3646 | + ino > EXT4_INODES_PER_GROUP(sb)) { | |
3647 | + spin_unlock(sb_bgl_lock(sbi, group)); | |
3648 | + ext4_error(sb, __func__, | |
3649 | + "reserved inode or inode > inodes count - " | |
3650 | + "block_group = %u, inode=%lu", group, | |
3651 | + ino + group * EXT4_INODES_PER_GROUP(sb)); | |
3652 | + return 1; | |
3653 | + } | |
3654 | + /* If we didn't allocate from within the initialized part of the inode | |
3655 | + * table then we need to initialize up to this inode. */ | |
3656 | + if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) { | |
3657 | + | |
3658 | + if (gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)) { | |
3659 | + gdp->bg_flags &= cpu_to_le16(~EXT4_BG_INODE_UNINIT); | |
3660 | + /* When marking the block group with | |
3661 | + * ~EXT4_BG_INODE_UNINIT we don't want to depend | |
3662 | + * on the value of bg_itable_unused even though | |
3663 | + * mke2fs could have initialized the same for us. | |
3664 | + * Instead we calculated the value below | |
3665 | + */ | |
3666 | + | |
3667 | + free = 0; | |
3668 | + } else { | |
3669 | + free = EXT4_INODES_PER_GROUP(sb) - | |
3670 | + ext4_itable_unused_count(sb, gdp); | |
3671 | + } | |
3672 | + | |
3673 | + /* | |
3674 | + * Check the relative inode number against the last used | |
3675 | + * relative inode number in this group. if it is greater | |
3676 | + * we need to update the bg_itable_unused count | |
3677 | + * | |
3678 | + */ | |
3679 | + if (ino > free) | |
3680 | + ext4_itable_unused_set(sb, gdp, | |
3681 | + (EXT4_INODES_PER_GROUP(sb) - ino)); | |
3682 | + } | |
3683 | + count = ext4_free_inodes_count(sb, gdp) - 1; | |
3684 | + ext4_free_inodes_set(sb, gdp, count); | |
3685 | + if (S_ISDIR(mode)) { | |
3686 | + count = ext4_used_dirs_count(sb, gdp) + 1; | |
3687 | + ext4_used_dirs_set(sb, gdp, count); | |
3688 | + } | |
3689 | + gdp->bg_checksum = ext4_group_desc_csum(sbi, group, gdp); | |
3690 | +err_ret: | |
3691 | + spin_unlock(sb_bgl_lock(sbi, group)); | |
3692 | + return retval; | |
3693 | +} | |
3694 | + | |
3695 | +/* | |
3696 | * There are two policies for allocating an inode. If the new inode is | |
3697 | * a directory, then a forward search is made for a block group with both | |
3698 | * free space and a low directory-to-inode ratio; if that fails, then of | |
3699 | @@ -576,16 +675,16 @@ static int find_group_other(struct super | |
3700 | * For other inodes, search forward from the parent directory's block | |
3701 | * group to find a free inode. | |
3702 | */ | |
3703 | -struct inode *ext4_new_inode(handle_t *handle, struct inode * dir, int mode) | |
3704 | +struct inode *ext4_new_inode(handle_t *handle, struct inode *dir, int mode) | |
3705 | { | |
3706 | struct super_block *sb; | |
3707 | - struct buffer_head *bitmap_bh = NULL; | |
3708 | - struct buffer_head *bh2; | |
3709 | + struct buffer_head *inode_bitmap_bh = NULL; | |
3710 | + struct buffer_head *group_desc_bh; | |
3711 | ext4_group_t group = 0; | |
3712 | unsigned long ino = 0; | |
3713 | - struct inode * inode; | |
3714 | - struct ext4_group_desc * gdp = NULL; | |
3715 | - struct ext4_super_block * es; | |
3716 | + struct inode *inode; | |
3717 | + struct ext4_group_desc *gdp = NULL; | |
3718 | + struct ext4_super_block *es; | |
3719 | struct ext4_inode_info *ei; | |
3720 | struct ext4_sb_info *sbi; | |
3721 | int ret2, err = 0; | |
3722 | @@ -613,7 +712,7 @@ struct inode *ext4_new_inode(handle_t *h | |
3723 | } | |
3724 | ||
3725 | if (S_ISDIR(mode)) { | |
3726 | - if (test_opt (sb, OLDALLOC)) | |
3727 | + if (test_opt(sb, OLDALLOC)) | |
3728 | ret2 = find_group_dir(sb, dir, &group); | |
3729 | else | |
3730 | ret2 = find_group_orlov(sb, dir, &group); | |
3731 | @@ -628,40 +727,50 @@ got_group: | |
3732 | for (i = 0; i < sbi->s_groups_count; i++) { | |
3733 | err = -EIO; | |
3734 | ||
3735 | - gdp = ext4_get_group_desc(sb, group, &bh2); | |
3736 | + gdp = ext4_get_group_desc(sb, group, &group_desc_bh); | |
3737 | if (!gdp) | |
3738 | goto fail; | |
3739 | ||
3740 | - brelse(bitmap_bh); | |
3741 | - bitmap_bh = ext4_read_inode_bitmap(sb, group); | |
3742 | - if (!bitmap_bh) | |
3743 | + brelse(inode_bitmap_bh); | |
3744 | + inode_bitmap_bh = ext4_read_inode_bitmap(sb, group); | |
3745 | + if (!inode_bitmap_bh) | |
3746 | goto fail; | |
3747 | ||
3748 | ino = 0; | |
3749 | ||
3750 | repeat_in_this_group: | |
3751 | ino = ext4_find_next_zero_bit((unsigned long *) | |
3752 | - bitmap_bh->b_data, EXT4_INODES_PER_GROUP(sb), ino); | |
3753 | + inode_bitmap_bh->b_data, | |
3754 | + EXT4_INODES_PER_GROUP(sb), ino); | |
3755 | if (ino < EXT4_INODES_PER_GROUP(sb)) { | |
3756 | ||
3757 | - BUFFER_TRACE(bitmap_bh, "get_write_access"); | |
3758 | - err = ext4_journal_get_write_access(handle, bitmap_bh); | |
3759 | + BUFFER_TRACE(inode_bitmap_bh, "get_write_access"); | |
3760 | + err = ext4_journal_get_write_access(handle, | |
3761 | + inode_bitmap_bh); | |
3762 | if (err) | |
3763 | goto fail; | |
3764 | ||
3765 | - if (!ext4_set_bit_atomic(sb_bgl_lock(sbi, group), | |
3766 | - ino, bitmap_bh->b_data)) { | |
3767 | + BUFFER_TRACE(group_desc_bh, "get_write_access"); | |
3768 | + err = ext4_journal_get_write_access(handle, | |
3769 | + group_desc_bh); | |
3770 | + if (err) | |
3771 | + goto fail; | |
3772 | + if (!ext4_claim_inode(sb, inode_bitmap_bh, | |
3773 | + ino, group, mode)) { | |
3774 | /* we won it */ | |
3775 | - BUFFER_TRACE(bitmap_bh, | |
3776 | + BUFFER_TRACE(inode_bitmap_bh, | |
3777 | "call ext4_journal_dirty_metadata"); | |
3778 | err = ext4_journal_dirty_metadata(handle, | |
3779 | - bitmap_bh); | |
3780 | + inode_bitmap_bh); | |
3781 | if (err) | |
3782 | goto fail; | |
3783 | + /* zero bit is inode number 1*/ | |
3784 | + ino++; | |
3785 | goto got; | |
3786 | } | |
3787 | /* we lost it */ | |
3788 | - jbd2_journal_release_buffer(handle, bitmap_bh); | |
3789 | + jbd2_journal_release_buffer(handle, inode_bitmap_bh); | |
3790 | + jbd2_journal_release_buffer(handle, group_desc_bh); | |
3791 | ||
3792 | if (++ino < EXT4_INODES_PER_GROUP(sb)) | |
3793 | goto repeat_in_this_group; | |
3794 | @@ -681,30 +790,16 @@ repeat_in_this_group: | |
3795 | goto out; | |
3796 | ||
3797 | got: | |
3798 | - ino++; | |
3799 | - if ((group == 0 && ino < EXT4_FIRST_INO(sb)) || | |
3800 | - ino > EXT4_INODES_PER_GROUP(sb)) { | |
3801 | - ext4_error(sb, __func__, | |
3802 | - "reserved inode or inode > inodes count - " | |
3803 | - "block_group = %lu, inode=%lu", group, | |
3804 | - ino + group * EXT4_INODES_PER_GROUP(sb)); | |
3805 | - err = -EIO; | |
3806 | - goto fail; | |
3807 | - } | |
3808 | - | |
3809 | - BUFFER_TRACE(bh2, "get_write_access"); | |
3810 | - err = ext4_journal_get_write_access(handle, bh2); | |
3811 | - if (err) goto fail; | |
3812 | - | |
3813 | /* We may have to initialize the block bitmap if it isn't already */ | |
3814 | if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM) && | |
3815 | gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { | |
3816 | - struct buffer_head *block_bh = ext4_read_block_bitmap(sb, group); | |
3817 | + struct buffer_head *block_bitmap_bh; | |
3818 | ||
3819 | - BUFFER_TRACE(block_bh, "get block bitmap access"); | |
3820 | - err = ext4_journal_get_write_access(handle, block_bh); | |
3821 | + block_bitmap_bh = ext4_read_block_bitmap(sb, group); | |
3822 | + BUFFER_TRACE(block_bitmap_bh, "get block bitmap access"); | |
3823 | + err = ext4_journal_get_write_access(handle, block_bitmap_bh); | |
3824 | if (err) { | |
3825 | - brelse(block_bh); | |
3826 | + brelse(block_bitmap_bh); | |
3827 | goto fail; | |
3828 | } | |
3829 | ||
3830 | @@ -712,9 +807,9 @@ got: | |
3831 | spin_lock(sb_bgl_lock(sbi, group)); | |
3832 | /* recheck and clear flag under lock if we still need to */ | |
3833 | if (gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { | |
3834 | - gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT); | |
3835 | free = ext4_free_blocks_after_init(sb, group, gdp); | |
3836 | - gdp->bg_free_blocks_count = cpu_to_le16(free); | |
3837 | + gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT); | |
3838 | + ext4_free_blks_set(sb, gdp, free); | |
3839 | gdp->bg_checksum = ext4_group_desc_csum(sbi, group, | |
3840 | gdp); | |
3841 | } | |
3842 | @@ -722,55 +817,19 @@ got: | |
3843 | ||
3844 | /* Don't need to dirty bitmap block if we didn't change it */ | |
3845 | if (free) { | |
3846 | - BUFFER_TRACE(block_bh, "dirty block bitmap"); | |
3847 | - err = ext4_journal_dirty_metadata(handle, block_bh); | |
3848 | + BUFFER_TRACE(block_bitmap_bh, "dirty block bitmap"); | |
3849 | + err = ext4_journal_dirty_metadata(handle, | |
3850 | + block_bitmap_bh); | |
3851 | } | |
3852 | ||
3853 | - brelse(block_bh); | |
3854 | + brelse(block_bitmap_bh); | |
3855 | if (err) | |
3856 | goto fail; | |
3857 | } | |
3858 | - | |
3859 | - spin_lock(sb_bgl_lock(sbi, group)); | |
3860 | - /* If we didn't allocate from within the initialized part of the inode | |
3861 | - * table then we need to initialize up to this inode. */ | |
3862 | - if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) { | |
3863 | - if (gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)) { | |
3864 | - gdp->bg_flags &= cpu_to_le16(~EXT4_BG_INODE_UNINIT); | |
3865 | - | |
3866 | - /* When marking the block group with | |
3867 | - * ~EXT4_BG_INODE_UNINIT we don't want to depend | |
3868 | - * on the value of bg_itable_unused even though | |
3869 | - * mke2fs could have initialized the same for us. | |
3870 | - * Instead we calculated the value below | |
3871 | - */ | |
3872 | - | |
3873 | - free = 0; | |
3874 | - } else { | |
3875 | - free = EXT4_INODES_PER_GROUP(sb) - | |
3876 | - le16_to_cpu(gdp->bg_itable_unused); | |
3877 | - } | |
3878 | - | |
3879 | - /* | |
3880 | - * Check the relative inode number against the last used | |
3881 | - * relative inode number in this group. if it is greater | |
3882 | - * we need to update the bg_itable_unused count | |
3883 | - * | |
3884 | - */ | |
3885 | - if (ino > free) | |
3886 | - gdp->bg_itable_unused = | |
3887 | - cpu_to_le16(EXT4_INODES_PER_GROUP(sb) - ino); | |
3888 | - } | |
3889 | - | |
3890 | - le16_add_cpu(&gdp->bg_free_inodes_count, -1); | |
3891 | - if (S_ISDIR(mode)) { | |
3892 | - le16_add_cpu(&gdp->bg_used_dirs_count, 1); | |
3893 | - } | |
3894 | - gdp->bg_checksum = ext4_group_desc_csum(sbi, group, gdp); | |
3895 | - spin_unlock(sb_bgl_lock(sbi, group)); | |
3896 | - BUFFER_TRACE(bh2, "call ext4_journal_dirty_metadata"); | |
3897 | - err = ext4_journal_dirty_metadata(handle, bh2); | |
3898 | - if (err) goto fail; | |
3899 | + BUFFER_TRACE(group_desc_bh, "call ext4_journal_dirty_metadata"); | |
3900 | + err = ext4_journal_dirty_metadata(handle, group_desc_bh); | |
3901 | + if (err) | |
3902 | + goto fail; | |
3903 | ||
3904 | percpu_counter_dec(&sbi->s_freeinodes_counter); | |
3905 | if (S_ISDIR(mode)) | |
3906 | @@ -784,15 +843,15 @@ got: | |
3907 | spin_unlock(sb_bgl_lock(sbi, flex_group)); | |
3908 | } | |
3909 | ||
3910 | - inode->i_uid = current->fsuid; | |
3911 | - if (test_opt (sb, GRPID)) | |
3912 | + inode->i_uid = current_fsuid(); | |
3913 | + if (test_opt(sb, GRPID)) | |
3914 | inode->i_gid = dir->i_gid; | |
3915 | else if (dir->i_mode & S_ISGID) { | |
3916 | inode->i_gid = dir->i_gid; | |
3917 | if (S_ISDIR(mode)) | |
3918 | mode |= S_ISGID; | |
3919 | } else | |
3920 | - inode->i_gid = current->fsgid; | |
3921 | + inode->i_gid = current_fsgid(); | |
3922 | inode->i_mode = mode; | |
3923 | ||
3924 | inode->i_ino = ino + group * EXT4_INODES_PER_GROUP(sb); | |
3925 | @@ -818,7 +877,6 @@ got: | |
3926 | ei->i_flags &= ~EXT4_DIRSYNC_FL; | |
3927 | ei->i_file_acl = 0; | |
3928 | ei->i_dtime = 0; | |
3929 | - ei->i_block_alloc_info = NULL; | |
3930 | ei->i_block_group = group; | |
3931 | ||
3932 | ext4_set_inode_flags(inode); | |
3933 | @@ -834,7 +892,7 @@ got: | |
3934 | ei->i_extra_isize = EXT4_SB(sb)->s_want_extra_isize; | |
3935 | ||
3936 | ret = inode; | |
3937 | - if(DQUOT_ALLOC_INODE(inode)) { | |
3938 | + if (DQUOT_ALLOC_INODE(inode)) { | |
3939 | err = -EDQUOT; | |
3940 | goto fail_drop; | |
3941 | } | |
3942 | @@ -843,7 +901,7 @@ got: | |
3943 | if (err) | |
3944 | goto fail_free_drop; | |
3945 | ||
3946 | - err = ext4_init_security(handle,inode, dir); | |
3947 | + err = ext4_init_security(handle, inode, dir); | |
3948 | if (err) | |
3949 | goto fail_free_drop; | |
3950 | ||
3951 | @@ -869,7 +927,7 @@ out: | |
3952 | iput(inode); | |
3953 | ret = ERR_PTR(err); | |
3954 | really_out: | |
3955 | - brelse(bitmap_bh); | |
3956 | + brelse(inode_bitmap_bh); | |
3957 | return ret; | |
3958 | ||
3959 | fail_free_drop: | |
3960 | @@ -880,7 +938,7 @@ fail_drop: | |
3961 | inode->i_flags |= S_NOQUOTA; | |
3962 | inode->i_nlink = 0; | |
3963 | iput(inode); | |
3964 | - brelse(bitmap_bh); | |
3965 | + brelse(inode_bitmap_bh); | |
3966 | return ERR_PTR(err); | |
3967 | } | |
3968 | ||
3969 | @@ -961,7 +1019,7 @@ error: | |
3970 | return ERR_PTR(err); | |
3971 | } | |
3972 | ||
3973 | -unsigned long ext4_count_free_inodes (struct super_block * sb) | |
3974 | +unsigned long ext4_count_free_inodes(struct super_block *sb) | |
3975 | { | |
3976 | unsigned long desc_count; | |
3977 | struct ext4_group_desc *gdp; | |
3978 | @@ -976,10 +1034,10 @@ unsigned long ext4_count_free_inodes (st | |
3979 | bitmap_count = 0; | |
3980 | gdp = NULL; | |
3981 | for (i = 0; i < EXT4_SB(sb)->s_groups_count; i++) { | |
3982 | - gdp = ext4_get_group_desc (sb, i, NULL); | |
3983 | + gdp = ext4_get_group_desc(sb, i, NULL); | |
3984 | if (!gdp) | |
3985 | continue; | |
3986 | - desc_count += le16_to_cpu(gdp->bg_free_inodes_count); | |
3987 | + desc_count += ext4_free_inodes_count(sb, gdp); | |
3988 | brelse(bitmap_bh); | |
3989 | bitmap_bh = ext4_read_inode_bitmap(sb, i); | |
3990 | if (!bitmap_bh) | |
3991 | @@ -987,20 +1045,21 @@ unsigned long ext4_count_free_inodes (st | |
3992 | ||
3993 | x = ext4_count_free(bitmap_bh, EXT4_INODES_PER_GROUP(sb) / 8); | |
3994 | printk(KERN_DEBUG "group %lu: stored = %d, counted = %lu\n", | |
3995 | - i, le16_to_cpu(gdp->bg_free_inodes_count), x); | |
3996 | + i, ext4_free_inodes_count(sb, gdp), x); | |
3997 | bitmap_count += x; | |
3998 | } | |
3999 | brelse(bitmap_bh); | |
4000 | - printk("ext4_count_free_inodes: stored = %u, computed = %lu, %lu\n", | |
4001 | - le32_to_cpu(es->s_free_inodes_count), desc_count, bitmap_count); | |
4002 | + printk(KERN_DEBUG "ext4_count_free_inodes: " | |
4003 | + "stored = %u, computed = %lu, %lu\n", | |
4004 | + le32_to_cpu(es->s_free_inodes_count), desc_count, bitmap_count); | |
4005 | return desc_count; | |
4006 | #else | |
4007 | desc_count = 0; | |
4008 | for (i = 0; i < EXT4_SB(sb)->s_groups_count; i++) { | |
4009 | - gdp = ext4_get_group_desc (sb, i, NULL); | |
4010 | + gdp = ext4_get_group_desc(sb, i, NULL); | |
4011 | if (!gdp) | |
4012 | continue; | |
4013 | - desc_count += le16_to_cpu(gdp->bg_free_inodes_count); | |
4014 | + desc_count += ext4_free_inodes_count(sb, gdp); | |
4015 | cond_resched(); | |
4016 | } | |
4017 | return desc_count; | |
4018 | @@ -1008,16 +1067,16 @@ unsigned long ext4_count_free_inodes (st | |
4019 | } | |
4020 | ||
4021 | /* Called at mount-time, super-block is locked */ | |
4022 | -unsigned long ext4_count_dirs (struct super_block * sb) | |
4023 | +unsigned long ext4_count_dirs(struct super_block * sb) | |
4024 | { | |
4025 | unsigned long count = 0; | |
4026 | ext4_group_t i; | |
4027 | ||
4028 | for (i = 0; i < EXT4_SB(sb)->s_groups_count; i++) { | |
4029 | - struct ext4_group_desc *gdp = ext4_get_group_desc (sb, i, NULL); | |
4030 | + struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); | |
4031 | if (!gdp) | |
4032 | continue; | |
4033 | - count += le16_to_cpu(gdp->bg_used_dirs_count); | |
4034 | + count += ext4_used_dirs_count(sb, gdp); | |
4035 | } | |
4036 | return count; | |
4037 | } | |
4038 | diff -rup b/fs/ext4//inode.c a/fs/ext4///inode.c | |
4039 | --- b/fs/ext4/inode.c 2009-02-11 14:37:58.000000000 +0100 | |
4040 | +++ a/fs/ext4/inode.c 2009-02-11 01:08:42.000000000 +0100 | |
4041 | @@ -190,7 +190,7 @@ static int ext4_journal_test_restart(han | |
4042 | /* | |
4043 | * Called at the last iput() if i_nlink is zero. | |
4044 | */ | |
4045 | -void ext4_delete_inode (struct inode * inode) | |
4046 | +void ext4_delete_inode(struct inode *inode) | |
4047 | { | |
4048 | handle_t *handle; | |
4049 | int err; | |
4050 | @@ -330,11 +330,11 @@ static int ext4_block_to_path(struct ino | |
4051 | int final = 0; | |
4052 | ||
4053 | if (i_block < 0) { | |
4054 | - ext4_warning (inode->i_sb, "ext4_block_to_path", "block < 0"); | |
4055 | + ext4_warning(inode->i_sb, "ext4_block_to_path", "block < 0"); | |
4056 | } else if (i_block < direct_blocks) { | |
4057 | offsets[n++] = i_block; | |
4058 | final = direct_blocks; | |
4059 | - } else if ( (i_block -= direct_blocks) < indirect_blocks) { | |
4060 | + } else if ((i_block -= direct_blocks) < indirect_blocks) { | |
4061 | offsets[n++] = EXT4_IND_BLOCK; | |
4062 | offsets[n++] = i_block; | |
4063 | final = ptrs; | |
4064 | @@ -400,14 +400,14 @@ static Indirect *ext4_get_branch(struct | |
4065 | ||
4066 | *err = 0; | |
4067 | /* i_data is not going away, no lock needed */ | |
4068 | - add_chain (chain, NULL, EXT4_I(inode)->i_data + *offsets); | |
4069 | + add_chain(chain, NULL, EXT4_I(inode)->i_data + *offsets); | |
4070 | if (!p->key) | |
4071 | goto no_block; | |
4072 | while (--depth) { | |
4073 | bh = sb_bread(sb, le32_to_cpu(p->key)); | |
4074 | if (!bh) | |
4075 | goto failure; | |
4076 | - add_chain(++p, bh, (__le32*)bh->b_data + *++offsets); | |
4077 | + add_chain(++p, bh, (__le32 *)bh->b_data + *++offsets); | |
4078 | /* Reader: end */ | |
4079 | if (!p->key) | |
4080 | goto no_block; | |
4081 | @@ -443,7 +443,7 @@ no_block: | |
4082 | static ext4_fsblk_t ext4_find_near(struct inode *inode, Indirect *ind) | |
4083 | { | |
4084 | struct ext4_inode_info *ei = EXT4_I(inode); | |
4085 | - __le32 *start = ind->bh ? (__le32*) ind->bh->b_data : ei->i_data; | |
4086 | + __le32 *start = ind->bh ? (__le32 *) ind->bh->b_data : ei->i_data; | |
4087 | __le32 *p; | |
4088 | ext4_fsblk_t bg_start; | |
4089 | ext4_fsblk_t last_block; | |
4090 | @@ -486,18 +486,9 @@ static ext4_fsblk_t ext4_find_near(struc | |
4091 | static ext4_fsblk_t ext4_find_goal(struct inode *inode, ext4_lblk_t block, | |
4092 | Indirect *partial) | |
4093 | { | |
4094 | - struct ext4_block_alloc_info *block_i; | |
4095 | - | |
4096 | - block_i = EXT4_I(inode)->i_block_alloc_info; | |
4097 | - | |
4098 | /* | |
4099 | - * try the heuristic for sequential allocation, | |
4100 | - * failing that at least try to get decent locality. | |
4101 | + * XXX need to get goal block from mballoc's data structures | |
4102 | */ | |
4103 | - if (block_i && (block == block_i->last_alloc_logical_block + 1) | |
4104 | - && (block_i->last_alloc_physical_block != 0)) { | |
4105 | - return block_i->last_alloc_physical_block + 1; | |
4106 | - } | |
4107 | ||
4108 | return ext4_find_near(inode, partial); | |
4109 | } | |
4110 | @@ -514,10 +505,10 @@ static ext4_fsblk_t ext4_find_goal(struc | |
4111 | * return the total number of blocks to be allocate, including the | |
4112 | * direct and indirect blocks. | |
4113 | */ | |
4114 | -static int ext4_blks_to_allocate(Indirect *branch, int k, unsigned long blks, | |
4115 | +static int ext4_blks_to_allocate(Indirect *branch, int k, unsigned int blks, | |
4116 | int blocks_to_boundary) | |
4117 | { | |
4118 | - unsigned long count = 0; | |
4119 | + unsigned int count = 0; | |
4120 | ||
4121 | /* | |
4122 | * Simple case, [t,d]Indirect block(s) has not allocated yet | |
4123 | @@ -555,6 +546,7 @@ static int ext4_alloc_blocks(handle_t *h | |
4124 | int indirect_blks, int blks, | |
4125 | ext4_fsblk_t new_blocks[4], int *err) | |
4126 | { | |
4127 | + struct ext4_allocation_request ar; | |
4128 | int target, i; | |
4129 | unsigned long count = 0, blk_allocated = 0; | |
4130 | int index = 0; | |
4131 | @@ -603,10 +595,15 @@ static int ext4_alloc_blocks(handle_t *h | |
4132 | if (!target) | |
4133 | goto allocated; | |
4134 | /* Now allocate data blocks */ | |
4135 | - count = target; | |
4136 | - /* allocating blocks for data blocks */ | |
4137 | - current_block = ext4_new_blocks(handle, inode, iblock, | |
4138 | - goal, &count, err); | |
4139 | + memset(&ar, 0, sizeof(ar)); | |
4140 | + ar.inode = inode; | |
4141 | + ar.goal = goal; | |
4142 | + ar.len = target; | |
4143 | + ar.logical = iblock; | |
4144 | + ar.flags = EXT4_MB_HINT_DATA; | |
4145 | + | |
4146 | + current_block = ext4_mb_new_blocks(handle, &ar, err); | |
4147 | + | |
4148 | if (*err && (target == blks)) { | |
4149 | /* | |
4150 | * if the allocation failed and we didn't allocate | |
4151 | @@ -622,7 +619,7 @@ static int ext4_alloc_blocks(handle_t *h | |
4152 | */ | |
4153 | new_blocks[index] = current_block; | |
4154 | } | |
4155 | - blk_allocated += count; | |
4156 | + blk_allocated += ar.len; | |
4157 | } | |
4158 | allocated: | |
4159 | /* total number of blocks allocated for direct blocks */ | |
4160 | @@ -630,7 +627,7 @@ allocated: | |
4161 | *err = 0; | |
4162 | return ret; | |
4163 | failed_out: | |
4164 | - for (i = 0; i <index; i++) | |
4165 | + for (i = 0; i < index; i++) | |
4166 | ext4_free_blocks(handle, inode, new_blocks[i], 1, 0); | |
4167 | return ret; | |
4168 | } | |
4169 | @@ -703,7 +700,7 @@ static int ext4_alloc_branch(handle_t *h | |
4170 | branch[n].p = (__le32 *) bh->b_data + offsets[n]; | |
4171 | branch[n].key = cpu_to_le32(new_blocks[n]); | |
4172 | *branch[n].p = branch[n].key; | |
4173 | - if ( n == indirect_blks) { | |
4174 | + if (n == indirect_blks) { | |
4175 | current_block = new_blocks[n]; | |
4176 | /* | |
4177 | * End of chain, update the last new metablock of | |
4178 | @@ -730,7 +727,7 @@ failed: | |
4179 | BUFFER_TRACE(branch[i].bh, "call jbd2_journal_forget"); | |
4180 | ext4_journal_forget(handle, branch[i].bh); | |
4181 | } | |
4182 | - for (i = 0; i <indirect_blks; i++) | |
4183 | + for (i = 0; i < indirect_blks; i++) | |
4184 | ext4_free_blocks(handle, inode, new_blocks[i], 1, 0); | |
4185 | ||
4186 | ext4_free_blocks(handle, inode, new_blocks[i], num, 0); | |
4187 | @@ -757,10 +754,8 @@ static int ext4_splice_branch(handle_t * | |
4188 | { | |
4189 | int i; | |
4190 | int err = 0; | |
4191 | - struct ext4_block_alloc_info *block_i; | |
4192 | ext4_fsblk_t current_block; | |
4193 | ||
4194 | - block_i = EXT4_I(inode)->i_block_alloc_info; | |
4195 | /* | |
4196 | * If we're splicing into a [td]indirect block (as opposed to the | |
4197 | * inode) then we need to get write access to the [td]indirect block | |
4198 | @@ -783,18 +778,7 @@ static int ext4_splice_branch(handle_t * | |
4199 | if (num == 0 && blks > 1) { | |
4200 | current_block = le32_to_cpu(where->key) + 1; | |
4201 | for (i = 1; i < blks; i++) | |
4202 | - *(where->p + i ) = cpu_to_le32(current_block++); | |
4203 | - } | |
4204 | - | |
4205 | - /* | |
4206 | - * update the most recently allocated logical & physical block | |
4207 | - * in i_block_alloc_info, to assist find the proper goal block for next | |
4208 | - * allocation | |
4209 | - */ | |
4210 | - if (block_i) { | |
4211 | - block_i->last_alloc_logical_block = block + blks - 1; | |
4212 | - block_i->last_alloc_physical_block = | |
4213 | - le32_to_cpu(where[num].key) + blks - 1; | |
4214 | + *(where->p + i) = cpu_to_le32(current_block++); | |
4215 | } | |
4216 | ||
4217 | /* We are done with atomic stuff, now do the rest of housekeeping */ | |
4218 | @@ -861,10 +845,10 @@ err_out: | |
4219 | * down_read(&EXT4_I(inode)->i_data_sem) if not allocating file system block | |
4220 | * (ie, create is zero). Otherwise down_write(&EXT4_I(inode)->i_data_sem) | |
4221 | */ | |
4222 | -int ext4_get_blocks_handle(handle_t *handle, struct inode *inode, | |
4223 | - ext4_lblk_t iblock, unsigned long maxblocks, | |
4224 | - struct buffer_head *bh_result, | |
4225 | - int create, int extend_disksize) | |
4226 | +static int ext4_get_blocks_handle(handle_t *handle, struct inode *inode, | |
4227 | + ext4_lblk_t iblock, unsigned int maxblocks, | |
4228 | + struct buffer_head *bh_result, | |
4229 | + int create, int extend_disksize) | |
4230 | { | |
4231 | int err = -EIO; | |
4232 | ext4_lblk_t offsets[4]; | |
4233 | @@ -914,12 +898,8 @@ int ext4_get_blocks_handle(handle_t *han | |
4234 | goto cleanup; | |
4235 | ||
4236 | /* | |
4237 | - * Okay, we need to do block allocation. Lazily initialize the block | |
4238 | - * allocation info here if necessary | |
4239 | + * Okay, we need to do block allocation. | |
4240 | */ | |
4241 | - if (S_ISREG(inode->i_mode) && (!ei->i_block_alloc_info)) | |
4242 | - ext4_init_block_alloc_info(inode); | |
4243 | - | |
4244 | goal = ext4_find_goal(inode, iblock, partial); | |
4245 | ||
4246 | /* the number of blocks need to allocate for [d,t]indirect blocks */ | |
4247 | @@ -1030,19 +1010,20 @@ static void ext4_da_update_reserve_space | |
4248 | BUG_ON(mdb > EXT4_I(inode)->i_reserved_meta_blocks); | |
4249 | mdb_free = EXT4_I(inode)->i_reserved_meta_blocks - mdb; | |
4250 | ||
4251 | - /* Account for allocated meta_blocks */ | |
4252 | - mdb_free -= EXT4_I(inode)->i_allocated_meta_blocks; | |
4253 | - | |
4254 | - /* update fs free blocks counter for truncate case */ | |
4255 | - percpu_counter_add(&sbi->s_freeblocks_counter, mdb_free); | |
4256 | + if (mdb_free) { | |
4257 | + /* Account for allocated meta_blocks */ | |
4258 | + mdb_free -= EXT4_I(inode)->i_allocated_meta_blocks; | |
4259 | + | |
4260 | + /* update fs dirty blocks counter */ | |
4261 | + percpu_counter_sub(&sbi->s_dirtyblocks_counter, mdb_free); | |
4262 | + EXT4_I(inode)->i_allocated_meta_blocks = 0; | |
4263 | + EXT4_I(inode)->i_reserved_meta_blocks = mdb; | |
4264 | + } | |
4265 | ||
4266 | /* update per-inode reservations */ | |
4267 | BUG_ON(used > EXT4_I(inode)->i_reserved_data_blocks); | |
4268 | EXT4_I(inode)->i_reserved_data_blocks -= used; | |
4269 | ||
4270 | - BUG_ON(mdb > EXT4_I(inode)->i_reserved_meta_blocks); | |
4271 | - EXT4_I(inode)->i_reserved_meta_blocks = mdb; | |
4272 | - EXT4_I(inode)->i_allocated_meta_blocks = 0; | |
4273 | spin_unlock(&EXT4_I(inode)->i_block_reservation_lock); | |
4274 | } | |
4275 | ||
4276 | @@ -1069,7 +1050,7 @@ static void ext4_da_update_reserve_space | |
4277 | * It returns the error in case of allocation failure. | |
4278 | */ | |
4279 | int ext4_get_blocks_wrap(handle_t *handle, struct inode *inode, sector_t block, | |
4280 | - unsigned long max_blocks, struct buffer_head *bh, | |
4281 | + unsigned int max_blocks, struct buffer_head *bh, | |
4282 | int create, int extend_disksize, int flag) | |
4283 | { | |
4284 | int retval; | |
4285 | @@ -1241,7 +1222,7 @@ struct buffer_head *ext4_getblk(handle_t | |
4286 | BUFFER_TRACE(bh, "call get_create_access"); | |
4287 | fatal = ext4_journal_get_create_access(handle, bh); | |
4288 | if (!fatal && !buffer_uptodate(bh)) { | |
4289 | - memset(bh->b_data,0,inode->i_sb->s_blocksize); | |
4290 | + memset(bh->b_data, 0, inode->i_sb->s_blocksize); | |
4291 | set_buffer_uptodate(bh); | |
4292 | } | |
4293 | unlock_buffer(bh); | |
4294 | @@ -1266,7 +1247,7 @@ err: | |
4295 | struct buffer_head *ext4_bread(handle_t *handle, struct inode *inode, | |
4296 | ext4_lblk_t block, int create, int *err) | |
4297 | { | |
4298 | - struct buffer_head * bh; | |
4299 | + struct buffer_head *bh; | |
4300 | ||
4301 | bh = ext4_getblk(handle, inode, block, create, err); | |
4302 | if (!bh) | |
4303 | @@ -1282,13 +1263,13 @@ struct buffer_head *ext4_bread(handle_t | |
4304 | return NULL; | |
4305 | } | |
4306 | ||
4307 | -static int walk_page_buffers( handle_t *handle, | |
4308 | - struct buffer_head *head, | |
4309 | - unsigned from, | |
4310 | - unsigned to, | |
4311 | - int *partial, | |
4312 | - int (*fn)( handle_t *handle, | |
4313 | - struct buffer_head *bh)) | |
4314 | +static int walk_page_buffers(handle_t *handle, | |
4315 | + struct buffer_head *head, | |
4316 | + unsigned from, | |
4317 | + unsigned to, | |
4318 | + int *partial, | |
4319 | + int (*fn)(handle_t *handle, | |
4320 | + struct buffer_head *bh)) | |
4321 | { | |
4322 | struct buffer_head *bh; | |
4323 | unsigned block_start, block_end; | |
4324 | @@ -1296,9 +1277,9 @@ static int walk_page_buffers( handle_t * | |
4325 | int err, ret = 0; | |
4326 | struct buffer_head *next; | |
4327 | ||
4328 | - for ( bh = head, block_start = 0; | |
4329 | - ret == 0 && (bh != head || !block_start); | |
4330 | - block_start = block_end, bh = next) | |
4331 | + for (bh = head, block_start = 0; | |
4332 | + ret == 0 && (bh != head || !block_start); | |
4333 | + block_start = block_end, bh = next) | |
4334 | { | |
4335 | next = bh->b_this_page; | |
4336 | block_end = block_start + blocksize; | |
4337 | @@ -1351,23 +1332,23 @@ static int ext4_write_begin(struct file | |
4338 | loff_t pos, unsigned len, unsigned flags, | |
4339 | struct page **pagep, void **fsdata) | |
4340 | { | |
4341 | - struct inode *inode = mapping->host; | |
4342 | + struct inode *inode = mapping->host; | |
4343 | int ret, needed_blocks = ext4_writepage_trans_blocks(inode); | |
4344 | handle_t *handle; | |
4345 | int retries = 0; | |
4346 | - struct page *page; | |
4347 | + struct page *page; | |
4348 | pgoff_t index; | |
4349 | - unsigned from, to; | |
4350 | + unsigned from, to; | |
4351 | ||
4352 | index = pos >> PAGE_CACHE_SHIFT; | |
4353 | - from = pos & (PAGE_CACHE_SIZE - 1); | |
4354 | - to = from + len; | |
4355 | + from = pos & (PAGE_CACHE_SIZE - 1); | |
4356 | + to = from + len; | |
4357 | ||
4358 | retry: | |
4359 | - handle = ext4_journal_start(inode, needed_blocks); | |
4360 | - if (IS_ERR(handle)) { | |
4361 | - ret = PTR_ERR(handle); | |
4362 | - goto out; | |
4363 | + handle = ext4_journal_start(inode, needed_blocks); | |
4364 | + if (IS_ERR(handle)) { | |
4365 | + ret = PTR_ERR(handle); | |
4366 | + goto out; | |
4367 | } | |
4368 | ||
4369 | page = grab_cache_page_write_begin(mapping, index, flags); | |
4370 | @@ -1387,9 +1368,16 @@ retry: | |
4371 | } | |
4372 | ||
4373 | if (ret) { | |
4374 | - unlock_page(page); | |
4375 | + unlock_page(page); | |
4376 | ext4_journal_stop(handle); | |
4377 | - page_cache_release(page); | |
4378 | + page_cache_release(page); | |
4379 | + /* | |
4380 | + * block_write_begin may have instantiated a few blocks | |
4381 | + * outside i_size. Trim these off again. Don't need | |
4382 | + * i_size_read because we hold i_mutex. | |
4383 | + */ | |
4384 | + if (pos + len > inode->i_size) | |
4385 | + vmtruncate(inode, inode->i_size); | |
4386 | } | |
4387 | ||
4388 | if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) | |
4389 | @@ -1426,16 +1414,18 @@ static int ext4_ordered_write_end(struct | |
4390 | ret = ext4_jbd2_file_inode(handle, inode); | |
4391 | ||
4392 | if (ret == 0) { | |
4393 | - /* | |
4394 | - * generic_write_end() will run mark_inode_dirty() if i_size | |
4395 | - * changes. So let's piggyback the i_disksize mark_inode_dirty | |
4396 | - * into that. | |
4397 | - */ | |
4398 | loff_t new_i_size; | |
4399 | ||
4400 | new_i_size = pos + copied; | |
4401 | - if (new_i_size > EXT4_I(inode)->i_disksize) | |
4402 | - EXT4_I(inode)->i_disksize = new_i_size; | |
4403 | + if (new_i_size > EXT4_I(inode)->i_disksize) { | |
4404 | + ext4_update_i_disksize(inode, new_i_size); | |
4405 | + /* We need to mark inode dirty even if | |
4406 | + * new_i_size is less that inode->i_size | |
4407 | + * bu greater than i_disksize.(hint delalloc) | |
4408 | + */ | |
4409 | + ext4_mark_inode_dirty(handle, inode); | |
4410 | + } | |
4411 | + | |
4412 | ret2 = generic_write_end(file, mapping, pos, len, copied, | |
4413 | page, fsdata); | |
4414 | copied = ret2; | |
4415 | @@ -1460,8 +1450,14 @@ static int ext4_writeback_write_end(stru | |
4416 | loff_t new_i_size; | |
4417 | ||
4418 | new_i_size = pos + copied; | |
4419 | - if (new_i_size > EXT4_I(inode)->i_disksize) | |
4420 | - EXT4_I(inode)->i_disksize = new_i_size; | |
4421 | + if (new_i_size > EXT4_I(inode)->i_disksize) { | |
4422 | + ext4_update_i_disksize(inode, new_i_size); | |
4423 | + /* We need to mark inode dirty even if | |
4424 | + * new_i_size is less that inode->i_size | |
4425 | + * bu greater than i_disksize.(hint delalloc) | |
4426 | + */ | |
4427 | + ext4_mark_inode_dirty(handle, inode); | |
4428 | + } | |
4429 | ||
4430 | ret2 = generic_write_end(file, mapping, pos, len, copied, | |
4431 | page, fsdata); | |
4432 | @@ -1486,6 +1482,7 @@ static int ext4_journalled_write_end(str | |
4433 | int ret = 0, ret2; | |
4434 | int partial = 0; | |
4435 | unsigned from, to; | |
4436 | + loff_t new_i_size; | |
4437 | ||
4438 | from = pos & (PAGE_CACHE_SIZE - 1); | |
4439 | to = from + len; | |
4440 | @@ -1500,11 +1497,12 @@ static int ext4_journalled_write_end(str | |
4441 | to, &partial, write_end_fn); | |
4442 | if (!partial) | |
4443 | SetPageUptodate(page); | |
4444 | - if (pos+copied > inode->i_size) | |
4445 | + new_i_size = pos + copied; | |
4446 | + if (new_i_size > inode->i_size) | |
4447 | i_size_write(inode, pos+copied); | |
4448 | EXT4_I(inode)->i_state |= EXT4_STATE_JDATA; | |
4449 | - if (inode->i_size > EXT4_I(inode)->i_disksize) { | |
4450 | - EXT4_I(inode)->i_disksize = inode->i_size; | |
4451 | + if (new_i_size > EXT4_I(inode)->i_disksize) { | |
4452 | + ext4_update_i_disksize(inode, new_i_size); | |
4453 | ret2 = ext4_mark_inode_dirty(handle, inode); | |
4454 | if (!ret) | |
4455 | ret = ret2; | |
4456 | @@ -1521,6 +1519,7 @@ static int ext4_journalled_write_end(str | |
4457 | ||
4458 | static int ext4_da_reserve_space(struct inode *inode, int nrblocks) | |
4459 | { | |
4460 | + int retries = 0; | |
4461 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
4462 | unsigned long md_needed, mdblocks, total = 0; | |
4463 | ||
4464 | @@ -1529,6 +1528,7 @@ static int ext4_da_reserve_space(struct | |
4465 | * in order to allocate nrblocks | |
4466 | * worse case is one extent per block | |
4467 | */ | |
4468 | +repeat: | |
4469 | spin_lock(&EXT4_I(inode)->i_block_reservation_lock); | |
4470 | total = EXT4_I(inode)->i_reserved_data_blocks + nrblocks; | |
4471 | mdblocks = ext4_calc_metadata_amount(inode, total); | |
4472 | @@ -1537,13 +1537,14 @@ static int ext4_da_reserve_space(struct | |
4473 | md_needed = mdblocks - EXT4_I(inode)->i_reserved_meta_blocks; | |
4474 | total = md_needed + nrblocks; | |
4475 | ||
4476 | - if (ext4_has_free_blocks(sbi, total) < total) { | |
4477 | + if (ext4_claim_free_blocks(sbi, total)) { | |
4478 | spin_unlock(&EXT4_I(inode)->i_block_reservation_lock); | |
4479 | + if (ext4_should_retry_alloc(inode->i_sb, &retries)) { | |
4480 | + yield(); | |
4481 | + goto repeat; | |
4482 | + } | |
4483 | return -ENOSPC; | |
4484 | } | |
4485 | - /* reduce fs free blocks counter */ | |
4486 | - percpu_counter_sub(&sbi->s_freeblocks_counter, total); | |
4487 | - | |
4488 | EXT4_I(inode)->i_reserved_data_blocks += nrblocks; | |
4489 | EXT4_I(inode)->i_reserved_meta_blocks = mdblocks; | |
4490 | ||
4491 | @@ -1585,8 +1586,8 @@ static void ext4_da_release_space(struct | |
4492 | ||
4493 | release = to_free + mdb_free; | |
4494 | ||
4495 | - /* update fs free blocks counter for truncate case */ | |
4496 | - percpu_counter_add(&sbi->s_freeblocks_counter, release); | |
4497 | + /* update fs dirty blocks counter for truncate case */ | |
4498 | + percpu_counter_sub(&sbi->s_dirtyblocks_counter, release); | |
4499 | ||
4500 | /* update per-inode reservations */ | |
4501 | BUG_ON(to_free > EXT4_I(inode)->i_reserved_data_blocks); | |
4502 | @@ -1629,7 +1630,8 @@ struct mpage_da_data { | |
4503 | get_block_t *get_block; | |
4504 | struct writeback_control *wbc; | |
4505 | int io_done; | |
4506 | - long pages_written; | |
4507 | + int pages_written; | |
4508 | + int retval; | |
4509 | }; | |
4510 | ||
4511 | /* | |
4512 | @@ -1648,18 +1650,25 @@ struct mpage_da_data { | |
4513 | */ | |
4514 | static int mpage_da_submit_io(struct mpage_da_data *mpd) | |
4515 | { | |
4516 | - struct address_space *mapping = mpd->inode->i_mapping; | |
4517 | - int ret = 0, err, nr_pages, i; | |
4518 | - unsigned long index, end; | |
4519 | + long pages_skipped; | |
4520 | struct pagevec pvec; | |
4521 | + unsigned long index, end; | |
4522 | + int ret = 0, err, nr_pages, i; | |
4523 | + struct inode *inode = mpd->inode; | |
4524 | + struct address_space *mapping = inode->i_mapping; | |
4525 | ||
4526 | BUG_ON(mpd->next_page <= mpd->first_page); | |
4527 | - pagevec_init(&pvec, 0); | |
4528 | + /* | |
4529 | + * We need to start from the first_page to the next_page - 1 | |
4530 | + * to make sure we also write the mapped dirty buffer_heads. | |
4531 | + * If we look at mpd->lbh.b_blocknr we would only be looking | |
4532 | + * at the currently mapped buffer_heads. | |
4533 | + */ | |
4534 | index = mpd->first_page; | |
4535 | end = mpd->next_page - 1; | |
4536 | ||
4537 | + pagevec_init(&pvec, 0); | |
4538 | while (index <= end) { | |
4539 | - /* XXX: optimize tail */ | |
4540 | nr_pages = pagevec_lookup(&pvec, mapping, index, PAGEVEC_SIZE); | |
4541 | if (nr_pages == 0) | |
4542 | break; | |
4543 | @@ -1671,8 +1680,16 @@ static int mpage_da_submit_io(struct mpa | |
4544 | break; | |
4545 | index++; | |
4546 | ||
4547 | + BUG_ON(!PageLocked(page)); | |
4548 | + BUG_ON(PageWriteback(page)); | |
4549 | + | |
4550 | + pages_skipped = mpd->wbc->pages_skipped; | |
4551 | err = mapping->a_ops->writepage(page, mpd->wbc); | |
4552 | - if (!err) | |
4553 | + if (!err && (pages_skipped == mpd->wbc->pages_skipped)) | |
4554 | + /* | |
4555 | + * have successfully written the page | |
4556 | + * without skipping the same | |
4557 | + */ | |
4558 | mpd->pages_written++; | |
4559 | /* | |
4560 | * In error case, we have to continue because | |
4561 | @@ -1783,6 +1800,57 @@ static inline void __unmap_underlying_bl | |
4562 | unmap_underlying_metadata(bdev, bh->b_blocknr + i); | |
4563 | } | |
4564 | ||
4565 | +static void ext4_da_block_invalidatepages(struct mpage_da_data *mpd, | |
4566 | + sector_t logical, long blk_cnt) | |
4567 | +{ | |
4568 | + int nr_pages, i; | |
4569 | + pgoff_t index, end; | |
4570 | + struct pagevec pvec; | |
4571 | + struct inode *inode = mpd->inode; | |
4572 | + struct address_space *mapping = inode->i_mapping; | |
4573 | + | |
4574 | + index = logical >> (PAGE_CACHE_SHIFT - inode->i_blkbits); | |
4575 | + end = (logical + blk_cnt - 1) >> | |
4576 | + (PAGE_CACHE_SHIFT - inode->i_blkbits); | |
4577 | + while (index <= end) { | |
4578 | + nr_pages = pagevec_lookup(&pvec, mapping, index, PAGEVEC_SIZE); | |
4579 | + if (nr_pages == 0) | |
4580 | + break; | |
4581 | + for (i = 0; i < nr_pages; i++) { | |
4582 | + struct page *page = pvec.pages[i]; | |
4583 | + index = page->index; | |
4584 | + if (index > end) | |
4585 | + break; | |
4586 | + index++; | |
4587 | + | |
4588 | + BUG_ON(!PageLocked(page)); | |
4589 | + BUG_ON(PageWriteback(page)); | |
4590 | + block_invalidatepage(page, 0); | |
4591 | + ClearPageUptodate(page); | |
4592 | + unlock_page(page); | |
4593 | + } | |
4594 | + } | |
4595 | + return; | |
4596 | +} | |
4597 | + | |
4598 | +static void ext4_print_free_blocks(struct inode *inode) | |
4599 | +{ | |
4600 | + struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
4601 | + printk(KERN_EMERG "Total free blocks count %lld\n", | |
4602 | + ext4_count_free_blocks(inode->i_sb)); | |
4603 | + printk(KERN_EMERG "Free/Dirty block details\n"); | |
4604 | + printk(KERN_EMERG "free_blocks=%lld\n", | |
4605 | + (long long)percpu_counter_sum(&sbi->s_freeblocks_counter)); | |
4606 | + printk(KERN_EMERG "dirty_blocks=%lld\n", | |
4607 | + (long long)percpu_counter_sum(&sbi->s_dirtyblocks_counter)); | |
4608 | + printk(KERN_EMERG "Block reservation details\n"); | |
4609 | + printk(KERN_EMERG "i_reserved_data_blocks=%u\n", | |
4610 | + EXT4_I(inode)->i_reserved_data_blocks); | |
4611 | + printk(KERN_EMERG "i_reserved_meta_blocks=%u\n", | |
4612 | + EXT4_I(inode)->i_reserved_meta_blocks); | |
4613 | + return; | |
4614 | +} | |
4615 | + | |
4616 | /* | |
4617 | * mpage_da_map_blocks - go through given space | |
4618 | * | |
4619 | @@ -1792,32 +1860,69 @@ static inline void __unmap_underlying_bl | |
4620 | * The function skips space we know is already mapped to disk blocks. | |
4621 | * | |
4622 | */ | |
4623 | -static void mpage_da_map_blocks(struct mpage_da_data *mpd) | |
4624 | +static int mpage_da_map_blocks(struct mpage_da_data *mpd) | |
4625 | { | |
4626 | int err = 0; | |
4627 | - struct buffer_head *lbh = &mpd->lbh; | |
4628 | - sector_t next = lbh->b_blocknr; | |
4629 | struct buffer_head new; | |
4630 | + struct buffer_head *lbh = &mpd->lbh; | |
4631 | + sector_t next; | |
4632 | ||
4633 | /* | |
4634 | * We consider only non-mapped and non-allocated blocks | |
4635 | */ | |
4636 | if (buffer_mapped(lbh) && !buffer_delay(lbh)) | |
4637 | - return; | |
4638 | - | |
4639 | + return 0; | |
4640 | new.b_state = lbh->b_state; | |
4641 | new.b_blocknr = 0; | |
4642 | new.b_size = lbh->b_size; | |
4643 | - | |
4644 | + next = lbh->b_blocknr; | |
4645 | /* | |
4646 | * If we didn't accumulate anything | |
4647 | * to write simply return | |
4648 | */ | |
4649 | if (!new.b_size) | |
4650 | - return; | |
4651 | + return 0; | |
4652 | err = mpd->get_block(mpd->inode, next, &new, 1); | |
4653 | - if (err) | |
4654 | - return; | |
4655 | + if (err) { | |
4656 | + | |
4657 | + /* If get block returns with error | |
4658 | + * we simply return. Later writepage | |
4659 | + * will redirty the page and writepages | |
4660 | + * will find the dirty page again | |
4661 | + */ | |
4662 | + if (err == -EAGAIN) | |
4663 | + return 0; | |
4664 | + | |
4665 | + if (err == -ENOSPC && | |
4666 | + ext4_count_free_blocks(mpd->inode->i_sb)) { | |
4667 | + mpd->retval = err; | |
4668 | + return 0; | |
4669 | + } | |
4670 | + | |
4671 | + /* | |
4672 | + * get block failure will cause us | |
4673 | + * to loop in writepages. Because | |
4674 | + * a_ops->writepage won't be able to | |
4675 | + * make progress. The page will be redirtied | |
4676 | + * by writepage and writepages will again | |
4677 | + * try to write the same. | |
4678 | + */ | |
4679 | + printk(KERN_EMERG "%s block allocation failed for inode %lu " | |
4680 | + "at logical offset %llu with max blocks " | |
4681 | + "%zd with error %d\n", | |
4682 | + __func__, mpd->inode->i_ino, | |
4683 | + (unsigned long long)next, | |
4684 | + lbh->b_size >> mpd->inode->i_blkbits, err); | |
4685 | + printk(KERN_EMERG "This should not happen.!! " | |
4686 | + "Data will be lost\n"); | |
4687 | + if (err == -ENOSPC) { | |
4688 | + ext4_print_free_blocks(mpd->inode); | |
4689 | + } | |
4690 | + /* invlaidate all the pages */ | |
4691 | + ext4_da_block_invalidatepages(mpd, next, | |
4692 | + lbh->b_size >> mpd->inode->i_blkbits); | |
4693 | + return err; | |
4694 | + } | |
4695 | BUG_ON(new.b_size == 0); | |
4696 | ||
4697 | if (buffer_new(&new)) | |
4698 | @@ -1830,7 +1935,7 @@ static void mpage_da_map_blocks(struct m | |
4699 | if (buffer_delay(lbh) || buffer_unwritten(lbh)) | |
4700 | mpage_put_bnr_to_bhs(mpd, next, &new); | |
4701 | ||
4702 | - return; | |
4703 | + return 0; | |
4704 | } | |
4705 | ||
4706 | #define BH_FLAGS ((1 << BH_Uptodate) | (1 << BH_Mapped) | \ | |
4707 | @@ -1899,8 +2004,8 @@ flush_it: | |
4708 | * We couldn't merge the block to our extent, so we | |
4709 | * need to flush current extent and start new one | |
4710 | */ | |
4711 | - mpage_da_map_blocks(mpd); | |
4712 | - mpage_da_submit_io(mpd); | |
4713 | + if (mpage_da_map_blocks(mpd) == 0) | |
4714 | + mpage_da_submit_io(mpd); | |
4715 | mpd->io_done = 1; | |
4716 | return; | |
4717 | } | |
4718 | @@ -1942,8 +2047,8 @@ static int __mpage_da_writepage(struct p | |
4719 | * and start IO on them using writepage() | |
4720 | */ | |
4721 | if (mpd->next_page != mpd->first_page) { | |
4722 | - mpage_da_map_blocks(mpd); | |
4723 | - mpage_da_submit_io(mpd); | |
4724 | + if (mpage_da_map_blocks(mpd) == 0) | |
4725 | + mpage_da_submit_io(mpd); | |
4726 | /* | |
4727 | * skip rest of the page in the page_vec | |
4728 | */ | |
4729 | @@ -1991,11 +2096,29 @@ static int __mpage_da_writepage(struct p | |
4730 | bh = head; | |
4731 | do { | |
4732 | BUG_ON(buffer_locked(bh)); | |
4733 | + /* | |
4734 | + * We need to try to allocate | |
4735 | + * unmapped blocks in the same page. | |
4736 | + * Otherwise we won't make progress | |
4737 | + * with the page in ext4_da_writepage | |
4738 | + */ | |
4739 | if (buffer_dirty(bh) && | |
4740 | (!buffer_mapped(bh) || buffer_delay(bh))) { | |
4741 | mpage_add_bh_to_extent(mpd, logical, bh); | |
4742 | if (mpd->io_done) | |
4743 | return MPAGE_DA_EXTENT_TAIL; | |
4744 | + } else if (buffer_dirty(bh) && (buffer_mapped(bh))) { | |
4745 | + /* | |
4746 | + * mapped dirty buffer. We need to update | |
4747 | + * the b_state because we look at | |
4748 | + * b_state in mpage_da_map_blocks. We don't | |
4749 | + * update b_size because if we find an | |
4750 | + * unmapped buffer_head later we need to | |
4751 | + * use the b_state flag of that buffer_head. | |
4752 | + */ | |
4753 | + if (mpd->lbh.b_size == 0) | |
4754 | + mpd->lbh.b_state = | |
4755 | + bh->b_state & BH_FLAGS; | |
4756 | } | |
4757 | logical++; | |
4758 | } while ((bh = bh->b_this_page) != head); | |
4759 | @@ -2018,39 +2141,42 @@ static int __mpage_da_writepage(struct p | |
4760 | */ | |
4761 | static int mpage_da_writepages(struct address_space *mapping, | |
4762 | struct writeback_control *wbc, | |
4763 | - get_block_t get_block) | |
4764 | + struct mpage_da_data *mpd) | |
4765 | { | |
4766 | - struct mpage_da_data mpd; | |
4767 | - long to_write; | |
4768 | int ret; | |
4769 | + long nr_to_write; | |
4770 | + pgoff_t index; | |
4771 | ||
4772 | - if (!get_block) | |
4773 | - return generic_writepages(mapping, wbc); | |
4774 | - | |
4775 | - mpd.wbc = wbc; | |
4776 | - mpd.inode = mapping->host; | |
4777 | - mpd.lbh.b_size = 0; | |
4778 | - mpd.lbh.b_state = 0; | |
4779 | - mpd.lbh.b_blocknr = 0; | |
4780 | - mpd.first_page = 0; | |
4781 | - mpd.next_page = 0; | |
4782 | - mpd.get_block = get_block; | |
4783 | - mpd.io_done = 0; | |
4784 | - mpd.pages_written = 0; | |
4785 | ||
4786 | - to_write = wbc->nr_to_write; | |
4787 | ||
4788 | - ret = write_cache_pages(mapping, wbc, __mpage_da_writepage, &mpd); | |
4789 | + if (!mpd->get_block) | |
4790 | + return generic_writepages(mapping, wbc); | |
4791 | ||
4792 | + mpd->lbh.b_size = 0; | |
4793 | + mpd->lbh.b_state = 0; | |
4794 | + mpd->lbh.b_blocknr = 0; | |
4795 | + mpd->first_page = 0; | |
4796 | + mpd->next_page = 0; | |
4797 | + mpd->io_done = 0; | |
4798 | + mpd->pages_written = 0; | |
4799 | + mpd->retval = 0; | |
4800 | + | |
4801 | + nr_to_write = wbc->nr_to_write; | |
4802 | + index = mapping->writeback_index; | |
4803 | + ret = write_cache_pages(mapping, wbc, __mpage_da_writepage, mpd); | |
4804 | + wbc->nr_to_write = nr_to_write; | |
4805 | + mapping->writeback_index = index; | |
4806 | /* | |
4807 | * Handle last extent of pages | |
4808 | */ | |
4809 | - if (!mpd.io_done && mpd.next_page != mpd.first_page) { | |
4810 | - mpage_da_map_blocks(&mpd); | |
4811 | - mpage_da_submit_io(&mpd); | |
4812 | - } | |
4813 | + if (!mpd->io_done && mpd->next_page != mpd->first_page) { | |
4814 | + if (mpage_da_map_blocks(mpd) == 0) | |
4815 | + mpage_da_submit_io(mpd); | |
4816 | ||
4817 | - wbc->nr_to_write = to_write - mpd.pages_written; | |
4818 | + mpd->io_done = 1; | |
4819 | + ret = MPAGE_DA_EXTENT_TAIL; | |
4820 | + } | |
4821 | + wbc->nr_to_write -= mpd->pages_written; | |
4822 | return ret; | |
4823 | } | |
4824 | ||
4825 | @@ -2103,18 +2229,24 @@ static int ext4_da_get_block_write(struc | |
4826 | handle_t *handle = NULL; | |
4827 | ||
4828 | handle = ext4_journal_current_handle(); | |
4829 | - if (!handle) { | |
4830 | - ret = ext4_get_blocks_wrap(handle, inode, iblock, max_blocks, | |
4831 | - bh_result, 0, 0, 0); | |
4832 | - BUG_ON(!ret); | |
4833 | - } else { | |
4834 | - ret = ext4_get_blocks_wrap(handle, inode, iblock, max_blocks, | |
4835 | - bh_result, create, 0, EXT4_DELALLOC_RSVED); | |
4836 | - } | |
4837 | - | |
4838 | + BUG_ON(!handle); | |
4839 | + ret = ext4_get_blocks_wrap(handle, inode, iblock, max_blocks, | |
4840 | + bh_result, create, 0, EXT4_DELALLOC_RSVED); | |
4841 | if (ret > 0) { | |
4842 | + | |
4843 | bh_result->b_size = (ret << inode->i_blkbits); | |
4844 | ||
4845 | + if (ext4_should_order_data(inode)) { | |
4846 | + int retval; | |
4847 | + retval = ext4_jbd2_file_inode(handle, inode); | |
4848 | + if (retval) | |
4849 | + /* | |
4850 | + * Failed to add inode for ordered | |
4851 | + * mode. Don't update file size | |
4852 | + */ | |
4853 | + return retval; | |
4854 | + } | |
4855 | + | |
4856 | /* | |
4857 | * Update on-disk size along with block allocation | |
4858 | * we don't use 'extend_disksize' as size may change | |
4859 | @@ -2124,18 +2256,9 @@ static int ext4_da_get_block_write(struc | |
4860 | if (disksize > i_size_read(inode)) | |
4861 | disksize = i_size_read(inode); | |
4862 | if (disksize > EXT4_I(inode)->i_disksize) { | |
4863 | - /* | |
4864 | - * XXX: replace with spinlock if seen contended -bzzz | |
4865 | - */ | |
4866 | - down_write(&EXT4_I(inode)->i_data_sem); | |
4867 | - if (disksize > EXT4_I(inode)->i_disksize) | |
4868 | - EXT4_I(inode)->i_disksize = disksize; | |
4869 | - up_write(&EXT4_I(inode)->i_data_sem); | |
4870 | - | |
4871 | - if (EXT4_I(inode)->i_disksize == disksize) { | |
4872 | - ret = ext4_mark_inode_dirty(handle, inode); | |
4873 | - return ret; | |
4874 | - } | |
4875 | + ext4_update_i_disksize(inode, disksize); | |
4876 | + ret = ext4_mark_inode_dirty(handle, inode); | |
4877 | + return ret; | |
4878 | } | |
4879 | ret = 0; | |
4880 | } | |
4881 | @@ -2181,7 +2304,7 @@ static int ext4_da_writepage(struct page | |
4882 | { | |
4883 | int ret = 0; | |
4884 | loff_t size; | |
4885 | - unsigned long len; | |
4886 | + unsigned int len; | |
4887 | struct buffer_head *page_bufs; | |
4888 | struct inode *inode = page->mapping->host; | |
4889 | ||
4890 | @@ -2284,11 +2407,14 @@ static int ext4_da_writepages_trans_bloc | |
4891 | static int ext4_da_writepages(struct address_space *mapping, | |
4892 | struct writeback_control *wbc) | |
4893 | { | |
4894 | + pgoff_t index; | |
4895 | + int range_whole = 0; | |
4896 | handle_t *handle = NULL; | |
4897 | - loff_t range_start = 0; | |
4898 | + struct mpage_da_data mpd; | |
4899 | struct inode *inode = mapping->host; | |
4900 | + int pages_written = 0; | |
4901 | + long pages_skipped; | |
4902 | int needed_blocks, ret = 0, nr_to_writebump = 0; | |
4903 | - long to_write, pages_skipped = 0; | |
4904 | struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb); | |
4905 | ||
4906 | /* | |
4907 | @@ -2298,6 +2424,20 @@ static int ext4_da_writepages(struct add | |
4908 | */ | |
4909 | if (!mapping->nrpages || !mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) | |
4910 | return 0; | |
4911 | + | |
4912 | + /* | |
4913 | + * If the filesystem has aborted, it is read-only, so return | |
4914 | + * right away instead of dumping stack traces later on that | |
4915 | + * will obscure the real source of the problem. We test | |
4916 | + * EXT4_MOUNT_ABORT instead of sb->s_flag's MS_RDONLY because | |
4917 | + * the latter could be true if the filesystem is mounted | |
4918 | + * read-only, and in that case, ext4_da_writepages should | |
4919 | + * *never* be called, so if that ever happens, we would want | |
4920 | + * the stack trace. | |
4921 | + */ | |
4922 | + if (unlikely(sbi->s_mount_opt & EXT4_MOUNT_ABORT)) | |
4923 | + return -EROFS; | |
4924 | + | |
4925 | /* | |
4926 | * Make sure nr_to_write is >= sbi->s_mb_stream_request | |
4927 | * This make sure small files blocks are allocated in | |
4928 | @@ -2308,20 +2448,24 @@ static int ext4_da_writepages(struct add | |
4929 | nr_to_writebump = sbi->s_mb_stream_request - wbc->nr_to_write; | |
4930 | wbc->nr_to_write = sbi->s_mb_stream_request; | |
4931 | } | |
4932 | + if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX) | |
4933 | + range_whole = 1; | |
4934 | ||
4935 | - if (!wbc->range_cyclic) | |
4936 | - /* | |
4937 | - * If range_cyclic is not set force range_cont | |
4938 | - * and save the old writeback_index | |
4939 | - */ | |
4940 | - wbc->range_cont = 1; | |
4941 | + if (wbc->range_cyclic) | |
4942 | + index = mapping->writeback_index; | |
4943 | + else | |
4944 | + index = wbc->range_start >> PAGE_CACHE_SHIFT; | |
4945 | ||
4946 | - range_start = wbc->range_start; | |
4947 | + mpd.wbc = wbc; | |
4948 | + mpd.inode = mapping->host; | |
4949 | + | |
4950 | + /* | |
4951 | + * we don't want write_cache_pages to update | |
4952 | + * nr_to_write and writeback_index | |
4953 | + */ | |
4954 | pages_skipped = wbc->pages_skipped; | |
4955 | ||
4956 | -restart_loop: | |
4957 | - to_write = wbc->nr_to_write; | |
4958 | - while (!ret && to_write > 0) { | |
4959 | + while (!ret && wbc->nr_to_write > 0) { | |
4960 | ||
4961 | /* | |
4962 | * we insert one extent at a time. So we need | |
4963 | @@ -2336,63 +2480,87 @@ restart_loop: | |
4964 | handle = ext4_journal_start(inode, needed_blocks); | |
4965 | if (IS_ERR(handle)) { | |
4966 | ret = PTR_ERR(handle); | |
4967 | - printk(KERN_EMERG "%s: jbd2_start: " | |
4968 | + printk(KERN_CRIT "%s: jbd2_start: " | |
4969 | "%ld pages, ino %lu; err %d\n", __func__, | |
4970 | wbc->nr_to_write, inode->i_ino, ret); | |
4971 | dump_stack(); | |
4972 | goto out_writepages; | |
4973 | } | |
4974 | - if (ext4_should_order_data(inode)) { | |
4975 | - /* | |
4976 | - * With ordered mode we need to add | |
4977 | - * the inode to the journal handl | |
4978 | - * when we do block allocation. | |
4979 | - */ | |
4980 | - ret = ext4_jbd2_file_inode(handle, inode); | |
4981 | - if (ret) { | |
4982 | - ext4_journal_stop(handle); | |
4983 | - goto out_writepages; | |
4984 | - } | |
4985 | - } | |
4986 | + mpd.get_block = ext4_da_get_block_write; | |
4987 | + ret = mpage_da_writepages(mapping, wbc, &mpd); | |
4988 | ||
4989 | - to_write -= wbc->nr_to_write; | |
4990 | - ret = mpage_da_writepages(mapping, wbc, | |
4991 | - ext4_da_get_block_write); | |
4992 | ext4_journal_stop(handle); | |
4993 | - if (ret == MPAGE_DA_EXTENT_TAIL) { | |
4994 | + | |
4995 | + if (mpd.retval == -ENOSPC) { | |
4996 | + /* commit the transaction which would | |
4997 | + * free blocks released in the transaction | |
4998 | + * and try again | |
4999 | + */ | |
5000 | + jbd2_journal_force_commit_nested(sbi->s_journal); | |
5001 | + wbc->pages_skipped = pages_skipped; | |
5002 | + ret = 0; | |
5003 | + } else if (ret == MPAGE_DA_EXTENT_TAIL) { | |
5004 | /* | |
5005 | * got one extent now try with | |
5006 | * rest of the pages | |
5007 | */ | |
5008 | - to_write += wbc->nr_to_write; | |
5009 | + pages_written += mpd.pages_written; | |
5010 | + wbc->pages_skipped = pages_skipped; | |
5011 | ret = 0; | |
5012 | - } else if (wbc->nr_to_write) { | |
5013 | + } else if (wbc->nr_to_write) | |
5014 | /* | |
5015 | * There is no more writeout needed | |
5016 | * or we requested for a noblocking writeout | |
5017 | * and we found the device congested | |
5018 | */ | |
5019 | - to_write += wbc->nr_to_write; | |
5020 | break; | |
5021 | - } | |
5022 | - wbc->nr_to_write = to_write; | |
5023 | - } | |
5024 | - | |
5025 | - if (wbc->range_cont && (pages_skipped != wbc->pages_skipped)) { | |
5026 | - /* We skipped pages in this loop */ | |
5027 | - wbc->range_start = range_start; | |
5028 | - wbc->nr_to_write = to_write + | |
5029 | - wbc->pages_skipped - pages_skipped; | |
5030 | - wbc->pages_skipped = pages_skipped; | |
5031 | - goto restart_loop; | |
5032 | } | |
5033 | + if (pages_skipped != wbc->pages_skipped) | |
5034 | + printk(KERN_EMERG "This should not happen leaving %s " | |
5035 | + "with nr_to_write = %ld ret = %d\n", | |
5036 | + __func__, wbc->nr_to_write, ret); | |
5037 | + | |
5038 | + /* Update index */ | |
5039 | + index += pages_written; | |
5040 | + if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) | |
5041 | + /* | |
5042 | + * set the writeback_index so that range_cyclic | |
5043 | + * mode will write it back later | |
5044 | + */ | |
5045 | + mapping->writeback_index = index; | |
5046 | ||
5047 | out_writepages: | |
5048 | - wbc->nr_to_write = to_write - nr_to_writebump; | |
5049 | - wbc->range_start = range_start; | |
5050 | + wbc->nr_to_write -= nr_to_writebump; | |
5051 | return ret; | |
5052 | } | |
5053 | ||
5054 | +#define FALL_BACK_TO_NONDELALLOC 1 | |
5055 | +static int ext4_nonda_switch(struct super_block *sb) | |
5056 | +{ | |
5057 | + s64 free_blocks, dirty_blocks; | |
5058 | + struct ext4_sb_info *sbi = EXT4_SB(sb); | |
5059 | + | |
5060 | + /* | |
5061 | + * switch to non delalloc mode if we are running low | |
5062 | + * on free block. The free block accounting via percpu | |
5063 | + * counters can get slightly wrong with FBC_BATCH getting | |
5064 | + * accumulated on each CPU without updating global counters | |
5065 | + * Delalloc need an accurate free block accounting. So switch | |
5066 | + * to non delalloc when we are near to error range. | |
5067 | + */ | |
5068 | + free_blocks = percpu_counter_read_positive(&sbi->s_freeblocks_counter); | |
5069 | + dirty_blocks = percpu_counter_read_positive(&sbi->s_dirtyblocks_counter); | |
5070 | + if (2 * free_blocks < 3 * dirty_blocks || | |
5071 | + free_blocks < (dirty_blocks + EXT4_FREEBLOCKS_WATERMARK)) { | |
5072 | + /* | |
5073 | + * free block count is less that 150% of dirty blocks | |
5074 | + * or free blocks is less that watermark | |
5075 | + */ | |
5076 | + return 1; | |
5077 | + } | |
5078 | + return 0; | |
5079 | +} | |
5080 | + | |
5081 | static int ext4_da_write_begin(struct file *file, struct address_space *mapping, | |
5082 | loff_t pos, unsigned len, unsigned flags, | |
5083 | struct page **pagep, void **fsdata) | |
5084 | @@ -2408,6 +2576,12 @@ static int ext4_da_write_begin(struct fi | |
5085 | from = pos & (PAGE_CACHE_SIZE - 1); | |
5086 | to = from + len; | |
5087 | ||
5088 | + if (ext4_nonda_switch(inode->i_sb)) { | |
5089 | + *fsdata = (void *)FALL_BACK_TO_NONDELALLOC; | |
5090 | + return ext4_write_begin(file, mapping, pos, | |
5091 | + len, flags, pagep, fsdata); | |
5092 | + } | |
5093 | + *fsdata = (void *)0; | |
5094 | retry: | |
5095 | /* | |
5096 | * With delayed allocation, we don't log the i_disksize update | |
5097 | @@ -2435,6 +2609,13 @@ retry: | |
5098 | unlock_page(page); | |
5099 | ext4_journal_stop(handle); | |
5100 | page_cache_release(page); | |
5101 | + /* | |
5102 | + * block_write_begin may have instantiated a few blocks | |
5103 | + * outside i_size. Trim these off again. Don't need | |
5104 | + * i_size_read because we hold i_mutex. | |
5105 | + */ | |
5106 | + if (pos + len > inode->i_size) | |
5107 | + vmtruncate(inode, inode->i_size); | |
5108 | } | |
5109 | ||
5110 | if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) | |
5111 | @@ -2458,7 +2639,7 @@ static int ext4_da_should_update_i_disks | |
5112 | bh = page_buffers(page); | |
5113 | idx = offset >> inode->i_blkbits; | |
5114 | ||
5115 | - for (i=0; i < idx; i++) | |
5116 | + for (i = 0; i < idx; i++) | |
5117 | bh = bh->b_this_page; | |
5118 | ||
5119 | if (!buffer_mapped(bh) || (buffer_delay(bh))) | |
5120 | @@ -2476,9 +2657,22 @@ static int ext4_da_write_end(struct file | |
5121 | handle_t *handle = ext4_journal_current_handle(); | |
5122 | loff_t new_i_size; | |
5123 | unsigned long start, end; | |
5124 | + int write_mode = (int)(unsigned long)fsdata; | |
5125 | + | |
5126 | + if (write_mode == FALL_BACK_TO_NONDELALLOC) { | |
5127 | + if (ext4_should_order_data(inode)) { | |
5128 | + return ext4_ordered_write_end(file, mapping, pos, | |
5129 | + len, copied, page, fsdata); | |
5130 | + } else if (ext4_should_writeback_data(inode)) { | |
5131 | + return ext4_writeback_write_end(file, mapping, pos, | |
5132 | + len, copied, page, fsdata); | |
5133 | + } else { | |
5134 | + BUG(); | |
5135 | + } | |
5136 | + } | |
5137 | ||
5138 | start = pos & (PAGE_CACHE_SIZE - 1); | |
5139 | - end = start + copied -1; | |
5140 | + end = start + copied - 1; | |
5141 | ||
5142 | /* | |
5143 | * generic_write_end() will run mark_inode_dirty() if i_size | |
5144 | @@ -2502,6 +2696,11 @@ static int ext4_da_write_end(struct file | |
5145 | EXT4_I(inode)->i_disksize = new_i_size; | |
5146 | } | |
5147 | up_write(&EXT4_I(inode)->i_data_sem); | |
5148 | + /* We need to mark inode dirty even if | |
5149 | + * new_i_size is less that inode->i_size | |
5150 | + * bu greater than i_disksize.(hint delalloc) | |
5151 | + */ | |
5152 | + ext4_mark_inode_dirty(handle, inode); | |
5153 | } | |
5154 | } | |
5155 | ret2 = generic_write_end(file, mapping, pos, len, copied, | |
5156 | @@ -2593,7 +2792,7 @@ static sector_t ext4_bmap(struct address | |
5157 | return 0; | |
5158 | } | |
5159 | ||
5160 | - return generic_block_bmap(mapping,block,ext4_get_block); | |
5161 | + return generic_block_bmap(mapping, block, ext4_get_block); | |
5162 | } | |
5163 | ||
5164 | static int bget_one(handle_t *handle, struct buffer_head *bh) | |
5165 | @@ -3199,7 +3398,7 @@ static Indirect *ext4_find_shared(struct | |
5166 | if (!partial->key && *partial->p) | |
5167 | /* Writer: end */ | |
5168 | goto no_top; | |
5169 | - for (p=partial; p>chain && all_zeroes((__le32*)p->bh->b_data,p->p); p--) | |
5170 | + for (p = partial; (p > chain) && all_zeroes((__le32 *) p->bh->b_data, p->p); p--) | |
5171 | ; | |
5172 | /* | |
5173 | * OK, we've found the last block that must survive. The rest of our | |
5174 | @@ -3218,7 +3417,7 @@ static Indirect *ext4_find_shared(struct | |
5175 | } | |
5176 | /* Writer: end */ | |
5177 | ||
5178 | - while(partial > p) { | |
5179 | + while (partial > p) { | |
5180 | brelse(partial->bh); | |
5181 | partial--; | |
5182 | } | |
5183 | @@ -3410,9 +3609,9 @@ static void ext4_free_branches(handle_t | |
5184 | /* This zaps the entire block. Bottom up. */ | |
5185 | BUFFER_TRACE(bh, "free child branches"); | |
5186 | ext4_free_branches(handle, inode, bh, | |
5187 | - (__le32*)bh->b_data, | |
5188 | - (__le32*)bh->b_data + addr_per_block, | |
5189 | - depth); | |
5190 | + (__le32 *) bh->b_data, | |
5191 | + (__le32 *) bh->b_data + addr_per_block, | |
5192 | + depth); | |
5193 | ||
5194 | /* | |
5195 | * We've probably journalled the indirect block several | |
5196 | @@ -3580,7 +3779,7 @@ void ext4_truncate(struct inode *inode) | |
5197 | */ | |
5198 | down_write(&ei->i_data_sem); | |
5199 | ||
5200 | - ext4_discard_reservation(inode); | |
5201 | + ext4_discard_preallocations(inode); | |
5202 | ||
5203 | /* | |
5204 | * The orphan list entry will now protect us from any crash which | |
5205 | @@ -3675,41 +3874,6 @@ out_stop: | |
5206 | ext4_journal_stop(handle); | |
5207 | } | |
5208 | ||
5209 | -static ext4_fsblk_t ext4_get_inode_block(struct super_block *sb, | |
5210 | - unsigned long ino, struct ext4_iloc *iloc) | |
5211 | -{ | |
5212 | - ext4_group_t block_group; | |
5213 | - unsigned long offset; | |
5214 | - ext4_fsblk_t block; | |
5215 | - struct ext4_group_desc *gdp; | |
5216 | - | |
5217 | - if (!ext4_valid_inum(sb, ino)) { | |
5218 | - /* | |
5219 | - * This error is already checked for in namei.c unless we are | |
5220 | - * looking at an NFS filehandle, in which case no error | |
5221 | - * report is needed | |
5222 | - */ | |
5223 | - return 0; | |
5224 | - } | |
5225 | - | |
5226 | - block_group = (ino - 1) / EXT4_INODES_PER_GROUP(sb); | |
5227 | - gdp = ext4_get_group_desc(sb, block_group, NULL); | |
5228 | - if (!gdp) | |
5229 | - return 0; | |
5230 | - | |
5231 | - /* | |
5232 | - * Figure out the offset within the block group inode table | |
5233 | - */ | |
5234 | - offset = ((ino - 1) % EXT4_INODES_PER_GROUP(sb)) * | |
5235 | - EXT4_INODE_SIZE(sb); | |
5236 | - block = ext4_inode_table(sb, gdp) + | |
5237 | - (offset >> EXT4_BLOCK_SIZE_BITS(sb)); | |
5238 | - | |
5239 | - iloc->block_group = block_group; | |
5240 | - iloc->offset = offset & (EXT4_BLOCK_SIZE(sb) - 1); | |
5241 | - return block; | |
5242 | -} | |
5243 | - | |
5244 | /* | |
5245 | * ext4_get_inode_loc returns with an extra refcount against the inode's | |
5246 | * underlying buffer_head on success. If 'in_mem' is true, we have all | |
5247 | @@ -3719,19 +3883,35 @@ static ext4_fsblk_t ext4_get_inode_block | |
5248 | static int __ext4_get_inode_loc(struct inode *inode, | |
5249 | struct ext4_iloc *iloc, int in_mem) | |
5250 | { | |
5251 | - ext4_fsblk_t block; | |
5252 | - struct buffer_head *bh; | |
5253 | + struct ext4_group_desc *gdp; | |
5254 | + struct buffer_head *bh; | |
5255 | + struct super_block *sb = inode->i_sb; | |
5256 | + ext4_fsblk_t block; | |
5257 | + int inodes_per_block, inode_offset; | |
5258 | + | |
5259 | + iloc->bh = NULL; | |
5260 | + if (!ext4_valid_inum(sb, inode->i_ino)) | |
5261 | + return -EIO; | |
5262 | ||
5263 | - block = ext4_get_inode_block(inode->i_sb, inode->i_ino, iloc); | |
5264 | - if (!block) | |
5265 | + iloc->block_group = (inode->i_ino - 1) / EXT4_INODES_PER_GROUP(sb); | |
5266 | + gdp = ext4_get_group_desc(sb, iloc->block_group, NULL); | |
5267 | + if (!gdp) | |
5268 | return -EIO; | |
5269 | ||
5270 | - bh = sb_getblk(inode->i_sb, block); | |
5271 | + /* | |
5272 | + * Figure out the offset within the block group inode table | |
5273 | + */ | |
5274 | + inodes_per_block = (EXT4_BLOCK_SIZE(sb) / EXT4_INODE_SIZE(sb)); | |
5275 | + inode_offset = ((inode->i_ino - 1) % | |
5276 | + EXT4_INODES_PER_GROUP(sb)); | |
5277 | + block = ext4_inode_table(sb, gdp) + (inode_offset / inodes_per_block); | |
5278 | + iloc->offset = (inode_offset % inodes_per_block) * EXT4_INODE_SIZE(sb); | |
5279 | + | |
5280 | + bh = sb_getblk(sb, block); | |
5281 | if (!bh) { | |
5282 | - ext4_error (inode->i_sb, "ext4_get_inode_loc", | |
5283 | - "unable to read inode block - " | |
5284 | - "inode=%lu, block=%llu", | |
5285 | - inode->i_ino, block); | |
5286 | + ext4_error(sb, "ext4_get_inode_loc", "unable to read " | |
5287 | + "inode block - inode=%lu, block=%llu", | |
5288 | + inode->i_ino, block); | |
5289 | return -EIO; | |
5290 | } | |
5291 | if (!buffer_uptodate(bh)) { | |
5292 | @@ -3759,28 +3939,12 @@ static int __ext4_get_inode_loc(struct i | |
5293 | */ | |
5294 | if (in_mem) { | |
5295 | struct buffer_head *bitmap_bh; | |
5296 | - struct ext4_group_desc *desc; | |
5297 | - int inodes_per_buffer; | |
5298 | - int inode_offset, i; | |
5299 | - ext4_group_t block_group; | |
5300 | - int start; | |
5301 | - | |
5302 | - block_group = (inode->i_ino - 1) / | |
5303 | - EXT4_INODES_PER_GROUP(inode->i_sb); | |
5304 | - inodes_per_buffer = bh->b_size / | |
5305 | - EXT4_INODE_SIZE(inode->i_sb); | |
5306 | - inode_offset = ((inode->i_ino - 1) % | |
5307 | - EXT4_INODES_PER_GROUP(inode->i_sb)); | |
5308 | - start = inode_offset & ~(inodes_per_buffer - 1); | |
5309 | + int i, start; | |
5310 | ||
5311 | - /* Is the inode bitmap in cache? */ | |
5312 | - desc = ext4_get_group_desc(inode->i_sb, | |
5313 | - block_group, NULL); | |
5314 | - if (!desc) | |
5315 | - goto make_io; | |
5316 | + start = inode_offset & ~(inodes_per_block - 1); | |
5317 | ||
5318 | - bitmap_bh = sb_getblk(inode->i_sb, | |
5319 | - ext4_inode_bitmap(inode->i_sb, desc)); | |
5320 | + /* Is the inode bitmap in cache? */ | |
5321 | + bitmap_bh = sb_getblk(sb, ext4_inode_bitmap(sb, gdp)); | |
5322 | if (!bitmap_bh) | |
5323 | goto make_io; | |
5324 | ||
5325 | @@ -3793,14 +3957,14 @@ static int __ext4_get_inode_loc(struct i | |
5326 | brelse(bitmap_bh); | |
5327 | goto make_io; | |
5328 | } | |
5329 | - for (i = start; i < start + inodes_per_buffer; i++) { | |
5330 | + for (i = start; i < start + inodes_per_block; i++) { | |
5331 | if (i == inode_offset) | |
5332 | continue; | |
5333 | if (ext4_test_bit(i, bitmap_bh->b_data)) | |
5334 | break; | |
5335 | } | |
5336 | brelse(bitmap_bh); | |
5337 | - if (i == start + inodes_per_buffer) { | |
5338 | + if (i == start + inodes_per_block) { | |
5339 | /* all other inodes are free, so skip I/O */ | |
5340 | memset(bh->b_data, 0, bh->b_size); | |
5341 | set_buffer_uptodate(bh); | |
5342 | @@ -3811,6 +3975,36 @@ static int __ext4_get_inode_loc(struct i | |
5343 | ||
5344 | make_io: | |
5345 | /* | |
5346 | + * If we need to do any I/O, try to pre-readahead extra | |
5347 | + * blocks from the inode table. | |
5348 | + */ | |
5349 | + if (EXT4_SB(sb)->s_inode_readahead_blks) { | |
5350 | + ext4_fsblk_t b, end, table; | |
5351 | + unsigned num; | |
5352 | + | |
5353 | + table = ext4_inode_table(sb, gdp); | |
5354 | + /* Make sure s_inode_readahead_blks is a power of 2 */ | |
5355 | + while (EXT4_SB(sb)->s_inode_readahead_blks & | |
5356 | + (EXT4_SB(sb)->s_inode_readahead_blks-1)) | |
5357 | + EXT4_SB(sb)->s_inode_readahead_blks = | |
5358 | + (EXT4_SB(sb)->s_inode_readahead_blks & | |
5359 | + (EXT4_SB(sb)->s_inode_readahead_blks-1)); | |
5360 | + b = block & ~(EXT4_SB(sb)->s_inode_readahead_blks-1); | |
5361 | + if (table > b) | |
5362 | + b = table; | |
5363 | + end = b + EXT4_SB(sb)->s_inode_readahead_blks; | |
5364 | + num = EXT4_INODES_PER_GROUP(sb); | |
5365 | + if (EXT4_HAS_RO_COMPAT_FEATURE(sb, | |
5366 | + EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) | |
5367 | + num -= ext4_itable_unused_count(sb, gdp); | |
5368 | + table += num / inodes_per_block; | |
5369 | + if (end > table) | |
5370 | + end = table; | |
5371 | + while (b <= end) | |
5372 | + sb_breadahead(sb, b++); | |
5373 | + } | |
5374 | + | |
5375 | + /* | |
5376 | * There are other valid inodes in the buffer, this inode | |
5377 | * has in-inode xattrs, or we don't have this inode in memory. | |
5378 | * Read the block from disk. | |
5379 | @@ -3820,10 +4014,9 @@ make_io: | |
5380 | submit_bh(READ_META, bh); | |
5381 | wait_on_buffer(bh); | |
5382 | if (!buffer_uptodate(bh)) { | |
5383 | - ext4_error(inode->i_sb, "ext4_get_inode_loc", | |
5384 | - "unable to read inode block - " | |
5385 | - "inode=%lu, block=%llu", | |
5386 | - inode->i_ino, block); | |
5387 | + ext4_error(sb, __func__, | |
5388 | + "unable to read inode block - inode=%lu, " | |
5389 | + "block=%llu", inode->i_ino, block); | |
5390 | brelse(bh); | |
5391 | return -EIO; | |
5392 | } | |
5393 | @@ -3915,11 +4108,10 @@ struct inode *ext4_iget(struct super_blo | |
5394 | return inode; | |
5395 | ||
5396 | ei = EXT4_I(inode); | |
5397 | -#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL | |
5398 | +#if defined(CONFIG_EXT4_FS_POSIX_ACL) || defined(CONFIG_EXT4DEV_FS_POSIX_ACL) | |
5399 | ei->i_acl = EXT4_ACL_NOT_CACHED; | |
5400 | ei->i_default_acl = EXT4_ACL_NOT_CACHED; | |
5401 | #endif | |
5402 | - ei->i_block_alloc_info = NULL; | |
5403 | ||
5404 | ret = __ext4_get_inode_loc(inode, &iloc, 0); | |
5405 | if (ret < 0) | |
5406 | @@ -3929,7 +4121,7 @@ struct inode *ext4_iget(struct super_blo | |
5407 | inode->i_mode = le16_to_cpu(raw_inode->i_mode); | |
5408 | inode->i_uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low); | |
5409 | inode->i_gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low); | |
5410 | - if(!(test_opt (inode->i_sb, NO_UID32))) { | |
5411 | + if (!(test_opt(inode->i_sb, NO_UID32))) { | |
5412 | inode->i_uid |= le16_to_cpu(raw_inode->i_uid_high) << 16; | |
5413 | inode->i_gid |= le16_to_cpu(raw_inode->i_gid_high) << 16; | |
5414 | } | |
5415 | @@ -3947,7 +4139,7 @@ struct inode *ext4_iget(struct super_blo | |
5416 | if (inode->i_mode == 0 || | |
5417 | !(EXT4_SB(inode->i_sb)->s_mount_state & EXT4_ORPHAN_FS)) { | |
5418 | /* this inode is deleted */ | |
5419 | - brelse (bh); | |
5420 | + brelse(bh); | |
5421 | ret = -ESTALE; | |
5422 | goto bad_inode; | |
5423 | } | |
5424 | @@ -3980,7 +4172,7 @@ struct inode *ext4_iget(struct super_blo | |
5425 | ei->i_extra_isize = le16_to_cpu(raw_inode->i_extra_isize); | |
5426 | if (EXT4_GOOD_OLD_INODE_SIZE + ei->i_extra_isize > | |
5427 | EXT4_INODE_SIZE(inode->i_sb)) { | |
5428 | - brelse (bh); | |
5429 | + brelse(bh); | |
5430 | ret = -EIO; | |
5431 | goto bad_inode; | |
5432 | } | |
5433 | @@ -4033,7 +4225,7 @@ struct inode *ext4_iget(struct super_blo | |
5434 | init_special_inode(inode, inode->i_mode, | |
5435 | new_decode_dev(le32_to_cpu(raw_inode->i_block[1]))); | |
5436 | } | |
5437 | - brelse (iloc.bh); | |
5438 | + brelse(iloc.bh); | |
5439 | ext4_set_inode_flags(inode); | |
5440 | unlock_new_inode(inode); | |
5441 | return inode; | |
5442 | @@ -4050,7 +4242,6 @@ static int ext4_inode_blocks_set(handle_ | |
5443 | struct inode *inode = &(ei->vfs_inode); | |
5444 | u64 i_blocks = inode->i_blocks; | |
5445 | struct super_block *sb = inode->i_sb; | |
5446 | - int err = 0; | |
5447 | ||
5448 | if (i_blocks <= ~0U) { | |
5449 | /* | |
5450 | @@ -4060,36 +4251,27 @@ static int ext4_inode_blocks_set(handle_ | |
5451 | raw_inode->i_blocks_lo = cpu_to_le32(i_blocks); | |
5452 | raw_inode->i_blocks_high = 0; | |
5453 | ei->i_flags &= ~EXT4_HUGE_FILE_FL; | |
5454 | - } else if (i_blocks <= 0xffffffffffffULL) { | |
5455 | + return 0; | |
5456 | + } | |
5457 | + if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) | |
5458 | + return -EFBIG; | |
5459 | + | |
5460 | + if (i_blocks <= 0xffffffffffffULL) { | |
5461 | /* | |
5462 | * i_blocks can be represented in a 48 bit variable | |
5463 | * as multiple of 512 bytes | |
5464 | */ | |
5465 | - err = ext4_update_rocompat_feature(handle, sb, | |
5466 | - EXT4_FEATURE_RO_COMPAT_HUGE_FILE); | |
5467 | - if (err) | |
5468 | - goto err_out; | |
5469 | - /* i_block is stored in the split 48 bit fields */ | |
5470 | raw_inode->i_blocks_lo = cpu_to_le32(i_blocks); | |
5471 | raw_inode->i_blocks_high = cpu_to_le16(i_blocks >> 32); | |
5472 | ei->i_flags &= ~EXT4_HUGE_FILE_FL; | |
5473 | } else { | |
5474 | - /* | |
5475 | - * i_blocks should be represented in a 48 bit variable | |
5476 | - * as multiple of file system block size | |
5477 | - */ | |
5478 | - err = ext4_update_rocompat_feature(handle, sb, | |
5479 | - EXT4_FEATURE_RO_COMPAT_HUGE_FILE); | |
5480 | - if (err) | |
5481 | - goto err_out; | |
5482 | ei->i_flags |= EXT4_HUGE_FILE_FL; | |
5483 | /* i_block is stored in file system block size */ | |
5484 | i_blocks = i_blocks >> (inode->i_blkbits - 9); | |
5485 | raw_inode->i_blocks_lo = cpu_to_le32(i_blocks); | |
5486 | raw_inode->i_blocks_high = cpu_to_le16(i_blocks >> 32); | |
5487 | } | |
5488 | -err_out: | |
5489 | - return err; | |
5490 | + return 0; | |
5491 | } | |
5492 | ||
5493 | /* | |
5494 | @@ -4115,14 +4297,14 @@ static int ext4_do_update_inode(handle_t | |
5495 | ||
5496 | ext4_get_inode_flags(ei); | |
5497 | raw_inode->i_mode = cpu_to_le16(inode->i_mode); | |
5498 | - if(!(test_opt(inode->i_sb, NO_UID32))) { | |
5499 | + if (!(test_opt(inode->i_sb, NO_UID32))) { | |
5500 | raw_inode->i_uid_low = cpu_to_le16(low_16_bits(inode->i_uid)); | |
5501 | raw_inode->i_gid_low = cpu_to_le16(low_16_bits(inode->i_gid)); | |
5502 | /* | |
5503 | * Fix up interoperability with old kernels. Otherwise, old inodes get | |
5504 | * re-used with the upper 16 bits of the uid/gid intact | |
5505 | */ | |
5506 | - if(!ei->i_dtime) { | |
5507 | + if (!ei->i_dtime) { | |
5508 | raw_inode->i_uid_high = | |
5509 | cpu_to_le16(high_16_bits(inode->i_uid)); | |
5510 | raw_inode->i_gid_high = | |
5511 | @@ -4210,7 +4392,7 @@ static int ext4_do_update_inode(handle_t | |
5512 | ei->i_state &= ~EXT4_STATE_NEW; | |
5513 | ||
5514 | out_brelse: | |
5515 | - brelse (bh); | |
5516 | + brelse(bh); | |
5517 | ext4_std_error(inode->i_sb, err); | |
5518 | return err; | |
5519 | } | |
5520 | @@ -4814,6 +4996,7 @@ int ext4_page_mkwrite(struct vm_area_str | |
5521 | loff_t size; | |
5522 | unsigned long len; | |
5523 | int ret = -EINVAL; | |
5524 | + void *fsdata; | |
5525 | struct file *file = vma->vm_file; | |
5526 | struct inode *inode = file->f_path.dentry->d_inode; | |
5527 | struct address_space *mapping = inode->i_mapping; | |
5528 | @@ -4852,11 +5035,11 @@ int ext4_page_mkwrite(struct vm_area_str | |
5529 | * on the same page though | |
5530 | */ | |
5531 | ret = mapping->a_ops->write_begin(file, mapping, page_offset(page), | |
5532 | - len, AOP_FLAG_UNINTERRUPTIBLE, &page, NULL); | |
5533 | + len, AOP_FLAG_UNINTERRUPTIBLE, &page, &fsdata); | |
5534 | if (ret < 0) | |
5535 | goto out_unlock; | |
5536 | ret = mapping->a_ops->write_end(file, mapping, page_offset(page), | |
5537 | - len, len, page, NULL); | |
5538 | + len, len, page, fsdata); | |
5539 | if (ret < 0) | |
5540 | goto out_unlock; | |
5541 | ret = 0; | |
5542 | diff -rup b/fs/ext4//ioctl.c a/fs/ext4///ioctl.c | |
5543 | --- b/fs/ext4/ioctl.c 2009-02-11 14:37:58.000000000 +0100 | |
5544 | +++ a/fs/ext4/ioctl.c 2009-02-10 21:40:11.000000000 +0100 | |
5545 | @@ -23,9 +23,8 @@ long ext4_ioctl(struct file *filp, unsig | |
5546 | struct inode *inode = filp->f_dentry->d_inode; | |
5547 | struct ext4_inode_info *ei = EXT4_I(inode); | |
5548 | unsigned int flags; | |
5549 | - unsigned short rsv_window_size; | |
5550 | ||
5551 | - ext4_debug ("cmd = %u, arg = %lu\n", cmd, arg); | |
5552 | + ext4_debug("cmd = %u, arg = %lu\n", cmd, arg); | |
5553 | ||
5554 | switch (cmd) { | |
5555 | case EXT4_IOC_GETFLAGS: | |
5556 | @@ -34,7 +33,7 @@ long ext4_ioctl(struct file *filp, unsig | |
5557 | return put_user(flags, (int __user *) arg); | |
5558 | case EXT4_IOC_SETFLAGS: { | |
5559 | handle_t *handle = NULL; | |
5560 | - int err; | |
5561 | + int err, migrate = 0; | |
5562 | struct ext4_iloc iloc; | |
5563 | unsigned int oldflags; | |
5564 | unsigned int jflag; | |
5565 | @@ -82,6 +81,17 @@ long ext4_ioctl(struct file *filp, unsig | |
5566 | if (!capable(CAP_SYS_RESOURCE)) | |
5567 | goto flags_out; | |
5568 | } | |
5569 | + if (oldflags & EXT4_EXTENTS_FL) { | |
5570 | + /* We don't support clearning extent flags */ | |
5571 | + if (!(flags & EXT4_EXTENTS_FL)) { | |
5572 | + err = -EOPNOTSUPP; | |
5573 | + goto flags_out; | |
5574 | + } | |
5575 | + } else if (flags & EXT4_EXTENTS_FL) { | |
5576 | + /* migrate the file */ | |
5577 | + migrate = 1; | |
5578 | + flags &= ~EXT4_EXTENTS_FL; | |
5579 | + } | |
5580 | ||
5581 | handle = ext4_journal_start(inode, 1); | |
5582 | if (IS_ERR(handle)) { | |
5583 | @@ -109,6 +119,10 @@ flags_err: | |
5584 | ||
5585 | if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) | |
5586 | err = ext4_change_inode_journal_flag(inode, jflag); | |
5587 | + if (err) | |
5588 | + goto flags_out; | |
5589 | + if (migrate) | |
5590 | + err = ext4_ext_migrate(inode); | |
5591 | flags_out: | |
5592 | mutex_unlock(&inode->i_mutex); | |
5593 | mnt_drop_write(filp->f_path.mnt); | |
5594 | @@ -175,53 +189,10 @@ setversion_out: | |
5595 | return ret; | |
5596 | } | |
5597 | #endif | |
5598 | - case EXT4_IOC_GETRSVSZ: | |
5599 | - if (test_opt(inode->i_sb, RESERVATION) | |
5600 | - && S_ISREG(inode->i_mode) | |
5601 | - && ei->i_block_alloc_info) { | |
5602 | - rsv_window_size = ei->i_block_alloc_info->rsv_window_node.rsv_goal_size; | |
5603 | - return put_user(rsv_window_size, (int __user *)arg); | |
5604 | - } | |
5605 | - return -ENOTTY; | |
5606 | - case EXT4_IOC_SETRSVSZ: { | |
5607 | - int err; | |
5608 | - | |
5609 | - if (!test_opt(inode->i_sb, RESERVATION) ||!S_ISREG(inode->i_mode)) | |
5610 | - return -ENOTTY; | |
5611 | - | |
5612 | - if (!is_owner_or_cap(inode)) | |
5613 | - return -EACCES; | |
5614 | - | |
5615 | - if (get_user(rsv_window_size, (int __user *)arg)) | |
5616 | - return -EFAULT; | |
5617 | - | |
5618 | - err = mnt_want_write(filp->f_path.mnt); | |
5619 | - if (err) | |
5620 | - return err; | |
5621 | - | |
5622 | - if (rsv_window_size > EXT4_MAX_RESERVE_BLOCKS) | |
5623 | - rsv_window_size = EXT4_MAX_RESERVE_BLOCKS; | |
5624 | - | |
5625 | - /* | |
5626 | - * need to allocate reservation structure for this inode | |
5627 | - * before set the window size | |
5628 | - */ | |
5629 | - down_write(&ei->i_data_sem); | |
5630 | - if (!ei->i_block_alloc_info) | |
5631 | - ext4_init_block_alloc_info(inode); | |
5632 | - | |
5633 | - if (ei->i_block_alloc_info){ | |
5634 | - struct ext4_reserve_window_node *rsv = &ei->i_block_alloc_info->rsv_window_node; | |
5635 | - rsv->rsv_goal_size = rsv_window_size; | |
5636 | - } | |
5637 | - up_write(&ei->i_data_sem); | |
5638 | - mnt_drop_write(filp->f_path.mnt); | |
5639 | - return 0; | |
5640 | - } | |
5641 | case EXT4_IOC_GROUP_EXTEND: { | |
5642 | ext4_fsblk_t n_blocks_count; | |
5643 | struct super_block *sb = inode->i_sb; | |
5644 | - int err; | |
5645 | + int err, err2; | |
5646 | ||
5647 | if (!capable(CAP_SYS_RESOURCE)) | |
5648 | return -EPERM; | |
5649 | @@ -235,8 +206,10 @@ setversion_out: | |
5650 | ||
5651 | err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count); | |
5652 | jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); | |
5653 | - jbd2_journal_flush(EXT4_SB(sb)->s_journal); | |
5654 | + err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal); | |
5655 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); | |
5656 | + if (err == 0) | |
5657 | + err = err2; | |
5658 | mnt_drop_write(filp->f_path.mnt); | |
5659 | ||
5660 | return err; | |
5661 | @@ -244,7 +217,7 @@ setversion_out: | |
5662 | case EXT4_IOC_GROUP_ADD: { | |
5663 | struct ext4_new_group_data input; | |
5664 | struct super_block *sb = inode->i_sb; | |
5665 | - int err; | |
5666 | + int err, err2; | |
5667 | ||
5668 | if (!capable(CAP_SYS_RESOURCE)) | |
5669 | return -EPERM; | |
5670 | @@ -259,8 +232,10 @@ setversion_out: | |
5671 | ||
5672 | err = ext4_group_add(sb, &input); | |
5673 | jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); | |
5674 | - jbd2_journal_flush(EXT4_SB(sb)->s_journal); | |
5675 | + err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal); | |
5676 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); | |
5677 | + if (err == 0) | |
5678 | + err = err2; | |
5679 | mnt_drop_write(filp->f_path.mnt); | |
5680 | ||
5681 | return err; | |
5682 | diff -rup b/fs/ext4//mballoc.c a/fs/ext4///mballoc.c | |
5683 | --- b/fs/ext4/mballoc.c 2009-02-11 14:37:58.000000000 +0100 | |
5684 | +++ a/fs/ext4/mballoc.c 2009-02-10 21:40:11.000000000 +0100 | |
5685 | @@ -100,7 +100,7 @@ | |
5686 | * inode as: | |
5687 | * | |
5688 | * { page } | |
5689 | - * [ group 0 buddy][ group 0 bitmap] [group 1][ group 1]... | |
5690 | + * [ group 0 bitmap][ group 0 buddy] [group 1][ group 1]... | |
5691 | * | |
5692 | * | |
5693 | * one block each for bitmap and buddy information. So for each group we | |
5694 | @@ -330,6 +330,20 @@ | |
5695 | * object | |
5696 | * | |
5697 | */ | |
5698 | +static struct kmem_cache *ext4_pspace_cachep; | |
5699 | +static struct kmem_cache *ext4_ac_cachep; | |
5700 | +static struct kmem_cache *ext4_free_ext_cachep; | |
5701 | +static void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap, | |
5702 | + ext4_group_t group); | |
5703 | +static void ext4_mb_generate_from_freelist(struct super_block *sb, void *bitmap, | |
5704 | + ext4_group_t group); | |
5705 | +static int ext4_mb_init_per_dev_proc(struct super_block *sb); | |
5706 | +static int ext4_mb_destroy_per_dev_proc(struct super_block *sb); | |
5707 | +static void ext4_mb_poll_new_transaction(struct super_block *, handle_t *); | |
5708 | +static void ext4_mb_free_committed_blocks(struct super_block *); | |
5709 | + | |
5710 | + | |
5711 | + | |
5712 | ||
5713 | static inline void *mb_correct_addr_and_bit(int *bit, void *addr) | |
5714 | { | |
5715 | @@ -445,9 +459,9 @@ static void mb_free_blocks_double(struct | |
5716 | blocknr += first + i; | |
5717 | blocknr += | |
5718 | le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block); | |
5719 | - | |
5720 | - ext4_error(sb, __func__, "double-free of inode" | |
5721 | - " %lu's block %llu(bit %u in group %lu)\n", | |
5722 | + ext4_grp_locked_error(sb, e4b->bd_group, | |
5723 | + __func__, "double-free of inode" | |
5724 | + " %lu's block %llu(bit %u in group %u)\n", | |
5725 | inode ? inode->i_ino : 0, blocknr, | |
5726 | first + i, e4b->bd_group); | |
5727 | } | |
5728 | @@ -477,9 +491,10 @@ static void mb_cmp_bitmaps(struct ext4_b | |
5729 | b2 = (unsigned char *) bitmap; | |
5730 | for (i = 0; i < e4b->bd_sb->s_blocksize; i++) { | |
5731 | if (b1[i] != b2[i]) { | |
5732 | - printk("corruption in group %lu at byte %u(%u):" | |
5733 | - " %x in copy != %x on disk/prealloc\n", | |
5734 | - e4b->bd_group, i, i * 8, b1[i], b2[i]); | |
5735 | + printk(KERN_ERR "corruption in group %u " | |
5736 | + "at byte %u(%u): %x in copy != %x " | |
5737 | + "on disk/prealloc\n", | |
5738 | + e4b->bd_group, i, i * 8, b1[i], b2[i]); | |
5739 | BUG(); | |
5740 | } | |
5741 | } | |
5742 | @@ -533,9 +548,6 @@ static int __mb_check_buddy(struct ext4_ | |
5743 | void *buddy; | |
5744 | void *buddy2; | |
5745 | ||
5746 | - if (!test_opt(sb, MBALLOC)) | |
5747 | - return 0; | |
5748 | - | |
5749 | { | |
5750 | static int mb_check_counter; | |
5751 | if (mb_check_counter++ % 100 != 0) | |
5752 | @@ -692,8 +704,8 @@ static void ext4_mb_generate_buddy(struc | |
5753 | grp->bb_fragments = fragments; | |
5754 | ||
5755 | if (free != grp->bb_free) { | |
5756 | - ext4_error(sb, __func__, | |
5757 | - "EXT4-fs: group %lu: %u blocks in bitmap, %u in gd\n", | |
5758 | + ext4_grp_locked_error(sb, group, __func__, | |
5759 | + "EXT4-fs: group %u: %u blocks in bitmap, %u in gd\n", | |
5760 | group, free, grp->bb_free); | |
5761 | /* | |
5762 | * If we intent to continue, we consider group descritor | |
5763 | @@ -718,7 +730,7 @@ static void ext4_mb_generate_buddy(struc | |
5764 | * stored in the inode as | |
5765 | * | |
5766 | * { page } | |
5767 | - * [ group 0 buddy][ group 0 bitmap] [group 1][ group 1]... | |
5768 | + * [ group 0 bitmap][ group 0 buddy] [group 1][ group 1]... | |
5769 | * | |
5770 | * | |
5771 | * one block each for bitmap and buddy information. | |
5772 | @@ -784,23 +796,45 @@ static int ext4_mb_init_cache(struct pag | |
5773 | if (bh[i] == NULL) | |
5774 | goto out; | |
5775 | ||
5776 | - if (bh_uptodate_or_lock(bh[i])) | |
5777 | + if (bitmap_uptodate(bh[i])) | |
5778 | continue; | |
5779 | ||
5780 | + lock_buffer(bh[i]); | |
5781 | + if (bitmap_uptodate(bh[i])) { | |
5782 | + unlock_buffer(bh[i]); | |
5783 | + continue; | |
5784 | + } | |
5785 | spin_lock(sb_bgl_lock(EXT4_SB(sb), first_group + i)); | |
5786 | if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { | |
5787 | ext4_init_block_bitmap(sb, bh[i], | |
5788 | first_group + i, desc); | |
5789 | + set_bitmap_uptodate(bh[i]); | |
5790 | set_buffer_uptodate(bh[i]); | |
5791 | - unlock_buffer(bh[i]); | |
5792 | spin_unlock(sb_bgl_lock(EXT4_SB(sb), first_group + i)); | |
5793 | + unlock_buffer(bh[i]); | |
5794 | continue; | |
5795 | } | |
5796 | spin_unlock(sb_bgl_lock(EXT4_SB(sb), first_group + i)); | |
5797 | + if (buffer_uptodate(bh[i])) { | |
5798 | + /* | |
5799 | + * if not uninit if bh is uptodate, | |
5800 | + * bitmap is also uptodate | |
5801 | + */ | |
5802 | + set_bitmap_uptodate(bh[i]); | |
5803 | + unlock_buffer(bh[i]); | |
5804 | + continue; | |
5805 | + } | |
5806 | get_bh(bh[i]); | |
5807 | + /* | |
5808 | + * submit the buffer_head for read. We can | |
5809 | + * safely mark the bitmap as uptodate now. | |
5810 | + * We do it here so the bitmap uptodate bit | |
5811 | + * get set with buffer lock held. | |
5812 | + */ | |
5813 | + set_bitmap_uptodate(bh[i]); | |
5814 | bh[i]->b_end_io = end_buffer_read_sync; | |
5815 | submit_bh(READ, bh[i]); | |
5816 | - mb_debug("read bitmap for group %lu\n", first_group + i); | |
5817 | + mb_debug("read bitmap for group %u\n", first_group + i); | |
5818 | } | |
5819 | ||
5820 | /* wait for I/O completion */ | |
5821 | @@ -814,6 +848,8 @@ static int ext4_mb_init_cache(struct pag | |
5822 | ||
5823 | err = 0; | |
5824 | first_block = page->index * blocks_per_page; | |
5825 | + /* init the page */ | |
5826 | + memset(page_address(page), 0xff, PAGE_CACHE_SIZE); | |
5827 | for (i = 0; i < blocks_per_page; i++) { | |
5828 | int group; | |
5829 | struct ext4_group_info *grinfo; | |
5830 | @@ -840,7 +876,6 @@ static int ext4_mb_init_cache(struct pag | |
5831 | BUG_ON(incore == NULL); | |
5832 | mb_debug("put buddy for group %u in page %lu/%x\n", | |
5833 | group, page->index, i * blocksize); | |
5834 | - memset(data, 0xff, blocksize); | |
5835 | grinfo = ext4_get_group_info(sb, group); | |
5836 | grinfo->bb_fragments = 0; | |
5837 | memset(grinfo->bb_counters, 0, | |
5838 | @@ -848,7 +883,9 @@ static int ext4_mb_init_cache(struct pag | |
5839 | /* | |
5840 | * incore got set to the group block bitmap below | |
5841 | */ | |
5842 | + ext4_lock_group(sb, group); | |
5843 | ext4_mb_generate_buddy(sb, data, incore, group); | |
5844 | + ext4_unlock_group(sb, group); | |
5845 | incore = NULL; | |
5846 | } else { | |
5847 | /* this is block of bitmap */ | |
5848 | @@ -862,6 +899,7 @@ static int ext4_mb_init_cache(struct pag | |
5849 | ||
5850 | /* mark all preallocated blks used in in-core bitmap */ | |
5851 | ext4_mb_generate_from_pa(sb, data, group); | |
5852 | + ext4_mb_generate_from_freelist(sb, data, group); | |
5853 | ext4_unlock_group(sb, group); | |
5854 | ||
5855 | /* set incore so that the buddy information can be | |
5856 | @@ -885,19 +923,22 @@ out: | |
5857 | static noinline_for_stack int | |
5858 | ext4_mb_load_buddy(struct super_block *sb, ext4_group_t group, | |
5859 | struct ext4_buddy *e4b) | |
5860 | +__acquires(e4b->alloc_semp) | |
5861 | { | |
5862 | - struct ext4_sb_info *sbi = EXT4_SB(sb); | |
5863 | - struct inode *inode = sbi->s_buddy_cache; | |
5864 | int blocks_per_page; | |
5865 | int block; | |
5866 | int pnum; | |
5867 | int poff; | |
5868 | struct page *page; | |
5869 | int ret; | |
5870 | + struct ext4_group_info *grp; | |
5871 | + struct ext4_sb_info *sbi = EXT4_SB(sb); | |
5872 | + struct inode *inode = sbi->s_buddy_cache; | |
5873 | ||
5874 | - mb_debug("load group %lu\n", group); | |
5875 | + mb_debug("load group %u\n", group); | |
5876 | ||
5877 | blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize; | |
5878 | + grp = ext4_get_group_info(sb, group); | |
5879 | ||
5880 | e4b->bd_blkbits = sb->s_blocksize_bits; | |
5881 | e4b->bd_info = ext4_get_group_info(sb, group); | |
5882 | @@ -905,6 +946,16 @@ ext4_mb_load_buddy(struct super_block *s | |
5883 | e4b->bd_group = group; | |
5884 | e4b->bd_buddy_page = NULL; | |
5885 | e4b->bd_bitmap_page = NULL; | |
5886 | + e4b->alloc_semp = &grp->alloc_sem; | |
5887 | + | |
5888 | + /* Take the read lock on the group alloc | |
5889 | + * sem. This would make sure a parallel | |
5890 | + * ext4_mb_init_group happening on other | |
5891 | + * groups mapped by the page is blocked | |
5892 | + * till we are done with allocation | |
5893 | + */ | |
5894 | + down_read(e4b->alloc_semp); | |
5895 | + __acquire(e4b->alloc_semp); | |
5896 | ||
5897 | /* | |
5898 | * the buddy cache inode stores the block bitmap | |
5899 | @@ -920,6 +971,14 @@ ext4_mb_load_buddy(struct super_block *s | |
5900 | page = find_get_page(inode->i_mapping, pnum); | |
5901 | if (page == NULL || !PageUptodate(page)) { | |
5902 | if (page) | |
5903 | + /* | |
5904 | + * drop the page reference and try | |
5905 | + * to get the page with lock. If we | |
5906 | + * are not uptodate that implies | |
5907 | + * somebody just created the page but | |
5908 | + * is yet to initialize the same. So | |
5909 | + * wait for it to initialize. | |
5910 | + */ | |
5911 | page_cache_release(page); | |
5912 | page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS); | |
5913 | if (page) { | |
5914 | @@ -985,15 +1044,23 @@ err: | |
5915 | page_cache_release(e4b->bd_buddy_page); | |
5916 | e4b->bd_buddy = NULL; | |
5917 | e4b->bd_bitmap = NULL; | |
5918 | + | |
5919 | + /* Done with the buddy cache */ | |
5920 | + up_read(e4b->alloc_semp); | |
5921 | return ret; | |
5922 | } | |
5923 | ||
5924 | static void ext4_mb_release_desc(struct ext4_buddy *e4b) | |
5925 | +__releases(e4b->alloc_semp) | |
5926 | { | |
5927 | if (e4b->bd_bitmap_page) | |
5928 | page_cache_release(e4b->bd_bitmap_page); | |
5929 | if (e4b->bd_buddy_page) | |
5930 | page_cache_release(e4b->bd_buddy_page); | |
5931 | + /* Done with the buddy cache */ | |
5932 | + if (e4b->alloc_semp) | |
5933 | + up_read(e4b->alloc_semp); | |
5934 | + __release(e4b->alloc_semp); | |
5935 | } | |
5936 | ||
5937 | ||
5938 | @@ -1031,7 +1098,10 @@ static void mb_clear_bits(spinlock_t *lo | |
5939 | cur += 32; | |
5940 | continue; | |
5941 | } | |
5942 | - mb_clear_bit_atomic(lock, cur, bm); | |
5943 | + if (lock) | |
5944 | + mb_clear_bit_atomic(lock, cur, bm); | |
5945 | + else | |
5946 | + mb_clear_bit(cur, bm); | |
5947 | cur++; | |
5948 | } | |
5949 | } | |
5950 | @@ -1049,7 +1119,10 @@ static void mb_set_bits(spinlock_t *lock | |
5951 | cur += 32; | |
5952 | continue; | |
5953 | } | |
5954 | - mb_set_bit_atomic(lock, cur, bm); | |
5955 | + if (lock) | |
5956 | + mb_set_bit_atomic(lock, cur, bm); | |
5957 | + else | |
5958 | + mb_set_bit(cur, bm); | |
5959 | cur++; | |
5960 | } | |
5961 | } | |
5962 | @@ -1094,12 +1167,11 @@ static void mb_free_blocks(struct inode | |
5963 | blocknr += block; | |
5964 | blocknr += | |
5965 | le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block); | |
5966 | - ext4_unlock_group(sb, e4b->bd_group); | |
5967 | - ext4_error(sb, __func__, "double-free of inode" | |
5968 | - " %lu's block %llu(bit %u in group %lu)\n", | |
5969 | + ext4_grp_locked_error(sb, e4b->bd_group, | |
5970 | + __func__, "double-free of inode" | |
5971 | + " %lu's block %llu(bit %u in group %u)\n", | |
5972 | inode ? inode->i_ino : 0, blocknr, block, | |
5973 | e4b->bd_group); | |
5974 | - ext4_lock_group(sb, e4b->bd_group); | |
5975 | } | |
5976 | mb_clear_bit(block, EXT4_MB_BITMAP(e4b)); | |
5977 | e4b->bd_info->bb_counters[order]++; | |
5978 | @@ -1296,13 +1368,20 @@ static void ext4_mb_use_best_found(struc | |
5979 | ac->ac_tail = ret & 0xffff; | |
5980 | ac->ac_buddy = ret >> 16; | |
5981 | ||
5982 | - /* XXXXXXX: SUCH A HORRIBLE **CK */ | |
5983 | - /*FIXME!! Why ? */ | |
5984 | + /* | |
5985 | + * take the page reference. We want the page to be pinned | |
5986 | + * so that we don't get a ext4_mb_init_cache_call for this | |
5987 | + * group until we update the bitmap. That would mean we | |
5988 | + * double allocate blocks. The reference is dropped | |
5989 | + * in ext4_mb_release_context | |
5990 | + */ | |
5991 | ac->ac_bitmap_page = e4b->bd_bitmap_page; | |
5992 | get_page(ac->ac_bitmap_page); | |
5993 | ac->ac_buddy_page = e4b->bd_buddy_page; | |
5994 | get_page(ac->ac_buddy_page); | |
5995 | - | |
5996 | + /* on allocation we use ac to track the held semaphore */ | |
5997 | + ac->alloc_semp = e4b->alloc_semp; | |
5998 | + e4b->alloc_semp = NULL; | |
5999 | /* store last allocated for subsequent stream allocation */ | |
6000 | if ((ac->ac_flags & EXT4_MB_HINT_DATA)) { | |
6001 | spin_lock(&sbi->s_md_lock); | |
6002 | @@ -1433,8 +1512,10 @@ static int ext4_mb_try_best_found(struct | |
6003 | ||
6004 | BUG_ON(ex.fe_len <= 0); | |
6005 | err = ext4_mb_load_buddy(ac->ac_sb, group, e4b); | |
6006 | - if (err) | |
6007 | + if (err) { | |
6008 | + __release(e4b->alloc_semp); | |
6009 | return err; | |
6010 | + } | |
6011 | ||
6012 | ext4_lock_group(ac->ac_sb, group); | |
6013 | max = mb_find_extent(e4b, 0, ex.fe_start, ex.fe_len, &ex); | |
6014 | @@ -1464,8 +1545,10 @@ static int ext4_mb_find_by_goal(struct e | |
6015 | return 0; | |
6016 | ||
6017 | err = ext4_mb_load_buddy(ac->ac_sb, group, e4b); | |
6018 | - if (err) | |
6019 | + if (err) { | |
6020 | + __release(e4b->alloc_semp); | |
6021 | return err; | |
6022 | + } | |
6023 | ||
6024 | ext4_lock_group(ac->ac_sb, group); | |
6025 | max = mb_find_extent(e4b, 0, ac->ac_g_ex.fe_start, | |
6026 | @@ -1575,7 +1658,8 @@ static void ext4_mb_complex_scan_group(s | |
6027 | * free blocks even though group info says we | |
6028 | * we have free blocks | |
6029 | */ | |
6030 | - ext4_error(sb, __func__, "%d free blocks as per " | |
6031 | + ext4_grp_locked_error(sb, e4b->bd_group, | |
6032 | + __func__, "%d free blocks as per " | |
6033 | "group info. But bitmap says 0\n", | |
6034 | free); | |
6035 | break; | |
6036 | @@ -1584,7 +1668,8 @@ static void ext4_mb_complex_scan_group(s | |
6037 | mb_find_extent(e4b, 0, i, ac->ac_g_ex.fe_len, &ex); | |
6038 | BUG_ON(ex.fe_len <= 0); | |
6039 | if (free < ex.fe_len) { | |
6040 | - ext4_error(sb, __func__, "%d free blocks as per " | |
6041 | + ext4_grp_locked_error(sb, e4b->bd_group, | |
6042 | + __func__, "%d free blocks as per " | |
6043 | "group info. But got %d blocks\n", | |
6044 | free, ex.fe_len); | |
6045 | /* | |
6046 | @@ -1692,6 +1777,173 @@ static int ext4_mb_good_group(struct ext | |
6047 | return 0; | |
6048 | } | |
6049 | ||
6050 | +/* | |
6051 | + * lock the group_info alloc_sem of all the groups | |
6052 | + * belonging to the same buddy cache page. This | |
6053 | + * make sure other parallel operation on the buddy | |
6054 | + * cache doesn't happen whild holding the buddy cache | |
6055 | + * lock | |
6056 | + */ | |
6057 | +int ext4_mb_get_buddy_cache_lock(struct super_block *sb, ext4_group_t group) | |
6058 | +{ | |
6059 | + int i; | |
6060 | + int block, pnum; | |
6061 | + int blocks_per_page; | |
6062 | + int groups_per_page; | |
6063 | + ext4_group_t first_group; | |
6064 | + struct ext4_group_info *grp; | |
6065 | + | |
6066 | + blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize; | |
6067 | + /* | |
6068 | + * the buddy cache inode stores the block bitmap | |
6069 | + * and buddy information in consecutive blocks. | |
6070 | + * So for each group we need two blocks. | |
6071 | + */ | |
6072 | + block = group * 2; | |
6073 | + pnum = block / blocks_per_page; | |
6074 | + first_group = pnum * blocks_per_page / 2; | |
6075 | + | |
6076 | + groups_per_page = blocks_per_page >> 1; | |
6077 | + if (groups_per_page == 0) | |
6078 | + groups_per_page = 1; | |
6079 | + /* read all groups the page covers into the cache */ | |
6080 | + for (i = 0; i < groups_per_page; i++) { | |
6081 | + | |
6082 | + if ((first_group + i) >= EXT4_SB(sb)->s_groups_count) | |
6083 | + break; | |
6084 | + grp = ext4_get_group_info(sb, first_group + i); | |
6085 | + /* take all groups write allocation | |
6086 | + * semaphore. This make sure there is | |
6087 | + * no block allocation going on in any | |
6088 | + * of that groups | |
6089 | + */ | |
6090 | + down_write_nested(&grp->alloc_sem, i); | |
6091 | + } | |
6092 | + return i; | |
6093 | +} | |
6094 | + | |
6095 | +void ext4_mb_put_buddy_cache_lock(struct super_block *sb, | |
6096 | + ext4_group_t group, int locked_group) | |
6097 | +{ | |
6098 | + int i; | |
6099 | + int block, pnum; | |
6100 | + int blocks_per_page; | |
6101 | + ext4_group_t first_group; | |
6102 | + struct ext4_group_info *grp; | |
6103 | + | |
6104 | + blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize; | |
6105 | + /* | |
6106 | + * the buddy cache inode stores the block bitmap | |
6107 | + * and buddy information in consecutive blocks. | |
6108 | + * So for each group we need two blocks. | |
6109 | + */ | |
6110 | + block = group * 2; | |
6111 | + pnum = block / blocks_per_page; | |
6112 | + first_group = pnum * blocks_per_page / 2; | |
6113 | + /* release locks on all the groups */ | |
6114 | + for (i = 0; i < locked_group; i++) { | |
6115 | + | |
6116 | + grp = ext4_get_group_info(sb, first_group + i); | |
6117 | + /* take all groups write allocation | |
6118 | + * semaphore. This make sure there is | |
6119 | + * no block allocation going on in any | |
6120 | + * of that groups | |
6121 | + */ | |
6122 | + up_write(&grp->alloc_sem); | |
6123 | + } | |
6124 | + | |
6125 | +} | |
6126 | + | |
6127 | +static int ext4_mb_init_group(struct super_block *sb, ext4_group_t group) | |
6128 | +{ | |
6129 | + | |
6130 | + int ret; | |
6131 | + void *bitmap; | |
6132 | + int blocks_per_page; | |
6133 | + int block, pnum, poff; | |
6134 | + int num_grp_locked = 0; | |
6135 | + struct ext4_group_info *this_grp; | |
6136 | + struct ext4_sb_info *sbi = EXT4_SB(sb); | |
6137 | + struct inode *inode = sbi->s_buddy_cache; | |
6138 | + struct page *page = NULL, *bitmap_page = NULL; | |
6139 | + | |
6140 | + mb_debug("init group %lu\n", group); | |
6141 | + blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize; | |
6142 | + this_grp = ext4_get_group_info(sb, group); | |
6143 | + /* | |
6144 | + * This ensures we don't add group | |
6145 | + * to this buddy cache via resize | |
6146 | + */ | |
6147 | + num_grp_locked = ext4_mb_get_buddy_cache_lock(sb, group); | |
6148 | + if (!EXT4_MB_GRP_NEED_INIT(this_grp)) { | |
6149 | + /* | |
6150 | + * somebody initialized the group | |
6151 | + * return without doing anything | |
6152 | + */ | |
6153 | + ret = 0; | |
6154 | + goto err; | |
6155 | + } | |
6156 | + /* | |
6157 | + * the buddy cache inode stores the block bitmap | |
6158 | + * and buddy information in consecutive blocks. | |
6159 | + * So for each group we need two blocks. | |
6160 | + */ | |
6161 | + block = group * 2; | |
6162 | + pnum = block / blocks_per_page; | |
6163 | + poff = block % blocks_per_page; | |
6164 | + page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS); | |
6165 | + if (page) { | |
6166 | + BUG_ON(page->mapping != inode->i_mapping); | |
6167 | + ret = ext4_mb_init_cache(page, NULL); | |
6168 | + if (ret) { | |
6169 | + unlock_page(page); | |
6170 | + goto err; | |
6171 | + } | |
6172 | + unlock_page(page); | |
6173 | + } | |
6174 | + if (page == NULL || !PageUptodate(page)) { | |
6175 | + ret = -EIO; | |
6176 | + goto err; | |
6177 | + } | |
6178 | + mark_page_accessed(page); | |
6179 | + bitmap_page = page; | |
6180 | + bitmap = page_address(page) + (poff * sb->s_blocksize); | |
6181 | + | |
6182 | + /* init buddy cache */ | |
6183 | + block++; | |
6184 | + pnum = block / blocks_per_page; | |
6185 | + poff = block % blocks_per_page; | |
6186 | + page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS); | |
6187 | + if (page == bitmap_page) { | |
6188 | + /* | |
6189 | + * If both the bitmap and buddy are in | |
6190 | + * the same page we don't need to force | |
6191 | + * init the buddy | |
6192 | + */ | |
6193 | + unlock_page(page); | |
6194 | + } else if (page) { | |
6195 | + BUG_ON(page->mapping != inode->i_mapping); | |
6196 | + ret = ext4_mb_init_cache(page, bitmap); | |
6197 | + if (ret) { | |
6198 | + unlock_page(page); | |
6199 | + goto err; | |
6200 | + } | |
6201 | + unlock_page(page); | |
6202 | + } | |
6203 | + if (page == NULL || !PageUptodate(page)) { | |
6204 | + ret = -EIO; | |
6205 | + goto err; | |
6206 | + } | |
6207 | + mark_page_accessed(page); | |
6208 | +err: | |
6209 | + ext4_mb_put_buddy_cache_lock(sb, group, num_grp_locked); | |
6210 | + if (bitmap_page) | |
6211 | + page_cache_release(bitmap_page); | |
6212 | + if (page) | |
6213 | + page_cache_release(page); | |
6214 | + return ret; | |
6215 | +} | |
6216 | + | |
6217 | static noinline_for_stack int | |
6218 | ext4_mb_regular_allocator(struct ext4_allocation_context *ac) | |
6219 | { | |
6220 | @@ -1775,7 +2027,7 @@ repeat: | |
6221 | group = 0; | |
6222 | ||
6223 | /* quick check to skip empty groups */ | |
6224 | - grp = ext4_get_group_info(ac->ac_sb, group); | |
6225 | + grp = ext4_get_group_info(sb, group); | |
6226 | if (grp->bb_free == 0) | |
6227 | continue; | |
6228 | ||
6229 | @@ -1788,10 +2040,9 @@ repeat: | |
6230 | * we need full data about the group | |
6231 | * to make a good selection | |
6232 | */ | |
6233 | - err = ext4_mb_load_buddy(sb, group, &e4b); | |
6234 | + err = ext4_mb_init_group(sb, group); | |
6235 | if (err) | |
6236 | goto out; | |
6237 | - ext4_mb_release_desc(&e4b); | |
6238 | } | |
6239 | ||
6240 | /* | |
6241 | @@ -1802,8 +2053,10 @@ repeat: | |
6242 | continue; | |
6243 | ||
6244 | err = ext4_mb_load_buddy(sb, group, &e4b); | |
6245 | - if (err) | |
6246 | + if (err) { | |
6247 | + __release(e4b->alloc_semp); | |
6248 | goto out; | |
6249 | + } | |
6250 | ||
6251 | ext4_lock_group(sb, group); | |
6252 | if (!ext4_mb_good_group(ac, group, cr)) { | |
6253 | @@ -1932,13 +2185,13 @@ static int ext4_mb_seq_history_show(stru | |
6254 | if (hs->op == EXT4_MB_HISTORY_ALLOC) { | |
6255 | fmt = "%-5u %-8u %-23s %-23s %-23s %-5u %-5u %-2u " | |
6256 | "%-5u %-5s %-5u %-6u\n"; | |
6257 | - sprintf(buf2, "%lu/%d/%u@%u", hs->result.fe_group, | |
6258 | + sprintf(buf2, "%u/%d/%u@%u", hs->result.fe_group, | |
6259 | hs->result.fe_start, hs->result.fe_len, | |
6260 | hs->result.fe_logical); | |
6261 | - sprintf(buf, "%lu/%d/%u@%u", hs->orig.fe_group, | |
6262 | + sprintf(buf, "%u/%d/%u@%u", hs->orig.fe_group, | |
6263 | hs->orig.fe_start, hs->orig.fe_len, | |
6264 | hs->orig.fe_logical); | |
6265 | - sprintf(buf3, "%lu/%d/%u@%u", hs->goal.fe_group, | |
6266 | + sprintf(buf3, "%u/%d/%u@%u", hs->goal.fe_group, | |
6267 | hs->goal.fe_start, hs->goal.fe_len, | |
6268 | hs->goal.fe_logical); | |
6269 | seq_printf(seq, fmt, hs->pid, hs->ino, buf, buf3, buf2, | |
6270 | @@ -1947,20 +2200,20 @@ static int ext4_mb_seq_history_show(stru | |
6271 | hs->buddy ? 1 << hs->buddy : 0); | |
6272 | } else if (hs->op == EXT4_MB_HISTORY_PREALLOC) { | |
6273 | fmt = "%-5u %-8u %-23s %-23s %-23s\n"; | |
6274 | - sprintf(buf2, "%lu/%d/%u@%u", hs->result.fe_group, | |
6275 | + sprintf(buf2, "%u/%d/%u@%u", hs->result.fe_group, | |
6276 | hs->result.fe_start, hs->result.fe_len, | |
6277 | hs->result.fe_logical); | |
6278 | - sprintf(buf, "%lu/%d/%u@%u", hs->orig.fe_group, | |
6279 | + sprintf(buf, "%u/%d/%u@%u", hs->orig.fe_group, | |
6280 | hs->orig.fe_start, hs->orig.fe_len, | |
6281 | hs->orig.fe_logical); | |
6282 | seq_printf(seq, fmt, hs->pid, hs->ino, buf, "", buf2); | |
6283 | } else if (hs->op == EXT4_MB_HISTORY_DISCARD) { | |
6284 | - sprintf(buf2, "%lu/%d/%u", hs->result.fe_group, | |
6285 | + sprintf(buf2, "%u/%d/%u", hs->result.fe_group, | |
6286 | hs->result.fe_start, hs->result.fe_len); | |
6287 | seq_printf(seq, "%-5u %-8u %-23s discard\n", | |
6288 | hs->pid, hs->ino, buf2); | |
6289 | } else if (hs->op == EXT4_MB_HISTORY_FREE) { | |
6290 | - sprintf(buf2, "%lu/%d/%u", hs->result.fe_group, | |
6291 | + sprintf(buf2, "%u/%d/%u", hs->result.fe_group, | |
6292 | hs->result.fe_start, hs->result.fe_len); | |
6293 | seq_printf(seq, "%-5u %-8u %-23s free\n", | |
6294 | hs->pid, hs->ino, buf2); | |
6295 | @@ -2073,7 +2326,7 @@ static void *ext4_mb_seq_groups_start(st | |
6296 | return NULL; | |
6297 | ||
6298 | group = *pos + 1; | |
6299 | - return (void *) group; | |
6300 | + return (void *) ((unsigned long) group); | |
6301 | } | |
6302 | ||
6303 | static void *ext4_mb_seq_groups_next(struct seq_file *seq, void *v, loff_t *pos) | |
6304 | @@ -2086,13 +2339,13 @@ static void *ext4_mb_seq_groups_next(str | |
6305 | if (*pos < 0 || *pos >= sbi->s_groups_count) | |
6306 | return NULL; | |
6307 | group = *pos + 1; | |
6308 | - return (void *) group;; | |
6309 | + return (void *) ((unsigned long) group); | |
6310 | } | |
6311 | ||
6312 | static int ext4_mb_seq_groups_show(struct seq_file *seq, void *v) | |
6313 | { | |
6314 | struct super_block *sb = seq->private; | |
6315 | - long group = (long) v; | |
6316 | + ext4_group_t group = (ext4_group_t) ((unsigned long) v); | |
6317 | int i; | |
6318 | int err; | |
6319 | struct ext4_buddy e4b; | |
6320 | @@ -2114,7 +2367,8 @@ static int ext4_mb_seq_groups_show(struc | |
6321 | sizeof(struct ext4_group_info); | |
6322 | err = ext4_mb_load_buddy(sb, group, &e4b); | |
6323 | if (err) { | |
6324 | - seq_printf(seq, "#%-5lu: I/O error\n", group); | |
6325 | + __release(e4b->alloc_semp); | |
6326 | + seq_printf(seq, "#%-5u: I/O error\n", group); | |
6327 | return 0; | |
6328 | } | |
6329 | ext4_lock_group(sb, group); | |
6330 | @@ -2122,7 +2376,7 @@ static int ext4_mb_seq_groups_show(struc | |
6331 | ext4_unlock_group(sb, group); | |
6332 | ext4_mb_release_desc(&e4b); | |
6333 | ||
6334 | - seq_printf(seq, "#%-5lu: %-5u %-5u %-5u [", group, sg.info.bb_free, | |
6335 | + seq_printf(seq, "#%-5u: %-5u %-5u %-5u [", group, sg.info.bb_free, | |
6336 | sg.info.bb_fragments, sg.info.bb_first_free); | |
6337 | for (i = 0; i <= 13; i++) | |
6338 | seq_printf(seq, " %-5u", i <= sb->s_blocksize_bits + 1 ? | |
6339 | @@ -2169,9 +2423,10 @@ static void ext4_mb_history_release(stru | |
6340 | { | |
6341 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
6342 | ||
6343 | - remove_proc_entry("mb_groups", sbi->s_mb_proc); | |
6344 | - remove_proc_entry("mb_history", sbi->s_mb_proc); | |
6345 | - | |
6346 | + if (sbi->s_proc != NULL) { | |
6347 | + remove_proc_entry("mb_groups", sbi->s_proc); | |
6348 | + remove_proc_entry("mb_history", sbi->s_proc); | |
6349 | + } | |
6350 | kfree(sbi->s_mb_history); | |
6351 | } | |
6352 | ||
6353 | @@ -2180,10 +2435,10 @@ static void ext4_mb_history_init(struct | |
6354 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
6355 | int i; | |
6356 | ||
6357 | - if (sbi->s_mb_proc != NULL) { | |
6358 | - proc_create_data("mb_history", S_IRUGO, sbi->s_mb_proc, | |
6359 | + if (sbi->s_proc != NULL) { | |
6360 | + proc_create_data("mb_history", S_IRUGO, sbi->s_proc, | |
6361 | &ext4_mb_seq_history_fops, sb); | |
6362 | - proc_create_data("mb_groups", S_IRUGO, sbi->s_mb_proc, | |
6363 | + proc_create_data("mb_groups", S_IRUGO, sbi->s_proc, | |
6364 | &ext4_mb_seq_groups_fops, sb); | |
6365 | } | |
6366 | ||
6367 | @@ -2295,10 +2550,12 @@ int ext4_mb_add_groupinfo(struct super_b | |
6368 | ext4_free_blocks_after_init(sb, group, desc); | |
6369 | } else { | |
6370 | meta_group_info[i]->bb_free = | |
6371 | - le16_to_cpu(desc->bg_free_blocks_count); | |
6372 | + ext4_free_blks_count(sb, desc); | |
6373 | } | |
6374 | ||
6375 | INIT_LIST_HEAD(&meta_group_info[i]->bb_prealloc_list); | |
6376 | + init_rwsem(&meta_group_info[i]->alloc_sem); | |
6377 | + meta_group_info[i]->bb_free_root.rb_node = NULL;; | |
6378 | ||
6379 | #ifdef DOUBLE_CHECK | |
6380 | { | |
6381 | @@ -2325,54 +2582,6 @@ exit_meta_group_info: | |
6382 | } /* ext4_mb_add_groupinfo */ | |
6383 | ||
6384 | /* | |
6385 | - * Add a group to the existing groups. | |
6386 | - * This function is used for online resize | |
6387 | - */ | |
6388 | -int ext4_mb_add_more_groupinfo(struct super_block *sb, ext4_group_t group, | |
6389 | - struct ext4_group_desc *desc) | |
6390 | -{ | |
6391 | - struct ext4_sb_info *sbi = EXT4_SB(sb); | |
6392 | - struct inode *inode = sbi->s_buddy_cache; | |
6393 | - int blocks_per_page; | |
6394 | - int block; | |
6395 | - int pnum; | |
6396 | - struct page *page; | |
6397 | - int err; | |
6398 | - | |
6399 | - /* Add group based on group descriptor*/ | |
6400 | - err = ext4_mb_add_groupinfo(sb, group, desc); | |
6401 | - if (err) | |
6402 | - return err; | |
6403 | - | |
6404 | - /* | |
6405 | - * Cache pages containing dynamic mb_alloc datas (buddy and bitmap | |
6406 | - * datas) are set not up to date so that they will be re-initilaized | |
6407 | - * during the next call to ext4_mb_load_buddy | |
6408 | - */ | |
6409 | - | |
6410 | - /* Set buddy page as not up to date */ | |
6411 | - blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize; | |
6412 | - block = group * 2; | |
6413 | - pnum = block / blocks_per_page; | |
6414 | - page = find_get_page(inode->i_mapping, pnum); | |
6415 | - if (page != NULL) { | |
6416 | - ClearPageUptodate(page); | |
6417 | - page_cache_release(page); | |
6418 | - } | |
6419 | - | |
6420 | - /* Set bitmap page as not up to date */ | |
6421 | - block++; | |
6422 | - pnum = block / blocks_per_page; | |
6423 | - page = find_get_page(inode->i_mapping, pnum); | |
6424 | - if (page != NULL) { | |
6425 | - ClearPageUptodate(page); | |
6426 | - page_cache_release(page); | |
6427 | - } | |
6428 | - | |
6429 | - return 0; | |
6430 | -} | |
6431 | - | |
6432 | -/* | |
6433 | * Update an existing group. | |
6434 | * This function is used for online resize | |
6435 | */ | |
6436 | @@ -2455,7 +2664,7 @@ static int ext4_mb_init_backend(struct s | |
6437 | desc = ext4_get_group_desc(sb, i, NULL); | |
6438 | if (desc == NULL) { | |
6439 | printk(KERN_ERR | |
6440 | - "EXT4-fs: can't read descriptor %lu\n", i); | |
6441 | + "EXT4-fs: can't read descriptor %u\n", i); | |
6442 | goto err_freebuddy; | |
6443 | } | |
6444 | if (ext4_mb_add_groupinfo(sb, i, desc) != 0) | |
6445 | @@ -2485,19 +2694,14 @@ int ext4_mb_init(struct super_block *sb, | |
6446 | unsigned max; | |
6447 | int ret; | |
6448 | ||
6449 | - if (!test_opt(sb, MBALLOC)) | |
6450 | - return 0; | |
6451 | - | |
6452 | i = (sb->s_blocksize_bits + 2) * sizeof(unsigned short); | |
6453 | ||
6454 | sbi->s_mb_offsets = kmalloc(i, GFP_KERNEL); | |
6455 | if (sbi->s_mb_offsets == NULL) { | |
6456 | - clear_opt(sbi->s_mount_opt, MBALLOC); | |
6457 | return -ENOMEM; | |
6458 | } | |
6459 | sbi->s_mb_maxs = kmalloc(i, GFP_KERNEL); | |
6460 | if (sbi->s_mb_maxs == NULL) { | |
6461 | - clear_opt(sbi->s_mount_opt, MBALLOC); | |
6462 | kfree(sbi->s_mb_maxs); | |
6463 | return -ENOMEM; | |
6464 | } | |
6465 | @@ -2520,7 +2724,6 @@ int ext4_mb_init(struct super_block *sb, | |
6466 | /* init file for buddy data */ | |
6467 | ret = ext4_mb_init_backend(sb); | |
6468 | if (ret != 0) { | |
6469 | - clear_opt(sbi->s_mount_opt, MBALLOC); | |
6470 | kfree(sbi->s_mb_offsets); | |
6471 | kfree(sbi->s_mb_maxs); | |
6472 | return ret; | |
6473 | @@ -2540,17 +2743,15 @@ int ext4_mb_init(struct super_block *sb, | |
6474 | sbi->s_mb_history_filter = EXT4_MB_HISTORY_DEFAULT; | |
6475 | sbi->s_mb_group_prealloc = MB_DEFAULT_GROUP_PREALLOC; | |
6476 | ||
6477 | - i = sizeof(struct ext4_locality_group) * nr_cpu_ids; | |
6478 | - sbi->s_locality_groups = kmalloc(i, GFP_KERNEL); | |
6479 | + sbi->s_locality_groups = alloc_percpu(struct ext4_locality_group); | |
6480 | if (sbi->s_locality_groups == NULL) { | |
6481 | - clear_opt(sbi->s_mount_opt, MBALLOC); | |
6482 | kfree(sbi->s_mb_offsets); | |
6483 | kfree(sbi->s_mb_maxs); | |
6484 | return -ENOMEM; | |
6485 | } | |
6486 | - for (i = 0; i < nr_cpu_ids; i++) { | |
6487 | + for_each_possible_cpu(i) { | |
6488 | struct ext4_locality_group *lg; | |
6489 | - lg = &sbi->s_locality_groups[i]; | |
6490 | + lg = per_cpu_ptr(sbi->s_locality_groups, i); | |
6491 | mutex_init(&lg->lg_mutex); | |
6492 | for (j = 0; j < PREALLOC_TB_SIZE; j++) | |
6493 | INIT_LIST_HEAD(&lg->lg_prealloc_list[j]); | |
6494 | @@ -2560,7 +2761,7 @@ int ext4_mb_init(struct super_block *sb, | |
6495 | ext4_mb_init_per_dev_proc(sb); | |
6496 | ext4_mb_history_init(sb); | |
6497 | ||
6498 | - printk("EXT4-fs: mballoc enabled\n"); | |
6499 | + printk(KERN_INFO "EXT4-fs: mballoc enabled\n"); | |
6500 | return 0; | |
6501 | } | |
6502 | ||
6503 | @@ -2589,9 +2790,6 @@ int ext4_mb_release(struct super_block * | |
6504 | struct ext4_group_info *grinfo; | |
6505 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
6506 | ||
6507 | - if (!test_opt(sb, MBALLOC)) | |
6508 | - return 0; | |
6509 | - | |
6510 | /* release freed, non-committed blocks */ | |
6511 | spin_lock(&sbi->s_md_lock); | |
6512 | list_splice_init(&sbi->s_closed_transaction, | |
6513 | @@ -2647,8 +2845,7 @@ int ext4_mb_release(struct super_block * | |
6514 | atomic_read(&sbi->s_mb_discarded)); | |
6515 | } | |
6516 | ||
6517 | - kfree(sbi->s_locality_groups); | |
6518 | - | |
6519 | + free_percpu(sbi->s_locality_groups); | |
6520 | ext4_mb_history_release(sb); | |
6521 | ext4_mb_destroy_per_dev_proc(sb); | |
6522 | ||
6523 | @@ -2658,13 +2855,11 @@ int ext4_mb_release(struct super_block * | |
6524 | static noinline_for_stack void | |
6525 | ext4_mb_free_committed_blocks(struct super_block *sb) | |
6526 | { | |
6527 | - struct ext4_sb_info *sbi = EXT4_SB(sb); | |
6528 | - int err; | |
6529 | - int i; | |
6530 | - int count = 0; | |
6531 | - int count2 = 0; | |
6532 | - struct ext4_free_metadata *md; | |
6533 | struct ext4_buddy e4b; | |
6534 | + struct ext4_group_info *db; | |
6535 | + struct ext4_sb_info *sbi = EXT4_SB(sb); | |
6536 | + int err, count = 0, count2 = 0; | |
6537 | + struct ext4_free_data *entry; | |
6538 | ||
6539 | if (list_empty(&sbi->s_committed_transaction)) | |
6540 | return; | |
6541 | @@ -2672,44 +2867,46 @@ ext4_mb_free_committed_blocks(struct sup | |
6542 | /* there is committed blocks to be freed yet */ | |
6543 | do { | |
6544 | /* get next array of blocks */ | |
6545 | - md = NULL; | |
6546 | + entry = NULL; | |
6547 | spin_lock(&sbi->s_md_lock); | |
6548 | if (!list_empty(&sbi->s_committed_transaction)) { | |
6549 | - md = list_entry(sbi->s_committed_transaction.next, | |
6550 | - struct ext4_free_metadata, list); | |
6551 | - list_del(&md->list); | |
6552 | + entry = list_entry(sbi->s_committed_transaction.next, | |
6553 | + struct ext4_free_data, list); | |
6554 | + list_del(&entry->list); | |
6555 | } | |
6556 | spin_unlock(&sbi->s_md_lock); | |
6557 | ||
6558 | - if (md == NULL) | |
6559 | + if (entry == NULL) | |
6560 | break; | |
6561 | ||
6562 | - mb_debug("gonna free %u blocks in group %lu (0x%p):", | |
6563 | - md->num, md->group, md); | |
6564 | + mb_debug("gonna free %u blocks in group %u (0x%p):", | |
6565 | + entry->count, entry->group, entry); | |
6566 | ||
6567 | - err = ext4_mb_load_buddy(sb, md->group, &e4b); | |
6568 | + err = ext4_mb_load_buddy(sb, entry->group, &e4b); | |
6569 | /* we expect to find existing buddy because it's pinned */ | |
6570 | BUG_ON(err != 0); | |
6571 | ||
6572 | + db = e4b.bd_info; | |
6573 | /* there are blocks to put in buddy to make them really free */ | |
6574 | - count += md->num; | |
6575 | + count += entry->count; | |
6576 | count2++; | |
6577 | - ext4_lock_group(sb, md->group); | |
6578 | - for (i = 0; i < md->num; i++) { | |
6579 | - mb_debug(" %u", md->blocks[i]); | |
6580 | - mb_free_blocks(NULL, &e4b, md->blocks[i], 1); | |
6581 | - } | |
6582 | - mb_debug("\n"); | |
6583 | - ext4_unlock_group(sb, md->group); | |
6584 | - | |
6585 | - /* balance refcounts from ext4_mb_free_metadata() */ | |
6586 | - page_cache_release(e4b.bd_buddy_page); | |
6587 | - page_cache_release(e4b.bd_bitmap_page); | |
6588 | + ext4_lock_group(sb, entry->group); | |
6589 | + /* Take it out of per group rb tree */ | |
6590 | + rb_erase(&entry->node, &(db->bb_free_root)); | |
6591 | + mb_free_blocks(NULL, &e4b, entry->start_blk, entry->count); | |
6592 | + | |
6593 | + if (!db->bb_free_root.rb_node) { | |
6594 | + /* No more items in the per group rb tree | |
6595 | + * balance refcounts from ext4_mb_free_metadata() | |
6596 | + */ | |
6597 | + page_cache_release(e4b.bd_buddy_page); | |
6598 | + page_cache_release(e4b.bd_bitmap_page); | |
6599 | + } | |
6600 | + ext4_unlock_group(sb, entry->group); | |
6601 | ||
6602 | - kfree(md); | |
6603 | + kmem_cache_free(ext4_free_ext_cachep, entry); | |
6604 | ext4_mb_release_desc(&e4b); | |
6605 | - | |
6606 | - } while (md); | |
6607 | + } while (1); | |
6608 | ||
6609 | mb_debug("freed %u blocks in %u structures\n", count, count2); | |
6610 | } | |
6611 | @@ -2721,129 +2918,52 @@ ext4_mb_free_committed_blocks(struct sup | |
6612 | #define EXT4_MB_STREAM_REQ "stream_req" | |
6613 | #define EXT4_MB_GROUP_PREALLOC "group_prealloc" | |
6614 | ||
6615 | - | |
6616 | - | |
6617 | -#define MB_PROC_FOPS(name) \ | |
6618 | -static int ext4_mb_##name##_proc_show(struct seq_file *m, void *v) \ | |
6619 | -{ \ | |
6620 | - struct ext4_sb_info *sbi = m->private; \ | |
6621 | - \ | |
6622 | - seq_printf(m, "%ld\n", sbi->s_mb_##name); \ | |
6623 | - return 0; \ | |
6624 | -} \ | |
6625 | - \ | |
6626 | -static int ext4_mb_##name##_proc_open(struct inode *inode, struct file *file)\ | |
6627 | -{ \ | |
6628 | - return single_open(file, ext4_mb_##name##_proc_show, PDE(inode)->data);\ | |
6629 | -} \ | |
6630 | - \ | |
6631 | -static ssize_t ext4_mb_##name##_proc_write(struct file *file, \ | |
6632 | - const char __user *buf, size_t cnt, loff_t *ppos) \ | |
6633 | -{ \ | |
6634 | - struct ext4_sb_info *sbi = PDE(file->f_path.dentry->d_inode)->data;\ | |
6635 | - char str[32]; \ | |
6636 | - long value; \ | |
6637 | - if (cnt >= sizeof(str)) \ | |
6638 | - return -EINVAL; \ | |
6639 | - if (copy_from_user(str, buf, cnt)) \ | |
6640 | - return -EFAULT; \ | |
6641 | - value = simple_strtol(str, NULL, 0); \ | |
6642 | - if (value <= 0) \ | |
6643 | - return -ERANGE; \ | |
6644 | - sbi->s_mb_##name = value; \ | |
6645 | - return cnt; \ | |
6646 | -} \ | |
6647 | - \ | |
6648 | -static const struct file_operations ext4_mb_##name##_proc_fops = { \ | |
6649 | - .owner = THIS_MODULE, \ | |
6650 | - .open = ext4_mb_##name##_proc_open, \ | |
6651 | - .read = seq_read, \ | |
6652 | - .llseek = seq_lseek, \ | |
6653 | - .release = single_release, \ | |
6654 | - .write = ext4_mb_##name##_proc_write, \ | |
6655 | -}; | |
6656 | - | |
6657 | -MB_PROC_FOPS(stats); | |
6658 | -MB_PROC_FOPS(max_to_scan); | |
6659 | -MB_PROC_FOPS(min_to_scan); | |
6660 | -MB_PROC_FOPS(order2_reqs); | |
6661 | -MB_PROC_FOPS(stream_request); | |
6662 | -MB_PROC_FOPS(group_prealloc); | |
6663 | - | |
6664 | -#define MB_PROC_HANDLER(name, var) \ | |
6665 | -do { \ | |
6666 | - proc = proc_create_data(name, mode, sbi->s_mb_proc, \ | |
6667 | - &ext4_mb_##var##_proc_fops, sbi); \ | |
6668 | - if (proc == NULL) { \ | |
6669 | - printk(KERN_ERR "EXT4-fs: can't to create %s\n", name); \ | |
6670 | - goto err_out; \ | |
6671 | - } \ | |
6672 | -} while (0) | |
6673 | - | |
6674 | static int ext4_mb_init_per_dev_proc(struct super_block *sb) | |
6675 | { | |
6676 | +#ifdef CONFIG_PROC_FS | |
6677 | mode_t mode = S_IFREG | S_IRUGO | S_IWUSR; | |
6678 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
6679 | struct proc_dir_entry *proc; | |
6680 | - char devname[BDEVNAME_SIZE], *p; | |
6681 | ||
6682 | - if (proc_root_ext4 == NULL) { | |
6683 | - sbi->s_mb_proc = NULL; | |
6684 | + if (sbi->s_proc == NULL) | |
6685 | return -EINVAL; | |
6686 | - } | |
6687 | - bdevname(sb->s_bdev, devname); | |
6688 | - p = devname; | |
6689 | - while ((p = strchr(p, '/'))) | |
6690 | - *p = '!'; | |
6691 | - | |
6692 | - sbi->s_mb_proc = proc_mkdir(devname, proc_root_ext4); | |
6693 | - if (!sbi->s_mb_proc) | |
6694 | - goto err_create_dir; | |
6695 | - | |
6696 | - MB_PROC_HANDLER(EXT4_MB_STATS_NAME, stats); | |
6697 | - MB_PROC_HANDLER(EXT4_MB_MAX_TO_SCAN_NAME, max_to_scan); | |
6698 | - MB_PROC_HANDLER(EXT4_MB_MIN_TO_SCAN_NAME, min_to_scan); | |
6699 | - MB_PROC_HANDLER(EXT4_MB_ORDER2_REQ, order2_reqs); | |
6700 | - MB_PROC_HANDLER(EXT4_MB_STREAM_REQ, stream_request); | |
6701 | - MB_PROC_HANDLER(EXT4_MB_GROUP_PREALLOC, group_prealloc); | |
6702 | ||
6703 | + EXT4_PROC_HANDLER(EXT4_MB_STATS_NAME, mb_stats); | |
6704 | + EXT4_PROC_HANDLER(EXT4_MB_MAX_TO_SCAN_NAME, mb_max_to_scan); | |
6705 | + EXT4_PROC_HANDLER(EXT4_MB_MIN_TO_SCAN_NAME, mb_min_to_scan); | |
6706 | + EXT4_PROC_HANDLER(EXT4_MB_ORDER2_REQ, mb_order2_reqs); | |
6707 | + EXT4_PROC_HANDLER(EXT4_MB_STREAM_REQ, mb_stream_request); | |
6708 | + EXT4_PROC_HANDLER(EXT4_MB_GROUP_PREALLOC, mb_group_prealloc); | |
6709 | return 0; | |
6710 | ||
6711 | err_out: | |
6712 | - remove_proc_entry(EXT4_MB_GROUP_PREALLOC, sbi->s_mb_proc); | |
6713 | - remove_proc_entry(EXT4_MB_STREAM_REQ, sbi->s_mb_proc); | |
6714 | - remove_proc_entry(EXT4_MB_ORDER2_REQ, sbi->s_mb_proc); | |
6715 | - remove_proc_entry(EXT4_MB_MIN_TO_SCAN_NAME, sbi->s_mb_proc); | |
6716 | - remove_proc_entry(EXT4_MB_MAX_TO_SCAN_NAME, sbi->s_mb_proc); | |
6717 | - remove_proc_entry(EXT4_MB_STATS_NAME, sbi->s_mb_proc); | |
6718 | - remove_proc_entry(devname, proc_root_ext4); | |
6719 | - sbi->s_mb_proc = NULL; | |
6720 | -err_create_dir: | |
6721 | - printk(KERN_ERR "EXT4-fs: Unable to create %s\n", devname); | |
6722 | - | |
6723 | + remove_proc_entry(EXT4_MB_GROUP_PREALLOC, sbi->s_proc); | |
6724 | + remove_proc_entry(EXT4_MB_STREAM_REQ, sbi->s_proc); | |
6725 | + remove_proc_entry(EXT4_MB_ORDER2_REQ, sbi->s_proc); | |
6726 | + remove_proc_entry(EXT4_MB_MIN_TO_SCAN_NAME, sbi->s_proc); | |
6727 | + remove_proc_entry(EXT4_MB_MAX_TO_SCAN_NAME, sbi->s_proc); | |
6728 | + remove_proc_entry(EXT4_MB_STATS_NAME, sbi->s_proc); | |
6729 | return -ENOMEM; | |
6730 | +#else | |
6731 | + return 0; | |
6732 | +#endif | |
6733 | } | |
6734 | ||
6735 | static int ext4_mb_destroy_per_dev_proc(struct super_block *sb) | |
6736 | { | |
6737 | +#ifdef CONFIG_PROC_FS | |
6738 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
6739 | - char devname[BDEVNAME_SIZE], *p; | |
6740 | ||
6741 | - if (sbi->s_mb_proc == NULL) | |
6742 | + if (sbi->s_proc == NULL) | |
6743 | return -EINVAL; | |
6744 | ||
6745 | - bdevname(sb->s_bdev, devname); | |
6746 | - p = devname; | |
6747 | - while ((p = strchr(p, '/'))) | |
6748 | - *p = '!'; | |
6749 | - remove_proc_entry(EXT4_MB_GROUP_PREALLOC, sbi->s_mb_proc); | |
6750 | - remove_proc_entry(EXT4_MB_STREAM_REQ, sbi->s_mb_proc); | |
6751 | - remove_proc_entry(EXT4_MB_ORDER2_REQ, sbi->s_mb_proc); | |
6752 | - remove_proc_entry(EXT4_MB_MIN_TO_SCAN_NAME, sbi->s_mb_proc); | |
6753 | - remove_proc_entry(EXT4_MB_MAX_TO_SCAN_NAME, sbi->s_mb_proc); | |
6754 | - remove_proc_entry(EXT4_MB_STATS_NAME, sbi->s_mb_proc); | |
6755 | - remove_proc_entry(devname, proc_root_ext4); | |
6756 | - | |
6757 | + remove_proc_entry(EXT4_MB_GROUP_PREALLOC, sbi->s_proc); | |
6758 | + remove_proc_entry(EXT4_MB_STREAM_REQ, sbi->s_proc); | |
6759 | + remove_proc_entry(EXT4_MB_ORDER2_REQ, sbi->s_proc); | |
6760 | + remove_proc_entry(EXT4_MB_MIN_TO_SCAN_NAME, sbi->s_proc); | |
6761 | + remove_proc_entry(EXT4_MB_MAX_TO_SCAN_NAME, sbi->s_proc); | |
6762 | + remove_proc_entry(EXT4_MB_STATS_NAME, sbi->s_proc); | |
6763 | +#endif | |
6764 | return 0; | |
6765 | } | |
6766 | ||
6767 | @@ -2864,11 +2984,16 @@ int __init init_ext4_mballoc(void) | |
6768 | kmem_cache_destroy(ext4_pspace_cachep); | |
6769 | return -ENOMEM; | |
6770 | } | |
6771 | -#ifdef CONFIG_PROC_FS | |
6772 | - proc_root_ext4 = proc_mkdir("fs/ext4", NULL); | |
6773 | - if (proc_root_ext4 == NULL) | |
6774 | - printk(KERN_ERR "EXT4-fs: Unable to create fs/ext4\n"); | |
6775 | -#endif | |
6776 | + | |
6777 | + ext4_free_ext_cachep = | |
6778 | + kmem_cache_create("ext4_free_block_extents", | |
6779 | + sizeof(struct ext4_free_data), | |
6780 | + 0, SLAB_RECLAIM_ACCOUNT, NULL); | |
6781 | + if (ext4_free_ext_cachep == NULL) { | |
6782 | + kmem_cache_destroy(ext4_pspace_cachep); | |
6783 | + kmem_cache_destroy(ext4_ac_cachep); | |
6784 | + return -ENOMEM; | |
6785 | + } | |
6786 | return 0; | |
6787 | } | |
6788 | ||
6789 | @@ -2877,9 +3002,7 @@ void exit_ext4_mballoc(void) | |
6790 | /* XXX: synchronize_rcu(); */ | |
6791 | kmem_cache_destroy(ext4_pspace_cachep); | |
6792 | kmem_cache_destroy(ext4_ac_cachep); | |
6793 | -#ifdef CONFIG_PROC_FS | |
6794 | - remove_proc_entry("fs/ext4", NULL); | |
6795 | -#endif | |
6796 | + kmem_cache_destroy(ext4_free_ext_cachep); | |
6797 | } | |
6798 | ||
6799 | ||
6800 | @@ -2889,7 +3012,7 @@ void exit_ext4_mballoc(void) | |
6801 | */ | |
6802 | static noinline_for_stack int | |
6803 | ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac, | |
6804 | - handle_t *handle) | |
6805 | + handle_t *handle, unsigned int reserv_blks) | |
6806 | { | |
6807 | struct buffer_head *bitmap_bh = NULL; | |
6808 | struct ext4_super_block *es; | |
6809 | @@ -2922,7 +3045,7 @@ ext4_mb_mark_diskspace_used(struct ext4_ | |
6810 | if (!gdp) | |
6811 | goto out_err; | |
6812 | ||
6813 | - ext4_debug("using block group %lu(%d)\n", ac->ac_b_ex.fe_group, | |
6814 | + ext4_debug("using block group %u(%d)\n", ac->ac_b_ex.fe_group, | |
6815 | gdp->bg_free_blocks_count); | |
6816 | ||
6817 | err = ext4_journal_get_write_access(handle, gdp_bh); | |
6818 | @@ -2941,8 +3064,8 @@ ext4_mb_mark_diskspace_used(struct ext4_ | |
6819 | in_range(block + len - 1, ext4_inode_table(sb, gdp), | |
6820 | EXT4_SB(sb)->s_itb_per_group)) { | |
6821 | ext4_error(sb, __func__, | |
6822 | - "Allocating block in system zone - block = %llu", | |
6823 | - block); | |
6824 | + "Allocating block %llu in system zone of %d group\n", | |
6825 | + block, ac->ac_b_ex.fe_group); | |
6826 | /* File system mounted not to panic on error | |
6827 | * Fix the bitmap and repeat the block allocation | |
6828 | * We leak some of the blocks here. | |
6829 | @@ -2964,29 +3087,29 @@ ext4_mb_mark_diskspace_used(struct ext4_ | |
6830 | } | |
6831 | } | |
6832 | #endif | |
6833 | - mb_set_bits(sb_bgl_lock(sbi, ac->ac_b_ex.fe_group), bitmap_bh->b_data, | |
6834 | - ac->ac_b_ex.fe_start, ac->ac_b_ex.fe_len); | |
6835 | - | |
6836 | spin_lock(sb_bgl_lock(sbi, ac->ac_b_ex.fe_group)); | |
6837 | + mb_set_bits(NULL, bitmap_bh->b_data, | |
6838 | + ac->ac_b_ex.fe_start, ac->ac_b_ex.fe_len); | |
6839 | if (gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { | |
6840 | gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT); | |
6841 | - gdp->bg_free_blocks_count = | |
6842 | - cpu_to_le16(ext4_free_blocks_after_init(sb, | |
6843 | - ac->ac_b_ex.fe_group, | |
6844 | - gdp)); | |
6845 | + ext4_free_blks_set(sb, gdp, | |
6846 | + ext4_free_blocks_after_init(sb, | |
6847 | + ac->ac_b_ex.fe_group, gdp)); | |
6848 | } | |
6849 | - le16_add_cpu(&gdp->bg_free_blocks_count, -ac->ac_b_ex.fe_len); | |
6850 | + len = ext4_free_blks_count(sb, gdp) - ac->ac_b_ex.fe_len; | |
6851 | + ext4_free_blks_set(sb, gdp, len); | |
6852 | gdp->bg_checksum = ext4_group_desc_csum(sbi, ac->ac_b_ex.fe_group, gdp); | |
6853 | spin_unlock(sb_bgl_lock(sbi, ac->ac_b_ex.fe_group)); | |
6854 | - | |
6855 | + percpu_counter_sub(&sbi->s_freeblocks_counter, ac->ac_b_ex.fe_len); | |
6856 | /* | |
6857 | - * free blocks account has already be reduced/reserved | |
6858 | - * at write_begin() time for delayed allocation | |
6859 | - * do not double accounting | |
6860 | + * Now reduce the dirty block count also. Should not go negative | |
6861 | */ | |
6862 | if (!(ac->ac_flags & EXT4_MB_DELALLOC_RESERVED)) | |
6863 | - percpu_counter_sub(&sbi->s_freeblocks_counter, | |
6864 | - ac->ac_b_ex.fe_len); | |
6865 | + /* release all the reserved blocks if non delalloc */ | |
6866 | + percpu_counter_sub(&sbi->s_dirtyblocks_counter, reserv_blks); | |
6867 | + else | |
6868 | + percpu_counter_sub(&sbi->s_dirtyblocks_counter, | |
6869 | + ac->ac_b_ex.fe_len); | |
6870 | ||
6871 | if (sbi->s_log_groups_per_flex) { | |
6872 | ext4_group_t flex_group = ext4_flex_group(sbi, | |
6873 | @@ -3128,7 +3251,7 @@ ext4_mb_normalize_request(struct ext4_al | |
6874 | /* check we don't cross already preallocated blocks */ | |
6875 | rcu_read_lock(); | |
6876 | list_for_each_entry_rcu(pa, &ei->i_prealloc_list, pa_inode_list) { | |
6877 | - unsigned long pa_end; | |
6878 | + ext4_lblk_t pa_end; | |
6879 | ||
6880 | if (pa->pa_deleted) | |
6881 | continue; | |
6882 | @@ -3172,7 +3295,7 @@ ext4_mb_normalize_request(struct ext4_al | |
6883 | /* XXX: extra loop to check we really don't overlap preallocations */ | |
6884 | rcu_read_lock(); | |
6885 | list_for_each_entry_rcu(pa, &ei->i_prealloc_list, pa_inode_list) { | |
6886 | - unsigned long pa_end; | |
6887 | + ext4_lblk_t pa_end; | |
6888 | spin_lock(&pa->pa_lock); | |
6889 | if (pa->pa_deleted == 0) { | |
6890 | pa_end = pa->pa_lstart + pa->pa_len; | |
6891 | @@ -3404,6 +3527,32 @@ ext4_mb_use_preallocated(struct ext4_all | |
6892 | } | |
6893 | ||
6894 | /* | |
6895 | + * the function goes through all block freed in the group | |
6896 | + * but not yet committed and marks them used in in-core bitmap. | |
6897 | + * buddy must be generated from this bitmap | |
6898 | + * Need to be called with ext4 group lock (ext4_lock_group) | |
6899 | + */ | |
6900 | +static void ext4_mb_generate_from_freelist(struct super_block *sb, void *bitmap, | |
6901 | + ext4_group_t group) | |
6902 | +{ | |
6903 | + struct rb_node *n; | |
6904 | + struct ext4_group_info *grp; | |
6905 | + struct ext4_free_data *entry; | |
6906 | + | |
6907 | + grp = ext4_get_group_info(sb, group); | |
6908 | + n = rb_first(&(grp->bb_free_root)); | |
6909 | + | |
6910 | + while (n) { | |
6911 | + entry = rb_entry(n, struct ext4_free_data, node); | |
6912 | + mb_set_bits(sb_bgl_lock(EXT4_SB(sb), group), | |
6913 | + bitmap, entry->start_blk, | |
6914 | + entry->count); | |
6915 | + n = rb_next(n); | |
6916 | + } | |
6917 | + return; | |
6918 | +} | |
6919 | + | |
6920 | +/* | |
6921 | * the function goes through all preallocation in this group and marks them | |
6922 | * used in in-core bitmap. buddy must be generated from this bitmap | |
6923 | * Need to be called with ext4 group lock (ext4_lock_group) | |
6924 | @@ -3443,7 +3592,7 @@ static void ext4_mb_generate_from_pa(str | |
6925 | preallocated += len; | |
6926 | count++; | |
6927 | } | |
6928 | - mb_debug("prellocated %u for group %lu\n", preallocated, group); | |
6929 | + mb_debug("prellocated %u for group %u\n", preallocated, group); | |
6930 | } | |
6931 | ||
6932 | static void ext4_mb_pa_callback(struct rcu_head *head) | |
6933 | @@ -3460,7 +3609,7 @@ static void ext4_mb_pa_callback(struct r | |
6934 | static void ext4_mb_put_pa(struct ext4_allocation_context *ac, | |
6935 | struct super_block *sb, struct ext4_prealloc_space *pa) | |
6936 | { | |
6937 | - unsigned long grp; | |
6938 | + ext4_group_t grp; | |
6939 | ||
6940 | if (!atomic_dec_and_test(&pa->pa_count) || pa->pa_free != 0) | |
6941 | return; | |
6942 | @@ -3676,8 +3825,8 @@ ext4_mb_release_inode_pa(struct ext4_bud | |
6943 | { | |
6944 | struct super_block *sb = e4b->bd_sb; | |
6945 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
6946 | - unsigned long end; | |
6947 | - unsigned long next; | |
6948 | + unsigned int end; | |
6949 | + unsigned int next; | |
6950 | ext4_group_t group; | |
6951 | ext4_grpblk_t bit; | |
6952 | sector_t start; | |
6953 | @@ -3723,8 +3872,9 @@ ext4_mb_release_inode_pa(struct ext4_bud | |
6954 | pa, (unsigned long) pa->pa_lstart, | |
6955 | (unsigned long) pa->pa_pstart, | |
6956 | (unsigned long) pa->pa_len); | |
6957 | - ext4_error(sb, __func__, "free %u, pa_free %u\n", | |
6958 | - free, pa->pa_free); | |
6959 | + ext4_grp_locked_error(sb, group, | |
6960 | + __func__, "free %u, pa_free %u\n", | |
6961 | + free, pa->pa_free); | |
6962 | /* | |
6963 | * pa is already deleted so we use the value obtained | |
6964 | * from the bitmap and continue. | |
6965 | @@ -3789,7 +3939,7 @@ ext4_mb_discard_group_preallocations(str | |
6966 | int busy = 0; | |
6967 | int free = 0; | |
6968 | ||
6969 | - mb_debug("discard preallocation for group %lu\n", group); | |
6970 | + mb_debug("discard preallocation for group %u\n", group); | |
6971 | ||
6972 | if (list_empty(&grp->bb_prealloc_list)) | |
6973 | return 0; | |
6974 | @@ -3797,14 +3947,15 @@ ext4_mb_discard_group_preallocations(str | |
6975 | bitmap_bh = ext4_read_block_bitmap(sb, group); | |
6976 | if (bitmap_bh == NULL) { | |
6977 | ext4_error(sb, __func__, "Error in reading block " | |
6978 | - "bitmap for %lu\n", group); | |
6979 | + "bitmap for %u\n", group); | |
6980 | return 0; | |
6981 | } | |
6982 | ||
6983 | err = ext4_mb_load_buddy(sb, group, &e4b); | |
6984 | if (err) { | |
6985 | + __release(e4b->alloc_semp); | |
6986 | ext4_error(sb, __func__, "Error in loading buddy " | |
6987 | - "information for %lu\n", group); | |
6988 | + "information for %u\n", group); | |
6989 | put_bh(bitmap_bh); | |
6990 | return 0; | |
6991 | } | |
6992 | @@ -3894,7 +4045,7 @@ out: | |
6993 | * | |
6994 | * FIXME!! Make sure it is valid at all the call sites | |
6995 | */ | |
6996 | -void ext4_mb_discard_inode_preallocations(struct inode *inode) | |
6997 | +void ext4_discard_preallocations(struct inode *inode) | |
6998 | { | |
6999 | struct ext4_inode_info *ei = EXT4_I(inode); | |
7000 | struct super_block *sb = inode->i_sb; | |
7001 | @@ -3906,7 +4057,7 @@ void ext4_mb_discard_inode_preallocation | |
7002 | struct ext4_buddy e4b; | |
7003 | int err; | |
7004 | ||
7005 | - if (!test_opt(sb, MBALLOC) || !S_ISREG(inode->i_mode)) { | |
7006 | + if (!S_ISREG(inode->i_mode)) { | |
7007 | /*BUG_ON(!list_empty(&ei->i_prealloc_list));*/ | |
7008 | return; | |
7009 | } | |
7010 | @@ -3970,15 +4121,16 @@ repeat: | |
7011 | ||
7012 | err = ext4_mb_load_buddy(sb, group, &e4b); | |
7013 | if (err) { | |
7014 | + __release(e4b->alloc_semp); | |
7015 | ext4_error(sb, __func__, "Error in loading buddy " | |
7016 | - "information for %lu\n", group); | |
7017 | + "information for %u\n", group); | |
7018 | continue; | |
7019 | } | |
7020 | ||
7021 | bitmap_bh = ext4_read_block_bitmap(sb, group); | |
7022 | if (bitmap_bh == NULL) { | |
7023 | ext4_error(sb, __func__, "Error in reading block " | |
7024 | - "bitmap for %lu\n", group); | |
7025 | + "bitmap for %u\n", group); | |
7026 | ext4_mb_release_desc(&e4b); | |
7027 | continue; | |
7028 | } | |
7029 | @@ -4104,8 +4256,7 @@ static void ext4_mb_group_or_file(struct | |
7030 | * per cpu locality group is to reduce the contention between block | |
7031 | * request from multiple CPUs. | |
7032 | */ | |
7033 | - ac->ac_lg = &sbi->s_locality_groups[get_cpu()]; | |
7034 | - put_cpu(); | |
7035 | + ac->ac_lg = per_cpu_ptr(sbi->s_locality_groups, raw_smp_processor_id()); | |
7036 | ||
7037 | /* we're going to use group allocation */ | |
7038 | ac->ac_flags |= EXT4_MB_HINT_GROUP_ALLOC; | |
7039 | @@ -4122,8 +4273,8 @@ ext4_mb_initialize_context(struct ext4_a | |
7040 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
7041 | struct ext4_super_block *es = sbi->s_es; | |
7042 | ext4_group_t group; | |
7043 | - unsigned long len; | |
7044 | - unsigned long goal; | |
7045 | + unsigned int len; | |
7046 | + ext4_fsblk_t goal; | |
7047 | ext4_grpblk_t block; | |
7048 | ||
7049 | /* we can't allocate > group size */ | |
7050 | @@ -4166,6 +4317,7 @@ ext4_mb_initialize_context(struct ext4_a | |
7051 | ac->ac_pa = NULL; | |
7052 | ac->ac_bitmap_page = NULL; | |
7053 | ac->ac_buddy_page = NULL; | |
7054 | + ac->alloc_semp = NULL; | |
7055 | ac->ac_lg = NULL; | |
7056 | ||
7057 | /* we have to define context: we'll we work with a file or | |
7058 | @@ -4243,8 +4395,9 @@ ext4_mb_discard_lg_preallocations(struct | |
7059 | ||
7060 | ext4_get_group_no_and_offset(sb, pa->pa_pstart, &group, NULL); | |
7061 | if (ext4_mb_load_buddy(sb, group, &e4b)) { | |
7062 | + __release(e4b->alloc_semp); | |
7063 | ext4_error(sb, __func__, "Error in loading buddy " | |
7064 | - "information for %lu\n", group); | |
7065 | + "information for %u\n", group); | |
7066 | continue; | |
7067 | } | |
7068 | ext4_lock_group(sb, group); | |
7069 | @@ -4346,6 +4499,8 @@ static int ext4_mb_release_context(struc | |
7070 | } | |
7071 | ext4_mb_put_pa(ac, ac->ac_sb, pa); | |
7072 | } | |
7073 | + if (ac->alloc_semp) | |
7074 | + up_read(ac->alloc_semp); | |
7075 | if (ac->ac_bitmap_page) | |
7076 | page_cache_release(ac->ac_bitmap_page); | |
7077 | if (ac->ac_buddy_page) | |
7078 | @@ -4379,40 +4534,39 @@ static int ext4_mb_discard_preallocation | |
7079 | ext4_fsblk_t ext4_mb_new_blocks(handle_t *handle, | |
7080 | struct ext4_allocation_request *ar, int *errp) | |
7081 | { | |
7082 | + int freed; | |
7083 | struct ext4_allocation_context *ac = NULL; | |
7084 | struct ext4_sb_info *sbi; | |
7085 | struct super_block *sb; | |
7086 | ext4_fsblk_t block = 0; | |
7087 | - int freed; | |
7088 | - int inquota; | |
7089 | + unsigned int inquota; | |
7090 | + unsigned int reserv_blks = 0; | |
7091 | ||
7092 | sb = ar->inode->i_sb; | |
7093 | sbi = EXT4_SB(sb); | |
7094 | ||
7095 | - if (!test_opt(sb, MBALLOC)) { | |
7096 | - block = ext4_old_new_blocks(handle, ar->inode, ar->goal, | |
7097 | - &(ar->len), errp); | |
7098 | - return block; | |
7099 | - } | |
7100 | if (!EXT4_I(ar->inode)->i_delalloc_reserved_flag) { | |
7101 | /* | |
7102 | * With delalloc we already reserved the blocks | |
7103 | */ | |
7104 | - ar->len = ext4_has_free_blocks(sbi, ar->len); | |
7105 | - } | |
7106 | - | |
7107 | - if (ar->len == 0) { | |
7108 | - *errp = -ENOSPC; | |
7109 | - return 0; | |
7110 | + while (ar->len && ext4_claim_free_blocks(sbi, ar->len)) { | |
7111 | + /* let others to free the space */ | |
7112 | + yield(); | |
7113 | + ar->len = ar->len >> 1; | |
7114 | + } | |
7115 | + if (!ar->len) { | |
7116 | + *errp = -ENOSPC; | |
7117 | + return 0; | |
7118 | + } | |
7119 | + reserv_blks = ar->len; | |
7120 | } | |
7121 | - | |
7122 | while (ar->len && DQUOT_ALLOC_BLOCK(ar->inode, ar->len)) { | |
7123 | ar->flags |= EXT4_MB_HINT_NOPREALLOC; | |
7124 | ar->len--; | |
7125 | } | |
7126 | if (ar->len == 0) { | |
7127 | *errp = -EDQUOT; | |
7128 | - return 0; | |
7129 | + goto out3; | |
7130 | } | |
7131 | inquota = ar->len; | |
7132 | ||
7133 | @@ -4449,10 +4603,14 @@ repeat: | |
7134 | ac->ac_o_ex.fe_len < ac->ac_b_ex.fe_len) | |
7135 | ext4_mb_new_preallocation(ac); | |
7136 | } | |
7137 | - | |
7138 | if (likely(ac->ac_status == AC_STATUS_FOUND)) { | |
7139 | - *errp = ext4_mb_mark_diskspace_used(ac, handle); | |
7140 | + *errp = ext4_mb_mark_diskspace_used(ac, handle, reserv_blks); | |
7141 | if (*errp == -EAGAIN) { | |
7142 | + /* | |
7143 | + * drop the reference that we took | |
7144 | + * in ext4_mb_use_best_found | |
7145 | + */ | |
7146 | + ext4_mb_release_context(ac); | |
7147 | ac->ac_b_ex.fe_group = 0; | |
7148 | ac->ac_b_ex.fe_start = 0; | |
7149 | ac->ac_b_ex.fe_len = 0; | |
7150 | @@ -4483,6 +4641,13 @@ out2: | |
7151 | out1: | |
7152 | if (ar->len < inquota) | |
7153 | DQUOT_FREE_BLOCK(ar->inode, inquota - ar->len); | |
7154 | +out3: | |
7155 | + if (!ar->len) { | |
7156 | + if (!EXT4_I(ar->inode)->i_delalloc_reserved_flag) | |
7157 | + /* release all the reserved blocks if non delalloc */ | |
7158 | + percpu_counter_sub(&sbi->s_dirtyblocks_counter, | |
7159 | + reserv_blks); | |
7160 | + } | |
7161 | ||
7162 | return block; | |
7163 | } | |
7164 | @@ -4517,65 +4682,97 @@ static void ext4_mb_poll_new_transaction | |
7165 | ext4_mb_free_committed_blocks(sb); | |
7166 | } | |
7167 | ||
7168 | +/* | |
7169 | + * We can merge two free data extents only if the physical blocks | |
7170 | + * are contiguous, AND the extents were freed by the same transaction, | |
7171 | + * AND the blocks are associated with the same group. | |
7172 | + */ | |
7173 | +static int can_merge(struct ext4_free_data *entry1, | |
7174 | + struct ext4_free_data *entry2) | |
7175 | +{ | |
7176 | + if ((entry1->t_tid == entry2->t_tid) && | |
7177 | + (entry1->group == entry2->group) && | |
7178 | + ((entry1->start_blk + entry1->count) == entry2->start_blk)) | |
7179 | + return 1; | |
7180 | + return 0; | |
7181 | +} | |
7182 | + | |
7183 | static noinline_for_stack int | |
7184 | ext4_mb_free_metadata(handle_t *handle, struct ext4_buddy *e4b, | |
7185 | - ext4_group_t group, ext4_grpblk_t block, int count) | |
7186 | + struct ext4_free_data *new_entry) | |
7187 | { | |
7188 | + ext4_grpblk_t block; | |
7189 | + struct ext4_free_data *entry; | |
7190 | struct ext4_group_info *db = e4b->bd_info; | |
7191 | struct super_block *sb = e4b->bd_sb; | |
7192 | struct ext4_sb_info *sbi = EXT4_SB(sb); | |
7193 | - struct ext4_free_metadata *md; | |
7194 | - int i; | |
7195 | + struct rb_node **n = &db->bb_free_root.rb_node, *node; | |
7196 | + struct rb_node *parent = NULL, *new_node; | |
7197 | ||
7198 | BUG_ON(e4b->bd_bitmap_page == NULL); | |
7199 | BUG_ON(e4b->bd_buddy_page == NULL); | |
7200 | ||
7201 | - ext4_lock_group(sb, group); | |
7202 | - for (i = 0; i < count; i++) { | |
7203 | - md = db->bb_md_cur; | |
7204 | - if (md && db->bb_tid != handle->h_transaction->t_tid) { | |
7205 | - db->bb_md_cur = NULL; | |
7206 | - md = NULL; | |
7207 | + new_node = &new_entry->node; | |
7208 | + block = new_entry->start_blk; | |
7209 | + | |
7210 | + if (!*n) { | |
7211 | + /* first free block exent. We need to | |
7212 | + protect buddy cache from being freed, | |
7213 | + * otherwise we'll refresh it from | |
7214 | + * on-disk bitmap and lose not-yet-available | |
7215 | + * blocks */ | |
7216 | + page_cache_get(e4b->bd_buddy_page); | |
7217 | + page_cache_get(e4b->bd_bitmap_page); | |
7218 | + } | |
7219 | + while (*n) { | |
7220 | + parent = *n; | |
7221 | + entry = rb_entry(parent, struct ext4_free_data, node); | |
7222 | + if (block < entry->start_blk) | |
7223 | + n = &(*n)->rb_left; | |
7224 | + else if (block >= (entry->start_blk + entry->count)) | |
7225 | + n = &(*n)->rb_right; | |
7226 | + else { | |
7227 | + ext4_grp_locked_error(sb, e4b->bd_group, __func__, | |
7228 | + "Double free of blocks %d (%d %d)\n", | |
7229 | + block, entry->start_blk, entry->count); | |
7230 | + return 0; | |
7231 | } | |
7232 | + } | |
7233 | ||
7234 | - if (md == NULL) { | |
7235 | - ext4_unlock_group(sb, group); | |
7236 | - md = kmalloc(sizeof(*md), GFP_NOFS); | |
7237 | - if (md == NULL) | |
7238 | - return -ENOMEM; | |
7239 | - md->num = 0; | |
7240 | - md->group = group; | |
7241 | + rb_link_node(new_node, parent, n); | |
7242 | + rb_insert_color(new_node, &db->bb_free_root); | |
7243 | ||
7244 | - ext4_lock_group(sb, group); | |
7245 | - if (db->bb_md_cur == NULL) { | |
7246 | - spin_lock(&sbi->s_md_lock); | |
7247 | - list_add(&md->list, &sbi->s_active_transaction); | |
7248 | - spin_unlock(&sbi->s_md_lock); | |
7249 | - /* protect buddy cache from being freed, | |
7250 | - * otherwise we'll refresh it from | |
7251 | - * on-disk bitmap and lose not-yet-available | |
7252 | - * blocks */ | |
7253 | - page_cache_get(e4b->bd_buddy_page); | |
7254 | - page_cache_get(e4b->bd_bitmap_page); | |
7255 | - db->bb_md_cur = md; | |
7256 | - db->bb_tid = handle->h_transaction->t_tid; | |
7257 | - mb_debug("new md 0x%p for group %lu\n", | |
7258 | - md, md->group); | |
7259 | - } else { | |
7260 | - kfree(md); | |
7261 | - md = db->bb_md_cur; | |
7262 | - } | |
7263 | + /* Now try to see the extent can be merged to left and right */ | |
7264 | + node = rb_prev(new_node); | |
7265 | + if (node) { | |
7266 | + entry = rb_entry(node, struct ext4_free_data, node); | |
7267 | + if (can_merge(entry, new_entry)) { | |
7268 | + new_entry->start_blk = entry->start_blk; | |
7269 | + new_entry->count += entry->count; | |
7270 | + rb_erase(node, &(db->bb_free_root)); | |
7271 | + spin_lock(&sbi->s_md_lock); | |
7272 | + list_del(&entry->list); | |
7273 | + spin_unlock(&sbi->s_md_lock); | |
7274 | + kmem_cache_free(ext4_free_ext_cachep, entry); | |
7275 | } | |
7276 | + } | |
7277 | ||
7278 | - BUG_ON(md->num >= EXT4_BB_MAX_BLOCKS); | |
7279 | - md->blocks[md->num] = block + i; | |
7280 | - md->num++; | |
7281 | - if (md->num == EXT4_BB_MAX_BLOCKS) { | |
7282 | - /* no more space, put full container on a sb's list */ | |
7283 | - db->bb_md_cur = NULL; | |
7284 | + node = rb_next(new_node); | |
7285 | + if (node) { | |
7286 | + entry = rb_entry(node, struct ext4_free_data, node); | |
7287 | + if (can_merge(new_entry, entry)) { | |
7288 | + new_entry->count += entry->count; | |
7289 | + rb_erase(node, &(db->bb_free_root)); | |
7290 | + spin_lock(&sbi->s_md_lock); | |
7291 | + list_del(&entry->list); | |
7292 | + spin_unlock(&sbi->s_md_lock); | |
7293 | + kmem_cache_free(ext4_free_ext_cachep, entry); | |
7294 | } | |
7295 | } | |
7296 | - ext4_unlock_group(sb, group); | |
7297 | + /* Add the extent to active_transaction list */ | |
7298 | + spin_lock(&sbi->s_md_lock); | |
7299 | + list_add(&new_entry->list, &sbi->s_active_transaction); | |
7300 | + spin_unlock(&sbi->s_md_lock); | |
7301 | return 0; | |
7302 | } | |
7303 | ||
7304 | @@ -4591,7 +4788,7 @@ void ext4_mb_free_blocks(handle_t *handl | |
7305 | struct ext4_allocation_context *ac = NULL; | |
7306 | struct ext4_group_desc *gdp; | |
7307 | struct ext4_super_block *es; | |
7308 | - unsigned long overflow; | |
7309 | + unsigned int overflow; | |
7310 | ext4_grpblk_t bit; | |
7311 | struct buffer_head *gd_bh; | |
7312 | ext4_group_t block_group; | |
7313 | @@ -4675,11 +4872,6 @@ do_more: | |
7314 | err = ext4_journal_get_write_access(handle, gd_bh); | |
7315 | if (err) | |
7316 | goto error_return; | |
7317 | - | |
7318 | - err = ext4_mb_load_buddy(sb, block_group, &e4b); | |
7319 | - if (err) | |
7320 | - goto error_return; | |
7321 | - | |
7322 | #ifdef AGGRESSIVE_CHECK | |
7323 | { | |
7324 | int i; | |
7325 | @@ -4687,13 +4879,6 @@ do_more: | |
7326 | BUG_ON(!mb_test_bit(bit + i, bitmap_bh->b_data)); | |
7327 | } | |
7328 | #endif | |
7329 | - mb_clear_bits(sb_bgl_lock(sbi, block_group), bitmap_bh->b_data, | |
7330 | - bit, count); | |
7331 | - | |
7332 | - /* We dirtied the bitmap block */ | |
7333 | - BUFFER_TRACE(bitmap_bh, "dirtied bitmap block"); | |
7334 | - err = ext4_journal_dirty_metadata(handle, bitmap_bh); | |
7335 | - | |
7336 | if (ac) { | |
7337 | ac->ac_b_ex.fe_group = block_group; | |
7338 | ac->ac_b_ex.fe_start = bit; | |
7339 | @@ -4701,19 +4886,43 @@ do_more: | |
7340 | ext4_mb_store_history(ac); | |
7341 | } | |
7342 | ||
7343 | + err = ext4_mb_load_buddy(sb, block_group, &e4b); | |
7344 | + if (err) { | |
7345 | + __release(e4b->alloc_semp); | |
7346 | + goto error_return; | |
7347 | + } | |
7348 | if (metadata) { | |
7349 | - /* blocks being freed are metadata. these blocks shouldn't | |
7350 | - * be used until this transaction is committed */ | |
7351 | - ext4_mb_free_metadata(handle, &e4b, block_group, bit, count); | |
7352 | + struct ext4_free_data *new_entry; | |
7353 | + /* | |
7354 | + * blocks being freed are metadata. these blocks shouldn't | |
7355 | + * be used until this transaction is committed | |
7356 | + */ | |
7357 | + new_entry = kmem_cache_alloc(ext4_free_ext_cachep, GFP_NOFS); | |
7358 | + new_entry->start_blk = bit; | |
7359 | + new_entry->group = block_group; | |
7360 | + new_entry->count = count; | |
7361 | + new_entry->t_tid = handle->h_transaction->t_tid; | |
7362 | + ext4_lock_group(sb, block_group); | |
7363 | + mb_clear_bits(sb_bgl_lock(sbi, block_group), bitmap_bh->b_data, | |
7364 | + bit, count); | |
7365 | + ext4_mb_free_metadata(handle, &e4b, new_entry); | |
7366 | + ext4_unlock_group(sb, block_group); | |
7367 | } else { | |
7368 | ext4_lock_group(sb, block_group); | |
7369 | + /* need to update group_info->bb_free and bitmap | |
7370 | + * with group lock held. generate_buddy look at | |
7371 | + * them with group lock_held | |
7372 | + */ | |
7373 | + mb_clear_bits(sb_bgl_lock(sbi, block_group), bitmap_bh->b_data, | |
7374 | + bit, count); | |
7375 | mb_free_blocks(inode, &e4b, bit, count); | |
7376 | ext4_mb_return_to_preallocation(inode, &e4b, block, count); | |
7377 | ext4_unlock_group(sb, block_group); | |
7378 | } | |
7379 | ||
7380 | spin_lock(sb_bgl_lock(sbi, block_group)); | |
7381 | - le16_add_cpu(&gdp->bg_free_blocks_count, count); | |
7382 | + ret = ext4_free_blks_count(sb, gdp) + count; | |
7383 | + ext4_free_blks_set(sb, gdp, ret); | |
7384 | gdp->bg_checksum = ext4_group_desc_csum(sbi, block_group, gdp); | |
7385 | spin_unlock(sb_bgl_lock(sbi, block_group)); | |
7386 | percpu_counter_add(&sbi->s_freeblocks_counter, count); | |
7387 | @@ -4729,6 +4938,10 @@ do_more: | |
7388 | ||
7389 | *freed += count; | |
7390 | ||
7391 | + /* We dirtied the bitmap block */ | |
7392 | + BUFFER_TRACE(bitmap_bh, "dirtied bitmap block"); | |
7393 | + err = ext4_journal_dirty_metadata(handle, bitmap_bh); | |
7394 | + | |
7395 | /* And the group descriptor block */ | |
7396 | BUFFER_TRACE(gd_bh, "dirtied group descriptor block"); | |
7397 | ret = ext4_journal_dirty_metadata(handle, gd_bh); | |
7398 | diff -rup b/fs/ext4//mballoc.h a/fs/ext4///mballoc.h | |
7399 | --- b/fs/ext4/mballoc.h 2009-02-11 14:37:58.000000000 +0100 | |
7400 | +++ a/fs/ext4/mballoc.h 2009-02-10 21:40:14.000000000 +0100 | |
7401 | @@ -18,6 +18,9 @@ | |
7402 | #include <linux/pagemap.h> | |
7403 | #include <linux/seq_file.h> | |
7404 | #include <linux/version.h> | |
7405 | +#include <linux/blkdev.h> | |
7406 | +#include <linux/marker.h> | |
7407 | +#include <linux/mutex.h> | |
7408 | #include "ext4_jbd2.h" | |
7409 | #include "ext4.h" | |
7410 | #include "group.h" | |
7411 | @@ -96,41 +99,24 @@ | |
7412 | */ | |
7413 | #define MB_DEFAULT_GROUP_PREALLOC 512 | |
7414 | ||
7415 | -static struct kmem_cache *ext4_pspace_cachep; | |
7416 | -static struct kmem_cache *ext4_ac_cachep; | |
7417 | ||
7418 | -#ifdef EXT4_BB_MAX_BLOCKS | |
7419 | -#undef EXT4_BB_MAX_BLOCKS | |
7420 | -#endif | |
7421 | -#define EXT4_BB_MAX_BLOCKS 30 | |
7422 | +struct ext4_free_data { | |
7423 | + /* this links the free block information from group_info */ | |
7424 | + struct rb_node node; | |
7425 | ||
7426 | -struct ext4_free_metadata { | |
7427 | - ext4_group_t group; | |
7428 | - unsigned short num; | |
7429 | - ext4_grpblk_t blocks[EXT4_BB_MAX_BLOCKS]; | |
7430 | + /* this links the free block information from ext4_sb_info */ | |
7431 | struct list_head list; | |
7432 | -}; | |
7433 | - | |
7434 | -struct ext4_group_info { | |
7435 | - unsigned long bb_state; | |
7436 | - unsigned long bb_tid; | |
7437 | - struct ext4_free_metadata *bb_md_cur; | |
7438 | - unsigned short bb_first_free; | |
7439 | - unsigned short bb_free; | |
7440 | - unsigned short bb_fragments; | |
7441 | - struct list_head bb_prealloc_list; | |
7442 | -#ifdef DOUBLE_CHECK | |
7443 | - void *bb_bitmap; | |
7444 | -#endif | |
7445 | - unsigned short bb_counters[]; | |
7446 | -}; | |
7447 | ||
7448 | -#define EXT4_GROUP_INFO_NEED_INIT_BIT 0 | |
7449 | -#define EXT4_GROUP_INFO_LOCKED_BIT 1 | |
7450 | + /* group which free block extent belongs */ | |
7451 | + ext4_group_t group; | |
7452 | ||
7453 | -#define EXT4_MB_GRP_NEED_INIT(grp) \ | |
7454 | - (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state))) | |
7455 | + /* free block extent */ | |
7456 | + ext4_grpblk_t start_blk; | |
7457 | + ext4_grpblk_t count; | |
7458 | ||
7459 | + /* transaction which freed this extent */ | |
7460 | + tid_t t_tid; | |
7461 | +}; | |
7462 | ||
7463 | struct ext4_prealloc_space { | |
7464 | struct list_head pa_inode_list; | |
7465 | @@ -209,6 +195,11 @@ struct ext4_allocation_context { | |
7466 | __u8 ac_op; /* operation, for history only */ | |
7467 | struct page *ac_bitmap_page; | |
7468 | struct page *ac_buddy_page; | |
7469 | + /* | |
7470 | + * pointer to the held semaphore upon successful | |
7471 | + * block allocation | |
7472 | + */ | |
7473 | + struct rw_semaphore *alloc_semp; | |
7474 | struct ext4_prealloc_space *ac_pa; | |
7475 | struct ext4_locality_group *ac_lg; | |
7476 | }; | |
7477 | @@ -242,6 +233,7 @@ struct ext4_buddy { | |
7478 | struct super_block *bd_sb; | |
7479 | __u16 bd_blkbits; | |
7480 | ext4_group_t bd_group; | |
7481 | + struct rw_semaphore *alloc_semp; | |
7482 | }; | |
7483 | #define EXT4_MB_BITMAP(e4b) ((e4b)->bd_bitmap) | |
7484 | #define EXT4_MB_BUDDY(e4b) ((e4b)->bd_buddy) | |
7485 | @@ -251,53 +243,12 @@ static inline void ext4_mb_store_history | |
7486 | { | |
7487 | return; | |
7488 | } | |
7489 | -#else | |
7490 | -static void ext4_mb_store_history(struct ext4_allocation_context *ac); | |
7491 | #endif | |
7492 | ||
7493 | #define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1) | |
7494 | ||
7495 | -static struct proc_dir_entry *proc_root_ext4; | |
7496 | struct buffer_head *read_block_bitmap(struct super_block *, ext4_group_t); | |
7497 | - | |
7498 | -static void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap, | |
7499 | - ext4_group_t group); | |
7500 | -static void ext4_mb_poll_new_transaction(struct super_block *, handle_t *); | |
7501 | -static void ext4_mb_free_committed_blocks(struct super_block *); | |
7502 | -static void ext4_mb_return_to_preallocation(struct inode *inode, | |
7503 | - struct ext4_buddy *e4b, sector_t block, | |
7504 | - int count); | |
7505 | -static void ext4_mb_put_pa(struct ext4_allocation_context *, | |
7506 | - struct super_block *, struct ext4_prealloc_space *pa); | |
7507 | -static int ext4_mb_init_per_dev_proc(struct super_block *sb); | |
7508 | -static int ext4_mb_destroy_per_dev_proc(struct super_block *sb); | |
7509 | - | |
7510 | - | |
7511 | -static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group) | |
7512 | -{ | |
7513 | - struct ext4_group_info *grinfo = ext4_get_group_info(sb, group); | |
7514 | - | |
7515 | - bit_spin_lock(EXT4_GROUP_INFO_LOCKED_BIT, &(grinfo->bb_state)); | |
7516 | -} | |
7517 | - | |
7518 | -static inline void ext4_unlock_group(struct super_block *sb, | |
7519 | - ext4_group_t group) | |
7520 | -{ | |
7521 | - struct ext4_group_info *grinfo = ext4_get_group_info(sb, group); | |
7522 | - | |
7523 | - bit_spin_unlock(EXT4_GROUP_INFO_LOCKED_BIT, &(grinfo->bb_state)); | |
7524 | -} | |
7525 | - | |
7526 | -static inline int ext4_is_group_locked(struct super_block *sb, | |
7527 | - ext4_group_t group) | |
7528 | -{ | |
7529 | - struct ext4_group_info *grinfo = ext4_get_group_info(sb, group); | |
7530 | - | |
7531 | - return bit_spin_is_locked(EXT4_GROUP_INFO_LOCKED_BIT, | |
7532 | - &(grinfo->bb_state)); | |
7533 | -} | |
7534 | - | |
7535 | -static ext4_fsblk_t ext4_grp_offs_to_block(struct super_block *sb, | |
7536 | +static inline ext4_fsblk_t ext4_grp_offs_to_block(struct super_block *sb, | |
7537 | struct ext4_free_extent *fex) | |
7538 | { | |
7539 | ext4_fsblk_t block; | |
7540 | diff -rup b/fs/ext4//namei.c a/fs/ext4///namei.c | |
7541 | --- b/fs/ext4/namei.c 2009-02-11 14:37:58.000000000 +0100 | |
7542 | +++ a/fs/ext4/namei.c 2009-02-10 21:40:11.000000000 +0100 | |
7543 | @@ -151,34 +151,36 @@ struct dx_map_entry | |
7544 | ||
7545 | static inline ext4_lblk_t dx_get_block(struct dx_entry *entry); | |
7546 | static void dx_set_block(struct dx_entry *entry, ext4_lblk_t value); | |
7547 | -static inline unsigned dx_get_hash (struct dx_entry *entry); | |
7548 | -static void dx_set_hash (struct dx_entry *entry, unsigned value); | |
7549 | -static unsigned dx_get_count (struct dx_entry *entries); | |
7550 | -static unsigned dx_get_limit (struct dx_entry *entries); | |
7551 | -static void dx_set_count (struct dx_entry *entries, unsigned value); | |
7552 | -static void dx_set_limit (struct dx_entry *entries, unsigned value); | |
7553 | -static unsigned dx_root_limit (struct inode *dir, unsigned infosize); | |
7554 | -static unsigned dx_node_limit (struct inode *dir); | |
7555 | -static struct dx_frame *dx_probe(struct dentry *dentry, | |
7556 | +static inline unsigned dx_get_hash(struct dx_entry *entry); | |
7557 | +static void dx_set_hash(struct dx_entry *entry, unsigned value); | |
7558 | +static unsigned dx_get_count(struct dx_entry *entries); | |
7559 | +static unsigned dx_get_limit(struct dx_entry *entries); | |
7560 | +static void dx_set_count(struct dx_entry *entries, unsigned value); | |
7561 | +static void dx_set_limit(struct dx_entry *entries, unsigned value); | |
7562 | +static unsigned dx_root_limit(struct inode *dir, unsigned infosize); | |
7563 | +static unsigned dx_node_limit(struct inode *dir); | |
7564 | +static struct dx_frame *dx_probe(const struct qstr *d_name, | |
7565 | struct inode *dir, | |
7566 | struct dx_hash_info *hinfo, | |
7567 | struct dx_frame *frame, | |
7568 | int *err); | |
7569 | -static void dx_release (struct dx_frame *frames); | |
7570 | -static int dx_make_map (struct ext4_dir_entry_2 *de, int size, | |
7571 | - struct dx_hash_info *hinfo, struct dx_map_entry map[]); | |
7572 | +static void dx_release(struct dx_frame *frames); | |
7573 | +static int dx_make_map(struct ext4_dir_entry_2 *de, int size, | |
7574 | + struct dx_hash_info *hinfo, struct dx_map_entry map[]); | |
7575 | static void dx_sort_map(struct dx_map_entry *map, unsigned count); | |
7576 | -static struct ext4_dir_entry_2 *dx_move_dirents (char *from, char *to, | |
7577 | +static struct ext4_dir_entry_2 *dx_move_dirents(char *from, char *to, | |
7578 | struct dx_map_entry *offsets, int count); | |
7579 | -static struct ext4_dir_entry_2* dx_pack_dirents (char *base, int size); | |
7580 | +static struct ext4_dir_entry_2* dx_pack_dirents(char *base, int size); | |
7581 | static void dx_insert_block(struct dx_frame *frame, | |
7582 | u32 hash, ext4_lblk_t block); | |
7583 | static int ext4_htree_next_block(struct inode *dir, __u32 hash, | |
7584 | struct dx_frame *frame, | |
7585 | struct dx_frame *frames, | |
7586 | __u32 *start_hash); | |
7587 | -static struct buffer_head * ext4_dx_find_entry(struct dentry *dentry, | |
7588 | - struct ext4_dir_entry_2 **res_dir, int *err); | |
7589 | +static struct buffer_head * ext4_dx_find_entry(struct inode *dir, | |
7590 | + const struct qstr *d_name, | |
7591 | + struct ext4_dir_entry_2 **res_dir, | |
7592 | + int *err); | |
7593 | static int ext4_dx_add_entry(handle_t *handle, struct dentry *dentry, | |
7594 | struct inode *inode); | |
7595 | ||
7596 | @@ -207,44 +209,44 @@ static inline void dx_set_block(struct d | |
7597 | entry->block = cpu_to_le32(value); | |
7598 | } | |
7599 | ||
7600 | -static inline unsigned dx_get_hash (struct dx_entry *entry) | |
7601 | +static inline unsigned dx_get_hash(struct dx_entry *entry) | |
7602 | { | |
7603 | return le32_to_cpu(entry->hash); | |
7604 | } | |
7605 | ||
7606 | -static inline void dx_set_hash (struct dx_entry *entry, unsigned value) | |
7607 | +static inline void dx_set_hash(struct dx_entry *entry, unsigned value) | |
7608 | { | |
7609 | entry->hash = cpu_to_le32(value); | |
7610 | } | |
7611 | ||
7612 | -static inline unsigned dx_get_count (struct dx_entry *entries) | |
7613 | +static inline unsigned dx_get_count(struct dx_entry *entries) | |
7614 | { | |
7615 | return le16_to_cpu(((struct dx_countlimit *) entries)->count); | |
7616 | } | |
7617 | ||
7618 | -static inline unsigned dx_get_limit (struct dx_entry *entries) | |
7619 | +static inline unsigned dx_get_limit(struct dx_entry *entries) | |
7620 | { | |
7621 | return le16_to_cpu(((struct dx_countlimit *) entries)->limit); | |
7622 | } | |
7623 | ||
7624 | -static inline void dx_set_count (struct dx_entry *entries, unsigned value) | |
7625 | +static inline void dx_set_count(struct dx_entry *entries, unsigned value) | |
7626 | { | |
7627 | ((struct dx_countlimit *) entries)->count = cpu_to_le16(value); | |
7628 | } | |
7629 | ||
7630 | -static inline void dx_set_limit (struct dx_entry *entries, unsigned value) | |
7631 | +static inline void dx_set_limit(struct dx_entry *entries, unsigned value) | |
7632 | { | |
7633 | ((struct dx_countlimit *) entries)->limit = cpu_to_le16(value); | |
7634 | } | |
7635 | ||
7636 | -static inline unsigned dx_root_limit (struct inode *dir, unsigned infosize) | |
7637 | +static inline unsigned dx_root_limit(struct inode *dir, unsigned infosize) | |
7638 | { | |
7639 | unsigned entry_space = dir->i_sb->s_blocksize - EXT4_DIR_REC_LEN(1) - | |
7640 | EXT4_DIR_REC_LEN(2) - infosize; | |
7641 | return entry_space / sizeof(struct dx_entry); | |
7642 | } | |
7643 | ||
7644 | -static inline unsigned dx_node_limit (struct inode *dir) | |
7645 | +static inline unsigned dx_node_limit(struct inode *dir) | |
7646 | { | |
7647 | unsigned entry_space = dir->i_sb->s_blocksize - EXT4_DIR_REC_LEN(0); | |
7648 | return entry_space / sizeof(struct dx_entry); | |
7649 | @@ -254,12 +256,12 @@ static inline unsigned dx_node_limit (st | |
7650 | * Debug | |
7651 | */ | |
7652 | #ifdef DX_DEBUG | |
7653 | -static void dx_show_index (char * label, struct dx_entry *entries) | |
7654 | +static void dx_show_index(char * label, struct dx_entry *entries) | |
7655 | { | |
7656 | int i, n = dx_get_count (entries); | |
7657 | - printk("%s index ", label); | |
7658 | + printk(KERN_DEBUG "%s index ", label); | |
7659 | for (i = 0; i < n; i++) { | |
7660 | - printk("%x->%lu ", i? dx_get_hash(entries + i) : | |
7661 | + printk("%x->%lu ", i ? dx_get_hash(entries + i) : | |
7662 | 0, (unsigned long)dx_get_block(entries + i)); | |
7663 | } | |
7664 | printk("\n"); | |
7665 | @@ -306,7 +308,7 @@ struct stats dx_show_entries(struct dx_h | |
7666 | struct dx_entry *entries, int levels) | |
7667 | { | |
7668 | unsigned blocksize = dir->i_sb->s_blocksize; | |
7669 | - unsigned count = dx_get_count (entries), names = 0, space = 0, i; | |
7670 | + unsigned count = dx_get_count(entries), names = 0, space = 0, i; | |
7671 | unsigned bcount = 0; | |
7672 | struct buffer_head *bh; | |
7673 | int err; | |
7674 | @@ -325,11 +327,12 @@ struct stats dx_show_entries(struct dx_h | |
7675 | names += stats.names; | |
7676 | space += stats.space; | |
7677 | bcount += stats.bcount; | |
7678 | - brelse (bh); | |
7679 | + brelse(bh); | |
7680 | } | |
7681 | if (bcount) | |
7682 | - printk("%snames %u, fullness %u (%u%%)\n", levels?"":" ", | |
7683 | - names, space/bcount,(space/bcount)*100/blocksize); | |
7684 | + printk(KERN_DEBUG "%snames %u, fullness %u (%u%%)\n", | |
7685 | + levels ? "" : " ", names, space/bcount, | |
7686 | + (space/bcount)*100/blocksize); | |
7687 | return (struct stats) { names, space, bcount}; | |
7688 | } | |
7689 | #endif /* DX_DEBUG */ | |
7690 | @@ -344,7 +347,7 @@ struct stats dx_show_entries(struct dx_h | |
7691 | * back to userspace. | |
7692 | */ | |
7693 | static struct dx_frame * | |
7694 | -dx_probe(struct dentry *dentry, struct inode *dir, | |
7695 | +dx_probe(const struct qstr *d_name, struct inode *dir, | |
7696 | struct dx_hash_info *hinfo, struct dx_frame *frame_in, int *err) | |
7697 | { | |
7698 | unsigned count, indirect; | |
7699 | @@ -355,8 +358,6 @@ dx_probe(struct dentry *dentry, struct i | |
7700 | u32 hash; | |
7701 | ||
7702 | frame->bh = NULL; | |
7703 | - if (dentry) | |
7704 | - dir = dentry->d_parent->d_inode; | |
7705 | if (!(bh = ext4_bread (NULL,dir, 0, 0, err))) | |
7706 | goto fail; | |
7707 | root = (struct dx_root *) bh->b_data; | |
7708 | @@ -371,9 +372,11 @@ dx_probe(struct dentry *dentry, struct i | |
7709 | goto fail; | |
7710 | } | |
7711 | hinfo->hash_version = root->info.hash_version; | |
7712 | + if (hinfo->hash_version <= DX_HASH_TEA) | |
7713 | + hinfo->hash_version += EXT4_SB(dir->i_sb)->s_hash_unsigned; | |
7714 | hinfo->seed = EXT4_SB(dir->i_sb)->s_hash_seed; | |
7715 | - if (dentry) | |
7716 | - ext4fs_dirhash(dentry->d_name.name, dentry->d_name.len, hinfo); | |
7717 | + if (d_name) | |
7718 | + ext4fs_dirhash(d_name->name, d_name->len, hinfo); | |
7719 | hash = hinfo->hash; | |
7720 | ||
7721 | if (root->info.unused_flags & 1) { | |
7722 | @@ -406,7 +409,7 @@ dx_probe(struct dentry *dentry, struct i | |
7723 | goto fail; | |
7724 | } | |
7725 | ||
7726 | - dxtrace (printk("Look up %x", hash)); | |
7727 | + dxtrace(printk("Look up %x", hash)); | |
7728 | while (1) | |
7729 | { | |
7730 | count = dx_get_count(entries); | |
7731 | @@ -555,7 +558,7 @@ static int ext4_htree_next_block(struct | |
7732 | 0, &err))) | |
7733 | return err; /* Failure */ | |
7734 | p++; | |
7735 | - brelse (p->bh); | |
7736 | + brelse(p->bh); | |
7737 | p->bh = bh; | |
7738 | p->at = p->entries = ((struct dx_node *) bh->b_data)->entries; | |
7739 | } | |
7740 | @@ -593,7 +596,7 @@ static int htree_dirblock_to_tree(struct | |
7741 | /* On error, skip the f_pos to the next block. */ | |
7742 | dir_file->f_pos = (dir_file->f_pos | | |
7743 | (dir->i_sb->s_blocksize - 1)) + 1; | |
7744 | - brelse (bh); | |
7745 | + brelse(bh); | |
7746 | return count; | |
7747 | } | |
7748 | ext4fs_dirhash(de->name, de->name_len, hinfo); | |
7749 | @@ -635,11 +638,14 @@ int ext4_htree_fill_tree(struct file *di | |
7750 | int ret, err; | |
7751 | __u32 hashval; | |
7752 | ||
7753 | - dxtrace(printk("In htree_fill_tree, start hash: %x:%x\n", start_hash, | |
7754 | - start_minor_hash)); | |
7755 | + dxtrace(printk(KERN_DEBUG "In htree_fill_tree, start hash: %x:%x\n", | |
7756 | + start_hash, start_minor_hash)); | |
7757 | dir = dir_file->f_path.dentry->d_inode; | |
7758 | if (!(EXT4_I(dir)->i_flags & EXT4_INDEX_FL)) { | |
7759 | hinfo.hash_version = EXT4_SB(dir->i_sb)->s_def_hash_version; | |
7760 | + if (hinfo.hash_version <= DX_HASH_TEA) | |
7761 | + hinfo.hash_version += | |
7762 | + EXT4_SB(dir->i_sb)->s_hash_unsigned; | |
7763 | hinfo.seed = EXT4_SB(dir->i_sb)->s_hash_seed; | |
7764 | count = htree_dirblock_to_tree(dir_file, dir, 0, &hinfo, | |
7765 | start_hash, start_minor_hash); | |
7766 | @@ -648,7 +654,7 @@ int ext4_htree_fill_tree(struct file *di | |
7767 | } | |
7768 | hinfo.hash = start_hash; | |
7769 | hinfo.minor_hash = 0; | |
7770 | - frame = dx_probe(NULL, dir_file->f_path.dentry->d_inode, &hinfo, frames, &err); | |
7771 | + frame = dx_probe(NULL, dir, &hinfo, frames, &err); | |
7772 | if (!frame) | |
7773 | return err; | |
7774 | ||
7775 | @@ -694,8 +700,8 @@ int ext4_htree_fill_tree(struct file *di | |
7776 | break; | |
7777 | } | |
7778 | dx_release(frames); | |
7779 | - dxtrace(printk("Fill tree: returned %d entries, next hash: %x\n", | |
7780 | - count, *next_hash)); | |
7781 | + dxtrace(printk(KERN_DEBUG "Fill tree: returned %d entries, " | |
7782 | + "next hash: %x\n", count, *next_hash)); | |
7783 | return count; | |
7784 | errout: | |
7785 | dx_release(frames); | |
7786 | @@ -802,17 +808,17 @@ static inline int ext4_match (int len, c | |
7787 | /* | |
7788 | * Returns 0 if not found, -1 on failure, and 1 on success | |
7789 | */ | |
7790 | -static inline int search_dirblock(struct buffer_head * bh, | |
7791 | +static inline int search_dirblock(struct buffer_head *bh, | |
7792 | struct inode *dir, | |
7793 | - struct dentry *dentry, | |
7794 | - unsigned long offset, | |
7795 | + const struct qstr *d_name, | |
7796 | + unsigned int offset, | |
7797 | struct ext4_dir_entry_2 ** res_dir) | |
7798 | { | |
7799 | struct ext4_dir_entry_2 * de; | |
7800 | char * dlimit; | |
7801 | int de_len; | |
7802 | - const char *name = dentry->d_name.name; | |
7803 | - int namelen = dentry->d_name.len; | |
7804 | + const char *name = d_name->name; | |
7805 | + int namelen = d_name->len; | |
7806 | ||
7807 | de = (struct ext4_dir_entry_2 *) bh->b_data; | |
7808 | dlimit = bh->b_data + dir->i_sb->s_blocksize; | |
7809 | @@ -851,12 +857,13 @@ static inline int search_dirblock(struct | |
7810 | * The returned buffer_head has ->b_count elevated. The caller is expected | |
7811 | * to brelse() it when appropriate. | |
7812 | */ | |
7813 | -static struct buffer_head * ext4_find_entry (struct dentry *dentry, | |
7814 | +static struct buffer_head * ext4_find_entry (struct inode *dir, | |
7815 | + const struct qstr *d_name, | |
7816 | struct ext4_dir_entry_2 ** res_dir) | |
7817 | { | |
7818 | - struct super_block * sb; | |
7819 | - struct buffer_head * bh_use[NAMEI_RA_SIZE]; | |
7820 | - struct buffer_head * bh, *ret = NULL; | |
7821 | + struct super_block *sb; | |
7822 | + struct buffer_head *bh_use[NAMEI_RA_SIZE]; | |
7823 | + struct buffer_head *bh, *ret = NULL; | |
7824 | ext4_lblk_t start, block, b; | |
7825 | int ra_max = 0; /* Number of bh's in the readahead | |
7826 | buffer, bh_use[] */ | |
7827 | @@ -865,16 +872,15 @@ static struct buffer_head * ext4_find_en | |
7828 | int num = 0; | |
7829 | ext4_lblk_t nblocks; | |
7830 | int i, err; | |
7831 | - struct inode *dir = dentry->d_parent->d_inode; | |
7832 | int namelen; | |
7833 | ||
7834 | *res_dir = NULL; | |
7835 | sb = dir->i_sb; | |
7836 | - namelen = dentry->d_name.len; | |
7837 | + namelen = d_name->len; | |
7838 | if (namelen > EXT4_NAME_LEN) | |
7839 | return NULL; | |
7840 | if (is_dx(dir)) { | |
7841 | - bh = ext4_dx_find_entry(dentry, res_dir, &err); | |
7842 | + bh = ext4_dx_find_entry(dir, d_name, res_dir, &err); | |
7843 | /* | |
7844 | * On success, or if the error was file not found, | |
7845 | * return. Otherwise, fall back to doing a search the | |
7846 | @@ -882,7 +888,8 @@ static struct buffer_head * ext4_find_en | |
7847 | */ | |
7848 | if (bh || (err != ERR_BAD_DX_DIR)) | |
7849 | return bh; | |
7850 | - dxtrace(printk("ext4_find_entry: dx failed, falling back\n")); | |
7851 | + dxtrace(printk(KERN_DEBUG "ext4_find_entry: dx failed, " | |
7852 | + "falling back\n")); | |
7853 | } | |
7854 | nblocks = dir->i_size >> EXT4_BLOCK_SIZE_BITS(sb); | |
7855 | start = EXT4_I(dir)->i_dir_start_lookup; | |
7856 | @@ -926,7 +933,7 @@ restart: | |
7857 | brelse(bh); | |
7858 | goto next; | |
7859 | } | |
7860 | - i = search_dirblock(bh, dir, dentry, | |
7861 | + i = search_dirblock(bh, dir, d_name, | |
7862 | block << EXT4_BLOCK_SIZE_BITS(sb), res_dir); | |
7863 | if (i == 1) { | |
7864 | EXT4_I(dir)->i_dir_start_lookup = block; | |
7865 | @@ -956,11 +963,11 @@ restart: | |
7866 | cleanup_and_exit: | |
7867 | /* Clean up the read-ahead blocks */ | |
7868 | for (; ra_ptr < ra_max; ra_ptr++) | |
7869 | - brelse (bh_use[ra_ptr]); | |
7870 | + brelse(bh_use[ra_ptr]); | |
7871 | return ret; | |
7872 | } | |
7873 | ||
7874 | -static struct buffer_head * ext4_dx_find_entry(struct dentry *dentry, | |
7875 | +static struct buffer_head * ext4_dx_find_entry(struct inode *dir, const struct qstr *d_name, | |
7876 | struct ext4_dir_entry_2 **res_dir, int *err) | |
7877 | { | |
7878 | struct super_block * sb; | |
7879 | @@ -971,14 +978,13 @@ static struct buffer_head * ext4_dx_find | |
7880 | struct buffer_head *bh; | |
7881 | ext4_lblk_t block; | |
7882 | int retval; | |
7883 | - int namelen = dentry->d_name.len; | |
7884 | - const u8 *name = dentry->d_name.name; | |
7885 | - struct inode *dir = dentry->d_parent->d_inode; | |
7886 | + int namelen = d_name->len; | |
7887 | + const u8 *name = d_name->name; | |
7888 | ||
7889 | sb = dir->i_sb; | |
7890 | /* NFS may look up ".." - look at dx_root directory block */ | |
7891 | if (namelen > 2 || name[0] != '.'||(name[1] != '.' && name[1] != '\0')){ | |
7892 | - if (!(frame = dx_probe(dentry, NULL, &hinfo, frames, err))) | |
7893 | + if (!(frame = dx_probe(d_name, dir, &hinfo, frames, err))) | |
7894 | return NULL; | |
7895 | } else { | |
7896 | frame = frames; | |
7897 | @@ -1010,7 +1016,7 @@ static struct buffer_head * ext4_dx_find | |
7898 | return bh; | |
7899 | } | |
7900 | } | |
7901 | - brelse (bh); | |
7902 | + brelse(bh); | |
7903 | /* Check to see if we should continue to search */ | |
7904 | retval = ext4_htree_next_block(dir, hash, frame, | |
7905 | frames, NULL); | |
7906 | @@ -1025,28 +1031,28 @@ static struct buffer_head * ext4_dx_find | |
7907 | ||
7908 | *err = -ENOENT; | |
7909 | errout: | |
7910 | - dxtrace(printk("%s not found\n", name)); | |
7911 | + dxtrace(printk(KERN_DEBUG "%s not found\n", name)); | |
7912 | dx_release (frames); | |
7913 | return NULL; | |
7914 | } | |
7915 | ||
7916 | -static struct dentry *ext4_lookup(struct inode * dir, struct dentry *dentry, struct nameidata *nd) | |
7917 | +static struct dentry *ext4_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd) | |
7918 | { | |
7919 | - struct inode * inode; | |
7920 | - struct ext4_dir_entry_2 * de; | |
7921 | - struct buffer_head * bh; | |
7922 | + struct inode *inode; | |
7923 | + struct ext4_dir_entry_2 *de; | |
7924 | + struct buffer_head *bh; | |
7925 | ||
7926 | if (dentry->d_name.len > EXT4_NAME_LEN) | |
7927 | return ERR_PTR(-ENAMETOOLONG); | |
7928 | ||
7929 | - bh = ext4_find_entry(dentry, &de); | |
7930 | + bh = ext4_find_entry(dir, &dentry->d_name, &de); | |
7931 | inode = NULL; | |
7932 | if (bh) { | |
7933 | - unsigned long ino = le32_to_cpu(de->inode); | |
7934 | - brelse (bh); | |
7935 | + __u32 ino = le32_to_cpu(de->inode); | |
7936 | + brelse(bh); | |
7937 | if (!ext4_valid_inum(dir->i_sb, ino)) { | |
7938 | ext4_error(dir->i_sb, "ext4_lookup", | |
7939 | - "bad inode number: %lu", ino); | |
7940 | + "bad inode number: %u", ino); | |
7941 | return ERR_PTR(-EIO); | |
7942 | } | |
7943 | inode = ext4_iget(dir->i_sb, ino); | |
7944 | @@ -1059,18 +1065,17 @@ static struct dentry *ext4_lookup(struct | |
7945 | ||
7946 | struct dentry *ext4_get_parent(struct dentry *child) | |
7947 | { | |
7948 | - unsigned long ino; | |
7949 | + __u32 ino; | |
7950 | struct dentry *parent; | |
7951 | struct inode *inode; | |
7952 | - struct dentry dotdot; | |
7953 | + static const struct qstr dotdot = { | |
7954 | + .name = "..", | |
7955 | + .len = 2, | |
7956 | + }; | |
7957 | struct ext4_dir_entry_2 * de; | |
7958 | struct buffer_head *bh; | |
7959 | ||
7960 | - dotdot.d_name.name = ".."; | |
7961 | - dotdot.d_name.len = 2; | |
7962 | - dotdot.d_parent = child; /* confusing, isn't it! */ | |
7963 | - | |
7964 | - bh = ext4_find_entry(&dotdot, &de); | |
7965 | + bh = ext4_find_entry(child->d_inode, &dotdot, &de); | |
7966 | inode = NULL; | |
7967 | if (!bh) | |
7968 | return ERR_PTR(-ENOENT); | |
7969 | @@ -1079,7 +1084,7 @@ struct dentry *ext4_get_parent(struct de | |
7970 | ||
7971 | if (!ext4_valid_inum(child->d_inode->i_sb, ino)) { | |
7972 | ext4_error(child->d_inode->i_sb, "ext4_get_parent", | |
7973 | - "bad inode number: %lu", ino); | |
7974 | + "bad inode number: %u", ino); | |
7975 | return ERR_PTR(-EIO); | |
7976 | } | |
7977 | ||
7978 | @@ -1176,9 +1181,9 @@ static struct ext4_dir_entry_2 *do_split | |
7979 | u32 hash2; | |
7980 | struct dx_map_entry *map; | |
7981 | char *data1 = (*bh)->b_data, *data2; | |
7982 | - unsigned split, move, size, i; | |
7983 | + unsigned split, move, size; | |
7984 | struct ext4_dir_entry_2 *de = NULL, *de2; | |
7985 | - int err = 0; | |
7986 | + int err = 0, i; | |
7987 | ||
7988 | bh2 = ext4_append (handle, dir, &newblock, &err); | |
7989 | if (!(bh2)) { | |
7990 | @@ -1201,10 +1206,10 @@ static struct ext4_dir_entry_2 *do_split | |
7991 | ||
7992 | /* create map in the end of data2 block */ | |
7993 | map = (struct dx_map_entry *) (data2 + blocksize); | |
7994 | - count = dx_make_map ((struct ext4_dir_entry_2 *) data1, | |
7995 | + count = dx_make_map((struct ext4_dir_entry_2 *) data1, | |
7996 | blocksize, hinfo, map); | |
7997 | map -= count; | |
7998 | - dx_sort_map (map, count); | |
7999 | + dx_sort_map(map, count); | |
8000 | /* Split the existing block in the middle, size-wise */ | |
8001 | size = 0; | |
8002 | move = 0; | |
8003 | @@ -1225,7 +1230,7 @@ static struct ext4_dir_entry_2 *do_split | |
8004 | ||
8005 | /* Fancy dance to stay within two buffers */ | |
8006 | de2 = dx_move_dirents(data1, data2, map + split, count - split); | |
8007 | - de = dx_pack_dirents(data1,blocksize); | |
8008 | + de = dx_pack_dirents(data1, blocksize); | |
8009 | de->rec_len = ext4_rec_len_to_disk(data1 + blocksize - (char *) de); | |
8010 | de2->rec_len = ext4_rec_len_to_disk(data2 + blocksize - (char *) de2); | |
8011 | dxtrace(dx_show_leaf (hinfo, (struct ext4_dir_entry_2 *) data1, blocksize, 1)); | |
8012 | @@ -1237,15 +1242,15 @@ static struct ext4_dir_entry_2 *do_split | |
8013 | swap(*bh, bh2); | |
8014 | de = de2; | |
8015 | } | |
8016 | - dx_insert_block (frame, hash2 + continued, newblock); | |
8017 | - err = ext4_journal_dirty_metadata (handle, bh2); | |
8018 | + dx_insert_block(frame, hash2 + continued, newblock); | |
8019 | + err = ext4_journal_dirty_metadata(handle, bh2); | |
8020 | if (err) | |
8021 | goto journal_error; | |
8022 | - err = ext4_journal_dirty_metadata (handle, frame->bh); | |
8023 | + err = ext4_journal_dirty_metadata(handle, frame->bh); | |
8024 | if (err) | |
8025 | goto journal_error; | |
8026 | - brelse (bh2); | |
8027 | - dxtrace(dx_show_index ("frame", frame->entries)); | |
8028 | + brelse(bh2); | |
8029 | + dxtrace(dx_show_index("frame", frame->entries)); | |
8030 | return de; | |
8031 | ||
8032 | journal_error: | |
8033 | @@ -1271,12 +1276,12 @@ errout: | |
8034 | */ | |
8035 | static int add_dirent_to_buf(handle_t *handle, struct dentry *dentry, | |
8036 | struct inode *inode, struct ext4_dir_entry_2 *de, | |
8037 | - struct buffer_head * bh) | |
8038 | + struct buffer_head *bh) | |
8039 | { | |
8040 | struct inode *dir = dentry->d_parent->d_inode; | |
8041 | const char *name = dentry->d_name.name; | |
8042 | int namelen = dentry->d_name.len; | |
8043 | - unsigned long offset = 0; | |
8044 | + unsigned int offset = 0; | |
8045 | unsigned short reclen; | |
8046 | int nlen, rlen, err; | |
8047 | char *top; | |
8048 | @@ -1288,11 +1293,11 @@ static int add_dirent_to_buf(handle_t *h | |
8049 | while ((char *) de <= top) { | |
8050 | if (!ext4_check_dir_entry("ext4_add_entry", dir, de, | |
8051 | bh, offset)) { | |
8052 | - brelse (bh); | |
8053 | + brelse(bh); | |
8054 | return -EIO; | |
8055 | } | |
8056 | - if (ext4_match (namelen, name, de)) { | |
8057 | - brelse (bh); | |
8058 | + if (ext4_match(namelen, name, de)) { | |
8059 | + brelse(bh); | |
8060 | return -EEXIST; | |
8061 | } | |
8062 | nlen = EXT4_DIR_REC_LEN(de->name_len); | |
8063 | @@ -1329,7 +1334,7 @@ static int add_dirent_to_buf(handle_t *h | |
8064 | } else | |
8065 | de->inode = 0; | |
8066 | de->name_len = namelen; | |
8067 | - memcpy (de->name, name, namelen); | |
8068 | + memcpy(de->name, name, namelen); | |
8069 | /* | |
8070 | * XXX shouldn't update any times until successful | |
8071 | * completion of syscall, but too many callers depend | |
8072 | @@ -1377,7 +1382,7 @@ static int make_indexed_dir(handle_t *ha | |
8073 | struct fake_dirent *fde; | |
8074 | ||
8075 | blocksize = dir->i_sb->s_blocksize; | |
8076 | - dxtrace(printk("Creating index\n")); | |
8077 | + dxtrace(printk(KERN_DEBUG "Creating index\n")); | |
8078 | retval = ext4_journal_get_write_access(handle, bh); | |
8079 | if (retval) { | |
8080 | ext4_std_error(dir->i_sb, retval); | |
8081 | @@ -1386,7 +1391,7 @@ static int make_indexed_dir(handle_t *ha | |
8082 | } | |
8083 | root = (struct dx_root *) bh->b_data; | |
8084 | ||
8085 | - bh2 = ext4_append (handle, dir, &block, &retval); | |
8086 | + bh2 = ext4_append(handle, dir, &block, &retval); | |
8087 | if (!(bh2)) { | |
8088 | brelse(bh); | |
8089 | return retval; | |
8090 | @@ -1412,12 +1417,14 @@ static int make_indexed_dir(handle_t *ha | |
8091 | root->info.info_length = sizeof(root->info); | |
8092 | root->info.hash_version = EXT4_SB(dir->i_sb)->s_def_hash_version; | |
8093 | entries = root->entries; | |
8094 | - dx_set_block (entries, 1); | |
8095 | - dx_set_count (entries, 1); | |
8096 | - dx_set_limit (entries, dx_root_limit(dir, sizeof(root->info))); | |
8097 | + dx_set_block(entries, 1); | |
8098 | + dx_set_count(entries, 1); | |
8099 | + dx_set_limit(entries, dx_root_limit(dir, sizeof(root->info))); | |
8100 | ||
8101 | /* Initialize as for dx_probe */ | |
8102 | hinfo.hash_version = root->info.hash_version; | |
8103 | + if (hinfo.hash_version <= DX_HASH_TEA) | |
8104 | + hinfo.hash_version += EXT4_SB(dir->i_sb)->s_hash_unsigned; | |
8105 | hinfo.seed = EXT4_SB(dir->i_sb)->s_hash_seed; | |
8106 | ext4fs_dirhash(name, namelen, &hinfo); | |
8107 | frame = frames; | |
8108 | @@ -1443,14 +1450,13 @@ static int make_indexed_dir(handle_t *ha | |
8109 | * may not sleep between calling this and putting something into | |
8110 | * the entry, as someone else might have used it while you slept. | |
8111 | */ | |
8112 | -static int ext4_add_entry (handle_t *handle, struct dentry *dentry, | |
8113 | - struct inode *inode) | |
8114 | +static int ext4_add_entry(handle_t *handle, struct dentry *dentry, | |
8115 | + struct inode *inode) | |
8116 | { | |
8117 | struct inode *dir = dentry->d_parent->d_inode; | |
8118 | - unsigned long offset; | |
8119 | - struct buffer_head * bh; | |
8120 | + struct buffer_head *bh; | |
8121 | struct ext4_dir_entry_2 *de; | |
8122 | - struct super_block * sb; | |
8123 | + struct super_block *sb; | |
8124 | int retval; | |
8125 | int dx_fallback=0; | |
8126 | unsigned blocksize; | |
8127 | @@ -1469,7 +1475,7 @@ static int ext4_add_entry (handle_t *han | |
8128 | ext4_mark_inode_dirty(handle, dir); | |
8129 | } | |
8130 | blocks = dir->i_size >> sb->s_blocksize_bits; | |
8131 | - for (block = 0, offset = 0; block < blocks; block++) { | |
8132 | + for (block = 0; block < blocks; block++) { | |
8133 | bh = ext4_bread(handle, dir, block, 0, &retval); | |
8134 | if(!bh) | |
8135 | return retval; | |
8136 | @@ -1500,13 +1506,13 @@ static int ext4_dx_add_entry(handle_t *h | |
8137 | struct dx_frame frames[2], *frame; | |
8138 | struct dx_entry *entries, *at; | |
8139 | struct dx_hash_info hinfo; | |
8140 | - struct buffer_head * bh; | |
8141 | + struct buffer_head *bh; | |
8142 | struct inode *dir = dentry->d_parent->d_inode; | |
8143 | - struct super_block * sb = dir->i_sb; | |
8144 | + struct super_block *sb = dir->i_sb; | |
8145 | struct ext4_dir_entry_2 *de; | |
8146 | int err; | |
8147 | ||
8148 | - frame = dx_probe(dentry, NULL, &hinfo, frames, &err); | |
8149 | + frame = dx_probe(&dentry->d_name, dir, &hinfo, frames, &err); | |
8150 | if (!frame) | |
8151 | return err; | |
8152 | entries = frame->entries; | |
8153 | @@ -1527,7 +1533,7 @@ static int ext4_dx_add_entry(handle_t *h | |
8154 | } | |
8155 | ||
8156 | /* Block full, should compress but for now just split */ | |
8157 | - dxtrace(printk("using %u of %u node entries\n", | |
8158 | + dxtrace(printk(KERN_DEBUG "using %u of %u node entries\n", | |
8159 | dx_get_count(entries), dx_get_limit(entries))); | |
8160 | /* Need to split index? */ | |
8161 | if (dx_get_count(entries) == dx_get_limit(entries)) { | |
8162 | @@ -1559,7 +1565,8 @@ static int ext4_dx_add_entry(handle_t *h | |
8163 | if (levels) { | |
8164 | unsigned icount1 = icount/2, icount2 = icount - icount1; | |
8165 | unsigned hash2 = dx_get_hash(entries + icount1); | |
8166 | - dxtrace(printk("Split index %i/%i\n", icount1, icount2)); | |
8167 | + dxtrace(printk(KERN_DEBUG "Split index %i/%i\n", | |
8168 | + icount1, icount2)); | |
8169 | ||
8170 | BUFFER_TRACE(frame->bh, "get_write_access"); /* index root */ | |
8171 | err = ext4_journal_get_write_access(handle, | |
8172 | @@ -1567,11 +1574,11 @@ static int ext4_dx_add_entry(handle_t *h | |
8173 | if (err) | |
8174 | goto journal_error; | |
8175 | ||
8176 | - memcpy ((char *) entries2, (char *) (entries + icount1), | |
8177 | - icount2 * sizeof(struct dx_entry)); | |
8178 | - dx_set_count (entries, icount1); | |
8179 | - dx_set_count (entries2, icount2); | |
8180 | - dx_set_limit (entries2, dx_node_limit(dir)); | |
8181 | + memcpy((char *) entries2, (char *) (entries + icount1), | |
8182 | + icount2 * sizeof(struct dx_entry)); | |
8183 | + dx_set_count(entries, icount1); | |
8184 | + dx_set_count(entries2, icount2); | |
8185 | + dx_set_limit(entries2, dx_node_limit(dir)); | |
8186 | ||
8187 | /* Which index block gets the new entry? */ | |
8188 | if (at - entries >= icount1) { | |
8189 | @@ -1579,16 +1586,17 @@ static int ext4_dx_add_entry(handle_t *h | |
8190 | frame->entries = entries = entries2; | |
8191 | swap(frame->bh, bh2); | |
8192 | } | |
8193 | - dx_insert_block (frames + 0, hash2, newblock); | |
8194 | - dxtrace(dx_show_index ("node", frames[1].entries)); | |
8195 | - dxtrace(dx_show_index ("node", | |
8196 | + dx_insert_block(frames + 0, hash2, newblock); | |
8197 | + dxtrace(dx_show_index("node", frames[1].entries)); | |
8198 | + dxtrace(dx_show_index("node", | |
8199 | ((struct dx_node *) bh2->b_data)->entries)); | |
8200 | err = ext4_journal_dirty_metadata(handle, bh2); | |
8201 | if (err) | |
8202 | goto journal_error; | |
8203 | brelse (bh2); | |
8204 | } else { | |
8205 | - dxtrace(printk("Creating second level index...\n")); | |
8206 | + dxtrace(printk(KERN_DEBUG | |
8207 | + "Creating second level index...\n")); | |
8208 | memcpy((char *) entries2, (char *) entries, | |
8209 | icount * sizeof(struct dx_entry)); | |
8210 | dx_set_limit(entries2, dx_node_limit(dir)); | |
8211 | @@ -1630,12 +1638,12 @@ cleanup: | |
8212 | * ext4_delete_entry deletes a directory entry by merging it with the | |
8213 | * previous entry | |
8214 | */ | |
8215 | -static int ext4_delete_entry (handle_t *handle, | |
8216 | - struct inode * dir, | |
8217 | - struct ext4_dir_entry_2 * de_del, | |
8218 | - struct buffer_head * bh) | |
8219 | +static int ext4_delete_entry(handle_t *handle, | |
8220 | + struct inode *dir, | |
8221 | + struct ext4_dir_entry_2 *de_del, | |
8222 | + struct buffer_head *bh) | |
8223 | { | |
8224 | - struct ext4_dir_entry_2 * de, * pde; | |
8225 | + struct ext4_dir_entry_2 *de, *pde; | |
8226 | int i; | |
8227 | ||
8228 | i = 0; | |
8229 | @@ -1716,11 +1724,11 @@ static int ext4_add_nondir(handle_t *han | |
8230 | * If the create succeeds, we fill in the inode information | |
8231 | * with d_instantiate(). | |
8232 | */ | |
8233 | -static int ext4_create (struct inode * dir, struct dentry * dentry, int mode, | |
8234 | - struct nameidata *nd) | |
8235 | +static int ext4_create(struct inode *dir, struct dentry *dentry, int mode, | |
8236 | + struct nameidata *nd) | |
8237 | { | |
8238 | handle_t *handle; | |
8239 | - struct inode * inode; | |
8240 | + struct inode *inode; | |
8241 | int err, retries = 0; | |
8242 | ||
8243 | retry: | |
8244 | @@ -1747,8 +1755,8 @@ retry: | |
8245 | return err; | |
8246 | } | |
8247 | ||
8248 | -static int ext4_mknod (struct inode * dir, struct dentry *dentry, | |
8249 | - int mode, dev_t rdev) | |
8250 | +static int ext4_mknod(struct inode *dir, struct dentry *dentry, | |
8251 | + int mode, dev_t rdev) | |
8252 | { | |
8253 | handle_t *handle; | |
8254 | struct inode *inode; | |
8255 | @@ -1767,11 +1775,11 @@ retry: | |
8256 | if (IS_DIRSYNC(dir)) | |
8257 | handle->h_sync = 1; | |
8258 | ||
8259 | - inode = ext4_new_inode (handle, dir, mode); | |
8260 | + inode = ext4_new_inode(handle, dir, mode); | |
8261 | err = PTR_ERR(inode); | |
8262 | if (!IS_ERR(inode)) { | |
8263 | init_special_inode(inode, inode->i_mode, rdev); | |
8264 | -#ifdef CONFIG_EXT4DEV_FS_XATTR | |
8265 | +#if defined(CONFIG_EXT4_FS_XATTR) || defined(CONFIG_EXT4DEV_FS_XATTR) | |
8266 | inode->i_op = &ext4_special_inode_operations; | |
8267 | #endif | |
8268 | err = ext4_add_nondir(handle, dentry, inode); | |
8269 | @@ -1782,12 +1790,12 @@ retry: | |
8270 | return err; | |
8271 | } | |
8272 | ||
8273 | -static int ext4_mkdir(struct inode * dir, struct dentry * dentry, int mode) | |
8274 | +static int ext4_mkdir(struct inode *dir, struct dentry *dentry, int mode) | |
8275 | { | |
8276 | handle_t *handle; | |
8277 | - struct inode * inode; | |
8278 | - struct buffer_head * dir_block; | |
8279 | - struct ext4_dir_entry_2 * de; | |
8280 | + struct inode *inode; | |
8281 | + struct buffer_head *dir_block; | |
8282 | + struct ext4_dir_entry_2 *de; | |
8283 | int err, retries = 0; | |
8284 | ||
8285 | if (EXT4_DIR_LINK_MAX(dir)) | |
8286 | @@ -1803,7 +1811,7 @@ retry: | |
8287 | if (IS_DIRSYNC(dir)) | |
8288 | handle->h_sync = 1; | |
8289 | ||
8290 | - inode = ext4_new_inode (handle, dir, S_IFDIR | mode); | |
8291 | + inode = ext4_new_inode(handle, dir, S_IFDIR | mode); | |
8292 | err = PTR_ERR(inode); | |
8293 | if (IS_ERR(inode)) | |
8294 | goto out_stop; | |
8295 | @@ -1811,7 +1819,7 @@ retry: | |
8296 | inode->i_op = &ext4_dir_inode_operations; | |
8297 | inode->i_fop = &ext4_dir_operations; | |
8298 | inode->i_size = EXT4_I(inode)->i_disksize = inode->i_sb->s_blocksize; | |
8299 | - dir_block = ext4_bread (handle, inode, 0, 1, &err); | |
8300 | + dir_block = ext4_bread(handle, inode, 0, 1, &err); | |
8301 | if (!dir_block) | |
8302 | goto out_clear_inode; | |
8303 | BUFFER_TRACE(dir_block, "get_write_access"); | |
8304 | @@ -1820,26 +1828,26 @@ retry: | |
8305 | de->inode = cpu_to_le32(inode->i_ino); | |
8306 | de->name_len = 1; | |
8307 | de->rec_len = ext4_rec_len_to_disk(EXT4_DIR_REC_LEN(de->name_len)); | |
8308 | - strcpy (de->name, "."); | |
8309 | + strcpy(de->name, "."); | |
8310 | ext4_set_de_type(dir->i_sb, de, S_IFDIR); | |
8311 | de = ext4_next_entry(de); | |
8312 | de->inode = cpu_to_le32(dir->i_ino); | |
8313 | de->rec_len = ext4_rec_len_to_disk(inode->i_sb->s_blocksize - | |
8314 | EXT4_DIR_REC_LEN(1)); | |
8315 | de->name_len = 2; | |
8316 | - strcpy (de->name, ".."); | |
8317 | + strcpy(de->name, ".."); | |
8318 | ext4_set_de_type(dir->i_sb, de, S_IFDIR); | |
8319 | inode->i_nlink = 2; | |
8320 | BUFFER_TRACE(dir_block, "call ext4_journal_dirty_metadata"); | |
8321 | ext4_journal_dirty_metadata(handle, dir_block); | |
8322 | - brelse (dir_block); | |
8323 | + brelse(dir_block); | |
8324 | ext4_mark_inode_dirty(handle, inode); | |
8325 | - err = ext4_add_entry (handle, dentry, inode); | |
8326 | + err = ext4_add_entry(handle, dentry, inode); | |
8327 | if (err) { | |
8328 | out_clear_inode: | |
8329 | clear_nlink(inode); | |
8330 | ext4_mark_inode_dirty(handle, inode); | |
8331 | - iput (inode); | |
8332 | + iput(inode); | |
8333 | goto out_stop; | |
8334 | } | |
8335 | ext4_inc_count(handle, dir); | |
8336 | @@ -1856,17 +1864,17 @@ out_stop: | |
8337 | /* | |
8338 | * routine to check that the specified directory is empty (for rmdir) | |
8339 | */ | |
8340 | -static int empty_dir (struct inode * inode) | |
8341 | +static int empty_dir(struct inode *inode) | |
8342 | { | |
8343 | - unsigned long offset; | |
8344 | - struct buffer_head * bh; | |
8345 | - struct ext4_dir_entry_2 * de, * de1; | |
8346 | - struct super_block * sb; | |
8347 | + unsigned int offset; | |
8348 | + struct buffer_head *bh; | |
8349 | + struct ext4_dir_entry_2 *de, *de1; | |
8350 | + struct super_block *sb; | |
8351 | int err = 0; | |
8352 | ||
8353 | sb = inode->i_sb; | |
8354 | if (inode->i_size < EXT4_DIR_REC_LEN(1) + EXT4_DIR_REC_LEN(2) || | |
8355 | - !(bh = ext4_bread (NULL, inode, 0, 0, &err))) { | |
8356 | + !(bh = ext4_bread(NULL, inode, 0, 0, &err))) { | |
8357 | if (err) | |
8358 | ext4_error(inode->i_sb, __func__, | |
8359 | "error %d reading directory #%lu offset 0", | |
8360 | @@ -1881,29 +1889,29 @@ static int empty_dir (struct inode * ino | |
8361 | de1 = ext4_next_entry(de); | |
8362 | if (le32_to_cpu(de->inode) != inode->i_ino || | |
8363 | !le32_to_cpu(de1->inode) || | |
8364 | - strcmp (".", de->name) || | |
8365 | - strcmp ("..", de1->name)) { | |
8366 | - ext4_warning (inode->i_sb, "empty_dir", | |
8367 | - "bad directory (dir #%lu) - no `.' or `..'", | |
8368 | - inode->i_ino); | |
8369 | - brelse (bh); | |
8370 | + strcmp(".", de->name) || | |
8371 | + strcmp("..", de1->name)) { | |
8372 | + ext4_warning(inode->i_sb, "empty_dir", | |
8373 | + "bad directory (dir #%lu) - no `.' or `..'", | |
8374 | + inode->i_ino); | |
8375 | + brelse(bh); | |
8376 | return 1; | |
8377 | } | |
8378 | offset = ext4_rec_len_from_disk(de->rec_len) + | |
8379 | ext4_rec_len_from_disk(de1->rec_len); | |
8380 | de = ext4_next_entry(de1); | |
8381 | - while (offset < inode->i_size ) { | |
8382 | + while (offset < inode->i_size) { | |
8383 | if (!bh || | |
8384 | (void *) de >= (void *) (bh->b_data+sb->s_blocksize)) { | |
8385 | err = 0; | |
8386 | - brelse (bh); | |
8387 | - bh = ext4_bread (NULL, inode, | |
8388 | + brelse(bh); | |
8389 | + bh = ext4_bread(NULL, inode, | |
8390 | offset >> EXT4_BLOCK_SIZE_BITS(sb), 0, &err); | |
8391 | if (!bh) { | |
8392 | if (err) | |
8393 | ext4_error(sb, __func__, | |
8394 | "error %d reading directory" | |
8395 | - " #%lu offset %lu", | |
8396 | + " #%lu offset %u", | |
8397 | err, inode->i_ino, offset); | |
8398 | offset += sb->s_blocksize; | |
8399 | continue; | |
8400 | @@ -1917,13 +1925,13 @@ static int empty_dir (struct inode * ino | |
8401 | continue; | |
8402 | } | |
8403 | if (le32_to_cpu(de->inode)) { | |
8404 | - brelse (bh); | |
8405 | + brelse(bh); | |
8406 | return 0; | |
8407 | } | |
8408 | offset += ext4_rec_len_from_disk(de->rec_len); | |
8409 | de = ext4_next_entry(de); | |
8410 | } | |
8411 | - brelse (bh); | |
8412 | + brelse(bh); | |
8413 | return 1; | |
8414 | } | |
8415 | ||
8416 | @@ -1954,8 +1962,8 @@ int ext4_orphan_add(handle_t *handle, st | |
8417 | * ->i_nlink. For, say it, character device. Not a regular file, | |
8418 | * not a directory, not a symlink and ->i_nlink > 0. | |
8419 | */ | |
8420 | - J_ASSERT ((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || | |
8421 | - S_ISLNK(inode->i_mode)) || inode->i_nlink == 0); | |
8422 | + J_ASSERT((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || | |
8423 | + S_ISLNK(inode->i_mode)) || inode->i_nlink == 0); | |
8424 | ||
8425 | BUFFER_TRACE(EXT4_SB(sb)->s_sbh, "get_write_access"); | |
8426 | err = ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh); | |
8427 | @@ -2003,7 +2011,7 @@ int ext4_orphan_del(handle_t *handle, st | |
8428 | struct list_head *prev; | |
8429 | struct ext4_inode_info *ei = EXT4_I(inode); | |
8430 | struct ext4_sb_info *sbi; | |
8431 | - unsigned long ino_next; | |
8432 | + __u32 ino_next; | |
8433 | struct ext4_iloc iloc; | |
8434 | int err = 0; | |
8435 | ||
8436 | @@ -2033,7 +2041,7 @@ int ext4_orphan_del(handle_t *handle, st | |
8437 | goto out_err; | |
8438 | ||
8439 | if (prev == &sbi->s_orphan) { | |
8440 | - jbd_debug(4, "superblock will point to %lu\n", ino_next); | |
8441 | + jbd_debug(4, "superblock will point to %u\n", ino_next); | |
8442 | BUFFER_TRACE(sbi->s_sbh, "get_write_access"); | |
8443 | err = ext4_journal_get_write_access(handle, sbi->s_sbh); | |
8444 | if (err) | |
8445 | @@ -2045,7 +2053,7 @@ int ext4_orphan_del(handle_t *handle, st | |
8446 | struct inode *i_prev = | |
8447 | &list_entry(prev, struct ext4_inode_info, i_orphan)->vfs_inode; | |
8448 | ||
8449 | - jbd_debug(4, "orphan inode %lu will point to %lu\n", | |
8450 | + jbd_debug(4, "orphan inode %lu will point to %u\n", | |
8451 | i_prev->i_ino, ino_next); | |
8452 | err = ext4_reserve_inode_write(handle, i_prev, &iloc2); | |
8453 | if (err) | |
8454 | @@ -2069,12 +2077,12 @@ out_brelse: | |
8455 | goto out_err; | |
8456 | } | |
8457 | ||
8458 | -static int ext4_rmdir (struct inode * dir, struct dentry *dentry) | |
8459 | +static int ext4_rmdir(struct inode *dir, struct dentry *dentry) | |
8460 | { | |
8461 | int retval; | |
8462 | - struct inode * inode; | |
8463 | - struct buffer_head * bh; | |
8464 | - struct ext4_dir_entry_2 * de; | |
8465 | + struct inode *inode; | |
8466 | + struct buffer_head *bh; | |
8467 | + struct ext4_dir_entry_2 *de; | |
8468 | handle_t *handle; | |
8469 | ||
8470 | /* Initialize quotas before so that eventual writes go in | |
8471 | @@ -2085,7 +2093,7 @@ static int ext4_rmdir (struct inode * di | |
8472 | return PTR_ERR(handle); | |
8473 | ||
8474 | retval = -ENOENT; | |
8475 | - bh = ext4_find_entry (dentry, &de); | |
8476 | + bh = ext4_find_entry(dir, &dentry->d_name, &de); | |
8477 | if (!bh) | |
8478 | goto end_rmdir; | |
8479 | ||
8480 | @@ -2099,16 +2107,16 @@ static int ext4_rmdir (struct inode * di | |
8481 | goto end_rmdir; | |
8482 | ||
8483 | retval = -ENOTEMPTY; | |
8484 | - if (!empty_dir (inode)) | |
8485 | + if (!empty_dir(inode)) | |
8486 | goto end_rmdir; | |
8487 | ||
8488 | retval = ext4_delete_entry(handle, dir, de, bh); | |
8489 | if (retval) | |
8490 | goto end_rmdir; | |
8491 | if (!EXT4_DIR_LINK_EMPTY(inode)) | |
8492 | - ext4_warning (inode->i_sb, "ext4_rmdir", | |
8493 | - "empty directory has too many links (%d)", | |
8494 | - inode->i_nlink); | |
8495 | + ext4_warning(inode->i_sb, "ext4_rmdir", | |
8496 | + "empty directory has too many links (%d)", | |
8497 | + inode->i_nlink); | |
8498 | inode->i_version++; | |
8499 | clear_nlink(inode); | |
8500 | /* There's no need to set i_disksize: the fact that i_nlink is | |
8501 | @@ -2124,16 +2132,16 @@ static int ext4_rmdir (struct inode * di | |
8502 | ||
8503 | end_rmdir: | |
8504 | ext4_journal_stop(handle); | |
8505 | - brelse (bh); | |
8506 | + brelse(bh); | |
8507 | return retval; | |
8508 | } | |
8509 | ||
8510 | -static int ext4_unlink(struct inode * dir, struct dentry *dentry) | |
8511 | +static int ext4_unlink(struct inode *dir, struct dentry *dentry) | |
8512 | { | |
8513 | int retval; | |
8514 | - struct inode * inode; | |
8515 | - struct buffer_head * bh; | |
8516 | - struct ext4_dir_entry_2 * de; | |
8517 | + struct inode *inode; | |
8518 | + struct buffer_head *bh; | |
8519 | + struct ext4_dir_entry_2 *de; | |
8520 | handle_t *handle; | |
8521 | ||
8522 | /* Initialize quotas before so that eventual writes go | |
8523 | @@ -2147,7 +2155,7 @@ static int ext4_unlink(struct inode * di | |
8524 | handle->h_sync = 1; | |
8525 | ||
8526 | retval = -ENOENT; | |
8527 | - bh = ext4_find_entry (dentry, &de); | |
8528 | + bh = ext4_find_entry(dir, &dentry->d_name, &de); | |
8529 | if (!bh) | |
8530 | goto end_unlink; | |
8531 | ||
8532 | @@ -2158,9 +2166,9 @@ static int ext4_unlink(struct inode * di | |
8533 | goto end_unlink; | |
8534 | ||
8535 | if (!inode->i_nlink) { | |
8536 | - ext4_warning (inode->i_sb, "ext4_unlink", | |
8537 | - "Deleting nonexistent file (%lu), %d", | |
8538 | - inode->i_ino, inode->i_nlink); | |
8539 | + ext4_warning(inode->i_sb, "ext4_unlink", | |
8540 | + "Deleting nonexistent file (%lu), %d", | |
8541 | + inode->i_ino, inode->i_nlink); | |
8542 | inode->i_nlink = 1; | |
8543 | } | |
8544 | retval = ext4_delete_entry(handle, dir, de, bh); | |
8545 | @@ -2178,15 +2186,15 @@ static int ext4_unlink(struct inode * di | |
8546 | ||
8547 | end_unlink: | |
8548 | ext4_journal_stop(handle); | |
8549 | - brelse (bh); | |
8550 | + brelse(bh); | |
8551 | return retval; | |
8552 | } | |
8553 | ||
8554 | -static int ext4_symlink (struct inode * dir, | |
8555 | - struct dentry *dentry, const char * symname) | |
8556 | +static int ext4_symlink(struct inode *dir, | |
8557 | + struct dentry *dentry, const char *symname) | |
8558 | { | |
8559 | handle_t *handle; | |
8560 | - struct inode * inode; | |
8561 | + struct inode *inode; | |
8562 | int l, err, retries = 0; | |
8563 | ||
8564 | l = strlen(symname)+1; | |
8565 | @@ -2203,12 +2211,12 @@ retry: | |
8566 | if (IS_DIRSYNC(dir)) | |
8567 | handle->h_sync = 1; | |
8568 | ||
8569 | - inode = ext4_new_inode (handle, dir, S_IFLNK|S_IRWXUGO); | |
8570 | + inode = ext4_new_inode(handle, dir, S_IFLNK|S_IRWXUGO); | |
8571 | err = PTR_ERR(inode); | |
8572 | if (IS_ERR(inode)) | |
8573 | goto out_stop; | |
8574 | ||
8575 | - if (l > sizeof (EXT4_I(inode)->i_data)) { | |
8576 | + if (l > sizeof(EXT4_I(inode)->i_data)) { | |
8577 | inode->i_op = &ext4_symlink_inode_operations; | |
8578 | ext4_set_aops(inode); | |
8579 | /* | |
8580 | @@ -2216,18 +2224,19 @@ retry: | |
8581 | * We have a transaction open. All is sweetness. It also sets | |
8582 | * i_size in generic_commit_write(). | |
8583 | */ | |
8584 | - err = __page_symlink(inode, symname, l, 1); | |
8585 | + err = __page_symlink(inode, symname, l, | |
8586 | + mapping_gfp_mask(inode->i_mapping) & ~__GFP_FS); | |
8587 | if (err) { | |
8588 | clear_nlink(inode); | |
8589 | ext4_mark_inode_dirty(handle, inode); | |
8590 | - iput (inode); | |
8591 | + iput(inode); | |
8592 | goto out_stop; | |
8593 | } | |
8594 | } else { | |
8595 | /* clear the extent format for fast symlink */ | |
8596 | EXT4_I(inode)->i_flags &= ~EXT4_EXTENTS_FL; | |
8597 | inode->i_op = &ext4_fast_symlink_inode_operations; | |
8598 | - memcpy((char*)&EXT4_I(inode)->i_data,symname,l); | |
8599 | + memcpy((char *)&EXT4_I(inode)->i_data, symname, l); | |
8600 | inode->i_size = l-1; | |
8601 | } | |
8602 | EXT4_I(inode)->i_disksize = inode->i_size; | |
8603 | @@ -2239,8 +2248,8 @@ out_stop: | |
8604 | return err; | |
8605 | } | |
8606 | ||
8607 | -static int ext4_link (struct dentry * old_dentry, | |
8608 | - struct inode * dir, struct dentry *dentry) | |
8609 | +static int ext4_link(struct dentry *old_dentry, | |
8610 | + struct inode *dir, struct dentry *dentry) | |
8611 | { | |
8612 | handle_t *handle; | |
8613 | struct inode *inode = old_dentry->d_inode; | |
8614 | @@ -2283,13 +2292,13 @@ retry: | |
8615 | * Anybody can rename anything with this: the permission checks are left to the | |
8616 | * higher-level routines. | |
8617 | */ | |
8618 | -static int ext4_rename (struct inode * old_dir, struct dentry *old_dentry, | |
8619 | - struct inode * new_dir,struct dentry *new_dentry) | |
8620 | +static int ext4_rename(struct inode *old_dir, struct dentry *old_dentry, | |
8621 | + struct inode *new_dir, struct dentry *new_dentry) | |
8622 | { | |
8623 | handle_t *handle; | |
8624 | - struct inode * old_inode, * new_inode; | |
8625 | - struct buffer_head * old_bh, * new_bh, * dir_bh; | |
8626 | - struct ext4_dir_entry_2 * old_de, * new_de; | |
8627 | + struct inode *old_inode, *new_inode; | |
8628 | + struct buffer_head *old_bh, *new_bh, *dir_bh; | |
8629 | + struct ext4_dir_entry_2 *old_de, *new_de; | |
8630 | int retval; | |
8631 | ||
8632 | old_bh = new_bh = dir_bh = NULL; | |
8633 | @@ -2307,7 +2316,7 @@ static int ext4_rename (struct inode * o | |
8634 | if (IS_DIRSYNC(old_dir) || IS_DIRSYNC(new_dir)) | |
8635 | handle->h_sync = 1; | |
8636 | ||
8637 | - old_bh = ext4_find_entry (old_dentry, &old_de); | |
8638 | + old_bh = ext4_find_entry(old_dir, &old_dentry->d_name, &old_de); | |
8639 | /* | |
8640 | * Check for inode number is _not_ due to possible IO errors. | |
8641 | * We might rmdir the source, keep it as pwd of some process | |
8642 | @@ -2320,32 +2329,32 @@ static int ext4_rename (struct inode * o | |
8643 | goto end_rename; | |
8644 | ||
8645 | new_inode = new_dentry->d_inode; | |
8646 | - new_bh = ext4_find_entry (new_dentry, &new_de); | |
8647 | + new_bh = ext4_find_entry(new_dir, &new_dentry->d_name, &new_de); | |
8648 | if (new_bh) { | |
8649 | if (!new_inode) { | |
8650 | - brelse (new_bh); | |
8651 | + brelse(new_bh); | |
8652 | new_bh = NULL; | |
8653 | } | |
8654 | } | |
8655 | if (S_ISDIR(old_inode->i_mode)) { | |
8656 | if (new_inode) { | |
8657 | retval = -ENOTEMPTY; | |
8658 | - if (!empty_dir (new_inode)) | |
8659 | + if (!empty_dir(new_inode)) | |
8660 | goto end_rename; | |
8661 | } | |
8662 | retval = -EIO; | |
8663 | - dir_bh = ext4_bread (handle, old_inode, 0, 0, &retval); | |
8664 | + dir_bh = ext4_bread(handle, old_inode, 0, 0, &retval); | |
8665 | if (!dir_bh) | |
8666 | goto end_rename; | |
8667 | if (le32_to_cpu(PARENT_INO(dir_bh->b_data)) != old_dir->i_ino) | |
8668 | goto end_rename; | |
8669 | retval = -EMLINK; | |
8670 | - if (!new_inode && new_dir!=old_dir && | |
8671 | + if (!new_inode && new_dir != old_dir && | |
8672 | new_dir->i_nlink >= EXT4_LINK_MAX) | |
8673 | goto end_rename; | |
8674 | } | |
8675 | if (!new_bh) { | |
8676 | - retval = ext4_add_entry (handle, new_dentry, old_inode); | |
8677 | + retval = ext4_add_entry(handle, new_dentry, old_inode); | |
8678 | if (retval) | |
8679 | goto end_rename; | |
8680 | } else { | |
8681 | @@ -2387,7 +2396,7 @@ static int ext4_rename (struct inode * o | |
8682 | struct buffer_head *old_bh2; | |
8683 | struct ext4_dir_entry_2 *old_de2; | |
8684 | ||
8685 | - old_bh2 = ext4_find_entry(old_dentry, &old_de2); | |
8686 | + old_bh2 = ext4_find_entry(old_dir, &old_dentry->d_name, &old_de2); | |
8687 | if (old_bh2) { | |
8688 | retval = ext4_delete_entry(handle, old_dir, | |
8689 | old_de2, old_bh2); | |
8690 | @@ -2432,9 +2441,9 @@ static int ext4_rename (struct inode * o | |
8691 | retval = 0; | |
8692 | ||
8693 | end_rename: | |
8694 | - brelse (dir_bh); | |
8695 | - brelse (old_bh); | |
8696 | - brelse (new_bh); | |
8697 | + brelse(dir_bh); | |
8698 | + brelse(old_bh); | |
8699 | + brelse(new_bh); | |
8700 | ext4_journal_stop(handle); | |
8701 | return retval; | |
8702 | } | |
8703 | @@ -2453,7 +2462,7 @@ const struct inode_operations ext4_dir_i | |
8704 | .mknod = ext4_mknod, | |
8705 | .rename = ext4_rename, | |
8706 | .setattr = ext4_setattr, | |
8707 | -#ifdef CONFIG_EXT4DEV_FS_XATTR | |
8708 | +#if defined(CONFIG_EXT4_FS_XATTR) || defined(CONFIG_EXT4DEV_FS_XATTR) | |
8709 | .setxattr = generic_setxattr, | |
8710 | .getxattr = generic_getxattr, | |
8711 | .listxattr = ext4_listxattr, | |
8712 | @@ -2464,7 +2473,7 @@ const struct inode_operations ext4_dir_i | |
8713 | ||
8714 | const struct inode_operations ext4_special_inode_operations = { | |
8715 | .setattr = ext4_setattr, | |
8716 | -#ifdef CONFIG_EXT4DEV_FS_XATTR | |
8717 | +#if defined(CONFIG_EXT4_FS_XATTR) || defined(CONFIG_EXT4DEV_FS_XATTR) | |
8718 | .setxattr = generic_setxattr, | |
8719 | .getxattr = generic_getxattr, | |
8720 | .listxattr = ext4_listxattr, | |
8721 | diff -rup b/fs/ext4//resize.c a/fs/ext4///resize.c | |
8722 | --- b/fs/ext4/resize.c 2009-02-11 14:37:58.000000000 +0100 | |
8723 | +++ a/fs/ext4/resize.c 2009-02-10 21:40:11.000000000 +0100 | |
8724 | @@ -50,7 +50,7 @@ static int verify_group_input(struct sup | |
8725 | ext4_get_group_no_and_offset(sb, start, NULL, &offset); | |
8726 | if (group != sbi->s_groups_count) | |
8727 | ext4_warning(sb, __func__, | |
8728 | - "Cannot add at group %u (only %lu groups)", | |
8729 | + "Cannot add at group %u (only %u groups)", | |
8730 | input->group, sbi->s_groups_count); | |
8731 | else if (offset != 0) | |
8732 | ext4_warning(sb, __func__, "Last group not full"); | |
8733 | @@ -284,11 +284,9 @@ static int setup_new_group_blocks(struct | |
8734 | if ((err = extend_or_restart_transaction(handle, 2, bh))) | |
8735 | goto exit_bh; | |
8736 | ||
8737 | - mark_bitmap_end(input->blocks_count, EXT4_BLOCKS_PER_GROUP(sb), | |
8738 | - bh->b_data); | |
8739 | + mark_bitmap_end(input->blocks_count, sb->s_blocksize * 8, bh->b_data); | |
8740 | ext4_journal_dirty_metadata(handle, bh); | |
8741 | brelse(bh); | |
8742 | - | |
8743 | /* Mark unused entries in inode bitmap used */ | |
8744 | ext4_debug("clear inode bitmap %#04llx (+%llu)\n", | |
8745 | input->inode_bitmap, input->inode_bitmap - start); | |
8746 | @@ -297,7 +295,7 @@ static int setup_new_group_blocks(struct | |
8747 | goto exit_journal; | |
8748 | } | |
8749 | ||
8750 | - mark_bitmap_end(EXT4_INODES_PER_GROUP(sb), EXT4_BLOCKS_PER_GROUP(sb), | |
8751 | + mark_bitmap_end(EXT4_INODES_PER_GROUP(sb), sb->s_blocksize * 8, | |
8752 | bh->b_data); | |
8753 | ext4_journal_dirty_metadata(handle, bh); | |
8754 | exit_bh: | |
8755 | @@ -416,8 +414,8 @@ static int add_new_gdb(handle_t *handle, | |
8756 | "EXT4-fs: ext4_add_new_gdb: adding group block %lu\n", | |
8757 | gdb_num); | |
8758 | ||
8759 | - /* | |
8760 | - * If we are not using the primary superblock/GDT copy don't resize, | |
8761 | + /* | |
8762 | + * If we are not using the primary superblock/GDT copy don't resize, | |
8763 | * because the user tools have no way of handling this. Probably a | |
8764 | * bad time to do it anyways. | |
8765 | */ | |
8766 | @@ -715,7 +713,7 @@ static void update_backups(struct super_ | |
8767 | exit_err: | |
8768 | if (err) { | |
8769 | ext4_warning(sb, __func__, | |
8770 | - "can't update backup for group %lu (err %d), " | |
8771 | + "can't update backup for group %u (err %d), " | |
8772 | "forcing fsck on next reboot", group, err); | |
8773 | sbi->s_mount_state &= ~EXT4_VALID_FS; | |
8774 | sbi->s_es->s_state &= cpu_to_le16(~EXT4_VALID_FS); | |
8775 | @@ -747,6 +745,7 @@ int ext4_group_add(struct super_block *s | |
8776 | struct inode *inode = NULL; | |
8777 | handle_t *handle; | |
8778 | int gdb_off, gdb_num; | |
8779 | + int num_grp_locked = 0; | |
8780 | int err, err2; | |
8781 | ||
8782 | gdb_num = input->group / EXT4_DESC_PER_BLOCK(sb); | |
8783 | @@ -787,6 +786,7 @@ int ext4_group_add(struct super_block *s | |
8784 | } | |
8785 | } | |
8786 | ||
8787 | + | |
8788 | if ((err = verify_group_input(sb, input))) | |
8789 | goto exit_put; | |
8790 | ||
8791 | @@ -855,6 +855,7 @@ int ext4_group_add(struct super_block *s | |
8792 | * using the new disk blocks. | |
8793 | */ | |
8794 | ||
8795 | + num_grp_locked = ext4_mb_get_buddy_cache_lock(sb, input->group); | |
8796 | /* Update group descriptor block for new group */ | |
8797 | gdp = (struct ext4_group_desc *)((char *)primary->b_data + | |
8798 | gdb_off * EXT4_DESC_SIZE(sb)); | |
8799 | @@ -862,19 +863,21 @@ int ext4_group_add(struct super_block *s | |
8800 | ext4_block_bitmap_set(sb, gdp, input->block_bitmap); /* LV FIXME */ | |
8801 | ext4_inode_bitmap_set(sb, gdp, input->inode_bitmap); /* LV FIXME */ | |
8802 | ext4_inode_table_set(sb, gdp, input->inode_table); /* LV FIXME */ | |
8803 | - gdp->bg_free_blocks_count = cpu_to_le16(input->free_blocks_count); | |
8804 | - gdp->bg_free_inodes_count = cpu_to_le16(EXT4_INODES_PER_GROUP(sb)); | |
8805 | + ext4_free_blks_set(sb, gdp, input->free_blocks_count); | |
8806 | + ext4_free_inodes_set(sb, gdp, EXT4_INODES_PER_GROUP(sb)); | |
8807 | + gdp->bg_flags |= cpu_to_le16(EXT4_BG_INODE_ZEROED); | |
8808 | gdp->bg_checksum = ext4_group_desc_csum(sbi, input->group, gdp); | |
8809 | ||
8810 | /* | |
8811 | * We can allocate memory for mb_alloc based on the new group | |
8812 | * descriptor | |
8813 | */ | |
8814 | - if (test_opt(sb, MBALLOC)) { | |
8815 | - err = ext4_mb_add_more_groupinfo(sb, input->group, gdp); | |
8816 | - if (err) | |
8817 | - goto exit_journal; | |
8818 | + err = ext4_mb_add_groupinfo(sb, input->group, gdp); | |
8819 | + if (err) { | |
8820 | + ext4_mb_put_buddy_cache_lock(sb, input->group, num_grp_locked); | |
8821 | + goto exit_journal; | |
8822 | } | |
8823 | + | |
8824 | /* | |
8825 | * Make the new blocks and inodes valid next. We do this before | |
8826 | * increasing the group count so that once the group is enabled, | |
8827 | @@ -915,6 +918,7 @@ int ext4_group_add(struct super_block *s | |
8828 | ||
8829 | /* Update the global fs size fields */ | |
8830 | sbi->s_groups_count++; | |
8831 | + ext4_mb_put_buddy_cache_lock(sb, input->group, num_grp_locked); | |
8832 | ||
8833 | ext4_journal_dirty_metadata(handle, primary); | |
8834 | ||
8835 | @@ -973,12 +977,10 @@ int ext4_group_extend(struct super_block | |
8836 | ext4_group_t o_groups_count; | |
8837 | ext4_grpblk_t last; | |
8838 | ext4_grpblk_t add; | |
8839 | - struct buffer_head * bh; | |
8840 | + struct buffer_head *bh; | |
8841 | handle_t *handle; | |
8842 | int err; | |
8843 | - unsigned long freed_blocks; | |
8844 | ext4_group_t group; | |
8845 | - struct ext4_group_info *grp; | |
8846 | ||
8847 | /* We don't need to worry about locking wrt other resizers just | |
8848 | * yet: we're going to revalidate es->s_blocks_count after | |
8849 | @@ -1077,50 +1079,13 @@ int ext4_group_extend(struct super_block | |
8850 | unlock_super(sb); | |
8851 | ext4_debug("freeing blocks %llu through %llu\n", o_blocks_count, | |
8852 | o_blocks_count + add); | |
8853 | - ext4_free_blocks_sb(handle, sb, o_blocks_count, add, &freed_blocks); | |
8854 | + /* We add the blocks to the bitmap and set the group need init bit */ | |
8855 | + ext4_add_groupblocks(handle, sb, o_blocks_count, add); | |
8856 | ext4_debug("freed blocks %llu through %llu\n", o_blocks_count, | |
8857 | o_blocks_count + add); | |
8858 | if ((err = ext4_journal_stop(handle))) | |
8859 | goto exit_put; | |
8860 | ||
8861 | - /* | |
8862 | - * Mark mballoc pages as not up to date so that they will be updated | |
8863 | - * next time they are loaded by ext4_mb_load_buddy. | |
8864 | - */ | |
8865 | - if (test_opt(sb, MBALLOC)) { | |
8866 | - struct ext4_sb_info *sbi = EXT4_SB(sb); | |
8867 | - struct inode *inode = sbi->s_buddy_cache; | |
8868 | - int blocks_per_page; | |
8869 | - int block; | |
8870 | - int pnum; | |
8871 | - struct page *page; | |
8872 | - | |
8873 | - /* Set buddy page as not up to date */ | |
8874 | - blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize; | |
8875 | - block = group * 2; | |
8876 | - pnum = block / blocks_per_page; | |
8877 | - page = find_get_page(inode->i_mapping, pnum); | |
8878 | - if (page != NULL) { | |
8879 | - ClearPageUptodate(page); | |
8880 | - page_cache_release(page); | |
8881 | - } | |
8882 | - | |
8883 | - /* Set bitmap page as not up to date */ | |
8884 | - block++; | |
8885 | - pnum = block / blocks_per_page; | |
8886 | - page = find_get_page(inode->i_mapping, pnum); | |
8887 | - if (page != NULL) { | |
8888 | - ClearPageUptodate(page); | |
8889 | - page_cache_release(page); | |
8890 | - } | |
8891 | - | |
8892 | - /* Get the info on the last group */ | |
8893 | - grp = ext4_get_group_info(sb, group); | |
8894 | - | |
8895 | - /* Update free blocks in group info */ | |
8896 | - ext4_mb_update_group_info(grp, add); | |
8897 | - } | |
8898 | - | |
8899 | if (test_opt(sb, DEBUG)) | |
8900 | printk(KERN_DEBUG "EXT4-fs: extended group to %llu blocks\n", | |
8901 | ext4_blocks_count(es)); | |
8902 | diff -rup b/fs/ext4//super.c a/fs/ext4///super.c | |
8903 | --- b/fs/ext4/super.c 2009-02-11 14:37:58.000000000 +0100 | |
8904 | +++ a/fs/ext4/super.c 2009-02-11 13:47:04.000000000 +0100 | |
8905 | @@ -34,6 +34,8 @@ | |
8906 | #include <linux/namei.h> | |
8907 | #include <linux/quotaops.h> | |
8908 | #include <linux/seq_file.h> | |
8909 | +#include <linux/proc_fs.h> | |
8910 | +#include <linux/marker.h> | |
8911 | #include <linux/log2.h> | |
8912 | #include <linux/crc16.h> | |
8913 | #include <asm/uaccess.h> | |
8914 | @@ -45,6 +47,8 @@ | |
8915 | #include "namei.h" | |
8916 | #include "group.h" | |
8917 | ||
8918 | +struct proc_dir_entry *ext4_proc_root; | |
8919 | + | |
8920 | static int ext4_load_journal(struct super_block *, struct ext4_super_block *, | |
8921 | unsigned long journal_devnum); | |
8922 | static int ext4_create_journal(struct super_block *, struct ext4_super_block *, | |
8923 | @@ -89,6 +93,38 @@ ext4_fsblk_t ext4_inode_table(struct sup | |
8924 | (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); | |
8925 | } | |
8926 | ||
8927 | +__u32 ext4_free_blks_count(struct super_block *sb, | |
8928 | + struct ext4_group_desc *bg) | |
8929 | +{ | |
8930 | + return le16_to_cpu(bg->bg_free_blocks_count_lo) | | |
8931 | + (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
8932 | + (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); | |
8933 | +} | |
8934 | + | |
8935 | +__u32 ext4_free_inodes_count(struct super_block *sb, | |
8936 | + struct ext4_group_desc *bg) | |
8937 | +{ | |
8938 | + return le16_to_cpu(bg->bg_free_inodes_count_lo) | | |
8939 | + (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
8940 | + (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); | |
8941 | +} | |
8942 | + | |
8943 | +__u32 ext4_used_dirs_count(struct super_block *sb, | |
8944 | + struct ext4_group_desc *bg) | |
8945 | +{ | |
8946 | + return le16_to_cpu(bg->bg_used_dirs_count_lo) | | |
8947 | + (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
8948 | + (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); | |
8949 | +} | |
8950 | + | |
8951 | +__u32 ext4_itable_unused_count(struct super_block *sb, | |
8952 | + struct ext4_group_desc *bg) | |
8953 | +{ | |
8954 | + return le16_to_cpu(bg->bg_itable_unused_lo) | | |
8955 | + (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? | |
8956 | + (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); | |
8957 | +} | |
8958 | + | |
8959 | void ext4_block_bitmap_set(struct super_block *sb, | |
8960 | struct ext4_group_desc *bg, ext4_fsblk_t blk) | |
8961 | { | |
8962 | @@ -113,6 +149,38 @@ void ext4_inode_table_set(struct super_b | |
8963 | bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); | |
8964 | } | |
8965 | ||
8966 | +void ext4_free_blks_set(struct super_block *sb, | |
8967 | + struct ext4_group_desc *bg, __u32 count) | |
8968 | +{ | |
8969 | + bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); | |
8970 | + if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
8971 | + bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); | |
8972 | +} | |
8973 | + | |
8974 | +void ext4_free_inodes_set(struct super_block *sb, | |
8975 | + struct ext4_group_desc *bg, __u32 count) | |
8976 | +{ | |
8977 | + bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); | |
8978 | + if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
8979 | + bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); | |
8980 | +} | |
8981 | + | |
8982 | +void ext4_used_dirs_set(struct super_block *sb, | |
8983 | + struct ext4_group_desc *bg, __u32 count) | |
8984 | +{ | |
8985 | + bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); | |
8986 | + if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
8987 | + bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); | |
8988 | +} | |
8989 | + | |
8990 | +void ext4_itable_unused_set(struct super_block *sb, | |
8991 | + struct ext4_group_desc *bg, __u32 count) | |
8992 | +{ | |
8993 | + bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); | |
8994 | + if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) | |
8995 | + bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); | |
8996 | +} | |
8997 | + | |
8998 | /* | |
8999 | * Wrappers for jbd2_journal_start/end. | |
9000 | * | |
9001 | @@ -329,7 +397,8 @@ void ext4_abort(struct super_block *sb, | |
9002 | EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; | |
9003 | sb->s_flags |= MS_RDONLY; | |
9004 | EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT; | |
9005 | - jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); | |
9006 | + if (EXT4_SB(sb)->s_journal) | |
9007 | + jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); | |
9008 | } | |
9009 | ||
9010 | void ext4_warning(struct super_block *sb, const char *function, | |
9011 | @@ -345,6 +414,44 @@ void ext4_warning(struct super_block *sb | |
9012 | va_end(args); | |
9013 | } | |
9014 | ||
9015 | +void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp, | |
9016 | + const char *function, const char *fmt, ...) | |
9017 | +__releases(bitlock) | |
9018 | +__acquires(bitlock) | |
9019 | +{ | |
9020 | + va_list args; | |
9021 | + struct ext4_super_block *es = EXT4_SB(sb)->s_es; | |
9022 | + | |
9023 | + va_start(args, fmt); | |
9024 | + printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function); | |
9025 | + vprintk(fmt, args); | |
9026 | + printk("\n"); | |
9027 | + va_end(args); | |
9028 | + | |
9029 | + if (test_opt(sb, ERRORS_CONT)) { | |
9030 | + EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; | |
9031 | + es->s_state |= cpu_to_le16(EXT4_ERROR_FS); | |
9032 | + ext4_commit_super(sb, es, 0); | |
9033 | + return; | |
9034 | + } | |
9035 | + ext4_unlock_group(sb, grp); | |
9036 | + ext4_handle_error(sb); | |
9037 | + /* | |
9038 | + * We only get here in the ERRORS_RO case; relocking the group | |
9039 | + * may be dangerous, but nothing bad will happen since the | |
9040 | + * filesystem will have already been marked read/only and the | |
9041 | + * journal has been aborted. We return 1 as a hint to callers | |
9042 | + * who might what to use the return value from | |
9043 | + * ext4_grp_locked_error() to distinguish beween the | |
9044 | + * ERRORS_CONT and ERRORS_RO case, and perhaps return more | |
9045 | + * aggressively from the ext4 function in question, with a | |
9046 | + * more appropriate error code. | |
9047 | + */ | |
9048 | + ext4_lock_group(sb, grp); | |
9049 | + return; | |
9050 | +} | |
9051 | + | |
9052 | + | |
9053 | void ext4_update_dynamic_rev(struct super_block *sb) | |
9054 | { | |
9055 | struct ext4_super_block *es = EXT4_SB(sb)->s_es; | |
9056 | @@ -370,66 +477,6 @@ void ext4_update_dynamic_rev(struct supe | |
9057 | */ | |
9058 | } | |
9059 | ||
9060 | -int ext4_update_compat_feature(handle_t *handle, | |
9061 | - struct super_block *sb, __u32 compat) | |
9062 | -{ | |
9063 | - int err = 0; | |
9064 | - if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) { | |
9065 | - err = ext4_journal_get_write_access(handle, | |
9066 | - EXT4_SB(sb)->s_sbh); | |
9067 | - if (err) | |
9068 | - return err; | |
9069 | - EXT4_SET_COMPAT_FEATURE(sb, compat); | |
9070 | - sb->s_dirt = 1; | |
9071 | - handle->h_sync = 1; | |
9072 | - BUFFER_TRACE(EXT4_SB(sb)->s_sbh, | |
9073 | - "call ext4_journal_dirty_met adata"); | |
9074 | - err = ext4_journal_dirty_metadata(handle, | |
9075 | - EXT4_SB(sb)->s_sbh); | |
9076 | - } | |
9077 | - return err; | |
9078 | -} | |
9079 | - | |
9080 | -int ext4_update_rocompat_feature(handle_t *handle, | |
9081 | - struct super_block *sb, __u32 rocompat) | |
9082 | -{ | |
9083 | - int err = 0; | |
9084 | - if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) { | |
9085 | - err = ext4_journal_get_write_access(handle, | |
9086 | - EXT4_SB(sb)->s_sbh); | |
9087 | - if (err) | |
9088 | - return err; | |
9089 | - EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat); | |
9090 | - sb->s_dirt = 1; | |
9091 | - handle->h_sync = 1; | |
9092 | - BUFFER_TRACE(EXT4_SB(sb)->s_sbh, | |
9093 | - "call ext4_journal_dirty_met adata"); | |
9094 | - err = ext4_journal_dirty_metadata(handle, | |
9095 | - EXT4_SB(sb)->s_sbh); | |
9096 | - } | |
9097 | - return err; | |
9098 | -} | |
9099 | - | |
9100 | -int ext4_update_incompat_feature(handle_t *handle, | |
9101 | - struct super_block *sb, __u32 incompat) | |
9102 | -{ | |
9103 | - int err = 0; | |
9104 | - if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) { | |
9105 | - err = ext4_journal_get_write_access(handle, | |
9106 | - EXT4_SB(sb)->s_sbh); | |
9107 | - if (err) | |
9108 | - return err; | |
9109 | - EXT4_SET_INCOMPAT_FEATURE(sb, incompat); | |
9110 | - sb->s_dirt = 1; | |
9111 | - handle->h_sync = 1; | |
9112 | - BUFFER_TRACE(EXT4_SB(sb)->s_sbh, | |
9113 | - "call ext4_journal_dirty_met adata"); | |
9114 | - err = ext4_journal_dirty_metadata(handle, | |
9115 | - EXT4_SB(sb)->s_sbh); | |
9116 | - } | |
9117 | - return err; | |
9118 | -} | |
9119 | - | |
9120 | /* | |
9121 | * Open the external journal device | |
9122 | */ | |
9123 | @@ -505,13 +552,16 @@ static void ext4_put_super(struct super_ | |
9124 | ext4_xattr_put_super(sb); | |
9125 | jbd2_journal_destroy(sbi->s_journal); | |
9126 | sbi->s_journal = NULL; | |
9127 | + | |
9128 | if (!(sb->s_flags & MS_RDONLY)) { | |
9129 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
9130 | es->s_state = cpu_to_le16(sbi->s_mount_state); | |
9131 | - BUFFER_TRACE(sbi->s_sbh, "marking dirty"); | |
9132 | - mark_buffer_dirty(sbi->s_sbh); | |
9133 | ext4_commit_super(sb, es, 1); | |
9134 | } | |
9135 | + if (sbi->s_proc) { | |
9136 | + remove_proc_entry("inode_readahead_blks", sbi->s_proc); | |
9137 | + remove_proc_entry(sb->s_id, ext4_proc_root); | |
9138 | + } | |
9139 | ||
9140 | for (i = 0; i < sbi->s_gdb_count; i++) | |
9141 | brelse(sbi->s_group_desc[i]); | |
9142 | @@ -520,6 +570,7 @@ static void ext4_put_super(struct super_ | |
9143 | percpu_counter_destroy(&sbi->s_freeblocks_counter); | |
9144 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
9145 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
9146 | + percpu_counter_destroy(&sbi->s_dirtyblocks_counter); | |
9147 | brelse(sbi->s_sbh); | |
9148 | #ifdef CONFIG_QUOTA | |
9149 | for (i = 0; i < MAXQUOTAS; i++) | |
9150 | @@ -562,11 +613,10 @@ static struct inode *ext4_alloc_inode(st | |
9151 | ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); | |
9152 | if (!ei) | |
9153 | return NULL; | |
9154 | -#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL | |
9155 | +#if defined(CONFIG_EXT4_FS_POSIX_ACL) || defined(CONFIG_EXT4DEV_FS_POSIX_ACL) | |
9156 | ei->i_acl = EXT4_ACL_NOT_CACHED; | |
9157 | ei->i_default_acl = EXT4_ACL_NOT_CACHED; | |
9158 | #endif | |
9159 | - ei->i_block_alloc_info = NULL; | |
9160 | ei->vfs_inode.i_version = 1; | |
9161 | ei->vfs_inode.i_data.writeback_index = 0; | |
9162 | memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache)); | |
9163 | @@ -599,7 +649,7 @@ static void init_once(void *foo) | |
9164 | struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; | |
9165 | ||
9166 | INIT_LIST_HEAD(&ei->i_orphan); | |
9167 | -#ifdef CONFIG_EXT4DEV_FS_XATTR | |
9168 | +#if defined(CONFIG_EXT4_FS_XATTR) || defined(CONFIG_EXT4DEV_FS_XATTR) | |
9169 | init_rwsem(&ei->xattr_sem); | |
9170 | #endif | |
9171 | init_rwsem(&ei->i_data_sem); | |
9172 | @@ -625,8 +675,7 @@ static void destroy_inodecache(void) | |
9173 | ||
9174 | static void ext4_clear_inode(struct inode *inode) | |
9175 | { | |
9176 | - struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info; | |
9177 | -#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL | |
9178 | +#if defined(CONFIG_EXT4_FS_POSIX_ACL) || defined(CONFIG_EXT4DEV_FS_POSIX_ACL) | |
9179 | if (EXT4_I(inode)->i_acl && | |
9180 | EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) { | |
9181 | posix_acl_release(EXT4_I(inode)->i_acl); | |
9182 | @@ -638,10 +687,7 @@ static void ext4_clear_inode(struct inod | |
9183 | EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED; | |
9184 | } | |
9185 | #endif | |
9186 | - ext4_discard_reservation(inode); | |
9187 | - EXT4_I(inode)->i_block_alloc_info = NULL; | |
9188 | - if (unlikely(rsv)) | |
9189 | - kfree(rsv); | |
9190 | + ext4_discard_preallocations(inode); | |
9191 | jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal, | |
9192 | &EXT4_I(inode)->jinode); | |
9193 | } | |
9194 | @@ -654,7 +700,7 @@ static inline void ext4_show_quota_optio | |
9195 | ||
9196 | if (sbi->s_jquota_fmt) | |
9197 | seq_printf(seq, ",jqfmt=%s", | |
9198 | - (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0"); | |
9199 | + (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0"); | |
9200 | ||
9201 | if (sbi->s_qf_names[USRQUOTA]) | |
9202 | seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]); | |
9203 | @@ -718,7 +764,7 @@ static int ext4_show_options(struct seq_ | |
9204 | seq_puts(seq, ",debug"); | |
9205 | if (test_opt(sb, OLDALLOC)) | |
9206 | seq_puts(seq, ",oldalloc"); | |
9207 | -#ifdef CONFIG_EXT4DEV_FS_XATTR | |
9208 | +#if defined(CONFIG_EXT4_FS_XATTR) || defined(CONFIG_EXT4DEV_FS_XATTR) | |
9209 | if (test_opt(sb, XATTR_USER) && | |
9210 | !(def_mount_opts & EXT4_DEFM_XATTR_USER)) | |
9211 | seq_puts(seq, ",user_xattr"); | |
9212 | @@ -727,7 +773,7 @@ static int ext4_show_options(struct seq_ | |
9213 | seq_puts(seq, ",nouser_xattr"); | |
9214 | } | |
9215 | #endif | |
9216 | -#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL | |
9217 | +#if defined(CONFIG_EXT4_FS_POSIX_ACL) || defined(CONFIG_EXT4DEV_FS_POSIX_ACL) | |
9218 | if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL)) | |
9219 | seq_puts(seq, ",acl"); | |
9220 | if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL)) | |
9221 | @@ -752,8 +798,6 @@ static int ext4_show_options(struct seq_ | |
9222 | seq_puts(seq, ",nobh"); | |
9223 | if (!test_opt(sb, EXTENTS)) | |
9224 | seq_puts(seq, ",noextents"); | |
9225 | - if (!test_opt(sb, MBALLOC)) | |
9226 | - seq_puts(seq, ",nomballoc"); | |
9227 | if (test_opt(sb, I_VERSION)) | |
9228 | seq_puts(seq, ",i_version"); | |
9229 | if (!test_opt(sb, DELALLOC)) | |
9230 | @@ -773,6 +817,10 @@ static int ext4_show_options(struct seq_ | |
9231 | else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) | |
9232 | seq_puts(seq, ",data=writeback"); | |
9233 | ||
9234 | + if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) | |
9235 | + seq_printf(seq, ",inode_readahead_blks=%u", | |
9236 | + sbi->s_inode_readahead_blks); | |
9237 | + | |
9238 | ext4_show_quota_options(seq, sb); | |
9239 | return 0; | |
9240 | } | |
9241 | @@ -822,7 +870,7 @@ static struct dentry *ext4_fh_to_parent( | |
9242 | } | |
9243 | ||
9244 | #ifdef CONFIG_QUOTA | |
9245 | -#define QTYPE2NAME(t) ((t) == USRQUOTA?"user":"group") | |
9246 | +#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group") | |
9247 | #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA)) | |
9248 | ||
9249 | static int ext4_dquot_initialize(struct inode *inode, int type); | |
9250 | @@ -896,7 +944,7 @@ static const struct export_operations ex | |
9251 | enum { | |
9252 | Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, | |
9253 | Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, | |
9254 | - Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov, | |
9255 | + Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov, | |
9256 | Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, | |
9257 | Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh, | |
9258 | Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev, | |
9259 | @@ -906,10 +954,11 @@ enum { | |
9260 | Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota, | |
9261 | Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota, | |
9262 | Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version, | |
9263 | - Opt_mballoc, Opt_nomballoc, Opt_stripe, Opt_delalloc, Opt_nodelalloc, | |
9264 | + Opt_stripe, Opt_delalloc, Opt_nodelalloc, | |
9265 | + Opt_inode_readahead_blks | |
9266 | }; | |
9267 | ||
9268 | -static match_table_t tokens = { | |
9269 | +static const match_table_t tokens = { | |
9270 | {Opt_bsd_df, "bsddf"}, | |
9271 | {Opt_minix_df, "minixdf"}, | |
9272 | {Opt_grpid, "grpid"}, | |
9273 | @@ -923,8 +972,6 @@ static match_table_t tokens = { | |
9274 | {Opt_err_panic, "errors=panic"}, | |
9275 | {Opt_err_ro, "errors=remount-ro"}, | |
9276 | {Opt_nouid32, "nouid32"}, | |
9277 | - {Opt_nocheck, "nocheck"}, | |
9278 | - {Opt_nocheck, "check=none"}, | |
9279 | {Opt_debug, "debug"}, | |
9280 | {Opt_oldalloc, "oldalloc"}, | |
9281 | {Opt_orlov, "orlov"}, | |
9282 | @@ -961,12 +1008,11 @@ static match_table_t tokens = { | |
9283 | {Opt_extents, "extents"}, | |
9284 | {Opt_noextents, "noextents"}, | |
9285 | {Opt_i_version, "i_version"}, | |
9286 | - {Opt_mballoc, "mballoc"}, | |
9287 | - {Opt_nomballoc, "nomballoc"}, | |
9288 | {Opt_stripe, "stripe=%u"}, | |
9289 | {Opt_resize, "resize"}, | |
9290 | {Opt_delalloc, "delalloc"}, | |
9291 | {Opt_nodelalloc, "nodelalloc"}, | |
9292 | + {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, | |
9293 | {Opt_err, NULL}, | |
9294 | }; | |
9295 | ||
9296 | @@ -981,7 +1027,7 @@ static ext4_fsblk_t get_sb_block(void ** | |
9297 | /*todo: use simple_strtoll with >32bit ext4 */ | |
9298 | sb_block = simple_strtoul(options, &options, 0); | |
9299 | if (*options && *options != ',') { | |
9300 | - printk("EXT4-fs: Invalid sb specification: %s\n", | |
9301 | + printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", | |
9302 | (char *) *data); | |
9303 | return 1; | |
9304 | } | |
9305 | @@ -1060,9 +1106,6 @@ static int parse_options(char *options, | |
9306 | case Opt_nouid32: | |
9307 | set_opt(sbi->s_mount_opt, NO_UID32); | |
9308 | break; | |
9309 | - case Opt_nocheck: | |
9310 | - clear_opt(sbi->s_mount_opt, CHECK); | |
9311 | - break; | |
9312 | case Opt_debug: | |
9313 | set_opt(sbi->s_mount_opt, DEBUG); | |
9314 | break; | |
9315 | @@ -1072,7 +1115,7 @@ static int parse_options(char *options, | |
9316 | case Opt_orlov: | |
9317 | clear_opt(sbi->s_mount_opt, OLDALLOC); | |
9318 | break; | |
9319 | -#ifdef CONFIG_EXT4DEV_FS_XATTR | |
9320 | +#if defined(CONFIG_EXT4_FS_XATTR) || defined(CONFIG_EXT4DEV_FS_XATTR) | |
9321 | case Opt_user_xattr: | |
9322 | set_opt(sbi->s_mount_opt, XATTR_USER); | |
9323 | break; | |
9324 | @@ -1082,10 +1125,11 @@ static int parse_options(char *options, | |
9325 | #else | |
9326 | case Opt_user_xattr: | |
9327 | case Opt_nouser_xattr: | |
9328 | - printk("EXT4 (no)user_xattr options not supported\n"); | |
9329 | + printk(KERN_ERR "EXT4 (no)user_xattr options " | |
9330 | + "not supported\n"); | |
9331 | break; | |
9332 | #endif | |
9333 | -#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL | |
9334 | +#if defined(CONFIG_EXT4_FS_POSIX_ACL) || defined(CONFIG_EXT4DEV_FS_POSIX_ACL) | |
9335 | case Opt_acl: | |
9336 | set_opt(sbi->s_mount_opt, POSIX_ACL); | |
9337 | break; | |
9338 | @@ -1095,7 +1139,8 @@ static int parse_options(char *options, | |
9339 | #else | |
9340 | case Opt_acl: | |
9341 | case Opt_noacl: | |
9342 | - printk("EXT4 (no)acl options not supported\n"); | |
9343 | + printk(KERN_ERR "EXT4 (no)acl options " | |
9344 | + "not supported\n"); | |
9345 | break; | |
9346 | #endif | |
9347 | case Opt_reservation: | |
9348 | @@ -1185,12 +1230,11 @@ static int parse_options(char *options, | |
9349 | case Opt_grpjquota: | |
9350 | qtype = GRPQUOTA; | |
9351 | set_qf_name: | |
9352 | - if ((sb_any_quota_enabled(sb) || | |
9353 | - sb_any_quota_suspended(sb)) && | |
9354 | + if (sb_any_quota_loaded(sb) && | |
9355 | !sbi->s_qf_names[qtype]) { | |
9356 | printk(KERN_ERR | |
9357 | - "EXT4-fs: Cannot change journaled " | |
9358 | - "quota options when quota turned on.\n"); | |
9359 | + "EXT4-fs: Cannot change journaled " | |
9360 | + "quota options when quota turned on.\n"); | |
9361 | return 0; | |
9362 | } | |
9363 | qname = match_strdup(&args[0]); | |
9364 | @@ -1225,8 +1269,7 @@ set_qf_name: | |
9365 | case Opt_offgrpjquota: | |
9366 | qtype = GRPQUOTA; | |
9367 | clear_qf_name: | |
9368 | - if ((sb_any_quota_enabled(sb) || | |
9369 | - sb_any_quota_suspended(sb)) && | |
9370 | + if (sb_any_quota_loaded(sb) && | |
9371 | sbi->s_qf_names[qtype]) { | |
9372 | printk(KERN_ERR "EXT4-fs: Cannot change " | |
9373 | "journaled quota options when " | |
9374 | @@ -1245,8 +1288,7 @@ clear_qf_name: | |
9375 | case Opt_jqfmt_vfsv0: | |
9376 | qfmt = QFMT_VFS_V0; | |
9377 | set_qf_format: | |
9378 | - if ((sb_any_quota_enabled(sb) || | |
9379 | - sb_any_quota_suspended(sb)) && | |
9380 | + if (sb_any_quota_loaded(sb) && | |
9381 | sbi->s_jquota_fmt != qfmt) { | |
9382 | printk(KERN_ERR "EXT4-fs: Cannot change " | |
9383 | "journaled quota options when " | |
9384 | @@ -1265,7 +1307,7 @@ set_qf_format: | |
9385 | set_opt(sbi->s_mount_opt, GRPQUOTA); | |
9386 | break; | |
9387 | case Opt_noquota: | |
9388 | - if (sb_any_quota_enabled(sb)) { | |
9389 | + if (sb_any_quota_loaded(sb)) { | |
9390 | printk(KERN_ERR "EXT4-fs: Cannot change quota " | |
9391 | "options when quota turned on.\n"); | |
9392 | return 0; | |
9393 | @@ -1357,12 +1399,6 @@ set_qf_format: | |
9394 | case Opt_nodelalloc: | |
9395 | clear_opt(sbi->s_mount_opt, DELALLOC); | |
9396 | break; | |
9397 | - case Opt_mballoc: | |
9398 | - set_opt(sbi->s_mount_opt, MBALLOC); | |
9399 | - break; | |
9400 | - case Opt_nomballoc: | |
9401 | - clear_opt(sbi->s_mount_opt, MBALLOC); | |
9402 | - break; | |
9403 | case Opt_stripe: | |
9404 | if (match_int(&args[0], &option)) | |
9405 | return 0; | |
9406 | @@ -1373,6 +1409,13 @@ set_qf_format: | |
9407 | case Opt_delalloc: | |
9408 | set_opt(sbi->s_mount_opt, DELALLOC); | |
9409 | break; | |
9410 | + case Opt_inode_readahead_blks: | |
9411 | + if (match_int(&args[0], &option)) | |
9412 | + return 0; | |
9413 | + if (option < 0 || option > (1 << 30)) | |
9414 | + return 0; | |
9415 | + sbi->s_inode_readahead_blks = option; | |
9416 | + break; | |
9417 | default: | |
9418 | printk(KERN_ERR | |
9419 | "EXT4-fs: Unrecognized mount option \"%s\" " | |
9420 | @@ -1465,7 +1508,7 @@ static int ext4_setup_super(struct super | |
9421 | ||
9422 | ext4_commit_super(sb, es, 1); | |
9423 | if (test_opt(sb, DEBUG)) | |
9424 | - printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, " | |
9425 | + printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " | |
9426 | "bpg=%lu, ipg=%lu, mo=%04lx]\n", | |
9427 | sb->s_blocksize, | |
9428 | sbi->s_groups_count, | |
9429 | @@ -1473,14 +1516,14 @@ static int ext4_setup_super(struct super | |
9430 | EXT4_INODES_PER_GROUP(sb), | |
9431 | sbi->s_mount_opt); | |
9432 | ||
9433 | - printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id); | |
9434 | if (EXT4_SB(sb)->s_journal->j_inode == NULL) { | |
9435 | char b[BDEVNAME_SIZE]; | |
9436 | ||
9437 | - printk("external journal on %s\n", | |
9438 | - bdevname(EXT4_SB(sb)->s_journal->j_dev, b)); | |
9439 | + printk(KERN_INFO "EXT4 FS on %s, external journal on %s\n", | |
9440 | + sb->s_id, bdevname(EXT4_SB(sb)->s_journal->j_dev, b)); | |
9441 | } else { | |
9442 | - printk("internal journal\n"); | |
9443 | + printk(KERN_INFO "EXT4 FS on %s, internal journal\n", | |
9444 | + sb->s_id); | |
9445 | } | |
9446 | return res; | |
9447 | } | |
9448 | @@ -1493,7 +1536,6 @@ static int ext4_fill_flex_info(struct su | |
9449 | ext4_group_t flex_group_count; | |
9450 | ext4_group_t flex_group; | |
9451 | int groups_per_flex = 0; | |
9452 | - __u64 block_bitmap = 0; | |
9453 | int i; | |
9454 | ||
9455 | if (!sbi->s_es->s_log_groups_per_flex) { | |
9456 | @@ -1512,21 +1554,18 @@ static int ext4_fill_flex_info(struct su | |
9457 | sizeof(struct flex_groups), GFP_KERNEL); | |
9458 | if (sbi->s_flex_groups == NULL) { | |
9459 | printk(KERN_ERR "EXT4-fs: not enough memory for " | |
9460 | - "%lu flex groups\n", flex_group_count); | |
9461 | + "%u flex groups\n", flex_group_count); | |
9462 | goto failed; | |
9463 | } | |
9464 | ||
9465 | - gdp = ext4_get_group_desc(sb, 1, &bh); | |
9466 | - block_bitmap = ext4_block_bitmap(sb, gdp) - 1; | |
9467 | - | |
9468 | for (i = 0; i < sbi->s_groups_count; i++) { | |
9469 | gdp = ext4_get_group_desc(sb, i, &bh); | |
9470 | ||
9471 | flex_group = ext4_flex_group(sbi, i); | |
9472 | sbi->s_flex_groups[flex_group].free_inodes += | |
9473 | - le16_to_cpu(gdp->bg_free_inodes_count); | |
9474 | + ext4_free_inodes_count(sb, gdp); | |
9475 | sbi->s_flex_groups[flex_group].free_blocks += | |
9476 | - le16_to_cpu(gdp->bg_free_blocks_count); | |
9477 | + ext4_free_blks_count(sb, gdp); | |
9478 | } | |
9479 | ||
9480 | return 1; | |
9481 | @@ -1586,7 +1625,7 @@ static int ext4_check_descriptors(struct | |
9482 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) | |
9483 | flexbg_flag = 1; | |
9484 | ||
9485 | - ext4_debug ("Checking group descriptors"); | |
9486 | + ext4_debug("Checking group descriptors"); | |
9487 | ||
9488 | for (i = 0; i < sbi->s_groups_count; i++) { | |
9489 | struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); | |
9490 | @@ -1600,29 +1639,29 @@ static int ext4_check_descriptors(struct | |
9491 | block_bitmap = ext4_block_bitmap(sb, gdp); | |
9492 | if (block_bitmap < first_block || block_bitmap > last_block) { | |
9493 | printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: " | |
9494 | - "Block bitmap for group %lu not in group " | |
9495 | - "(block %llu)!", i, block_bitmap); | |
9496 | + "Block bitmap for group %u not in group " | |
9497 | + "(block %llu)!\n", i, block_bitmap); | |
9498 | return 0; | |
9499 | } | |
9500 | inode_bitmap = ext4_inode_bitmap(sb, gdp); | |
9501 | if (inode_bitmap < first_block || inode_bitmap > last_block) { | |
9502 | printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: " | |
9503 | - "Inode bitmap for group %lu not in group " | |
9504 | - "(block %llu)!", i, inode_bitmap); | |
9505 | + "Inode bitmap for group %u not in group " | |
9506 | + "(block %llu)!\n", i, inode_bitmap); | |
9507 | return 0; | |
9508 | } | |
9509 | inode_table = ext4_inode_table(sb, gdp); | |
9510 | if (inode_table < first_block || | |
9511 | inode_table + sbi->s_itb_per_group - 1 > last_block) { | |
9512 | printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: " | |
9513 | - "Inode table for group %lu not in group " | |
9514 | - "(block %llu)!", i, inode_table); | |
9515 | + "Inode table for group %u not in group " | |
9516 | + "(block %llu)!\n", i, inode_table); | |
9517 | return 0; | |
9518 | } | |
9519 | spin_lock(sb_bgl_lock(sbi, i)); | |
9520 | if (!ext4_group_desc_csum_verify(sbi, i, gdp)) { | |
9521 | printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: " | |
9522 | - "Checksum for group %lu failed (%u!=%u)\n", | |
9523 | + "Checksum for group %u failed (%u!=%u)\n", | |
9524 | i, le16_to_cpu(ext4_group_desc_csum(sbi, i, | |
9525 | gdp)), le16_to_cpu(gdp->bg_checksum)); | |
9526 | if (!(sb->s_flags & MS_RDONLY)) { | |
9527 | @@ -1718,9 +1757,9 @@ static void ext4_orphan_cleanup(struct s | |
9528 | DQUOT_INIT(inode); | |
9529 | if (inode->i_nlink) { | |
9530 | printk(KERN_DEBUG | |
9531 | - "%s: truncating inode %lu to %Ld bytes\n", | |
9532 | + "%s: truncating inode %lu to %lld bytes\n", | |
9533 | __func__, inode->i_ino, inode->i_size); | |
9534 | - jbd_debug(2, "truncating inode %lu to %Ld bytes\n", | |
9535 | + jbd_debug(2, "truncating inode %lu to %lld bytes\n", | |
9536 | inode->i_ino, inode->i_size); | |
9537 | ext4_truncate(inode); | |
9538 | nr_truncates++; | |
9539 | @@ -1761,13 +1800,13 @@ static void ext4_orphan_cleanup(struct s | |
9540 | * | |
9541 | * Note, this does *not* consider any metadata overhead for vfs i_blocks. | |
9542 | */ | |
9543 | -static loff_t ext4_max_size(int blkbits) | |
9544 | +static loff_t ext4_max_size(int blkbits, int has_huge_files) | |
9545 | { | |
9546 | loff_t res; | |
9547 | loff_t upper_limit = MAX_LFS_FILESIZE; | |
9548 | ||
9549 | /* small i_blocks in vfs inode? */ | |
9550 | - if (sizeof(blkcnt_t) < sizeof(u64)) { | |
9551 | + if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { | |
9552 | /* | |
9553 | * CONFIG_LSF is not enabled implies the inode | |
9554 | * i_block represent total blocks in 512 bytes | |
9555 | @@ -1797,7 +1836,7 @@ static loff_t ext4_max_size(int blkbits) | |
9556 | * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. | |
9557 | * We need to be 1 filesystem block less than the 2^48 sector limit. | |
9558 | */ | |
9559 | -static loff_t ext4_max_bitmap_size(int bits) | |
9560 | +static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) | |
9561 | { | |
9562 | loff_t res = EXT4_NDIR_BLOCKS; | |
9563 | int meta_blocks; | |
9564 | @@ -1810,11 +1849,11 @@ static loff_t ext4_max_bitmap_size(int b | |
9565 | * total number of 512 bytes blocks of the file | |
9566 | */ | |
9567 | ||
9568 | - if (sizeof(blkcnt_t) < sizeof(u64)) { | |
9569 | + if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { | |
9570 | /* | |
9571 | - * CONFIG_LSF is not enabled implies the inode | |
9572 | - * i_block represent total blocks in 512 bytes | |
9573 | - * 32 == size of vfs inode i_blocks * 8 | |
9574 | + * !has_huge_files or CONFIG_LSF is not enabled | |
9575 | + * implies the inode i_block represent total blocks in | |
9576 | + * 512 bytes 32 == size of vfs inode i_blocks * 8 | |
9577 | */ | |
9578 | upper_limit = (1LL << 32) - 1; | |
9579 | ||
9580 | @@ -1918,12 +1957,13 @@ static int ext4_fill_super(struct super_ | |
9581 | unsigned long journal_devnum = 0; | |
9582 | unsigned long def_mount_opts; | |
9583 | struct inode *root; | |
9584 | + char *cp; | |
9585 | int ret = -EINVAL; | |
9586 | int blocksize; | |
9587 | int db_count; | |
9588 | int i; | |
9589 | - int needs_recovery; | |
9590 | - __le32 features; | |
9591 | + int needs_recovery, has_huge_files; | |
9592 | + int features; | |
9593 | __u64 blocks_count; | |
9594 | int err; | |
9595 | ||
9596 | @@ -1934,10 +1974,15 @@ static int ext4_fill_super(struct super_ | |
9597 | sbi->s_mount_opt = 0; | |
9598 | sbi->s_resuid = EXT4_DEF_RESUID; | |
9599 | sbi->s_resgid = EXT4_DEF_RESGID; | |
9600 | + sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; | |
9601 | sbi->s_sb_block = sb_block; | |
9602 | ||
9603 | unlock_kernel(); | |
9604 | ||
9605 | + /* Cleanup superblock name */ | |
9606 | + for (cp = sb->s_id; (cp = strchr(cp, '/'));) | |
9607 | + *cp = '!'; | |
9608 | + | |
9609 | blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); | |
9610 | if (!blocksize) { | |
9611 | printk(KERN_ERR "EXT4-fs: unable to set blocksize\n"); | |
9612 | @@ -1977,11 +2022,11 @@ static int ext4_fill_super(struct super_ | |
9613 | set_opt(sbi->s_mount_opt, GRPID); | |
9614 | if (def_mount_opts & EXT4_DEFM_UID16) | |
9615 | set_opt(sbi->s_mount_opt, NO_UID32); | |
9616 | -#ifdef CONFIG_EXT4DEV_FS_XATTR | |
9617 | +#if defined(CONFIG_EXT4_FS_XATTR) || defined(CONFIG_EXT4DEV_FS_XATTR) | |
9618 | if (def_mount_opts & EXT4_DEFM_XATTR_USER) | |
9619 | set_opt(sbi->s_mount_opt, XATTR_USER); | |
9620 | #endif | |
9621 | -#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL | |
9622 | +#if defined(CONFIG_EXT4_FS_POSIX_ACL) || defined(CONFIG_EXT4DEV_FS_POSIX_ACL) | |
9623 | if (def_mount_opts & EXT4_DEFM_ACL) | |
9624 | set_opt(sbi->s_mount_opt, POSIX_ACL); | |
9625 | #endif | |
9626 | @@ -2016,11 +2061,6 @@ static int ext4_fill_super(struct super_ | |
9627 | ext4_warning(sb, __func__, | |
9628 | "extents feature not enabled on this filesystem, " | |
9629 | "use tune2fs.\n"); | |
9630 | - /* | |
9631 | - * turn on mballoc code by default in ext4 filesystem | |
9632 | - * Use -o nomballoc to turn it off | |
9633 | - */ | |
9634 | - set_opt(sbi->s_mount_opt, MBALLOC); | |
9635 | ||
9636 | /* | |
9637 | * enable delayed allocation by default | |
9638 | @@ -2045,16 +2085,6 @@ static int ext4_fill_super(struct super_ | |
9639 | "running e2fsck is recommended\n"); | |
9640 | ||
9641 | /* | |
9642 | - * Since ext4 is still considered development code, we require | |
9643 | - * that the TEST_FILESYS flag in s->flags be set. | |
9644 | - */ | |
9645 | - if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) { | |
9646 | - printk(KERN_WARNING "EXT4-fs: %s: not marked " | |
9647 | - "OK to use with test code.\n", sb->s_id); | |
9648 | - goto failed_mount; | |
9649 | - } | |
9650 | - | |
9651 | - /* | |
9652 | * Check feature flags regardless of the revision level, since we | |
9653 | * previously didn't change the revision level when setting the flags, | |
9654 | * so there is a chance incompat flags are set on a rev 0 filesystem. | |
9655 | @@ -2062,18 +2092,22 @@ static int ext4_fill_super(struct super_ | |
9656 | features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP); | |
9657 | if (features) { | |
9658 | printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of " | |
9659 | - "unsupported optional features (%x).\n", | |
9660 | - sb->s_id, le32_to_cpu(features)); | |
9661 | + "unsupported optional features (%x).\n", sb->s_id, | |
9662 | + (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & | |
9663 | + ~EXT4_FEATURE_INCOMPAT_SUPP)); | |
9664 | goto failed_mount; | |
9665 | } | |
9666 | features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP); | |
9667 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
9668 | printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of " | |
9669 | - "unsupported optional features (%x).\n", | |
9670 | - sb->s_id, le32_to_cpu(features)); | |
9671 | + "unsupported optional features (%x).\n", sb->s_id, | |
9672 | + (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & | |
9673 | + ~EXT4_FEATURE_RO_COMPAT_SUPP)); | |
9674 | goto failed_mount; | |
9675 | } | |
9676 | - if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) { | |
9677 | + has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb, | |
9678 | + EXT4_FEATURE_RO_COMPAT_HUGE_FILE); | |
9679 | + if (has_huge_files) { | |
9680 | /* | |
9681 | * Large file size enabled file system can only be | |
9682 | * mount if kernel is build with CONFIG_LSF | |
9683 | @@ -2123,8 +2157,9 @@ static int ext4_fill_super(struct super_ | |
9684 | } | |
9685 | } | |
9686 | ||
9687 | - sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits); | |
9688 | - sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits); | |
9689 | + sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, | |
9690 | + has_huge_files); | |
9691 | + sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); | |
9692 | ||
9693 | if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { | |
9694 | sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; | |
9695 | @@ -2172,6 +2207,18 @@ static int ext4_fill_super(struct super_ | |
9696 | for (i = 0; i < 4; i++) | |
9697 | sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); | |
9698 | sbi->s_def_hash_version = es->s_def_hash_version; | |
9699 | + i = le32_to_cpu(es->s_flags); | |
9700 | + if (i & EXT2_FLAGS_UNSIGNED_HASH) | |
9701 | + sbi->s_hash_unsigned = 3; | |
9702 | + else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { | |
9703 | +#ifdef __CHAR_UNSIGNED__ | |
9704 | + es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); | |
9705 | + sbi->s_hash_unsigned = 3; | |
9706 | +#else | |
9707 | + es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); | |
9708 | +#endif | |
9709 | + sb->s_dirt = 1; | |
9710 | + } | |
9711 | ||
9712 | if (sbi->s_blocks_per_group > blocksize * 8) { | |
9713 | printk(KERN_ERR | |
9714 | @@ -2223,6 +2270,16 @@ static int ext4_fill_super(struct super_ | |
9715 | goto failed_mount; | |
9716 | } | |
9717 | ||
9718 | +#ifdef CONFIG_PROC_FS | |
9719 | + if (ext4_proc_root) | |
9720 | + sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root); | |
9721 | + | |
9722 | + if (sbi->s_proc) | |
9723 | + proc_create_data("inode_readahead_blks", 0644, sbi->s_proc, | |
9724 | + &ext4_ui_proc_fops, | |
9725 | + &sbi->s_inode_readahead_blks); | |
9726 | +#endif | |
9727 | + | |
9728 | bgl_lock_init(&sbi->s_blockgroup_lock); | |
9729 | ||
9730 | for (i = 0; i < db_count; i++) { | |
9731 | @@ -2261,24 +2318,14 @@ static int ext4_fill_super(struct super_ | |
9732 | err = percpu_counter_init(&sbi->s_dirs_counter, | |
9733 | ext4_count_dirs(sb)); | |
9734 | } | |
9735 | + if (!err) { | |
9736 | + err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0); | |
9737 | + } | |
9738 | if (err) { | |
9739 | printk(KERN_ERR "EXT4-fs: insufficient memory\n"); | |
9740 | goto failed_mount3; | |
9741 | } | |
9742 | ||
9743 | - /* per fileystem reservation list head & lock */ | |
9744 | - spin_lock_init(&sbi->s_rsv_window_lock); | |
9745 | - sbi->s_rsv_window_root = RB_ROOT; | |
9746 | - /* Add a single, static dummy reservation to the start of the | |
9747 | - * reservation window list --- it gives us a placeholder for | |
9748 | - * append-at-start-of-list which makes the allocation logic | |
9749 | - * _much_ simpler. */ | |
9750 | - sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED; | |
9751 | - sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED; | |
9752 | - sbi->s_rsv_window_head.rsv_alloc_hit = 0; | |
9753 | - sbi->s_rsv_window_head.rsv_goal_size = 0; | |
9754 | - ext4_rsv_window_add(sb, &sbi->s_rsv_window_head); | |
9755 | - | |
9756 | sbi->s_stripe = ext4_get_stripe_size(sbi); | |
9757 | ||
9758 | /* | |
9759 | @@ -2498,11 +2545,16 @@ failed_mount3: | |
9760 | percpu_counter_destroy(&sbi->s_freeblocks_counter); | |
9761 | percpu_counter_destroy(&sbi->s_freeinodes_counter); | |
9762 | percpu_counter_destroy(&sbi->s_dirs_counter); | |
9763 | + percpu_counter_destroy(&sbi->s_dirtyblocks_counter); | |
9764 | failed_mount2: | |
9765 | for (i = 0; i < db_count; i++) | |
9766 | brelse(sbi->s_group_desc[i]); | |
9767 | kfree(sbi->s_group_desc); | |
9768 | failed_mount: | |
9769 | + if (sbi->s_proc) { | |
9770 | + remove_proc_entry("inode_readahead_blks", sbi->s_proc); | |
9771 | + remove_proc_entry(sb->s_id, ext4_proc_root); | |
9772 | + } | |
9773 | #ifdef CONFIG_QUOTA | |
9774 | for (i = 0; i < MAXQUOTAS; i++) | |
9775 | kfree(sbi->s_qf_names[i]); | |
9776 | @@ -2561,7 +2613,7 @@ static journal_t *ext4_get_journal(struc | |
9777 | return NULL; | |
9778 | } | |
9779 | ||
9780 | - jbd_debug(2, "Journal inode found at %p: %Ld bytes\n", | |
9781 | + jbd_debug(2, "Journal inode found at %p: %lld bytes\n", | |
9782 | journal_inode, journal_inode->i_size); | |
9783 | if (!S_ISREG(journal_inode->i_mode)) { | |
9784 | printk(KERN_ERR "EXT4-fs: invalid journal inode.\n"); | |
9785 | @@ -2724,6 +2776,11 @@ static int ext4_load_journal(struct supe | |
9786 | return -EINVAL; | |
9787 | } | |
9788 | ||
9789 | + if (journal->j_flags & JBD2_BARRIER) | |
9790 | + printk(KERN_INFO "EXT4-fs: barriers enabled\n"); | |
9791 | + else | |
9792 | + printk(KERN_INFO "EXT4-fs: barriers disabled\n"); | |
9793 | + | |
9794 | if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) { | |
9795 | err = jbd2_journal_update_format(journal); | |
9796 | if (err) { | |
9797 | @@ -2823,8 +2880,11 @@ static void ext4_commit_super(struct sup | |
9798 | set_buffer_uptodate(sbh); | |
9799 | } | |
9800 | es->s_wtime = cpu_to_le32(get_seconds()); | |
9801 | - ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb)); | |
9802 | - es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb)); | |
9803 | + ext4_free_blocks_count_set(es, percpu_counter_sum_positive( | |
9804 | + &EXT4_SB(sb)->s_freeblocks_counter)); | |
9805 | + es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive( | |
9806 | + &EXT4_SB(sb)->s_freeinodes_counter)); | |
9807 | + | |
9808 | BUFFER_TRACE(sbh, "marking dirty"); | |
9809 | mark_buffer_dirty(sbh); | |
9810 | if (sync) { | |
9811 | @@ -2850,7 +2910,9 @@ static void ext4_mark_recovery_complete( | |
9812 | journal_t *journal = EXT4_SB(sb)->s_journal; | |
9813 | ||
9814 | jbd2_journal_lock_updates(journal); | |
9815 | - jbd2_journal_flush(journal); | |
9816 | + if (jbd2_journal_flush(journal) < 0) | |
9817 | + goto out; | |
9818 | + | |
9819 | lock_super(sb); | |
9820 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) && | |
9821 | sb->s_flags & MS_RDONLY) { | |
9822 | @@ -2859,6 +2921,8 @@ static void ext4_mark_recovery_complete( | |
9823 | ext4_commit_super(sb, es, 1); | |
9824 | } | |
9825 | unlock_super(sb); | |
9826 | + | |
9827 | +out: | |
9828 | jbd2_journal_unlock_updates(journal); | |
9829 | } | |
9830 | ||
9831 | @@ -2934,6 +2998,7 @@ static int ext4_sync_fs(struct super_blo | |
9832 | { | |
9833 | int ret = 0; | |
9834 | ||
9835 | + trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait); | |
9836 | sb->s_dirt = 0; | |
9837 | if (wait) | |
9838 | ret = ext4_force_commit(sb); | |
9839 | @@ -2955,7 +3020,13 @@ static void ext4_write_super_lockfs(stru | |
9840 | ||
9841 | /* Now we set up the journal barrier. */ | |
9842 | jbd2_journal_lock_updates(journal); | |
9843 | - jbd2_journal_flush(journal); | |
9844 | + | |
9845 | + /* | |
9846 | + * We don't want to clear needs_recovery flag when we failed | |
9847 | + * to flush the journal. | |
9848 | + */ | |
9849 | + if (jbd2_journal_flush(journal) < 0) | |
9850 | + return; | |
9851 | ||
9852 | /* Journal blocked and flushed, clear needs_recovery flag. */ | |
9853 | EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); | |
9854 | @@ -3053,13 +3124,14 @@ static int ext4_remount(struct super_blo | |
9855 | ext4_mark_recovery_complete(sb, es); | |
9856 | lock_super(sb); | |
9857 | } else { | |
9858 | - __le32 ret; | |
9859 | + int ret; | |
9860 | if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb, | |
9861 | ~EXT4_FEATURE_RO_COMPAT_SUPP))) { | |
9862 | printk(KERN_WARNING "EXT4-fs: %s: couldn't " | |
9863 | "remount RDWR because of unsupported " | |
9864 | - "optional features (%x).\n", | |
9865 | - sb->s_id, le32_to_cpu(ret)); | |
9866 | + "optional features (%x).\n", sb->s_id, | |
9867 | + (le32_to_cpu(sbi->s_es->s_feature_ro_compat) & | |
9868 | + ~EXT4_FEATURE_RO_COMPAT_SUPP)); | |
9869 | err = -EROFS; | |
9870 | goto restore_opts; | |
9871 | } | |
9872 | @@ -3076,7 +3148,7 @@ static int ext4_remount(struct super_blo | |
9873 | if (!ext4_group_desc_csum_verify(sbi, g, gdp)) { | |
9874 | printk(KERN_ERR | |
9875 | "EXT4-fs: ext4_remount: " | |
9876 | - "Checksum for group %lu failed (%u!=%u)\n", | |
9877 | + "Checksum for group %u failed (%u!=%u)\n", | |
9878 | g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)), | |
9879 | le16_to_cpu(gdp->bg_checksum)); | |
9880 | err = -EINVAL; | |
9881 | @@ -3189,7 +3261,8 @@ static int ext4_statfs(struct dentry *de | |
9882 | buf->f_type = EXT4_SUPER_MAGIC; | |
9883 | buf->f_bsize = sb->s_blocksize; | |
9884 | buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last; | |
9885 | - buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter); | |
9886 | + buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) - | |
9887 | + percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter); | |
9888 | ext4_free_blocks_count_set(es, buf->f_bfree); | |
9889 | buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es); | |
9890 | if (buf->f_bfree < ext4_r_blocks_count(es)) | |
9891 | @@ -3394,8 +3467,12 @@ static int ext4_quota_on(struct super_bl | |
9892 | * otherwise be livelocked... | |
9893 | */ | |
9894 | jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); | |
9895 | - jbd2_journal_flush(EXT4_SB(sb)->s_journal); | |
9896 | + err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); | |
9897 | jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); | |
9898 | + if (err) { | |
9899 | + path_put(&nd.path); | |
9900 | + return err; | |
9901 | + } | |
9902 | } | |
9903 | ||
9904 | err = vfs_quota_on_path(sb, type, format_id, &nd.path); | |
9905 | @@ -3459,7 +3536,7 @@ static ssize_t ext4_quota_write(struct s | |
9906 | handle_t *handle = journal_current_handle(); | |
9907 | ||
9908 | if (!handle) { | |
9909 | - printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)" | |
9910 | + printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)" | |
9911 | " cancelled because transaction is not started.\n", | |
9912 | (unsigned long long)off, (unsigned long long)len); | |
9913 | return -EIO; | |
9914 | @@ -3520,18 +3597,73 @@ static int ext4_get_sb(struct file_syste | |
9915 | return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt); | |
9916 | } | |
9917 | ||
9918 | +#ifdef CONFIG_PROC_FS | |
9919 | +static int ext4_ui_proc_show(struct seq_file *m, void *v) | |
9920 | +{ | |
9921 | + unsigned int *p = m->private; | |
9922 | + | |
9923 | + seq_printf(m, "%u\n", *p); | |
9924 | + return 0; | |
9925 | +} | |
9926 | + | |
9927 | +static int ext4_ui_proc_open(struct inode *inode, struct file *file) | |
9928 | +{ | |
9929 | + return single_open(file, ext4_ui_proc_show, PDE(inode)->data); | |
9930 | +} | |
9931 | + | |
9932 | +static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf, | |
9933 | + size_t cnt, loff_t *ppos) | |
9934 | +{ | |
9935 | + unsigned long *p = PDE(file->f_path.dentry->d_inode)->data; | |
9936 | + char str[32]; | |
9937 | + | |
9938 | + if (cnt >= sizeof(str)) | |
9939 | + return -EINVAL; | |
9940 | + if (copy_from_user(str, buf, cnt)) | |
9941 | + return -EFAULT; | |
9942 | + | |
9943 | + *p = simple_strtoul(str, NULL, 0); | |
9944 | + return cnt; | |
9945 | +} | |
9946 | + | |
9947 | +const struct file_operations ext4_ui_proc_fops = { | |
9948 | + .owner = THIS_MODULE, | |
9949 | + .open = ext4_ui_proc_open, | |
9950 | + .read = seq_read, | |
9951 | + .llseek = seq_lseek, | |
9952 | + .release = single_release, | |
9953 | + .write = ext4_ui_proc_write, | |
9954 | +}; | |
9955 | +#endif | |
9956 | + | |
9957 | +static struct file_system_type ext4_fs_type = { | |
9958 | + .owner = THIS_MODULE, | |
9959 | + .name = "ext4", | |
9960 | + .get_sb = ext4_get_sb, | |
9961 | + .kill_sb = kill_block_super, | |
9962 | + .fs_flags = FS_REQUIRES_DEV, | |
9963 | +}; | |
9964 | + | |
9965 | +static int ext4dev_get_sb(struct file_system_type *fs_type, | |
9966 | + int flags, const char *dev_name, void *data, struct vfsmount *mnt) | |
9967 | +{ | |
9968 | + return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt); | |
9969 | +} | |
9970 | + | |
9971 | static struct file_system_type ext4dev_fs_type = { | |
9972 | .owner = THIS_MODULE, | |
9973 | .name = "ext4dev", | |
9974 | - .get_sb = ext4_get_sb, | |
9975 | + .get_sb = ext4dev_get_sb, | |
9976 | .kill_sb = kill_block_super, | |
9977 | .fs_flags = FS_REQUIRES_DEV, | |
9978 | }; | |
9979 | +MODULE_ALIAS("ext4dev"); | |
9980 | ||
9981 | static int __init init_ext4_fs(void) | |
9982 | { | |
9983 | int err; | |
9984 | ||
9985 | + ext4_proc_root = proc_mkdir("fs/ext4", NULL); | |
9986 | err = init_ext4_mballoc(); | |
9987 | if (err) | |
9988 | return err; | |
9989 | @@ -3542,9 +3674,14 @@ static int __init init_ext4_fs(void) | |
9990 | err = init_inodecache(); | |
9991 | if (err) | |
9992 | goto out1; | |
9993 | - err = register_filesystem(&ext4dev_fs_type); | |
9994 | + err = register_filesystem(&ext4_fs_type); | |
9995 | if (err) | |
9996 | goto out; | |
9997 | + err = register_filesystem(&ext4dev_fs_type); | |
9998 | + if (err) { | |
9999 | + unregister_filesystem(&ext4_fs_type); | |
10000 | + goto out; | |
10001 | + } | |
10002 | return 0; | |
10003 | out: | |
10004 | destroy_inodecache(); | |
10005 | @@ -3557,10 +3694,12 @@ out2: | |
10006 | ||
10007 | static void __exit exit_ext4_fs(void) | |
10008 | { | |
10009 | + unregister_filesystem(&ext4_fs_type); | |
10010 | unregister_filesystem(&ext4dev_fs_type); | |
10011 | destroy_inodecache(); | |
10012 | exit_ext4_xattr(); | |
10013 | exit_ext4_mballoc(); | |
10014 | + remove_proc_entry("fs/ext4", NULL); | |
10015 | } | |
10016 | ||
10017 | MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); | |
10018 | diff -rup b/fs/ext4//symlink.c a/fs/ext4///symlink.c | |
10019 | --- b/fs/ext4/symlink.c 2009-02-11 14:37:58.000000000 +0100 | |
10020 | +++ a/fs/ext4/symlink.c 2009-02-10 21:40:11.000000000 +0100 | |
10021 | @@ -23,10 +23,10 @@ | |
10022 | #include "ext4.h" | |
10023 | #include "xattr.h" | |
10024 | ||
10025 | -static void * ext4_follow_link(struct dentry *dentry, struct nameidata *nd) | |
10026 | +static void *ext4_follow_link(struct dentry *dentry, struct nameidata *nd) | |
10027 | { | |
10028 | struct ext4_inode_info *ei = EXT4_I(dentry->d_inode); | |
10029 | - nd_set_link(nd, (char*)ei->i_data); | |
10030 | + nd_set_link(nd, (char *) ei->i_data); | |
10031 | return NULL; | |
10032 | } | |
10033 | ||
10034 | @@ -34,7 +34,7 @@ const struct inode_operations ext4_symli | |
10035 | .readlink = generic_readlink, | |
10036 | .follow_link = page_follow_link_light, | |
10037 | .put_link = page_put_link, | |
10038 | -#ifdef CONFIG_EXT4DEV_FS_XATTR | |
10039 | +#if defined(CONFIG_EXT4_FS_XATTR) || defined(CONFIG_EXT4DEV_FS_XATTR) | |
10040 | .setxattr = generic_setxattr, | |
10041 | .getxattr = generic_getxattr, | |
10042 | .listxattr = ext4_listxattr, | |
10043 | @@ -45,7 +45,7 @@ const struct inode_operations ext4_symli | |
10044 | const struct inode_operations ext4_fast_symlink_inode_operations = { | |
10045 | .readlink = generic_readlink, | |
10046 | .follow_link = ext4_follow_link, | |
10047 | -#ifdef CONFIG_EXT4DEV_FS_XATTR | |
10048 | +#if defined(CONFIG_EXT4_FS_XATTR) || defined(CONFIG_EXT4DEV_FS_XATTR) | |
10049 | .setxattr = generic_setxattr, | |
10050 | .getxattr = generic_getxattr, | |
10051 | .listxattr = ext4_listxattr, | |
10052 | diff -rup b/fs/ext4//xattr.c a/fs/ext4///xattr.c | |
10053 | --- b/fs/ext4/xattr.c 2009-02-11 14:37:58.000000000 +0100 | |
10054 | +++ a/fs/ext4/xattr.c 2009-02-10 21:40:11.000000000 +0100 | |
10055 | @@ -99,12 +99,12 @@ static struct mb_cache *ext4_xattr_cache | |
10056 | ||
10057 | static struct xattr_handler *ext4_xattr_handler_map[] = { | |
10058 | [EXT4_XATTR_INDEX_USER] = &ext4_xattr_user_handler, | |
10059 | -#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL | |
10060 | +#if defined(CONFIG_EXT4_FS_POSIX_ACL) || defined(CONFIG_EXT4DEV_FS_POSIX_ACL) | |
10061 | [EXT4_XATTR_INDEX_POSIX_ACL_ACCESS] = &ext4_xattr_acl_access_handler, | |
10062 | [EXT4_XATTR_INDEX_POSIX_ACL_DEFAULT] = &ext4_xattr_acl_default_handler, | |
10063 | #endif | |
10064 | [EXT4_XATTR_INDEX_TRUSTED] = &ext4_xattr_trusted_handler, | |
10065 | -#ifdef CONFIG_EXT4DEV_FS_SECURITY | |
10066 | +#if defined(CONFIG_EXT4_FS_SECURITY) || defined(CONFIG_EXT4DEV_FS_SECURITY) | |
10067 | [EXT4_XATTR_INDEX_SECURITY] = &ext4_xattr_security_handler, | |
10068 | #endif | |
10069 | }; | |
10070 | @@ -112,11 +112,11 @@ static struct xattr_handler *ext4_xattr_ | |
10071 | struct xattr_handler *ext4_xattr_handlers[] = { | |
10072 | &ext4_xattr_user_handler, | |
10073 | &ext4_xattr_trusted_handler, | |
10074 | -#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL | |
10075 | +#if defined(CONFIG_EXT4_FS_POSIX_ACL) || defined(CONFIG_EXT4DEV_FS_POSIX_ACL) | |
10076 | &ext4_xattr_acl_access_handler, | |
10077 | &ext4_xattr_acl_default_handler, | |
10078 | #endif | |
10079 | -#ifdef CONFIG_EXT4DEV_FS_SECURITY | |
10080 | +#if defined(CONFIG_EXT4_FS_SECURITY) || defined(CONFIG_EXT4DEV_FS_SECURITY) | |
10081 | &ext4_xattr_security_handler, | |
10082 | #endif | |
10083 | NULL | |
10084 | @@ -810,8 +810,8 @@ inserted: | |
10085 | /* We need to allocate a new block */ | |
10086 | ext4_fsblk_t goal = ext4_group_first_block_no(sb, | |
10087 | EXT4_I(inode)->i_block_group); | |
10088 | - ext4_fsblk_t block = ext4_new_meta_block(handle, inode, | |
10089 | - goal, &error); | |
10090 | + ext4_fsblk_t block = ext4_new_meta_blocks(handle, inode, | |
10091 | + goal, NULL, &error); | |
10092 | if (error) | |
10093 | goto cleanup; | |
10094 | ea_idebug(inode, "creating block %d", block); | |
10095 | diff -rup b/fs/ext4//xattr.h a/fs/ext4///xattr.h | |
10096 | --- b/fs/ext4/xattr.h 2009-02-11 14:37:58.000000000 +0100 | |
10097 | +++ a/fs/ext4/xattr.h 2009-02-10 21:40:14.000000000 +0100 | |
10098 | @@ -51,8 +51,8 @@ struct ext4_xattr_entry { | |
10099 | (((name_len) + EXT4_XATTR_ROUND + \ | |
10100 | sizeof(struct ext4_xattr_entry)) & ~EXT4_XATTR_ROUND) | |
10101 | #define EXT4_XATTR_NEXT(entry) \ | |
10102 | - ( (struct ext4_xattr_entry *)( \ | |
10103 | - (char *)(entry) + EXT4_XATTR_LEN((entry)->e_name_len)) ) | |
10104 | + ((struct ext4_xattr_entry *)( \ | |
10105 | + (char *)(entry) + EXT4_XATTR_LEN((entry)->e_name_len))) | |
10106 | #define EXT4_XATTR_SIZE(size) \ | |
10107 | (((size) + EXT4_XATTR_ROUND) & ~EXT4_XATTR_ROUND) | |
10108 | ||
10109 | @@ -63,7 +63,7 @@ struct ext4_xattr_entry { | |
10110 | EXT4_I(inode)->i_extra_isize)) | |
10111 | #define IFIRST(hdr) ((struct ext4_xattr_entry *)((hdr)+1)) | |
10112 | ||
10113 | -# ifdef CONFIG_EXT4DEV_FS_XATTR | |
10114 | +# if defined(CONFIG_EXT4_FS_XATTR) || defined(CONFIG_EXT4DEV_FS_XATTR) | |
10115 | ||
10116 | extern struct xattr_handler ext4_xattr_user_handler; | |
10117 | extern struct xattr_handler ext4_xattr_trusted_handler; | |
10118 | @@ -88,7 +88,7 @@ extern void exit_ext4_xattr(void); | |
10119 | ||
10120 | extern struct xattr_handler *ext4_xattr_handlers[]; | |
10121 | ||
10122 | -# else /* CONFIG_EXT4DEV_FS_XATTR */ | |
10123 | +# else /* CONFIG_EXT4_FS_XATTR */ | |
10124 | ||
10125 | static inline int | |
10126 | ext4_xattr_get(struct inode *inode, int name_index, const char *name, | |
10127 | @@ -141,9 +141,9 @@ ext4_expand_extra_isize_ea(struct inode | |
10128 | ||
10129 | #define ext4_xattr_handlers NULL | |
10130 | ||
10131 | -# endif /* CONFIG_EXT4DEV_FS_XATTR */ | |
10132 | +# endif /* CONFIG_EXT4_FS_XATTR */ | |
10133 | ||
10134 | -#ifdef CONFIG_EXT4DEV_FS_SECURITY | |
10135 | +#if defined(CONFIG_EXT4_FS_SECURITY) || defined(CONFIG_EXT4DEV_FS_SECURITY) | |
10136 | extern int ext4_init_security(handle_t *handle, struct inode *inode, | |
10137 | struct inode *dir); | |
10138 | #else |