Blame view
fs/ext4/mballoc.c
139 KB
c9de560de
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 |
/* * Copyright (c) 2003-2006, Cluster File Systems, Inc, info@clusterfs.com * Written by Alex Tomas <alex@clusterfs.com> * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License version 2 as * published by the Free Software Foundation. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public Licens * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111- */ /* * mballoc.c contains the multiblocks allocation routines */ |
8f6e39a7a
|
23 |
#include "mballoc.h" |
6ba495e92
|
24 |
#include <linux/debugfs.h> |
5a0e3ad6a
|
25 |
#include <linux/slab.h> |
9bffad1ed
|
26 |
#include <trace/events/ext4.h> |
c9de560de
|
27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 |
/* * MUSTDO: * - test ext4_ext_search_left() and ext4_ext_search_right() * - search for metadata in few groups * * TODO v4: * - normalization should take into account whether file is still open * - discard preallocations if no free space left (policy?) * - don't normalize tails * - quota * - reservation for superuser * * TODO v3: * - bitmap read-ahead (proposed by Oleg Drokin aka green) * - track min/max extents in each group for better group selection * - mb_mark_used() may allocate chunk right after splitting buddy * - tree of groups sorted by number of free blocks * - error handling */ /* * The allocation request involve request for multiple number of blocks * near to the goal(block) value specified. * |
b713a5ec5
|
51 52 53 54 55 56 57 58 59 |
* During initialization phase of the allocator we decide to use the * group preallocation or inode preallocation depending on the size of * the file. The size of the file could be the resulting file size we * would have after allocation, or the current file size, which ever * is larger. If the size is less than sbi->s_mb_stream_request we * select to use the group preallocation. The default value of * s_mb_stream_request is 16 blocks. This can also be tuned via * /sys/fs/ext4/<partition>/mb_stream_req. The value is represented in * terms of number of blocks. |
c9de560de
|
60 61 |
* * The main motivation for having small file use group preallocation is to |
b713a5ec5
|
62 |
* ensure that we have small files closer together on the disk. |
c9de560de
|
63 |
* |
b713a5ec5
|
64 65 66 67 |
* First stage the allocator looks at the inode prealloc list, * ext4_inode_info->i_prealloc_list, which contains list of prealloc * spaces for this particular inode. The inode prealloc space is * represented as: |
c9de560de
|
68 69 70 |
* * pa_lstart -> the logical start block for this prealloc space * pa_pstart -> the physical start block for this prealloc space |
53accfa9f
|
71 72 |
* pa_len -> length for this prealloc space (in clusters) * pa_free -> free space available in this prealloc space (in clusters) |
c9de560de
|
73 74 75 |
* * The inode preallocation space is used looking at the _logical_ start * block. If only the logical file block falls within the range of prealloc |
caaf7a29d
|
76 77 |
* space we will consume the particular prealloc space. This makes sure that * we have contiguous physical blocks representing the file blocks |
c9de560de
|
78 79 80 81 82 83 84 |
* * The important thing to be noted in case of inode prealloc space is that * we don't modify the values associated to inode prealloc space except * pa_free. * * If we are not able to find blocks in the inode prealloc space and if we * have the group allocation flag set then we look at the locality group |
caaf7a29d
|
85 |
* prealloc space. These are per CPU prealloc list represented as |
c9de560de
|
86 87 88 89 90 91 92 |
* * ext4_sb_info.s_locality_groups[smp_processor_id()] * * The reason for having a per cpu locality group is to reduce the contention * between CPUs. It is possible to get scheduled at this point. * * The locality group prealloc space is used looking at whether we have |
25985edce
|
93 |
* enough free space (pa_free) within the prealloc space. |
c9de560de
|
94 95 96 97 98 99 100 101 102 103 104 105 |
* * If we can't allocate blocks via inode prealloc or/and locality group * prealloc then we look at the buddy cache. The buddy cache is represented * by ext4_sb_info.s_buddy_cache (struct inode) whose file offset gets * mapped to the buddy and bitmap information regarding different * groups. The buddy information is attached to buddy cache inode so that * we can access them through the page cache. The information regarding * each group is loaded via ext4_mb_load_buddy. The information involve * block bitmap and buddy information. The information are stored in the * inode as: * * { page } |
c3a326a65
|
106 |
* [ group 0 bitmap][ group 0 buddy] [group 1][ group 1]... |
c9de560de
|
107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 |
* * * one block each for bitmap and buddy information. So for each group we * take up 2 blocks. A page can contain blocks_per_page (PAGE_CACHE_SIZE / * blocksize) blocks. So it can have information regarding groups_per_page * which is blocks_per_page/2 * * The buddy cache inode is not stored on disk. The inode is thrown * away when the filesystem is unmounted. * * We look for count number of blocks in the buddy cache. If we were able * to locate that many free blocks we return with additional information * regarding rest of the contiguous physical block available * * Before allocating blocks via buddy cache we normalize the request * blocks. This ensure we ask for more blocks that we needed. The extra * blocks that we get after allocation is added to the respective prealloc * list. In case of inode preallocation we follow a list of heuristics * based on file size. This can be found in ext4_mb_normalize_request. If * we are doing a group prealloc we try to normalize the request to |
27baebb84
|
127 128 |
* sbi->s_mb_group_prealloc. The default value of s_mb_group_prealloc is * dependent on the cluster size; for non-bigalloc file systems, it is |
c9de560de
|
129 |
* 512 blocks. This can be tuned via |
d7a1fee13
|
130 |
* /sys/fs/ext4/<partition>/mb_group_prealloc. The value is represented in |
c9de560de
|
131 132 |
* terms of number of blocks. If we have mounted the file system with -O * stripe=<value> option the group prealloc request is normalized to the |
d7a1fee13
|
133 134 |
* the smallest multiple of the stripe value (sbi->s_stripe) which is * greater than the default mb_group_prealloc. |
c9de560de
|
135 |
* |
d7a1fee13
|
136 |
* The regular allocator (using the buddy cache) supports a few tunables. |
c9de560de
|
137 |
* |
b713a5ec5
|
138 139 140 |
* /sys/fs/ext4/<partition>/mb_min_to_scan * /sys/fs/ext4/<partition>/mb_max_to_scan * /sys/fs/ext4/<partition>/mb_order2_req |
c9de560de
|
141 |
* |
b713a5ec5
|
142 |
* The regular allocator uses buddy scan only if the request len is power of |
c9de560de
|
143 144 |
* 2 blocks and the order of allocation is >= sbi->s_mb_order2_reqs. The * value of s_mb_order2_reqs can be tuned via |
b713a5ec5
|
145 |
* /sys/fs/ext4/<partition>/mb_order2_req. If the request len is equal to |
af901ca18
|
146 |
* stripe size (sbi->s_stripe), we try to search for contiguous block in |
b713a5ec5
|
147 148 149 |
* stripe size. This should result in better allocation on RAID setups. If * not, we search in the specific group using bitmap for best extents. The * tunable min_to_scan and max_to_scan control the behaviour here. |
c9de560de
|
150 |
* min_to_scan indicate how long the mballoc __must__ look for a best |
b713a5ec5
|
151 |
* extent and max_to_scan indicates how long the mballoc __can__ look for a |
c9de560de
|
152 153 154 |
* best extent in the found extents. Searching for the blocks starts with * the group specified as the goal value in allocation context via * ac_g_ex. Each group is first checked based on the criteria whether it |
caaf7a29d
|
155 |
* can be used for allocation. ext4_mb_good_group explains how the groups are |
c9de560de
|
156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 |
* checked. * * Both the prealloc space are getting populated as above. So for the first * request we will hit the buddy cache which will result in this prealloc * space getting filled. The prealloc space is then later used for the * subsequent request. */ /* * mballoc operates on the following data: * - on-disk bitmap * - in-core buddy (actually includes buddy and bitmap) * - preallocation descriptors (PAs) * * there are two types of preallocations: * - inode * assiged to specific inode and can be used for this inode only. * it describes part of inode's space preallocated to specific * physical blocks. any block from that preallocated can be used * independent. the descriptor just tracks number of blocks left * unused. so, before taking some block from descriptor, one must * make sure corresponded logical block isn't allocated yet. this * also means that freeing any block within descriptor's range * must discard all preallocated blocks. * - locality group * assigned to specific locality group which does not translate to * permanent set of inodes: inode can join and leave group. space * from this type of preallocation can be used for any inode. thus * it's consumed from the beginning to the end. * * relation between them can be expressed as: * in-core buddy = on-disk bitmap + preallocation descriptors * * this mean blocks mballoc considers used are: * - allocated blocks (persistent) * - preallocated blocks (non-persistent) * * consistency in mballoc world means that at any time a block is either * free or used in ALL structures. notice: "any time" should not be read * literally -- time is discrete and delimited by locks. * * to keep it simple, we don't use block numbers, instead we count number of * blocks: how many blocks marked used/free in on-disk bitmap, buddy and PA. * * all operations can be expressed as: * - init buddy: buddy = on-disk + PAs * - new PA: buddy += N; PA = N * - use inode PA: on-disk += N; PA -= N * - discard inode PA buddy -= on-disk - PA; PA = 0 * - use locality group PA on-disk += N; PA -= N * - discard locality group PA buddy -= PA; PA = 0 * note: 'buddy -= on-disk - PA' is used to show that on-disk bitmap * is used in real operation because we can't know actual used * bits from PA, only from on-disk bitmap * * if we follow this strict logic, then all operations above should be atomic. * given some of them can block, we'd have to use something like semaphores * killing performance on high-end SMP hardware. let's try to relax it using * the following knowledge: * 1) if buddy is referenced, it's already initialized * 2) while block is used in buddy and the buddy is referenced, * nobody can re-allocate that block * 3) we work on bitmaps and '+' actually means 'set bits'. if on-disk has * bit set and PA claims same block, it's OK. IOW, one can set bit in * on-disk bitmap if buddy has same bit set or/and PA covers corresponded * block * * so, now we're building a concurrency table: * - init buddy vs. * - new PA * blocks for PA are allocated in the buddy, buddy must be referenced * until PA is linked to allocation group to avoid concurrent buddy init * - use inode PA * we need to make sure that either on-disk bitmap or PA has uptodate data * given (3) we care that PA-=N operation doesn't interfere with init * - discard inode PA * the simplest way would be to have buddy initialized by the discard * - use locality group PA * again PA-=N must be serialized with init * - discard locality group PA * the simplest way would be to have buddy initialized by the discard * - new PA vs. * - use inode PA * i_data_sem serializes them * - discard inode PA * discard process must wait until PA isn't used by another process * - use locality group PA * some mutex should serialize them * - discard locality group PA * discard process must wait until PA isn't used by another process * - use inode PA * - use inode PA * i_data_sem or another mutex should serializes them * - discard inode PA * discard process must wait until PA isn't used by another process * - use locality group PA * nothing wrong here -- they're different PAs covering different blocks * - discard locality group PA * discard process must wait until PA isn't used by another process * * now we're ready to make few consequences: * - PA is referenced and while it is no discard is possible * - PA is referenced until block isn't marked in on-disk bitmap * - PA changes only after on-disk bitmap * - discard must not compete with init. either init is done before * any discard or they're serialized somehow * - buddy init as sum of on-disk bitmap and PAs is done atomically * * a special case when we've used PA to emptiness. no need to modify buddy * in this case, but we should care about concurrent init * */ /* * Logic in few words: * * - allocation: * load group * find blocks * mark bits in on-disk bitmap * release group * * - use preallocation: * find proper PA (per-inode or group) * load group * mark bits in on-disk bitmap * release group * release PA * * - free: * load group * mark bits in on-disk bitmap * release group * * - discard preallocations in group: * mark PAs deleted * move them onto local list * load on-disk bitmap * load group * remove PA from object (inode or locality group) * mark free blocks in-core * * - discard inode's preallocations: */ /* * Locking rules * * Locks: * - bitlock on a group (group) * - object (inode/locality) (object) * - per-pa lock (pa) * * Paths: * - new pa * object * group * * - find and use pa: * pa * * - release consumed pa: * pa * group * object * * - generate in-core bitmap: * group * pa * * - discard all for given object (inode, locality group): * object * pa * group * * - discard all for given group: * group * pa * group * object * */ |
c3a326a65
|
338 339 340 |
static struct kmem_cache *ext4_pspace_cachep; static struct kmem_cache *ext4_ac_cachep; static struct kmem_cache *ext4_free_ext_cachep; |
fb1813f4a
|
341 342 343 344 |
/* We create slab caches for groupinfo data structures based on the * superblock block size. There will be one per mounted filesystem for * each unique s_blocksize_bits */ |
2892c15dd
|
345 |
#define NR_GRPINFO_CACHES 8 |
fb1813f4a
|
346 |
static struct kmem_cache *ext4_groupinfo_caches[NR_GRPINFO_CACHES]; |
2892c15dd
|
347 348 349 350 351 |
static const char *ext4_groupinfo_slab_names[NR_GRPINFO_CACHES] = { "ext4_groupinfo_1k", "ext4_groupinfo_2k", "ext4_groupinfo_4k", "ext4_groupinfo_8k", "ext4_groupinfo_16k", "ext4_groupinfo_32k", "ext4_groupinfo_64k", "ext4_groupinfo_128k" }; |
c3a326a65
|
352 353 |
static void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap, ext4_group_t group); |
7a2fcbf7f
|
354 355 |
static void ext4_mb_generate_from_freelist(struct super_block *sb, void *bitmap, ext4_group_t group); |
c3a326a65
|
356 |
static void release_blocks_on_commit(journal_t *journal, transaction_t *txn); |
ffad0a44b
|
357 358 |
static inline void *mb_correct_addr_and_bit(int *bit, void *addr) { |
c9de560de
|
359 |
#if BITS_PER_LONG == 64 |
ffad0a44b
|
360 361 |
*bit += ((unsigned long) addr & 7UL) << 3; addr = (void *) ((unsigned long) addr & ~7UL); |
c9de560de
|
362 |
#elif BITS_PER_LONG == 32 |
ffad0a44b
|
363 364 |
*bit += ((unsigned long) addr & 3UL) << 3; addr = (void *) ((unsigned long) addr & ~3UL); |
c9de560de
|
365 366 367 |
#else #error "how many bits you are?!" #endif |
ffad0a44b
|
368 369 |
return addr; } |
c9de560de
|
370 371 372 373 374 375 376 |
static inline int mb_test_bit(int bit, void *addr) { /* * ext4_test_bit on architecture like powerpc * needs unsigned long aligned address */ |
ffad0a44b
|
377 |
addr = mb_correct_addr_and_bit(&bit, addr); |
c9de560de
|
378 379 380 381 382 |
return ext4_test_bit(bit, addr); } static inline void mb_set_bit(int bit, void *addr) { |
ffad0a44b
|
383 |
addr = mb_correct_addr_and_bit(&bit, addr); |
c9de560de
|
384 385 |
ext4_set_bit(bit, addr); } |
c9de560de
|
386 387 |
static inline void mb_clear_bit(int bit, void *addr) { |
ffad0a44b
|
388 |
addr = mb_correct_addr_and_bit(&bit, addr); |
c9de560de
|
389 390 |
ext4_clear_bit(bit, addr); } |
ffad0a44b
|
391 392 |
static inline int mb_find_next_zero_bit(void *addr, int max, int start) { |
e7dfb2463
|
393 |
int fix = 0, ret, tmpmax; |
ffad0a44b
|
394 |
addr = mb_correct_addr_and_bit(&fix, addr); |
e7dfb2463
|
395 |
tmpmax = max + fix; |
ffad0a44b
|
396 |
start += fix; |
e7dfb2463
|
397 398 399 400 |
ret = ext4_find_next_zero_bit(addr, tmpmax, start) - fix; if (ret > max) return max; return ret; |
ffad0a44b
|
401 402 403 404 |
} static inline int mb_find_next_bit(void *addr, int max, int start) { |
e7dfb2463
|
405 |
int fix = 0, ret, tmpmax; |
ffad0a44b
|
406 |
addr = mb_correct_addr_and_bit(&fix, addr); |
e7dfb2463
|
407 |
tmpmax = max + fix; |
ffad0a44b
|
408 |
start += fix; |
e7dfb2463
|
409 410 411 412 |
ret = ext4_find_next_bit(addr, tmpmax, start) - fix; if (ret > max) return max; return ret; |
ffad0a44b
|
413 |
} |
c9de560de
|
414 415 416 |
static void *mb_find_buddy(struct ext4_buddy *e4b, int order, int *max) { char *bb; |
c9de560de
|
417 418 419 420 421 422 423 424 425 |
BUG_ON(EXT4_MB_BITMAP(e4b) == EXT4_MB_BUDDY(e4b)); BUG_ON(max == NULL); if (order > e4b->bd_blkbits + 1) { *max = 0; return NULL; } /* at order 0 we see each particular block */ |
84b775a35
|
426 427 |
if (order == 0) { *max = 1 << (e4b->bd_blkbits + 3); |
c9de560de
|
428 |
return EXT4_MB_BITMAP(e4b); |
84b775a35
|
429 |
} |
c9de560de
|
430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 |
bb = EXT4_MB_BUDDY(e4b) + EXT4_SB(e4b->bd_sb)->s_mb_offsets[order]; *max = EXT4_SB(e4b->bd_sb)->s_mb_maxs[order]; return bb; } #ifdef DOUBLE_CHECK static void mb_free_blocks_double(struct inode *inode, struct ext4_buddy *e4b, int first, int count) { int i; struct super_block *sb = e4b->bd_sb; if (unlikely(e4b->bd_info->bb_bitmap == NULL)) return; |
bc8e67409
|
446 |
assert_spin_locked(ext4_group_lock_ptr(sb, e4b->bd_group)); |
c9de560de
|
447 448 449 |
for (i = 0; i < count; i++) { if (!mb_test_bit(first + i, e4b->bd_info->bb_bitmap)) { ext4_fsblk_t blocknr; |
5661bd686
|
450 451 |
blocknr = ext4_group_first_block_no(sb, e4b->bd_group); |
53accfa9f
|
452 |
blocknr += EXT4_C2B(EXT4_SB(sb), first + i); |
5d1b1b3f4
|
453 |
ext4_grp_locked_error(sb, e4b->bd_group, |
e29136f80
|
454 455 456 457 458 |
inode ? inode->i_ino : 0, blocknr, "freeing block already freed " "(bit %u)", first + i); |
c9de560de
|
459 460 461 462 463 464 465 466 467 468 469 |
} mb_clear_bit(first + i, e4b->bd_info->bb_bitmap); } } static void mb_mark_used_double(struct ext4_buddy *e4b, int first, int count) { int i; if (unlikely(e4b->bd_info->bb_bitmap == NULL)) return; |
bc8e67409
|
470 |
assert_spin_locked(ext4_group_lock_ptr(e4b->bd_sb, e4b->bd_group)); |
c9de560de
|
471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 |
for (i = 0; i < count; i++) { BUG_ON(mb_test_bit(first + i, e4b->bd_info->bb_bitmap)); mb_set_bit(first + i, e4b->bd_info->bb_bitmap); } } static void mb_cmp_bitmaps(struct ext4_buddy *e4b, void *bitmap) { if (memcmp(e4b->bd_info->bb_bitmap, bitmap, e4b->bd_sb->s_blocksize)) { unsigned char *b1, *b2; int i; b1 = (unsigned char *) e4b->bd_info->bb_bitmap; b2 = (unsigned char *) bitmap; for (i = 0; i < e4b->bd_sb->s_blocksize; i++) { if (b1[i] != b2[i]) { |
9d8b9ec44
|
486 487 488 489 490 |
ext4_msg(e4b->bd_sb, KERN_ERR, "corruption in group %u " "at byte %u(%u): %x in copy != %x " "on disk/prealloc", e4b->bd_group, i, i * 8, b1[i], b2[i]); |
c9de560de
|
491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 |
BUG(); } } } } #else static inline void mb_free_blocks_double(struct inode *inode, struct ext4_buddy *e4b, int first, int count) { return; } static inline void mb_mark_used_double(struct ext4_buddy *e4b, int first, int count) { return; } static inline void mb_cmp_bitmaps(struct ext4_buddy *e4b, void *bitmap) { return; } #endif #ifdef AGGRESSIVE_CHECK #define MB_CHECK_ASSERT(assert) \ do { \ if (!(assert)) { \ printk(KERN_EMERG \ "Assertion failure in %s() at %s:%d: \"%s\" ", \ function, file, line, # assert); \ BUG(); \ } \ } while (0) static int __mb_check_buddy(struct ext4_buddy *e4b, char *file, const char *function, int line) { struct super_block *sb = e4b->bd_sb; int order = e4b->bd_blkbits + 1; int max; int max2; int i; int j; int k; int count; struct ext4_group_info *grp; int fragments = 0; int fstart; struct list_head *cur; void *buddy; void *buddy2; |
c9de560de
|
544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 |
{ static int mb_check_counter; if (mb_check_counter++ % 100 != 0) return 0; } while (order > 1) { buddy = mb_find_buddy(e4b, order, &max); MB_CHECK_ASSERT(buddy); buddy2 = mb_find_buddy(e4b, order - 1, &max2); MB_CHECK_ASSERT(buddy2); MB_CHECK_ASSERT(buddy != buddy2); MB_CHECK_ASSERT(max * 2 == max2); count = 0; for (i = 0; i < max; i++) { if (mb_test_bit(i, buddy)) { /* only single bit in buddy2 may be 1 */ if (!mb_test_bit(i << 1, buddy2)) { MB_CHECK_ASSERT( mb_test_bit((i<<1)+1, buddy2)); } else if (!mb_test_bit((i << 1) + 1, buddy2)) { MB_CHECK_ASSERT( mb_test_bit(i << 1, buddy2)); } continue; } |
0a10da73e
|
572 |
/* both bits in buddy2 must be 1 */ |
c9de560de
|
573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 |
MB_CHECK_ASSERT(mb_test_bit(i << 1, buddy2)); MB_CHECK_ASSERT(mb_test_bit((i << 1) + 1, buddy2)); for (j = 0; j < (1 << order); j++) { k = (i * (1 << order)) + j; MB_CHECK_ASSERT( !mb_test_bit(k, EXT4_MB_BITMAP(e4b))); } count++; } MB_CHECK_ASSERT(e4b->bd_info->bb_counters[order] == count); order--; } fstart = -1; buddy = mb_find_buddy(e4b, 0, &max); for (i = 0; i < max; i++) { if (!mb_test_bit(i, buddy)) { MB_CHECK_ASSERT(i >= e4b->bd_info->bb_first_free); if (fstart == -1) { fragments++; fstart = i; } continue; } fstart = -1; /* check used bits only */ for (j = 0; j < e4b->bd_blkbits + 1; j++) { buddy2 = mb_find_buddy(e4b, j, &max2); k = i >> j; MB_CHECK_ASSERT(k < max2); MB_CHECK_ASSERT(mb_test_bit(k, buddy2)); } } MB_CHECK_ASSERT(!EXT4_MB_GRP_NEED_INIT(e4b->bd_info)); MB_CHECK_ASSERT(e4b->bd_info->bb_fragments == fragments); grp = ext4_get_group_info(sb, e4b->bd_group); |
c9de560de
|
611 612 613 |
list_for_each(cur, &grp->bb_prealloc_list) { ext4_group_t groupnr; struct ext4_prealloc_space *pa; |
60bd63d19
|
614 615 |
pa = list_entry(cur, struct ext4_prealloc_space, pa_group_list); ext4_get_group_no_and_offset(sb, pa->pa_pstart, &groupnr, &k); |
c9de560de
|
616 |
MB_CHECK_ASSERT(groupnr == e4b->bd_group); |
60bd63d19
|
617 |
for (i = 0; i < pa->pa_len; i++) |
c9de560de
|
618 619 620 621 622 623 |
MB_CHECK_ASSERT(mb_test_bit(k + i, buddy)); } return 0; } #undef MB_CHECK_ASSERT #define mb_check_buddy(e4b) __mb_check_buddy(e4b, \ |
46e665e9d
|
624 |
__FILE__, __func__, __LINE__) |
c9de560de
|
625 626 627 |
#else #define mb_check_buddy(e4b) #endif |
7c7860592
|
628 629 630 631 632 633 |
/* * Divide blocks started from @first with length @len into * smaller chunks with power of 2 blocks. * Clear the bits in bitmap which the blocks of the chunk(s) covered, * then increase bb_counters[] for corresponded chunk size. */ |
c9de560de
|
634 |
static void ext4_mb_mark_free_simple(struct super_block *sb, |
a36b44988
|
635 |
void *buddy, ext4_grpblk_t first, ext4_grpblk_t len, |
c9de560de
|
636 637 638 |
struct ext4_group_info *grp) { struct ext4_sb_info *sbi = EXT4_SB(sb); |
a36b44988
|
639 640 641 |
ext4_grpblk_t min; ext4_grpblk_t max; ext4_grpblk_t chunk; |
c9de560de
|
642 |
unsigned short border; |
7137d7a48
|
643 |
BUG_ON(len > EXT4_CLUSTERS_PER_GROUP(sb)); |
c9de560de
|
644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 |
border = 2 << sb->s_blocksize_bits; while (len > 0) { /* find how many blocks can be covered since this position */ max = ffs(first | border) - 1; /* find how many blocks of power 2 we need to mark */ min = fls(len) - 1; if (max < min) min = max; chunk = 1 << min; /* mark multiblock chunks only */ grp->bb_counters[min]++; if (min > 0) mb_clear_bit(first >> min, buddy + sbi->s_mb_offsets[min]); len -= chunk; first += chunk; } } |
8a57d9d61
|
668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 |
/* * Cache the order of the largest free extent we have available in this block * group. */ static void mb_set_largest_free_order(struct super_block *sb, struct ext4_group_info *grp) { int i; int bits; grp->bb_largest_free_order = -1; /* uninit */ bits = sb->s_blocksize_bits + 1; for (i = bits; i >= 0; i--) { if (grp->bb_counters[i] > 0) { grp->bb_largest_free_order = i; break; } } } |
089ceecc1
|
688 689 |
static noinline_for_stack void ext4_mb_generate_buddy(struct super_block *sb, |
c9de560de
|
690 691 692 |
void *buddy, void *bitmap, ext4_group_t group) { struct ext4_group_info *grp = ext4_get_group_info(sb, group); |
7137d7a48
|
693 |
ext4_grpblk_t max = EXT4_CLUSTERS_PER_GROUP(sb); |
a36b44988
|
694 695 696 |
ext4_grpblk_t i = 0; ext4_grpblk_t first; ext4_grpblk_t len; |
c9de560de
|
697 698 699 700 701 702 |
unsigned free = 0; unsigned fragments = 0; unsigned long long period = get_cycles(); /* initialize buddy from bitmap which is aggregation * of on-disk bitmap and preallocations */ |
ffad0a44b
|
703 |
i = mb_find_next_zero_bit(bitmap, max, 0); |
c9de560de
|
704 705 706 707 |
grp->bb_first_free = i; while (i < max) { fragments++; first = i; |
ffad0a44b
|
708 |
i = mb_find_next_bit(bitmap, max, i); |
c9de560de
|
709 710 711 712 713 714 715 |
len = i - first; free += len; if (len > 1) ext4_mb_mark_free_simple(sb, buddy, first, len, grp); else grp->bb_counters[0]++; if (i < max) |
ffad0a44b
|
716 |
i = mb_find_next_zero_bit(bitmap, max, i); |
c9de560de
|
717 718 719 720 |
} grp->bb_fragments = fragments; if (free != grp->bb_free) { |
e29136f80
|
721 |
ext4_grp_locked_error(sb, group, 0, 0, |
53accfa9f
|
722 |
"%u clusters in bitmap, %u in gd", |
e29136f80
|
723 |
free, grp->bb_free); |
e56eb6590
|
724 725 726 727 |
/* * If we intent to continue, we consider group descritor * corrupt and update bb_free using bitmap value */ |
c9de560de
|
728 729 |
grp->bb_free = free; } |
8a57d9d61
|
730 |
mb_set_largest_free_order(sb, grp); |
c9de560de
|
731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 |
clear_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &(grp->bb_state)); period = get_cycles() - period; spin_lock(&EXT4_SB(sb)->s_bal_lock); EXT4_SB(sb)->s_mb_buddies_generated++; EXT4_SB(sb)->s_mb_generation_time += period; spin_unlock(&EXT4_SB(sb)->s_bal_lock); } /* The buddy information is attached the buddy cache inode * for convenience. The information regarding each group * is loaded via ext4_mb_load_buddy. The information involve * block bitmap and buddy information. The information are * stored in the inode as * * { page } |
c3a326a65
|
748 |
* [ group 0 bitmap][ group 0 buddy] [group 1][ group 1]... |
c9de560de
|
749 750 751 752 753 754 755 |
* * * one block each for bitmap and buddy information. * So for each group we take up 2 blocks. A page can * contain blocks_per_page (PAGE_CACHE_SIZE / blocksize) blocks. * So it can have information regarding groups_per_page which * is blocks_per_page/2 |
8a57d9d61
|
756 757 758 |
* * Locking note: This routine takes the block group lock of all groups * for this page; do not hold this lock when calling this routine! |
c9de560de
|
759 760 761 762 |
*/ static int ext4_mb_init_cache(struct page *page, char *incore) { |
8df9675f8
|
763 |
ext4_group_t ngroups; |
c9de560de
|
764 765 766 767 768 769 770 771 772 773 774 775 776 |
int blocksize; int blocks_per_page; int groups_per_page; int err = 0; int i; ext4_group_t first_group; int first_block; struct super_block *sb; struct buffer_head *bhs; struct buffer_head **bh; struct inode *inode; char *data; char *bitmap; |
9b8b7d353
|
777 |
struct ext4_group_info *grinfo; |
c9de560de
|
778 |
|
6ba495e92
|
779 780 |
mb_debug(1, "init page %lu ", page->index); |
c9de560de
|
781 782 783 |
inode = page->mapping->host; sb = inode->i_sb; |
8df9675f8
|
784 |
ngroups = ext4_get_groups_count(sb); |
c9de560de
|
785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 |
blocksize = 1 << inode->i_blkbits; blocks_per_page = PAGE_CACHE_SIZE / blocksize; groups_per_page = blocks_per_page >> 1; if (groups_per_page == 0) groups_per_page = 1; /* allocate buffer_heads to read bitmaps */ if (groups_per_page > 1) { err = -ENOMEM; i = sizeof(struct buffer_head *) * groups_per_page; bh = kzalloc(i, GFP_NOFS); if (bh == NULL) goto out; } else bh = &bhs; first_group = page->index * blocks_per_page / 2; /* read all groups the page covers into the cache */ for (i = 0; i < groups_per_page; i++) { struct ext4_group_desc *desc; |
8df9675f8
|
807 |
if (first_group + i >= ngroups) |
c9de560de
|
808 |
break; |
9b8b7d353
|
809 810 811 812 813 814 815 816 817 818 819 |
grinfo = ext4_get_group_info(sb, first_group + i); /* * If page is uptodate then we came here after online resize * which added some new uninitialized group info structs, so * we must skip all initialized uptodate buddies on the page, * which may be currently in use by an allocating task. */ if (PageUptodate(page) && !EXT4_MB_GRP_NEED_INIT(grinfo)) { bh[i] = NULL; continue; } |
c9de560de
|
820 821 822 823 824 825 826 827 828 |
err = -EIO; desc = ext4_get_group_desc(sb, first_group + i, NULL); if (desc == NULL) goto out; err = -ENOMEM; bh[i] = sb_getblk(sb, ext4_block_bitmap(sb, desc)); if (bh[i] == NULL) goto out; |
2ccb5fb9f
|
829 |
if (bitmap_uptodate(bh[i])) |
c9de560de
|
830 |
continue; |
c806e68f5
|
831 |
lock_buffer(bh[i]); |
2ccb5fb9f
|
832 833 834 835 |
if (bitmap_uptodate(bh[i])) { unlock_buffer(bh[i]); continue; } |
955ce5f5b
|
836 |
ext4_lock_group(sb, first_group + i); |
c9de560de
|
837 838 839 |
if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { ext4_init_block_bitmap(sb, bh[i], first_group + i, desc); |
2ccb5fb9f
|
840 |
set_bitmap_uptodate(bh[i]); |
c9de560de
|
841 |
set_buffer_uptodate(bh[i]); |
955ce5f5b
|
842 |
ext4_unlock_group(sb, first_group + i); |
3300beda5
|
843 |
unlock_buffer(bh[i]); |
c9de560de
|
844 845 |
continue; } |
955ce5f5b
|
846 |
ext4_unlock_group(sb, first_group + i); |
2ccb5fb9f
|
847 848 849 850 851 852 853 854 855 |
if (buffer_uptodate(bh[i])) { /* * if not uninit if bh is uptodate, * bitmap is also uptodate */ set_bitmap_uptodate(bh[i]); unlock_buffer(bh[i]); continue; } |
c9de560de
|
856 |
get_bh(bh[i]); |
2ccb5fb9f
|
857 858 859 860 861 862 863 |
/* * submit the buffer_head for read. We can * safely mark the bitmap as uptodate now. * We do it here so the bitmap uptodate bit * get set with buffer lock held. */ set_bitmap_uptodate(bh[i]); |
c9de560de
|
864 865 |
bh[i]->b_end_io = end_buffer_read_sync; submit_bh(READ, bh[i]); |
6ba495e92
|
866 867 |
mb_debug(1, "read bitmap for group %u ", first_group + i); |
c9de560de
|
868 869 870 |
} /* wait for I/O completion */ |
9b8b7d353
|
871 872 873 |
for (i = 0; i < groups_per_page; i++) if (bh[i]) wait_on_buffer(bh[i]); |
c9de560de
|
874 875 |
err = -EIO; |
9b8b7d353
|
876 877 |
for (i = 0; i < groups_per_page; i++) if (bh[i] && !buffer_uptodate(bh[i])) |
c9de560de
|
878 |
goto out; |
31b481dc7
|
879 |
err = 0; |
c9de560de
|
880 881 882 |
first_block = page->index * blocks_per_page; for (i = 0; i < blocks_per_page; i++) { int group; |
c9de560de
|
883 884 |
group = (first_block + i) >> 1; |
8df9675f8
|
885 |
if (group >= ngroups) |
c9de560de
|
886 |
break; |
9b8b7d353
|
887 888 889 |
if (!bh[group - first_group]) /* skip initialized uptodate buddy */ continue; |
c9de560de
|
890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 |
/* * data carry information regarding this * particular group in the format specified * above * */ data = page_address(page) + (i * blocksize); bitmap = bh[group - first_group]->b_data; /* * We place the buddy block and bitmap block * close together */ if ((first_block + i) & 1) { /* this is block of buddy */ BUG_ON(incore == NULL); |
6ba495e92
|
906 907 |
mb_debug(1, "put buddy for group %u in page %lu/%x ", |
c9de560de
|
908 |
group, page->index, i * blocksize); |
f307333e1
|
909 |
trace_ext4_mb_buddy_bitmap_load(sb, group); |
c9de560de
|
910 911 912 |
grinfo = ext4_get_group_info(sb, group); grinfo->bb_fragments = 0; memset(grinfo->bb_counters, 0, |
1927805e6
|
913 914 |
sizeof(*grinfo->bb_counters) * (sb->s_blocksize_bits+2)); |
c9de560de
|
915 916 917 |
/* * incore got set to the group block bitmap below */ |
7a2fcbf7f
|
918 |
ext4_lock_group(sb, group); |
9b8b7d353
|
919 920 |
/* init the buddy */ memset(data, 0xff, blocksize); |
c9de560de
|
921 |
ext4_mb_generate_buddy(sb, data, incore, group); |
7a2fcbf7f
|
922 |
ext4_unlock_group(sb, group); |
c9de560de
|
923 924 925 926 |
incore = NULL; } else { /* this is block of bitmap */ BUG_ON(incore != NULL); |
6ba495e92
|
927 928 |
mb_debug(1, "put bitmap for group %u in page %lu/%x ", |
c9de560de
|
929 |
group, page->index, i * blocksize); |
f307333e1
|
930 |
trace_ext4_mb_bitmap_load(sb, group); |
c9de560de
|
931 932 933 934 935 936 937 |
/* see comments in ext4_mb_put_pa() */ ext4_lock_group(sb, group); memcpy(data, bitmap, blocksize); /* mark all preallocated blks used in in-core bitmap */ ext4_mb_generate_from_pa(sb, data, group); |
7a2fcbf7f
|
938 |
ext4_mb_generate_from_freelist(sb, data, group); |
c9de560de
|
939 940 941 942 943 944 945 946 947 948 949 950 |
ext4_unlock_group(sb, group); /* set incore so that the buddy information can be * generated using this */ incore = data; } } SetPageUptodate(page); out: if (bh) { |
9b8b7d353
|
951 |
for (i = 0; i < groups_per_page; i++) |
c9de560de
|
952 953 954 955 956 957 |
brelse(bh[i]); if (bh != &bhs) kfree(bh); } return err; } |
8a57d9d61
|
958 |
/* |
2de8807b2
|
959 960 961 962 |
* Lock the buddy and bitmap pages. This make sure other parallel init_group * on the same buddy page doesn't happen whild holding the buddy page lock. * Return locked buddy and bitmap pages on e4b struct. If buddy and bitmap * are on the same page e4b->bd_buddy_page is NULL and return value is 0. |
eee4adc70
|
963 |
*/ |
2de8807b2
|
964 965 |
static int ext4_mb_get_buddy_page_lock(struct super_block *sb, ext4_group_t group, struct ext4_buddy *e4b) |
eee4adc70
|
966 |
{ |
2de8807b2
|
967 968 |
struct inode *inode = EXT4_SB(sb)->s_buddy_cache; int block, pnum, poff; |
eee4adc70
|
969 |
int blocks_per_page; |
2de8807b2
|
970 971 972 973 |
struct page *page; e4b->bd_buddy_page = NULL; e4b->bd_bitmap_page = NULL; |
eee4adc70
|
974 975 976 977 978 979 980 981 982 |
blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize; /* * the buddy cache inode stores the block bitmap * and buddy information in consecutive blocks. * So for each group we need two blocks. */ block = group * 2; pnum = block / blocks_per_page; |
2de8807b2
|
983 984 985 986 987 988 989 990 991 992 993 |
poff = block % blocks_per_page; page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS); if (!page) return -EIO; BUG_ON(page->mapping != inode->i_mapping); e4b->bd_bitmap_page = page; e4b->bd_bitmap = page_address(page) + (poff * sb->s_blocksize); if (blocks_per_page >= 2) { /* buddy and bitmap are on the same page */ return 0; |
eee4adc70
|
994 |
} |
2de8807b2
|
995 996 997 998 999 1000 1001 1002 1003 1004 |
block++; pnum = block / blocks_per_page; poff = block % blocks_per_page; page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS); if (!page) return -EIO; BUG_ON(page->mapping != inode->i_mapping); e4b->bd_buddy_page = page; return 0; |
eee4adc70
|
1005 |
} |
2de8807b2
|
1006 |
static void ext4_mb_put_buddy_page_lock(struct ext4_buddy *e4b) |
eee4adc70
|
1007 |
{ |
2de8807b2
|
1008 1009 1010 1011 1012 1013 1014 |
if (e4b->bd_bitmap_page) { unlock_page(e4b->bd_bitmap_page); page_cache_release(e4b->bd_bitmap_page); } if (e4b->bd_buddy_page) { unlock_page(e4b->bd_buddy_page); page_cache_release(e4b->bd_buddy_page); |
eee4adc70
|
1015 |
} |
eee4adc70
|
1016 1017 1018 |
} /* |
8a57d9d61
|
1019 1020 1021 1022 |
* Locking note: This routine calls ext4_mb_init_cache(), which takes the * block group lock of all groups for this page; do not hold the BG lock when * calling this routine! */ |
b6a758ec3
|
1023 1024 1025 |
static noinline_for_stack int ext4_mb_init_group(struct super_block *sb, ext4_group_t group) { |
b6a758ec3
|
1026 |
struct ext4_group_info *this_grp; |
2de8807b2
|
1027 1028 1029 |
struct ext4_buddy e4b; struct page *page; int ret = 0; |
b6a758ec3
|
1030 1031 1032 |
mb_debug(1, "init group %u ", group); |
b6a758ec3
|
1033 1034 |
this_grp = ext4_get_group_info(sb, group); /* |
08c3a8133
|
1035 1036 1037 1038 |
* This ensures that we don't reinit the buddy cache * page which map to the group from which we are already * allocating. If we are looking at the buddy cache we would * have taken a reference using ext4_mb_load_buddy and that |
2de8807b2
|
1039 |
* would have pinned buddy page to page cache. |
b6a758ec3
|
1040 |
*/ |
2de8807b2
|
1041 1042 |
ret = ext4_mb_get_buddy_page_lock(sb, group, &e4b); if (ret || !EXT4_MB_GRP_NEED_INIT(this_grp)) { |
b6a758ec3
|
1043 1044 1045 1046 |
/* * somebody initialized the group * return without doing anything */ |
b6a758ec3
|
1047 1048 |
goto err; } |
2de8807b2
|
1049 1050 1051 1052 1053 1054 |
page = e4b.bd_bitmap_page; ret = ext4_mb_init_cache(page, NULL); if (ret) goto err; if (!PageUptodate(page)) { |
b6a758ec3
|
1055 1056 1057 1058 |
ret = -EIO; goto err; } mark_page_accessed(page); |
b6a758ec3
|
1059 |
|
2de8807b2
|
1060 |
if (e4b.bd_buddy_page == NULL) { |
b6a758ec3
|
1061 1062 1063 1064 1065 |
/* * If both the bitmap and buddy are in * the same page we don't need to force * init the buddy */ |
2de8807b2
|
1066 1067 |
ret = 0; goto err; |
b6a758ec3
|
1068 |
} |
2de8807b2
|
1069 1070 1071 1072 1073 1074 |
/* init buddy cache */ page = e4b.bd_buddy_page; ret = ext4_mb_init_cache(page, e4b.bd_bitmap); if (ret) goto err; if (!PageUptodate(page)) { |
b6a758ec3
|
1075 1076 1077 1078 1079 |
ret = -EIO; goto err; } mark_page_accessed(page); err: |
2de8807b2
|
1080 |
ext4_mb_put_buddy_page_lock(&e4b); |
b6a758ec3
|
1081 1082 |
return ret; } |
8a57d9d61
|
1083 1084 1085 1086 1087 |
/* * Locking note: This routine calls ext4_mb_init_cache(), which takes the * block group lock of all groups for this page; do not hold the BG lock when * calling this routine! */ |
4ddfef7b4
|
1088 1089 1090 |
static noinline_for_stack int ext4_mb_load_buddy(struct super_block *sb, ext4_group_t group, struct ext4_buddy *e4b) |
c9de560de
|
1091 |
{ |
c9de560de
|
1092 1093 1094 1095 1096 |
int blocks_per_page; int block; int pnum; int poff; struct page *page; |
fdf6c7a76
|
1097 |
int ret; |
920313a72
|
1098 1099 1100 |
struct ext4_group_info *grp; struct ext4_sb_info *sbi = EXT4_SB(sb); struct inode *inode = sbi->s_buddy_cache; |
c9de560de
|
1101 |
|
6ba495e92
|
1102 1103 |
mb_debug(1, "load group %u ", group); |
c9de560de
|
1104 1105 |
blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize; |
920313a72
|
1106 |
grp = ext4_get_group_info(sb, group); |
c9de560de
|
1107 1108 |
e4b->bd_blkbits = sb->s_blocksize_bits; |
529da704a
|
1109 |
e4b->bd_info = grp; |
c9de560de
|
1110 1111 1112 1113 |
e4b->bd_sb = sb; e4b->bd_group = group; e4b->bd_buddy_page = NULL; e4b->bd_bitmap_page = NULL; |
f41c07505
|
1114 |
if (unlikely(EXT4_MB_GRP_NEED_INIT(grp))) { |
f41c07505
|
1115 1116 1117 1118 1119 1120 1121 |
/* * we need full data about the group * to make a good selection */ ret = ext4_mb_init_group(sb, group); if (ret) return ret; |
f41c07505
|
1122 |
} |
c9de560de
|
1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 |
/* * the buddy cache inode stores the block bitmap * and buddy information in consecutive blocks. * So for each group we need two blocks. */ block = group * 2; pnum = block / blocks_per_page; poff = block % blocks_per_page; /* we could use find_or_create_page(), but it locks page * what we'd like to avoid in fast path ... */ page = find_get_page(inode->i_mapping, pnum); if (page == NULL || !PageUptodate(page)) { if (page) |
920313a72
|
1137 1138 1139 1140 1141 1142 1143 1144 |
/* * drop the page reference and try * to get the page with lock. If we * are not uptodate that implies * somebody just created the page but * is yet to initialize the same. So * wait for it to initialize. */ |
c9de560de
|
1145 1146 1147 1148 1149 |
page_cache_release(page); page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS); if (page) { BUG_ON(page->mapping != inode->i_mapping); if (!PageUptodate(page)) { |
fdf6c7a76
|
1150 1151 1152 1153 1154 |
ret = ext4_mb_init_cache(page, NULL); if (ret) { unlock_page(page); goto err; } |
c9de560de
|
1155 1156 1157 1158 1159 1160 |
mb_cmp_bitmaps(e4b, page_address(page) + (poff * sb->s_blocksize)); } unlock_page(page); } } |
fdf6c7a76
|
1161 1162 |
if (page == NULL || !PageUptodate(page)) { ret = -EIO; |
c9de560de
|
1163 |
goto err; |
fdf6c7a76
|
1164 |
} |
c9de560de
|
1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 |
e4b->bd_bitmap_page = page; e4b->bd_bitmap = page_address(page) + (poff * sb->s_blocksize); mark_page_accessed(page); block++; pnum = block / blocks_per_page; poff = block % blocks_per_page; page = find_get_page(inode->i_mapping, pnum); if (page == NULL || !PageUptodate(page)) { if (page) page_cache_release(page); page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS); if (page) { BUG_ON(page->mapping != inode->i_mapping); |
fdf6c7a76
|
1180 1181 1182 1183 1184 1185 1186 |
if (!PageUptodate(page)) { ret = ext4_mb_init_cache(page, e4b->bd_bitmap); if (ret) { unlock_page(page); goto err; } } |
c9de560de
|
1187 1188 1189 |
unlock_page(page); } } |
fdf6c7a76
|
1190 1191 |
if (page == NULL || !PageUptodate(page)) { ret = -EIO; |
c9de560de
|
1192 |
goto err; |
fdf6c7a76
|
1193 |
} |
c9de560de
|
1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 |
e4b->bd_buddy_page = page; e4b->bd_buddy = page_address(page) + (poff * sb->s_blocksize); mark_page_accessed(page); BUG_ON(e4b->bd_bitmap_page == NULL); BUG_ON(e4b->bd_buddy_page == NULL); return 0; err: |
26626f117
|
1204 1205 |
if (page) page_cache_release(page); |
c9de560de
|
1206 1207 1208 1209 1210 1211 |
if (e4b->bd_bitmap_page) page_cache_release(e4b->bd_bitmap_page); if (e4b->bd_buddy_page) page_cache_release(e4b->bd_buddy_page); e4b->bd_buddy = NULL; e4b->bd_bitmap = NULL; |
fdf6c7a76
|
1212 |
return ret; |
c9de560de
|
1213 |
} |
e39e07fdf
|
1214 |
static void ext4_mb_unload_buddy(struct ext4_buddy *e4b) |
c9de560de
|
1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 |
{ if (e4b->bd_bitmap_page) page_cache_release(e4b->bd_bitmap_page); if (e4b->bd_buddy_page) page_cache_release(e4b->bd_buddy_page); } static int mb_find_order_for_block(struct ext4_buddy *e4b, int block) { int order = 1; void *bb; BUG_ON(EXT4_MB_BITMAP(e4b) == EXT4_MB_BUDDY(e4b)); BUG_ON(block >= (1 << (e4b->bd_blkbits + 3))); bb = EXT4_MB_BUDDY(e4b); while (order <= e4b->bd_blkbits + 1) { block = block >> 1; if (!mb_test_bit(block, bb)) { /* this block is part of buddy of order 'order' */ return order; } bb += 1 << (e4b->bd_blkbits - order); order++; } return 0; } |
955ce5f5b
|
1243 |
static void mb_clear_bits(void *bm, int cur, int len) |
c9de560de
|
1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 |
{ __u32 *addr; len = cur + len; while (cur < len) { if ((cur & 31) == 0 && (len - cur) >= 32) { /* fast path: clear whole word at once */ addr = bm + (cur >> 3); *addr = 0; cur += 32; continue; } |
955ce5f5b
|
1256 |
mb_clear_bit(cur, bm); |
c9de560de
|
1257 1258 1259 |
cur++; } } |
c3e94d1df
|
1260 |
void ext4_set_bits(void *bm, int cur, int len) |
c9de560de
|
1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 |
{ __u32 *addr; len = cur + len; while (cur < len) { if ((cur & 31) == 0 && (len - cur) >= 32) { /* fast path: set whole word at once */ addr = bm + (cur >> 3); *addr = 0xffffffff; cur += 32; continue; } |
955ce5f5b
|
1273 |
mb_set_bit(cur, bm); |
c9de560de
|
1274 1275 1276 |
cur++; } } |
7e5a8cdd8
|
1277 |
static void mb_free_blocks(struct inode *inode, struct ext4_buddy *e4b, |
c9de560de
|
1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 |
int first, int count) { int block = 0; int max = 0; int order; void *buddy; void *buddy2; struct super_block *sb = e4b->bd_sb; BUG_ON(first + count > (sb->s_blocksize << 3)); |
bc8e67409
|
1288 |
assert_spin_locked(ext4_group_lock_ptr(sb, e4b->bd_group)); |
c9de560de
|
1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 |
mb_check_buddy(e4b); mb_free_blocks_double(inode, e4b, first, count); e4b->bd_info->bb_free += count; if (first < e4b->bd_info->bb_first_free) e4b->bd_info->bb_first_free = first; /* let's maintain fragments counter */ if (first != 0) block = !mb_test_bit(first - 1, EXT4_MB_BITMAP(e4b)); if (first + count < EXT4_SB(sb)->s_mb_maxs[0]) max = !mb_test_bit(first + count, EXT4_MB_BITMAP(e4b)); if (block && max) e4b->bd_info->bb_fragments--; else if (!block && !max) e4b->bd_info->bb_fragments++; /* let's maintain buddy itself */ while (count-- > 0) { block = first++; order = 0; if (!mb_test_bit(block, EXT4_MB_BITMAP(e4b))) { ext4_fsblk_t blocknr; |
5661bd686
|
1313 1314 |
blocknr = ext4_group_first_block_no(sb, e4b->bd_group); |
53accfa9f
|
1315 |
blocknr += EXT4_C2B(EXT4_SB(sb), block); |
5d1b1b3f4
|
1316 |
ext4_grp_locked_error(sb, e4b->bd_group, |
e29136f80
|
1317 1318 1319 1320 |
inode ? inode->i_ino : 0, blocknr, "freeing already freed block " "(bit %u)", block); |
c9de560de
|
1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 |
} mb_clear_bit(block, EXT4_MB_BITMAP(e4b)); e4b->bd_info->bb_counters[order]++; /* start of the buddy */ buddy = mb_find_buddy(e4b, order, &max); do { block &= ~1UL; if (mb_test_bit(block, buddy) || mb_test_bit(block + 1, buddy)) break; /* both the buddies are free, try to coalesce them */ buddy2 = mb_find_buddy(e4b, order + 1, &max); if (!buddy2) break; if (order > 0) { /* for special purposes, we don't set * free bits in bitmap */ mb_set_bit(block, buddy); mb_set_bit(block + 1, buddy); } e4b->bd_info->bb_counters[order]--; e4b->bd_info->bb_counters[order]--; block = block >> 1; order++; e4b->bd_info->bb_counters[order]++; mb_clear_bit(block, buddy2); buddy = buddy2; } while (1); } |
8a57d9d61
|
1357 |
mb_set_largest_free_order(sb, e4b->bd_info); |
c9de560de
|
1358 |
mb_check_buddy(e4b); |
c9de560de
|
1359 1360 1361 1362 1363 1364 1365 |
} static int mb_find_extent(struct ext4_buddy *e4b, int order, int block, int needed, struct ext4_free_extent *ex) { int next = block; int max; |
c9de560de
|
1366 |
void *buddy; |
bc8e67409
|
1367 |
assert_spin_locked(ext4_group_lock_ptr(e4b->bd_sb, e4b->bd_group)); |
c9de560de
|
1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 |
BUG_ON(ex == NULL); buddy = mb_find_buddy(e4b, order, &max); BUG_ON(buddy == NULL); BUG_ON(block >= max); if (mb_test_bit(block, buddy)) { ex->fe_len = 0; ex->fe_start = 0; ex->fe_group = 0; return 0; } /* FIXME dorp order completely ? */ if (likely(order == 0)) { /* find actual order */ order = mb_find_order_for_block(e4b, block); block = block >> order; } ex->fe_len = 1 << order; ex->fe_start = block << order; ex->fe_group = e4b->bd_group; /* calc difference from given start */ next = next - ex->fe_start; ex->fe_len -= next; ex->fe_start += next; while (needed > ex->fe_len && (buddy = mb_find_buddy(e4b, order, &max))) { if (block + 1 >= max) break; next = (block + 1) * (1 << order); if (mb_test_bit(next, EXT4_MB_BITMAP(e4b))) break; |
b051d8dc4
|
1405 |
order = mb_find_order_for_block(e4b, next); |
c9de560de
|
1406 |
|
c9de560de
|
1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 |
block = next >> order; ex->fe_len += 1 << order; } BUG_ON(ex->fe_start + ex->fe_len > (1 << (e4b->bd_blkbits + 3))); return ex->fe_len; } static int mb_mark_used(struct ext4_buddy *e4b, struct ext4_free_extent *ex) { int ord; int mlen = 0; int max = 0; int cur; int start = ex->fe_start; int len = ex->fe_len; unsigned ret = 0; int len0 = len; void *buddy; BUG_ON(start + len > (e4b->bd_sb->s_blocksize << 3)); BUG_ON(e4b->bd_group != ex->fe_group); |
bc8e67409
|
1429 |
assert_spin_locked(ext4_group_lock_ptr(e4b->bd_sb, e4b->bd_group)); |
c9de560de
|
1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 |
mb_check_buddy(e4b); mb_mark_used_double(e4b, start, len); e4b->bd_info->bb_free -= len; if (e4b->bd_info->bb_first_free == start) e4b->bd_info->bb_first_free += len; /* let's maintain fragments counter */ if (start != 0) mlen = !mb_test_bit(start - 1, EXT4_MB_BITMAP(e4b)); if (start + len < EXT4_SB(e4b->bd_sb)->s_mb_maxs[0]) max = !mb_test_bit(start + len, EXT4_MB_BITMAP(e4b)); if (mlen && max) e4b->bd_info->bb_fragments++; else if (!mlen && !max) e4b->bd_info->bb_fragments--; /* let's maintain buddy itself */ while (len) { ord = mb_find_order_for_block(e4b, start); if (((start >> ord) << ord) == start && len >= (1 << ord)) { /* the whole chunk may be allocated at once! */ mlen = 1 << ord; buddy = mb_find_buddy(e4b, ord, &max); BUG_ON((start >> ord) >= max); mb_set_bit(start >> ord, buddy); e4b->bd_info->bb_counters[ord]--; start += mlen; len -= mlen; BUG_ON(len < 0); continue; } /* store for history */ if (ret == 0) ret = len | (ord << 16); /* we have to split large buddy */ BUG_ON(ord <= 0); buddy = mb_find_buddy(e4b, ord, &max); mb_set_bit(start >> ord, buddy); e4b->bd_info->bb_counters[ord]--; ord--; cur = (start >> ord) & ~1U; buddy = mb_find_buddy(e4b, ord, &max); mb_clear_bit(cur, buddy); mb_clear_bit(cur + 1, buddy); e4b->bd_info->bb_counters[ord]++; e4b->bd_info->bb_counters[ord]++; } |
8a57d9d61
|
1482 |
mb_set_largest_free_order(e4b->bd_sb, e4b->bd_info); |
c9de560de
|
1483 |
|
c3e94d1df
|
1484 |
ext4_set_bits(EXT4_MB_BITMAP(e4b), ex->fe_start, len0); |
c9de560de
|
1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 |
mb_check_buddy(e4b); return ret; } /* * Must be called under group lock! */ static void ext4_mb_use_best_found(struct ext4_allocation_context *ac, struct ext4_buddy *e4b) { struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb); int ret; BUG_ON(ac->ac_b_ex.fe_group != e4b->bd_group); BUG_ON(ac->ac_status == AC_STATUS_FOUND); ac->ac_b_ex.fe_len = min(ac->ac_b_ex.fe_len, ac->ac_g_ex.fe_len); ac->ac_b_ex.fe_logical = ac->ac_g_ex.fe_logical; ret = mb_mark_used(e4b, &ac->ac_b_ex); /* preallocation can change ac_b_ex, thus we store actually * allocated blocks for history */ ac->ac_f_ex = ac->ac_b_ex; ac->ac_status = AC_STATUS_FOUND; ac->ac_tail = ret & 0xffff; ac->ac_buddy = ret >> 16; |
c3a326a65
|
1513 1514 1515 1516 1517 1518 1519 |
/* * take the page reference. We want the page to be pinned * so that we don't get a ext4_mb_init_cache_call for this * group until we update the bitmap. That would mean we * double allocate blocks. The reference is dropped * in ext4_mb_release_context */ |
c9de560de
|
1520 1521 1522 1523 |
ac->ac_bitmap_page = e4b->bd_bitmap_page; get_page(ac->ac_bitmap_page); ac->ac_buddy_page = e4b->bd_buddy_page; get_page(ac->ac_buddy_page); |
c9de560de
|
1524 |
/* store last allocated for subsequent stream allocation */ |
4ba74d00a
|
1525 |
if (ac->ac_flags & EXT4_MB_STREAM_ALLOC) { |
c9de560de
|
1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 |
spin_lock(&sbi->s_md_lock); sbi->s_mb_last_group = ac->ac_f_ex.fe_group; sbi->s_mb_last_start = ac->ac_f_ex.fe_start; spin_unlock(&sbi->s_md_lock); } } /* * regular allocator, for general purposes allocation */ static void ext4_mb_check_limits(struct ext4_allocation_context *ac, struct ext4_buddy *e4b, int finish_group) { struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb); struct ext4_free_extent *bex = &ac->ac_b_ex; struct ext4_free_extent *gex = &ac->ac_g_ex; struct ext4_free_extent ex; int max; |
032115fce
|
1546 1547 |
if (ac->ac_status == AC_STATUS_FOUND) return; |
c9de560de
|
1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 |
/* * We don't want to scan for a whole year */ if (ac->ac_found > sbi->s_mb_max_to_scan && !(ac->ac_flags & EXT4_MB_HINT_FIRST)) { ac->ac_status = AC_STATUS_BREAK; return; } /* * Haven't found good chunk so far, let's continue */ if (bex->fe_len < gex->fe_len) return; if ((finish_group || ac->ac_found > sbi->s_mb_min_to_scan) && bex->fe_group == e4b->bd_group) { /* recheck chunk's availability - we don't know * when it was found (within this lock-unlock * period or not) */ max = mb_find_extent(e4b, 0, bex->fe_start, gex->fe_len, &ex); if (max >= gex->fe_len) { ext4_mb_use_best_found(ac, e4b); return; } } } /* * The routine checks whether found extent is good enough. If it is, * then the extent gets marked used and flag is set to the context * to stop scanning. Otherwise, the extent is compared with the * previous found extent and if new one is better, then it's stored * in the context. Later, the best found extent will be used, if * mballoc can't find good enough extent. * * FIXME: real allocation policy is to be designed yet! */ static void ext4_mb_measure_extent(struct ext4_allocation_context *ac, struct ext4_free_extent *ex, struct ext4_buddy *e4b) { struct ext4_free_extent *bex = &ac->ac_b_ex; struct ext4_free_extent *gex = &ac->ac_g_ex; BUG_ON(ex->fe_len <= 0); |
7137d7a48
|
1594 1595 |
BUG_ON(ex->fe_len > EXT4_CLUSTERS_PER_GROUP(ac->ac_sb)); BUG_ON(ex->fe_start >= EXT4_CLUSTERS_PER_GROUP(ac->ac_sb)); |
c9de560de
|
1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 |
BUG_ON(ac->ac_status != AC_STATUS_CONTINUE); ac->ac_found++; /* * The special case - take what you catch first */ if (unlikely(ac->ac_flags & EXT4_MB_HINT_FIRST)) { *bex = *ex; ext4_mb_use_best_found(ac, e4b); return; } /* * Let's check whether the chuck is good enough */ if (ex->fe_len == gex->fe_len) { *bex = *ex; ext4_mb_use_best_found(ac, e4b); return; } /* * If this is first found extent, just store it in the context */ if (bex->fe_len == 0) { *bex = *ex; return; } /* * If new found extent is better, store it in the context */ if (bex->fe_len < gex->fe_len) { /* if the request isn't satisfied, any found extent * larger than previous best one is better */ if (ex->fe_len > bex->fe_len) *bex = *ex; } else if (ex->fe_len > gex->fe_len) { /* if the request is satisfied, then we try to find * an extent that still satisfy the request, but is * smaller than previous one */ if (ex->fe_len < bex->fe_len) *bex = *ex; } ext4_mb_check_limits(ac, e4b, 0); } |
089ceecc1
|
1644 1645 |
static noinline_for_stack int ext4_mb_try_best_found(struct ext4_allocation_context *ac, |
c9de560de
|
1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 |
struct ext4_buddy *e4b) { struct ext4_free_extent ex = ac->ac_b_ex; ext4_group_t group = ex.fe_group; int max; int err; BUG_ON(ex.fe_len <= 0); err = ext4_mb_load_buddy(ac->ac_sb, group, e4b); if (err) return err; ext4_lock_group(ac->ac_sb, group); max = mb_find_extent(e4b, 0, ex.fe_start, ex.fe_len, &ex); if (max > 0) { ac->ac_b_ex = ex; ext4_mb_use_best_found(ac, e4b); } ext4_unlock_group(ac->ac_sb, group); |
e39e07fdf
|
1667 |
ext4_mb_unload_buddy(e4b); |
c9de560de
|
1668 1669 1670 |
return 0; } |
089ceecc1
|
1671 1672 |
static noinline_for_stack int ext4_mb_find_by_goal(struct ext4_allocation_context *ac, |
c9de560de
|
1673 1674 1675 1676 1677 1678 |
struct ext4_buddy *e4b) { ext4_group_t group = ac->ac_g_ex.fe_group; int max; int err; struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb); |
c9de560de
|
1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 |
struct ext4_free_extent ex; if (!(ac->ac_flags & EXT4_MB_HINT_TRY_GOAL)) return 0; err = ext4_mb_load_buddy(ac->ac_sb, group, e4b); if (err) return err; ext4_lock_group(ac->ac_sb, group); max = mb_find_extent(e4b, 0, ac->ac_g_ex.fe_start, ac->ac_g_ex.fe_len, &ex); if (max >= ac->ac_g_ex.fe_len && ac->ac_g_ex.fe_len == sbi->s_stripe) { ext4_fsblk_t start; |
5661bd686
|
1694 1695 |
start = ext4_group_first_block_no(ac->ac_sb, e4b->bd_group) + ex.fe_start; |
c9de560de
|
1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 |
/* use do_div to get remainder (would be 64-bit modulo) */ if (do_div(start, sbi->s_stripe) == 0) { ac->ac_found++; ac->ac_b_ex = ex; ext4_mb_use_best_found(ac, e4b); } } else if (max >= ac->ac_g_ex.fe_len) { BUG_ON(ex.fe_len <= 0); BUG_ON(ex.fe_group != ac->ac_g_ex.fe_group); BUG_ON(ex.fe_start != ac->ac_g_ex.fe_start); ac->ac_found++; ac->ac_b_ex = ex; ext4_mb_use_best_found(ac, e4b); } else if (max > 0 && (ac->ac_flags & EXT4_MB_HINT_MERGE)) { /* Sometimes, caller may want to merge even small * number of blocks to an existing extent */ BUG_ON(ex.fe_len <= 0); BUG_ON(ex.fe_group != ac->ac_g_ex.fe_group); BUG_ON(ex.fe_start != ac->ac_g_ex.fe_start); ac->ac_found++; ac->ac_b_ex = ex; ext4_mb_use_best_found(ac, e4b); } ext4_unlock_group(ac->ac_sb, group); |
e39e07fdf
|
1720 |
ext4_mb_unload_buddy(e4b); |
c9de560de
|
1721 1722 1723 1724 1725 1726 1727 1728 |
return 0; } /* * The routine scans buddy structures (not bitmap!) from given order * to max order and tries to find big enough chunk to satisfy the req */ |
089ceecc1
|
1729 1730 |
static noinline_for_stack void ext4_mb_simple_scan_group(struct ext4_allocation_context *ac, |
c9de560de
|
1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 |
struct ext4_buddy *e4b) { struct super_block *sb = ac->ac_sb; struct ext4_group_info *grp = e4b->bd_info; void *buddy; int i; int k; int max; BUG_ON(ac->ac_2order <= 0); for (i = ac->ac_2order; i <= sb->s_blocksize_bits + 1; i++) { if (grp->bb_counters[i] == 0) continue; buddy = mb_find_buddy(e4b, i, &max); BUG_ON(buddy == NULL); |
ffad0a44b
|
1747 |
k = mb_find_next_zero_bit(buddy, max, 0); |
c9de560de
|
1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 |
BUG_ON(k >= max); ac->ac_found++; ac->ac_b_ex.fe_len = 1 << i; ac->ac_b_ex.fe_start = k << i; ac->ac_b_ex.fe_group = e4b->bd_group; ext4_mb_use_best_found(ac, e4b); BUG_ON(ac->ac_b_ex.fe_len != ac->ac_g_ex.fe_len); if (EXT4_SB(sb)->s_mb_stats) atomic_inc(&EXT4_SB(sb)->s_bal_2orders); break; } } /* * The routine scans the group and measures all found extents. * In order to optimize scanning, caller must pass number of * free blocks in the group, so the routine can know upper limit. */ |
089ceecc1
|
1772 1773 |
static noinline_for_stack void ext4_mb_complex_scan_group(struct ext4_allocation_context *ac, |
c9de560de
|
1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 |
struct ext4_buddy *e4b) { struct super_block *sb = ac->ac_sb; void *bitmap = EXT4_MB_BITMAP(e4b); struct ext4_free_extent ex; int i; int free; free = e4b->bd_info->bb_free; BUG_ON(free <= 0); i = e4b->bd_info->bb_first_free; while (free && ac->ac_status == AC_STATUS_CONTINUE) { |
ffad0a44b
|
1788 |
i = mb_find_next_zero_bit(bitmap, |
7137d7a48
|
1789 1790 |
EXT4_CLUSTERS_PER_GROUP(sb), i); if (i >= EXT4_CLUSTERS_PER_GROUP(sb)) { |
26346ff68
|
1791 |
/* |
e56eb6590
|
1792 |
* IF we have corrupt bitmap, we won't find any |
26346ff68
|
1793 1794 1795 |
* free blocks even though group info says we * we have free blocks */ |
e29136f80
|
1796 |
ext4_grp_locked_error(sb, e4b->bd_group, 0, 0, |
53accfa9f
|
1797 |
"%d free clusters as per " |
fde4d95ad
|
1798 |
"group info. But bitmap says 0", |
26346ff68
|
1799 |
free); |
c9de560de
|
1800 1801 1802 1803 1804 |
break; } mb_find_extent(e4b, 0, i, ac->ac_g_ex.fe_len, &ex); BUG_ON(ex.fe_len <= 0); |
26346ff68
|
1805 |
if (free < ex.fe_len) { |
e29136f80
|
1806 |
ext4_grp_locked_error(sb, e4b->bd_group, 0, 0, |
53accfa9f
|
1807 |
"%d free clusters as per " |
fde4d95ad
|
1808 |
"group info. But got %d blocks", |
26346ff68
|
1809 |
free, ex.fe_len); |
e56eb6590
|
1810 1811 1812 1813 1814 1815 |
/* * The number of free blocks differs. This mostly * indicate that the bitmap is corrupt. So exit * without claiming the space. */ break; |
26346ff68
|
1816 |
} |
c9de560de
|
1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 |
ext4_mb_measure_extent(ac, &ex, e4b); i += ex.fe_len; free -= ex.fe_len; } ext4_mb_check_limits(ac, e4b, 1); } /* * This is a special case for storages like raid5 |
506bf2d82
|
1829 |
* we try to find stripe-aligned chunks for stripe-size-multiple requests |
c9de560de
|
1830 |
*/ |
089ceecc1
|
1831 1832 |
static noinline_for_stack void ext4_mb_scan_aligned(struct ext4_allocation_context *ac, |
c9de560de
|
1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 |
struct ext4_buddy *e4b) { struct super_block *sb = ac->ac_sb; struct ext4_sb_info *sbi = EXT4_SB(sb); void *bitmap = EXT4_MB_BITMAP(e4b); struct ext4_free_extent ex; ext4_fsblk_t first_group_block; ext4_fsblk_t a; ext4_grpblk_t i; int max; BUG_ON(sbi->s_stripe == 0); /* find first stripe-aligned block in group */ |
5661bd686
|
1847 |
first_group_block = ext4_group_first_block_no(sb, e4b->bd_group); |
c9de560de
|
1848 1849 1850 |
a = first_group_block + sbi->s_stripe - 1; do_div(a, sbi->s_stripe); i = (a * sbi->s_stripe) - first_group_block; |
7137d7a48
|
1851 |
while (i < EXT4_CLUSTERS_PER_GROUP(sb)) { |
c9de560de
|
1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 |
if (!mb_test_bit(i, bitmap)) { max = mb_find_extent(e4b, 0, i, sbi->s_stripe, &ex); if (max >= sbi->s_stripe) { ac->ac_found++; ac->ac_b_ex = ex; ext4_mb_use_best_found(ac, e4b); break; } } i += sbi->s_stripe; } } |
8a57d9d61
|
1864 |
/* This is now called BEFORE we load the buddy bitmap. */ |
c9de560de
|
1865 1866 1867 1868 |
static int ext4_mb_good_group(struct ext4_allocation_context *ac, ext4_group_t group, int cr) { unsigned free, fragments; |
a4912123b
|
1869 |
int flex_size = ext4_flex_bg_size(EXT4_SB(ac->ac_sb)); |
c9de560de
|
1870 1871 1872 |
struct ext4_group_info *grp = ext4_get_group_info(ac->ac_sb, group); BUG_ON(cr < 0 || cr >= 4); |
8a57d9d61
|
1873 1874 1875 1876 1877 1878 1879 |
/* We only do this if the grp has never been initialized */ if (unlikely(EXT4_MB_GRP_NEED_INIT(grp))) { int ret = ext4_mb_init_group(ac->ac_sb, group); if (ret) return 0; } |
c9de560de
|
1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 |
free = grp->bb_free; fragments = grp->bb_fragments; if (free == 0) return 0; if (fragments == 0) return 0; switch (cr) { case 0: BUG_ON(ac->ac_2order == 0); |
c9de560de
|
1891 |
|
8a57d9d61
|
1892 1893 |
if (grp->bb_largest_free_order < ac->ac_2order) return 0; |
a4912123b
|
1894 1895 1896 1897 1898 |
/* Avoid using the first bg of a flexgroup for data files */ if ((ac->ac_flags & EXT4_MB_HINT_DATA) && (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) && ((group % flex_size) == 0)) return 0; |
8a57d9d61
|
1899 |
return 1; |
c9de560de
|
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 |
case 1: if ((free / fragments) >= ac->ac_g_ex.fe_len) return 1; break; case 2: if (free >= ac->ac_g_ex.fe_len) return 1; break; case 3: return 1; default: BUG(); } return 0; } |
4ddfef7b4
|
1916 1917 |
static noinline_for_stack int ext4_mb_regular_allocator(struct ext4_allocation_context *ac) |
c9de560de
|
1918 |
{ |
8df9675f8
|
1919 |
ext4_group_t ngroups, group, i; |
c9de560de
|
1920 1921 |
int cr; int err = 0; |
c9de560de
|
1922 1923 1924 |
struct ext4_sb_info *sbi; struct super_block *sb; struct ext4_buddy e4b; |
c9de560de
|
1925 1926 1927 |
sb = ac->ac_sb; sbi = EXT4_SB(sb); |
8df9675f8
|
1928 |
ngroups = ext4_get_groups_count(sb); |
fb0a387dc
|
1929 |
/* non-extent files are limited to low blocks/groups */ |
12e9b8920
|
1930 |
if (!(ext4_test_inode_flag(ac->ac_inode, EXT4_INODE_EXTENTS))) |
fb0a387dc
|
1931 |
ngroups = sbi->s_blockfile_groups; |
c9de560de
|
1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 |
BUG_ON(ac->ac_status == AC_STATUS_FOUND); /* first, try the goal */ err = ext4_mb_find_by_goal(ac, &e4b); if (err || ac->ac_status == AC_STATUS_FOUND) goto out; if (unlikely(ac->ac_flags & EXT4_MB_HINT_GOAL_ONLY)) goto out; /* * ac->ac2_order is set only if the fe_len is a power of 2 * if ac2_order is set we also set criteria to 0 so that we * try exact allocation using buddy. */ i = fls(ac->ac_g_ex.fe_len); ac->ac_2order = 0; /* * We search using buddy data only if the order of the request * is greater than equal to the sbi_s_mb_order2_reqs |
b713a5ec5
|
1952 |
* You can tune it via /sys/fs/ext4/<partition>/mb_order2_req |
c9de560de
|
1953 1954 1955 1956 1957 1958 1959 1960 |
*/ if (i >= sbi->s_mb_order2_reqs) { /* * This should tell if fe_len is exactly power of 2 */ if ((ac->ac_g_ex.fe_len & (~(1 << (i - 1)))) == 0) ac->ac_2order = i - 1; } |
4ba74d00a
|
1961 1962 |
/* if stream allocation is enabled, use global goal */ if (ac->ac_flags & EXT4_MB_STREAM_ALLOC) { |
c9de560de
|
1963 1964 1965 1966 1967 1968 |
/* TBD: may be hot point */ spin_lock(&sbi->s_md_lock); ac->ac_g_ex.fe_group = sbi->s_mb_last_group; ac->ac_g_ex.fe_start = sbi->s_mb_last_start; spin_unlock(&sbi->s_md_lock); } |
4ba74d00a
|
1969 |
|
c9de560de
|
1970 1971 1972 1973 1974 1975 1976 1977 1978 |
/* Let's just scan groups to find more-less suitable blocks */ cr = ac->ac_2order ? 0 : 1; /* * cr == 0 try to get exact allocation, * cr == 3 try to get anything */ repeat: for (; cr < 4 && ac->ac_status == AC_STATUS_CONTINUE; cr++) { ac->ac_criteria = cr; |
ed8f9c751
|
1979 1980 1981 1982 1983 |
/* * searching for the right group start * from the goal value specified */ group = ac->ac_g_ex.fe_group; |
8df9675f8
|
1984 |
for (i = 0; i < ngroups; group++, i++) { |
8df9675f8
|
1985 |
if (group == ngroups) |
c9de560de
|
1986 |
group = 0; |
8a57d9d61
|
1987 1988 |
/* This now checks without needing the buddy page */ if (!ext4_mb_good_group(ac, group, cr)) |
c9de560de
|
1989 |
continue; |
c9de560de
|
1990 1991 1992 1993 1994 |
err = ext4_mb_load_buddy(sb, group, &e4b); if (err) goto out; ext4_lock_group(sb, group); |
8a57d9d61
|
1995 1996 1997 1998 1999 |
/* * We need to check again after locking the * block group */ |
c9de560de
|
2000 |
if (!ext4_mb_good_group(ac, group, cr)) { |
c9de560de
|
2001 |
ext4_unlock_group(sb, group); |
e39e07fdf
|
2002 |
ext4_mb_unload_buddy(&e4b); |
c9de560de
|
2003 2004 2005 2006 |
continue; } ac->ac_groups_scanned++; |
75507efb1
|
2007 |
if (cr == 0) |
c9de560de
|
2008 |
ext4_mb_simple_scan_group(ac, &e4b); |
506bf2d82
|
2009 2010 |
else if (cr == 1 && sbi->s_stripe && !(ac->ac_g_ex.fe_len % sbi->s_stripe)) |
c9de560de
|
2011 2012 2013 2014 2015 |
ext4_mb_scan_aligned(ac, &e4b); else ext4_mb_complex_scan_group(ac, &e4b); ext4_unlock_group(sb, group); |
e39e07fdf
|
2016 |
ext4_mb_unload_buddy(&e4b); |
c9de560de
|
2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 |
if (ac->ac_status != AC_STATUS_CONTINUE) break; } } if (ac->ac_b_ex.fe_len > 0 && ac->ac_status != AC_STATUS_FOUND && !(ac->ac_flags & EXT4_MB_HINT_FIRST)) { /* * We've been searching too long. Let's try to allocate * the best chunk we've found so far */ ext4_mb_try_best_found(ac, &e4b); if (ac->ac_status != AC_STATUS_FOUND) { /* * Someone more lucky has already allocated it. * The only thing we can do is just take first * found block(s) printk(KERN_DEBUG "EXT4-fs: someone won our chunk "); */ ac->ac_b_ex.fe_group = 0; ac->ac_b_ex.fe_start = 0; ac->ac_b_ex.fe_len = 0; ac->ac_status = AC_STATUS_CONTINUE; ac->ac_flags |= EXT4_MB_HINT_FIRST; cr = 3; atomic_inc(&sbi->s_mb_lost_chunks); goto repeat; } } out: return err; } |
c9de560de
|
2052 2053 2054 |
static void *ext4_mb_seq_groups_start(struct seq_file *seq, loff_t *pos) { struct super_block *sb = seq->private; |
c9de560de
|
2055 |
ext4_group_t group; |
8df9675f8
|
2056 |
if (*pos < 0 || *pos >= ext4_get_groups_count(sb)) |
c9de560de
|
2057 |
return NULL; |
c9de560de
|
2058 |
group = *pos + 1; |
a9df9a491
|
2059 |
return (void *) ((unsigned long) group); |
c9de560de
|
2060 2061 2062 2063 2064 |
} static void *ext4_mb_seq_groups_next(struct seq_file *seq, void *v, loff_t *pos) { struct super_block *sb = seq->private; |
c9de560de
|
2065 2066 2067 |
ext4_group_t group; ++*pos; |
8df9675f8
|
2068 |
if (*pos < 0 || *pos >= ext4_get_groups_count(sb)) |
c9de560de
|
2069 2070 |
return NULL; group = *pos + 1; |
a9df9a491
|
2071 |
return (void *) ((unsigned long) group); |
c9de560de
|
2072 2073 2074 2075 2076 |
} static int ext4_mb_seq_groups_show(struct seq_file *seq, void *v) { struct super_block *sb = seq->private; |
a9df9a491
|
2077 |
ext4_group_t group = (ext4_group_t) ((unsigned long) v); |
c9de560de
|
2078 2079 2080 2081 2082 |
int i; int err; struct ext4_buddy e4b; struct sg { struct ext4_group_info info; |
a36b44988
|
2083 |
ext4_grpblk_t counters[16]; |
c9de560de
|
2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 |
} sg; group--; if (group == 0) seq_printf(seq, "#%-5s: %-5s %-5s %-5s " "[ %-5s %-5s %-5s %-5s %-5s %-5s %-5s " "%-5s %-5s %-5s %-5s %-5s %-5s %-5s ] ", "group", "free", "frags", "first", "2^0", "2^1", "2^2", "2^3", "2^4", "2^5", "2^6", "2^7", "2^8", "2^9", "2^10", "2^11", "2^12", "2^13"); i = (sb->s_blocksize_bits + 2) * sizeof(sg.info.bb_counters[0]) + sizeof(struct ext4_group_info); err = ext4_mb_load_buddy(sb, group, &e4b); if (err) { |
a9df9a491
|
2100 2101 |
seq_printf(seq, "#%-5u: I/O error ", group); |
c9de560de
|
2102 2103 2104 2105 2106 |
return 0; } ext4_lock_group(sb, group); memcpy(&sg, ext4_get_group_info(sb, group), i); ext4_unlock_group(sb, group); |
e39e07fdf
|
2107 |
ext4_mb_unload_buddy(&e4b); |
c9de560de
|
2108 |
|
a9df9a491
|
2109 |
seq_printf(seq, "#%-5u: %-5u %-5u %-5u [", group, sg.info.bb_free, |
c9de560de
|
2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 |
sg.info.bb_fragments, sg.info.bb_first_free); for (i = 0; i <= 13; i++) seq_printf(seq, " %-5u", i <= sb->s_blocksize_bits + 1 ? sg.info.bb_counters[i] : 0); seq_printf(seq, " ] "); return 0; } static void ext4_mb_seq_groups_stop(struct seq_file *seq, void *v) { } |
7f1346a9d
|
2123 |
static const struct seq_operations ext4_mb_seq_groups_ops = { |
c9de560de
|
2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 |
.start = ext4_mb_seq_groups_start, .next = ext4_mb_seq_groups_next, .stop = ext4_mb_seq_groups_stop, .show = ext4_mb_seq_groups_show, }; static int ext4_mb_seq_groups_open(struct inode *inode, struct file *file) { struct super_block *sb = PDE(inode)->data; int rc; rc = seq_open(file, &ext4_mb_seq_groups_ops); if (rc == 0) { |
a271fe852
|
2137 |
struct seq_file *m = file->private_data; |
c9de560de
|
2138 2139 2140 2141 2142 |
m->private = sb; } return rc; } |
7f1346a9d
|
2143 |
static const struct file_operations ext4_mb_seq_groups_fops = { |
c9de560de
|
2144 2145 2146 2147 2148 2149 |
.owner = THIS_MODULE, .open = ext4_mb_seq_groups_open, .read = seq_read, .llseek = seq_lseek, .release = seq_release, }; |
fb1813f4a
|
2150 2151 2152 2153 2154 2155 2156 2157 |
static struct kmem_cache *get_groupinfo_cache(int blocksize_bits) { int cache_index = blocksize_bits - EXT4_MIN_BLOCK_LOG_SIZE; struct kmem_cache *cachep = ext4_groupinfo_caches[cache_index]; BUG_ON(!cachep); return cachep; } |
5f21b0e64
|
2158 2159 |
/* Create and initialize ext4_group_info data for the given group. */ |
920313a72
|
2160 |
int ext4_mb_add_groupinfo(struct super_block *sb, ext4_group_t group, |
5f21b0e64
|
2161 2162 |
struct ext4_group_desc *desc) { |
fb1813f4a
|
2163 |
int i; |
5f21b0e64
|
2164 2165 2166 |
int metalen = 0; struct ext4_sb_info *sbi = EXT4_SB(sb); struct ext4_group_info **meta_group_info; |
fb1813f4a
|
2167 |
struct kmem_cache *cachep = get_groupinfo_cache(sb->s_blocksize_bits); |
5f21b0e64
|
2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 |
/* * First check if this group is the first of a reserved block. * If it's true, we have to allocate a new table of pointers * to ext4_group_info structures */ if (group % EXT4_DESC_PER_BLOCK(sb) == 0) { metalen = sizeof(*meta_group_info) << EXT4_DESC_PER_BLOCK_BITS(sb); meta_group_info = kmalloc(metalen, GFP_KERNEL); if (meta_group_info == NULL) { |
9d8b9ec44
|
2179 2180 |
ext4_msg(sb, KERN_ERR, "EXT4-fs: can't allocate mem " "for a buddy group"); |
5f21b0e64
|
2181 2182 2183 2184 2185 |
goto exit_meta_group_info; } sbi->s_group_info[group >> EXT4_DESC_PER_BLOCK_BITS(sb)] = meta_group_info; } |
5f21b0e64
|
2186 2187 2188 |
meta_group_info = sbi->s_group_info[group >> EXT4_DESC_PER_BLOCK_BITS(sb)]; i = group & (EXT4_DESC_PER_BLOCK(sb) - 1); |
fb1813f4a
|
2189 |
meta_group_info[i] = kmem_cache_alloc(cachep, GFP_KERNEL); |
5f21b0e64
|
2190 |
if (meta_group_info[i] == NULL) { |
9d8b9ec44
|
2191 |
ext4_msg(sb, KERN_ERR, "EXT4-fs: can't allocate buddy mem"); |
5f21b0e64
|
2192 2193 |
goto exit_group_info; } |
fb1813f4a
|
2194 |
memset(meta_group_info[i], 0, kmem_cache_size(cachep)); |
5f21b0e64
|
2195 2196 2197 2198 2199 2200 2201 2202 2203 |
set_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &(meta_group_info[i]->bb_state)); /* * initialize bb_free to be able to skip * empty groups without initialization */ if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { meta_group_info[i]->bb_free = |
cff1dfd76
|
2204 |
ext4_free_clusters_after_init(sb, group, desc); |
5f21b0e64
|
2205 2206 |
} else { meta_group_info[i]->bb_free = |
021b65bb1
|
2207 |
ext4_free_group_clusters(sb, desc); |
5f21b0e64
|
2208 2209 2210 |
} INIT_LIST_HEAD(&meta_group_info[i]->bb_prealloc_list); |
920313a72
|
2211 |
init_rwsem(&meta_group_info[i]->alloc_sem); |
64e290ec6
|
2212 |
meta_group_info[i]->bb_free_root = RB_ROOT; |
8a57d9d61
|
2213 |
meta_group_info[i]->bb_largest_free_order = -1; /* uninit */ |
5f21b0e64
|
2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 |
#ifdef DOUBLE_CHECK { struct buffer_head *bh; meta_group_info[i]->bb_bitmap = kmalloc(sb->s_blocksize, GFP_KERNEL); BUG_ON(meta_group_info[i]->bb_bitmap == NULL); bh = ext4_read_block_bitmap(sb, group); BUG_ON(bh == NULL); memcpy(meta_group_info[i]->bb_bitmap, bh->b_data, sb->s_blocksize); put_bh(bh); } #endif return 0; exit_group_info: /* If a meta_group_info table has been allocated, release it now */ |
caaf7a29d
|
2233 |
if (group % EXT4_DESC_PER_BLOCK(sb) == 0) { |
5f21b0e64
|
2234 |
kfree(sbi->s_group_info[group >> EXT4_DESC_PER_BLOCK_BITS(sb)]); |
caaf7a29d
|
2235 2236 |
sbi->s_group_info[group >> EXT4_DESC_PER_BLOCK_BITS(sb)] = NULL; } |
5f21b0e64
|
2237 2238 2239 |
exit_meta_group_info: return -ENOMEM; } /* ext4_mb_add_groupinfo */ |
c9de560de
|
2240 2241 |
static int ext4_mb_init_backend(struct super_block *sb) { |
8df9675f8
|
2242 |
ext4_group_t ngroups = ext4_get_groups_count(sb); |
c9de560de
|
2243 |
ext4_group_t i; |
c9de560de
|
2244 |
struct ext4_sb_info *sbi = EXT4_SB(sb); |
5f21b0e64
|
2245 2246 2247 2248 |
struct ext4_super_block *es = sbi->s_es; int num_meta_group_infos; int num_meta_group_infos_max; int array_size; |
5f21b0e64
|
2249 |
struct ext4_group_desc *desc; |
fb1813f4a
|
2250 |
struct kmem_cache *cachep; |
5f21b0e64
|
2251 2252 |
/* This is the number of blocks used by GDT */ |
8df9675f8
|
2253 |
num_meta_group_infos = (ngroups + EXT4_DESC_PER_BLOCK(sb) - |
5f21b0e64
|
2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 |
1) >> EXT4_DESC_PER_BLOCK_BITS(sb); /* * This is the total number of blocks used by GDT including * the number of reserved blocks for GDT. * The s_group_info array is allocated with this value * to allow a clean online resize without a complex * manipulation of pointer. * The drawback is the unused memory when no resize * occurs but it's very low in terms of pages * (see comments below) * Need to handle this properly when META_BG resizing is allowed */ num_meta_group_infos_max = num_meta_group_infos + le16_to_cpu(es->s_reserved_gdt_blocks); |
c9de560de
|
2269 |
|
5f21b0e64
|
2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 |
/* * array_size is the size of s_group_info array. We round it * to the next power of two because this approximation is done * internally by kmalloc so we can have some more memory * for free here (e.g. may be used for META_BG resize). */ array_size = 1; while (array_size < sizeof(*sbi->s_group_info) * num_meta_group_infos_max) array_size = array_size << 1; |
c9de560de
|
2280 2281 2282 |
/* An 8TB filesystem with 64-bit pointers requires a 4096 byte * kmalloc. A 128kb malloc should suffice for a 256TB filesystem. * So a two level scheme suffices for now. */ |
f18a5f21c
|
2283 |
sbi->s_group_info = ext4_kvzalloc(array_size, GFP_KERNEL); |
c9de560de
|
2284 |
if (sbi->s_group_info == NULL) { |
9d8b9ec44
|
2285 |
ext4_msg(sb, KERN_ERR, "can't allocate buddy meta group"); |
c9de560de
|
2286 2287 2288 2289 |
return -ENOMEM; } sbi->s_buddy_cache = new_inode(sb); if (sbi->s_buddy_cache == NULL) { |
9d8b9ec44
|
2290 |
ext4_msg(sb, KERN_ERR, "can't get new inode"); |
c9de560de
|
2291 2292 |
goto err_freesgi; } |
48e6061bf
|
2293 2294 2295 2296 2297 |
/* To avoid potentially colliding with an valid on-disk inode number, * use EXT4_BAD_INO for the buddy cache inode number. This inode is * not in the inode hash, so it should never be found by iget(), but * this will avoid confusion if it ever shows up during debugging. */ sbi->s_buddy_cache->i_ino = EXT4_BAD_INO; |
c9de560de
|
2298 |
EXT4_I(sbi->s_buddy_cache)->i_disksize = 0; |
8df9675f8
|
2299 |
for (i = 0; i < ngroups; i++) { |
c9de560de
|
2300 2301 |
desc = ext4_get_group_desc(sb, i, NULL); if (desc == NULL) { |
9d8b9ec44
|
2302 |
ext4_msg(sb, KERN_ERR, "can't read descriptor %u", i); |
c9de560de
|
2303 2304 |
goto err_freebuddy; } |
5f21b0e64
|
2305 2306 |
if (ext4_mb_add_groupinfo(sb, i, desc) != 0) goto err_freebuddy; |
c9de560de
|
2307 2308 2309 2310 2311 |
} return 0; err_freebuddy: |
fb1813f4a
|
2312 |
cachep = get_groupinfo_cache(sb->s_blocksize_bits); |
f1fa3342e
|
2313 |
while (i-- > 0) |
fb1813f4a
|
2314 |
kmem_cache_free(cachep, ext4_get_group_info(sb, i)); |
c9de560de
|
2315 |
i = num_meta_group_infos; |
f1fa3342e
|
2316 |
while (i-- > 0) |
c9de560de
|
2317 2318 2319 |
kfree(sbi->s_group_info[i]); iput(sbi->s_buddy_cache); err_freesgi: |
f18a5f21c
|
2320 |
ext4_kvfree(sbi->s_group_info); |
c9de560de
|
2321 2322 |
return -ENOMEM; } |
2892c15dd
|
2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 |
static void ext4_groupinfo_destroy_slabs(void) { int i; for (i = 0; i < NR_GRPINFO_CACHES; i++) { if (ext4_groupinfo_caches[i]) kmem_cache_destroy(ext4_groupinfo_caches[i]); ext4_groupinfo_caches[i] = NULL; } } static int ext4_groupinfo_create_slab(size_t size) { static DEFINE_MUTEX(ext4_grpinfo_slab_create_mutex); int slab_size; int blocksize_bits = order_base_2(size); int cache_index = blocksize_bits - EXT4_MIN_BLOCK_LOG_SIZE; struct kmem_cache *cachep; if (cache_index >= NR_GRPINFO_CACHES) return -EINVAL; if (unlikely(cache_index < 0)) cache_index = 0; mutex_lock(&ext4_grpinfo_slab_create_mutex); if (ext4_groupinfo_caches[cache_index]) { mutex_unlock(&ext4_grpinfo_slab_create_mutex); return 0; /* Already created */ } slab_size = offsetof(struct ext4_group_info, bb_counters[blocksize_bits + 2]); cachep = kmem_cache_create(ext4_groupinfo_slab_names[cache_index], slab_size, 0, SLAB_RECLAIM_ACCOUNT, NULL); |
823ba01fc
|
2360 |
ext4_groupinfo_caches[cache_index] = cachep; |
2892c15dd
|
2361 2362 |
mutex_unlock(&ext4_grpinfo_slab_create_mutex); if (!cachep) { |
9d8b9ec44
|
2363 2364 2365 |
printk(KERN_EMERG "EXT4-fs: no memory for groupinfo slab cache "); |
2892c15dd
|
2366 2367 |
return -ENOMEM; } |
2892c15dd
|
2368 2369 |
return 0; } |
c9de560de
|
2370 2371 2372 |
int ext4_mb_init(struct super_block *sb, int needs_recovery) { struct ext4_sb_info *sbi = EXT4_SB(sb); |
6be2ded1d
|
2373 |
unsigned i, j; |
c9de560de
|
2374 2375 |
unsigned offset; unsigned max; |
74767c5a2
|
2376 |
int ret; |
c9de560de
|
2377 |
|
1927805e6
|
2378 |
i = (sb->s_blocksize_bits + 2) * sizeof(*sbi->s_mb_offsets); |
c9de560de
|
2379 2380 2381 |
sbi->s_mb_offsets = kmalloc(i, GFP_KERNEL); if (sbi->s_mb_offsets == NULL) { |
fb1813f4a
|
2382 2383 |
ret = -ENOMEM; goto out; |
c9de560de
|
2384 |
} |
ff7ef329b
|
2385 |
|
1927805e6
|
2386 |
i = (sb->s_blocksize_bits + 2) * sizeof(*sbi->s_mb_maxs); |
c9de560de
|
2387 2388 |
sbi->s_mb_maxs = kmalloc(i, GFP_KERNEL); if (sbi->s_mb_maxs == NULL) { |
fb1813f4a
|
2389 2390 2391 |
ret = -ENOMEM; goto out; } |
2892c15dd
|
2392 2393 2394 |
ret = ext4_groupinfo_create_slab(sb->s_blocksize); if (ret < 0) goto out; |
c9de560de
|
2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 |
/* order 0 is regular bitmap */ sbi->s_mb_maxs[0] = sb->s_blocksize << 3; sbi->s_mb_offsets[0] = 0; i = 1; offset = 0; max = sb->s_blocksize << 2; do { sbi->s_mb_offsets[i] = offset; sbi->s_mb_maxs[i] = max; offset += 1 << (sb->s_blocksize_bits - i); max = max >> 1; i++; } while (i <= sb->s_blocksize_bits + 1); |
c9de560de
|
2410 |
spin_lock_init(&sbi->s_md_lock); |
c9de560de
|
2411 2412 2413 2414 2415 2416 2417 |
spin_lock_init(&sbi->s_bal_lock); sbi->s_mb_max_to_scan = MB_DEFAULT_MAX_TO_SCAN; sbi->s_mb_min_to_scan = MB_DEFAULT_MIN_TO_SCAN; sbi->s_mb_stats = MB_DEFAULT_STATS; sbi->s_mb_stream_request = MB_DEFAULT_STREAM_THRESHOLD; sbi->s_mb_order2_reqs = MB_DEFAULT_ORDER2_REQS; |
27baebb84
|
2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 |
/* * The default group preallocation is 512, which for 4k block * sizes translates to 2 megabytes. However for bigalloc file * systems, this is probably too big (i.e, if the cluster size * is 1 megabyte, then group preallocation size becomes half a * gigabyte!). As a default, we will keep a two megabyte * group pralloc size for cluster sizes up to 64k, and after * that, we will force a minimum group preallocation size of * 32 clusters. This translates to 8 megs when the cluster * size is 256k, and 32 megs when the cluster size is 1 meg, * which seems reasonable as a default. */ sbi->s_mb_group_prealloc = max(MB_DEFAULT_GROUP_PREALLOC >> sbi->s_cluster_bits, 32); |
d7a1fee13
|
2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 |
/* * If there is a s_stripe > 1, then we set the s_mb_group_prealloc * to the lowest multiple of s_stripe which is bigger than * the s_mb_group_prealloc as determined above. We want * the preallocation size to be an exact multiple of the * RAID stripe size so that preallocations don't fragment * the stripes. */ if (sbi->s_stripe > 1) { sbi->s_mb_group_prealloc = roundup( sbi->s_mb_group_prealloc, sbi->s_stripe); } |
c9de560de
|
2444 |
|
730c213c7
|
2445 |
sbi->s_locality_groups = alloc_percpu(struct ext4_locality_group); |
c9de560de
|
2446 |
if (sbi->s_locality_groups == NULL) { |
fb1813f4a
|
2447 |
ret = -ENOMEM; |
7aa0baeab
|
2448 |
goto out_free_groupinfo_slab; |
c9de560de
|
2449 |
} |
730c213c7
|
2450 |
for_each_possible_cpu(i) { |
c9de560de
|
2451 |
struct ext4_locality_group *lg; |
730c213c7
|
2452 |
lg = per_cpu_ptr(sbi->s_locality_groups, i); |
c9de560de
|
2453 |
mutex_init(&lg->lg_mutex); |
6be2ded1d
|
2454 2455 |
for (j = 0; j < PREALLOC_TB_SIZE; j++) INIT_LIST_HEAD(&lg->lg_prealloc_list[j]); |
c9de560de
|
2456 2457 |
spin_lock_init(&lg->lg_prealloc_lock); } |
79a77c5ac
|
2458 2459 |
/* init file for buddy data */ ret = ext4_mb_init_backend(sb); |
7aa0baeab
|
2460 2461 |
if (ret != 0) goto out_free_locality_groups; |
79a77c5ac
|
2462 |
|
296c355cd
|
2463 2464 2465 |
if (sbi->s_proc) proc_create_data("mb_groups", S_IRUGO, sbi->s_proc, &ext4_mb_seq_groups_fops, sb); |
c9de560de
|
2466 |
|
0390131ba
|
2467 2468 |
if (sbi->s_journal) sbi->s_journal->j_commit_callback = release_blocks_on_commit; |
7aa0baeab
|
2469 2470 2471 2472 2473 2474 2475 2476 |
return 0; out_free_locality_groups: free_percpu(sbi->s_locality_groups); sbi->s_locality_groups = NULL; out_free_groupinfo_slab: ext4_groupinfo_destroy_slabs(); |
fb1813f4a
|
2477 |
out: |
7aa0baeab
|
2478 2479 2480 2481 |
kfree(sbi->s_mb_offsets); sbi->s_mb_offsets = NULL; kfree(sbi->s_mb_maxs); sbi->s_mb_maxs = NULL; |
fb1813f4a
|
2482 |
return ret; |
c9de560de
|
2483 |
} |
955ce5f5b
|
2484 |
/* need to called with the ext4 group lock held */ |
c9de560de
|
2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 |
static void ext4_mb_cleanup_pa(struct ext4_group_info *grp) { struct ext4_prealloc_space *pa; struct list_head *cur, *tmp; int count = 0; list_for_each_safe(cur, tmp, &grp->bb_prealloc_list) { pa = list_entry(cur, struct ext4_prealloc_space, pa_group_list); list_del(&pa->pa_group_list); count++; |
688f05a01
|
2495 |
kmem_cache_free(ext4_pspace_cachep, pa); |
c9de560de
|
2496 2497 |
} if (count) |
6ba495e92
|
2498 2499 |
mb_debug(1, "mballoc: %u PAs left ", count); |
c9de560de
|
2500 2501 2502 2503 2504 |
} int ext4_mb_release(struct super_block *sb) { |
8df9675f8
|
2505 |
ext4_group_t ngroups = ext4_get_groups_count(sb); |
c9de560de
|
2506 2507 2508 2509 |
ext4_group_t i; int num_meta_group_infos; struct ext4_group_info *grinfo; struct ext4_sb_info *sbi = EXT4_SB(sb); |
fb1813f4a
|
2510 |
struct kmem_cache *cachep = get_groupinfo_cache(sb->s_blocksize_bits); |
c9de560de
|
2511 |
|
c9de560de
|
2512 |
if (sbi->s_group_info) { |
8df9675f8
|
2513 |
for (i = 0; i < ngroups; i++) { |
c9de560de
|
2514 2515 2516 2517 2518 2519 2520 |
grinfo = ext4_get_group_info(sb, i); #ifdef DOUBLE_CHECK kfree(grinfo->bb_bitmap); #endif ext4_lock_group(sb, i); ext4_mb_cleanup_pa(grinfo); ext4_unlock_group(sb, i); |
fb1813f4a
|
2521 |
kmem_cache_free(cachep, grinfo); |
c9de560de
|
2522 |
} |
8df9675f8
|
2523 |
num_meta_group_infos = (ngroups + |
c9de560de
|
2524 2525 2526 2527 |
EXT4_DESC_PER_BLOCK(sb) - 1) >> EXT4_DESC_PER_BLOCK_BITS(sb); for (i = 0; i < num_meta_group_infos; i++) kfree(sbi->s_group_info[i]); |
f18a5f21c
|
2528 |
ext4_kvfree(sbi->s_group_info); |
c9de560de
|
2529 2530 2531 2532 2533 2534 |
} kfree(sbi->s_mb_offsets); kfree(sbi->s_mb_maxs); if (sbi->s_buddy_cache) iput(sbi->s_buddy_cache); if (sbi->s_mb_stats) { |
9d8b9ec44
|
2535 2536 |
ext4_msg(sb, KERN_INFO, "mballoc: %u blocks %u reqs (%u success)", |
c9de560de
|
2537 2538 2539 |
atomic_read(&sbi->s_bal_allocated), atomic_read(&sbi->s_bal_reqs), atomic_read(&sbi->s_bal_success)); |
9d8b9ec44
|
2540 2541 2542 |
ext4_msg(sb, KERN_INFO, "mballoc: %u extents scanned, %u goal hits, " "%u 2^N hits, %u breaks, %u lost", |
c9de560de
|
2543 2544 2545 2546 2547 |
atomic_read(&sbi->s_bal_ex_scanned), atomic_read(&sbi->s_bal_goals), atomic_read(&sbi->s_bal_2orders), atomic_read(&sbi->s_bal_breaks), atomic_read(&sbi->s_mb_lost_chunks)); |
9d8b9ec44
|
2548 2549 |
ext4_msg(sb, KERN_INFO, "mballoc: %lu generated and it took %Lu", |
ced156e46
|
2550 |
sbi->s_mb_buddies_generated, |
c9de560de
|
2551 |
sbi->s_mb_generation_time); |
9d8b9ec44
|
2552 2553 |
ext4_msg(sb, KERN_INFO, "mballoc: %u preallocated, %u discarded", |
c9de560de
|
2554 2555 2556 |
atomic_read(&sbi->s_mb_preallocated), atomic_read(&sbi->s_mb_discarded)); } |
730c213c7
|
2557 |
free_percpu(sbi->s_locality_groups); |
296c355cd
|
2558 2559 |
if (sbi->s_proc) remove_proc_entry("mb_groups", sbi->s_proc); |
c9de560de
|
2560 2561 2562 |
return 0; } |
77ca6cdf0
|
2563 |
static inline int ext4_issue_discard(struct super_block *sb, |
84130193e
|
2564 |
ext4_group_t block_group, ext4_grpblk_t cluster, int count) |
5c521830c
|
2565 |
{ |
5c521830c
|
2566 |
ext4_fsblk_t discard_block; |
84130193e
|
2567 2568 2569 |
discard_block = (EXT4_C2B(EXT4_SB(sb), cluster) + ext4_group_first_block_no(sb, block_group)); count = EXT4_C2B(EXT4_SB(sb), count); |
5c521830c
|
2570 2571 |
trace_ext4_discard_blocks(sb, (unsigned long long) discard_block, count); |
932596366
|
2572 |
return sb_issue_discard(sb, discard_block, count, GFP_NOFS, 0); |
5c521830c
|
2573 |
} |
3e624fc72
|
2574 2575 2576 2577 2578 |
/* * This function is called by the jbd2 layer once the commit has finished, * so we know we can free the blocks that were released with that commit. */ static void release_blocks_on_commit(journal_t *journal, transaction_t *txn) |
c9de560de
|
2579 |
{ |
3e624fc72
|
2580 |
struct super_block *sb = journal->j_private; |
c9de560de
|
2581 |
struct ext4_buddy e4b; |
c894058d6
|
2582 |
struct ext4_group_info *db; |
d9f34504e
|
2583 |
int err, count = 0, count2 = 0; |
c894058d6
|
2584 |
struct ext4_free_data *entry; |
3e624fc72
|
2585 |
struct list_head *l, *ltmp; |
c9de560de
|
2586 |
|
3e624fc72
|
2587 2588 |
list_for_each_safe(l, ltmp, &txn->t_private_list) { entry = list_entry(l, struct ext4_free_data, list); |
c9de560de
|
2589 |
|
6ba495e92
|
2590 |
mb_debug(1, "gonna free %u blocks in group %u (0x%p):", |
3e624fc72
|
2591 |
entry->count, entry->group, entry); |
c9de560de
|
2592 |
|
d9f34504e
|
2593 2594 |
if (test_opt(sb, DISCARD)) ext4_issue_discard(sb, entry->group, |
84130193e
|
2595 |
entry->start_cluster, entry->count); |
b90f68701
|
2596 |
|
c894058d6
|
2597 |
err = ext4_mb_load_buddy(sb, entry->group, &e4b); |
c9de560de
|
2598 2599 |
/* we expect to find existing buddy because it's pinned */ BUG_ON(err != 0); |
c894058d6
|
2600 |
db = e4b.bd_info; |
c9de560de
|
2601 |
/* there are blocks to put in buddy to make them really free */ |
c894058d6
|
2602 |
count += entry->count; |
c9de560de
|
2603 |
count2++; |
c894058d6
|
2604 2605 2606 |
ext4_lock_group(sb, entry->group); /* Take it out of per group rb tree */ rb_erase(&entry->node, &(db->bb_free_root)); |
84130193e
|
2607 |
mb_free_blocks(NULL, &e4b, entry->start_cluster, entry->count); |
c894058d6
|
2608 |
|
3d56b8d2c
|
2609 2610 2611 2612 2613 2614 2615 2616 |
/* * Clear the trimmed flag for the group so that the next * ext4_trim_fs can trim it. * If the volume is mounted with -o discard, online discard * is supported and the free blocks will be trimmed online. */ if (!test_opt(sb, DISCARD)) EXT4_MB_GRP_CLEAR_TRIMMED(db); |
c894058d6
|
2617 2618 2619 2620 2621 2622 |
if (!db->bb_free_root.rb_node) { /* No more items in the per group rb tree * balance refcounts from ext4_mb_free_metadata() */ page_cache_release(e4b.bd_buddy_page); page_cache_release(e4b.bd_bitmap_page); |
c9de560de
|
2623 |
} |
c894058d6
|
2624 |
ext4_unlock_group(sb, entry->group); |
c894058d6
|
2625 |
kmem_cache_free(ext4_free_ext_cachep, entry); |
e39e07fdf
|
2626 |
ext4_mb_unload_buddy(&e4b); |
3e624fc72
|
2627 |
} |
c9de560de
|
2628 |
|
6ba495e92
|
2629 2630 |
mb_debug(1, "freed %u blocks in %u structures ", count, count2); |
c9de560de
|
2631 |
} |
6ba495e92
|
2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 |
#ifdef CONFIG_EXT4_DEBUG u8 mb_enable_debug __read_mostly; static struct dentry *debugfs_dir; static struct dentry *debugfs_debug; static void __init ext4_create_debugfs_entry(void) { debugfs_dir = debugfs_create_dir("ext4", NULL); if (debugfs_dir) debugfs_debug = debugfs_create_u8("mballoc-debug", S_IRUGO | S_IWUSR, debugfs_dir, &mb_enable_debug); } static void ext4_remove_debugfs_entry(void) { debugfs_remove(debugfs_debug); debugfs_remove(debugfs_dir); } #else static void __init ext4_create_debugfs_entry(void) { } static void ext4_remove_debugfs_entry(void) { } #endif |
5dabfc78d
|
2665 |
int __init ext4_init_mballoc(void) |
c9de560de
|
2666 |
{ |
16828088f
|
2667 2668 |
ext4_pspace_cachep = KMEM_CACHE(ext4_prealloc_space, SLAB_RECLAIM_ACCOUNT); |
c9de560de
|
2669 2670 |
if (ext4_pspace_cachep == NULL) return -ENOMEM; |
16828088f
|
2671 2672 |
ext4_ac_cachep = KMEM_CACHE(ext4_allocation_context, SLAB_RECLAIM_ACCOUNT); |
256bdb497
|
2673 2674 2675 2676 |
if (ext4_ac_cachep == NULL) { kmem_cache_destroy(ext4_pspace_cachep); return -ENOMEM; } |
c894058d6
|
2677 |
|
16828088f
|
2678 2679 |
ext4_free_ext_cachep = KMEM_CACHE(ext4_free_data, SLAB_RECLAIM_ACCOUNT); |
c894058d6
|
2680 2681 2682 2683 2684 |
if (ext4_free_ext_cachep == NULL) { kmem_cache_destroy(ext4_pspace_cachep); kmem_cache_destroy(ext4_ac_cachep); return -ENOMEM; } |
6ba495e92
|
2685 |
ext4_create_debugfs_entry(); |
c9de560de
|
2686 2687 |
return 0; } |
5dabfc78d
|
2688 |
void ext4_exit_mballoc(void) |
c9de560de
|
2689 |
{ |
60e6679e2
|
2690 |
/* |
3e03f9ca6
|
2691 2692 2693 2694 |
* Wait for completion of call_rcu()'s on ext4_pspace_cachep * before destroying the slab cache. */ rcu_barrier(); |
c9de560de
|
2695 |
kmem_cache_destroy(ext4_pspace_cachep); |
256bdb497
|
2696 |
kmem_cache_destroy(ext4_ac_cachep); |
c894058d6
|
2697 |
kmem_cache_destroy(ext4_free_ext_cachep); |
2892c15dd
|
2698 |
ext4_groupinfo_destroy_slabs(); |
6ba495e92
|
2699 |
ext4_remove_debugfs_entry(); |
c9de560de
|
2700 2701 2702 2703 |
} /* |
73b2c7165
|
2704 |
* Check quota and mark chosen space (ac->ac_b_ex) non-free in bitmaps |
c9de560de
|
2705 2706 |
* Returns 0 if success or error code */ |
4ddfef7b4
|
2707 2708 |
static noinline_for_stack int ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac, |
53accfa9f
|
2709 |
handle_t *handle, unsigned int reserv_clstrs) |
c9de560de
|
2710 2711 |
{ struct buffer_head *bitmap_bh = NULL; |
c9de560de
|
2712 2713 2714 2715 2716 |
struct ext4_group_desc *gdp; struct buffer_head *gdp_bh; struct ext4_sb_info *sbi; struct super_block *sb; ext4_fsblk_t block; |
519deca04
|
2717 |
int err, len; |
c9de560de
|
2718 2719 2720 2721 2722 2723 |
BUG_ON(ac->ac_status != AC_STATUS_FOUND); BUG_ON(ac->ac_b_ex.fe_len <= 0); sb = ac->ac_sb; sbi = EXT4_SB(sb); |
c9de560de
|
2724 2725 |
err = -EIO; |
574ca174c
|
2726 |
bitmap_bh = ext4_read_block_bitmap(sb, ac->ac_b_ex.fe_group); |
c9de560de
|
2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 |
if (!bitmap_bh) goto out_err; err = ext4_journal_get_write_access(handle, bitmap_bh); if (err) goto out_err; err = -EIO; gdp = ext4_get_group_desc(sb, ac->ac_b_ex.fe_group, &gdp_bh); if (!gdp) goto out_err; |
a9df9a491
|
2738 2739 |
ext4_debug("using block group %u(%d) ", ac->ac_b_ex.fe_group, |
021b65bb1
|
2740 |
ext4_free_group_clusters(sb, gdp)); |
03cddb80e
|
2741 |
|
c9de560de
|
2742 2743 2744 |
err = ext4_journal_get_write_access(handle, gdp_bh); if (err) goto out_err; |
bda00de7e
|
2745 |
block = ext4_grp_offs_to_block(sb, &ac->ac_b_ex); |
c9de560de
|
2746 |
|
53accfa9f
|
2747 |
len = EXT4_C2B(sbi, ac->ac_b_ex.fe_len); |
6fd058f77
|
2748 |
if (!ext4_data_block_valid(sbi, block, len)) { |
12062dddd
|
2749 |
ext4_error(sb, "Allocating blocks %llu-%llu which overlap " |
6fd058f77
|
2750 2751 |
"fs metadata ", block, block+len); |
519deca04
|
2752 2753 2754 2755 |
/* File system mounted not to panic on error * Fix the bitmap and repeat the block allocation * We leak some of the blocks here. */ |
955ce5f5b
|
2756 |
ext4_lock_group(sb, ac->ac_b_ex.fe_group); |
c3e94d1df
|
2757 2758 |
ext4_set_bits(bitmap_bh->b_data, ac->ac_b_ex.fe_start, ac->ac_b_ex.fe_len); |
955ce5f5b
|
2759 |
ext4_unlock_group(sb, ac->ac_b_ex.fe_group); |
0390131ba
|
2760 |
err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh); |
519deca04
|
2761 2762 2763 |
if (!err) err = -EAGAIN; goto out_err; |
c9de560de
|
2764 |
} |
955ce5f5b
|
2765 2766 |
ext4_lock_group(sb, ac->ac_b_ex.fe_group); |
c9de560de
|
2767 2768 2769 2770 2771 2772 2773 2774 2775 |
#ifdef AGGRESSIVE_CHECK { int i; for (i = 0; i < ac->ac_b_ex.fe_len; i++) { BUG_ON(mb_test_bit(ac->ac_b_ex.fe_start + i, bitmap_bh->b_data)); } } #endif |
c3e94d1df
|
2776 2777 |
ext4_set_bits(bitmap_bh->b_data, ac->ac_b_ex.fe_start, ac->ac_b_ex.fe_len); |
c9de560de
|
2778 2779 |
if (gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) { gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT); |
021b65bb1
|
2780 |
ext4_free_group_clusters_set(sb, gdp, |
cff1dfd76
|
2781 |
ext4_free_clusters_after_init(sb, |
021b65bb1
|
2782 |
ac->ac_b_ex.fe_group, gdp)); |
c9de560de
|
2783 |
} |
021b65bb1
|
2784 2785 |
len = ext4_free_group_clusters(sb, gdp) - ac->ac_b_ex.fe_len; ext4_free_group_clusters_set(sb, gdp, len); |
c9de560de
|
2786 |
gdp->bg_checksum = ext4_group_desc_csum(sbi, ac->ac_b_ex.fe_group, gdp); |
955ce5f5b
|
2787 2788 |
ext4_unlock_group(sb, ac->ac_b_ex.fe_group); |
570426518
|
2789 |
percpu_counter_sub(&sbi->s_freeclusters_counter, ac->ac_b_ex.fe_len); |
d2a176379
|
2790 |
/* |
6bc6e63fc
|
2791 |
* Now reduce the dirty block count also. Should not go negative |
d2a176379
|
2792 |
*/ |
6bc6e63fc
|
2793 2794 |
if (!(ac->ac_flags & EXT4_MB_DELALLOC_RESERVED)) /* release all the reserved blocks if non delalloc */ |
570426518
|
2795 2796 |
percpu_counter_sub(&sbi->s_dirtyclusters_counter, reserv_clstrs); |
c9de560de
|
2797 |
|
772cb7c83
|
2798 2799 2800 |
if (sbi->s_log_groups_per_flex) { ext4_group_t flex_group = ext4_flex_group(sbi, ac->ac_b_ex.fe_group); |
9f24e4208
|
2801 |
atomic_sub(ac->ac_b_ex.fe_len, |
24aaa8ef4
|
2802 |
&sbi->s_flex_groups[flex_group].free_clusters); |
772cb7c83
|
2803 |
} |
0390131ba
|
2804 |
err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh); |
c9de560de
|
2805 2806 |
if (err) goto out_err; |
0390131ba
|
2807 |
err = ext4_handle_dirty_metadata(handle, NULL, gdp_bh); |
c9de560de
|
2808 2809 |
out_err: |
a0375156c
|
2810 |
ext4_mark_super_dirty(sb); |
42a10add8
|
2811 |
brelse(bitmap_bh); |
c9de560de
|
2812 2813 2814 2815 2816 |
return err; } /* * here we normalize request for locality group |
d7a1fee13
|
2817 2818 2819 |
* Group request are normalized to s_mb_group_prealloc, which goes to * s_strip if we set the same via mount option. * s_mb_group_prealloc can be configured via |
b713a5ec5
|
2820 |
* /sys/fs/ext4/<partition>/mb_group_prealloc |
c9de560de
|
2821 2822 2823 2824 2825 2826 2827 2828 2829 |
* * XXX: should we try to preallocate more than the group has now? */ static void ext4_mb_normalize_group_request(struct ext4_allocation_context *ac) { struct super_block *sb = ac->ac_sb; struct ext4_locality_group *lg = ac->ac_lg; BUG_ON(lg == NULL); |
d7a1fee13
|
2830 |
ac->ac_g_ex.fe_len = EXT4_SB(sb)->s_mb_group_prealloc; |
6ba495e92
|
2831 2832 |
mb_debug(1, "#%u: goal %u blocks for locality group ", |
c9de560de
|
2833 2834 2835 2836 2837 2838 2839 |
current->pid, ac->ac_g_ex.fe_len); } /* * Normalization means making request better in terms of * size and alignment */ |
4ddfef7b4
|
2840 2841 |
static noinline_for_stack void ext4_mb_normalize_request(struct ext4_allocation_context *ac, |
c9de560de
|
2842 2843 |
struct ext4_allocation_request *ar) { |
53accfa9f
|
2844 |
struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb); |
c9de560de
|
2845 2846 |
int bsbits, max; ext4_lblk_t end; |
c9de560de
|
2847 |
loff_t size, orig_size, start_off; |
5a0790c2c
|
2848 |
ext4_lblk_t start; |
c9de560de
|
2849 |
struct ext4_inode_info *ei = EXT4_I(ac->ac_inode); |
9a0762c5a
|
2850 |
struct ext4_prealloc_space *pa; |
c9de560de
|
2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 |
/* do normalize only data requests, metadata requests do not need preallocation */ if (!(ac->ac_flags & EXT4_MB_HINT_DATA)) return; /* sometime caller may want exact blocks */ if (unlikely(ac->ac_flags & EXT4_MB_HINT_GOAL_ONLY)) return; /* caller may indicate that preallocation isn't * required (it's a tail, for example) */ if (ac->ac_flags & EXT4_MB_HINT_NOPREALLOC) return; if (ac->ac_flags & EXT4_MB_HINT_GROUP_ALLOC) { ext4_mb_normalize_group_request(ac); return ; } bsbits = ac->ac_sb->s_blocksize_bits; /* first, let's learn actual file size * given current request is allocated */ |
53accfa9f
|
2875 |
size = ac->ac_o_ex.fe_logical + EXT4_C2B(sbi, ac->ac_o_ex.fe_len); |
c9de560de
|
2876 2877 2878 |
size = size << bsbits; if (size < i_size_read(ac->ac_inode)) size = i_size_read(ac->ac_inode); |
5a0790c2c
|
2879 |
orig_size = size; |
c9de560de
|
2880 |
|
1930479c4
|
2881 2882 |
/* max size of free chunks */ max = 2 << bsbits; |
c9de560de
|
2883 |
|
1930479c4
|
2884 2885 |
#define NRL_CHECK_SIZE(req, size, max, chunk_size) \ (req <= (size) || max <= (chunk_size)) |
c9de560de
|
2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 |
/* first, try to predict filesize */ /* XXX: should this table be tunable? */ start_off = 0; if (size <= 16 * 1024) { size = 16 * 1024; } else if (size <= 32 * 1024) { size = 32 * 1024; } else if (size <= 64 * 1024) { size = 64 * 1024; } else if (size <= 128 * 1024) { size = 128 * 1024; } else if (size <= 256 * 1024) { size = 256 * 1024; } else if (size <= 512 * 1024) { size = 512 * 1024; } else if (size <= 1024 * 1024) { size = 1024 * 1024; |
1930479c4
|
2904 |
} else if (NRL_CHECK_SIZE(size, 4 * 1024 * 1024, max, 2 * 1024)) { |
c9de560de
|
2905 |
start_off = ((loff_t)ac->ac_o_ex.fe_logical >> |
1930479c4
|
2906 2907 2908 |
(21 - bsbits)) << 21; size = 2 * 1024 * 1024; } else if (NRL_CHECK_SIZE(size, 8 * 1024 * 1024, max, 4 * 1024)) { |
c9de560de
|
2909 2910 2911 2912 |
start_off = ((loff_t)ac->ac_o_ex.fe_logical >> (22 - bsbits)) << 22; size = 4 * 1024 * 1024; } else if (NRL_CHECK_SIZE(ac->ac_o_ex.fe_len, |
1930479c4
|
2913 |
(8<<20)>>bsbits, max, 8 * 1024)) { |
c9de560de
|
2914 2915 2916 2917 2918 2919 2920 |
start_off = ((loff_t)ac->ac_o_ex.fe_logical >> (23 - bsbits)) << 23; size = 8 * 1024 * 1024; } else { start_off = (loff_t)ac->ac_o_ex.fe_logical << bsbits; size = ac->ac_o_ex.fe_len << bsbits; } |
5a0790c2c
|
2921 2922 |
size = size >> bsbits; start = start_off >> bsbits; |
c9de560de
|
2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 |
/* don't cover already allocated blocks in selected range */ if (ar->pleft && start <= ar->lleft) { size -= ar->lleft + 1 - start; start = ar->lleft + 1; } if (ar->pright && start + size - 1 >= ar->lright) size -= start + size - ar->lright; end = start + size; /* check we don't cross already preallocated blocks */ rcu_read_lock(); |
9a0762c5a
|
2936 |
list_for_each_entry_rcu(pa, &ei->i_prealloc_list, pa_inode_list) { |
498e5f241
|
2937 |
ext4_lblk_t pa_end; |
c9de560de
|
2938 |
|
c9de560de
|
2939 2940 2941 2942 2943 2944 2945 |
if (pa->pa_deleted) continue; spin_lock(&pa->pa_lock); if (pa->pa_deleted) { spin_unlock(&pa->pa_lock); continue; } |
53accfa9f
|
2946 2947 |
pa_end = pa->pa_lstart + EXT4_C2B(EXT4_SB(ac->ac_sb), pa->pa_len); |
c9de560de
|
2948 2949 2950 2951 |
/* PA must not overlap original request */ BUG_ON(!(ac->ac_o_ex.fe_logical >= pa_end || ac->ac_o_ex.fe_logical < pa->pa_lstart)); |
38877f4e8
|
2952 2953 |
/* skip PAs this normalized request doesn't overlap with */ if (pa->pa_lstart >= end || pa_end <= start) { |
c9de560de
|
2954 2955 2956 2957 |
spin_unlock(&pa->pa_lock); continue; } BUG_ON(pa->pa_lstart <= start && pa_end >= end); |
38877f4e8
|
2958 |
/* adjust start or end to be adjacent to this pa */ |
c9de560de
|
2959 2960 2961 |
if (pa_end <= ac->ac_o_ex.fe_logical) { BUG_ON(pa_end < start); start = pa_end; |
38877f4e8
|
2962 |
} else if (pa->pa_lstart > ac->ac_o_ex.fe_logical) { |
c9de560de
|
2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 |
BUG_ON(pa->pa_lstart > end); end = pa->pa_lstart; } spin_unlock(&pa->pa_lock); } rcu_read_unlock(); size = end - start; /* XXX: extra loop to check we really don't overlap preallocations */ rcu_read_lock(); |
9a0762c5a
|
2973 |
list_for_each_entry_rcu(pa, &ei->i_prealloc_list, pa_inode_list) { |
498e5f241
|
2974 |
ext4_lblk_t pa_end; |
53accfa9f
|
2975 |
|
c9de560de
|
2976 2977 |
spin_lock(&pa->pa_lock); if (pa->pa_deleted == 0) { |
53accfa9f
|
2978 2979 |
pa_end = pa->pa_lstart + EXT4_C2B(EXT4_SB(ac->ac_sb), pa->pa_len); |
c9de560de
|
2980 2981 2982 2983 2984 2985 2986 2987 |
BUG_ON(!(start >= pa_end || end <= pa->pa_lstart)); } spin_unlock(&pa->pa_lock); } rcu_read_unlock(); if (start + size <= ac->ac_o_ex.fe_logical && start > ac->ac_o_ex.fe_logical) { |
9d8b9ec44
|
2988 2989 2990 2991 |
ext4_msg(ac->ac_sb, KERN_ERR, "start %lu, size %lu, fe_logical %lu", (unsigned long) start, (unsigned long) size, (unsigned long) ac->ac_o_ex.fe_logical); |
c9de560de
|
2992 2993 2994 |
} BUG_ON(start + size <= ac->ac_o_ex.fe_logical && start > ac->ac_o_ex.fe_logical); |
7137d7a48
|
2995 |
BUG_ON(size <= 0 || size > EXT4_CLUSTERS_PER_GROUP(ac->ac_sb)); |
c9de560de
|
2996 2997 2998 2999 3000 3001 |
/* now prepare goal request */ /* XXX: is it better to align blocks WRT to logical * placement or satisfy big request as is */ ac->ac_g_ex.fe_logical = start; |
53accfa9f
|
3002 |
ac->ac_g_ex.fe_len = EXT4_NUM_B2C(sbi, size); |
c9de560de
|
3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 |
/* define goal start in order to merge */ if (ar->pright && (ar->lright == (start + size))) { /* merge to the right */ ext4_get_group_no_and_offset(ac->ac_sb, ar->pright - size, &ac->ac_f_ex.fe_group, &ac->ac_f_ex.fe_start); ac->ac_flags |= EXT4_MB_HINT_TRY_GOAL; } if (ar->pleft && (ar->lleft + 1 == start)) { /* merge to the left */ ext4_get_group_no_and_offset(ac->ac_sb, ar->pleft + 1, &ac->ac_f_ex.fe_group, &ac->ac_f_ex.fe_start); ac->ac_flags |= EXT4_MB_HINT_TRY_GOAL; } |
6ba495e92
|
3019 3020 |
mb_debug(1, "goal: %u(was %u) blocks at %u ", (unsigned) size, |
c9de560de
|
3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 |
(unsigned) orig_size, (unsigned) start); } static void ext4_mb_collect_stats(struct ext4_allocation_context *ac) { struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb); if (sbi->s_mb_stats && ac->ac_g_ex.fe_len > 1) { atomic_inc(&sbi->s_bal_reqs); atomic_add(ac->ac_b_ex.fe_len, &sbi->s_bal_allocated); |
291dae472
|
3031 |
if (ac->ac_b_ex.fe_len >= ac->ac_o_ex.fe_len) |
c9de560de
|
3032 3033 3034 3035 3036 3037 3038 3039 |
atomic_inc(&sbi->s_bal_success); atomic_add(ac->ac_found, &sbi->s_bal_ex_scanned); if (ac->ac_g_ex.fe_start == ac->ac_b_ex.fe_start && ac->ac_g_ex.fe_group == ac->ac_b_ex.fe_group) atomic_inc(&sbi->s_bal_goals); if (ac->ac_found > sbi->s_mb_max_to_scan) atomic_inc(&sbi->s_bal_breaks); } |
296c355cd
|
3040 3041 3042 3043 |
if (ac->ac_op == EXT4_MB_HISTORY_ALLOC) trace_ext4_mballoc_alloc(ac); else trace_ext4_mballoc_prealloc(ac); |
c9de560de
|
3044 3045 3046 |
} /* |
b844167ed
|
3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 |
* Called on failure; free up any blocks from the inode PA for this * context. We don't need this for MB_GROUP_PA because we only change * pa_free in ext4_mb_release_context(), but on failure, we've already * zeroed out ac->ac_b_ex.fe_len, so group_pa->pa_free is not changed. */ static void ext4_discard_allocated_blocks(struct ext4_allocation_context *ac) { struct ext4_prealloc_space *pa = ac->ac_pa; int len; if (pa && pa->pa_type == MB_INODE_PA) { len = ac->ac_b_ex.fe_len; pa->pa_free += len; } } /* |
c9de560de
|
3065 3066 3067 3068 3069 |
* use blocks preallocated to inode */ static void ext4_mb_use_inode_pa(struct ext4_allocation_context *ac, struct ext4_prealloc_space *pa) { |
53accfa9f
|
3070 |
struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb); |
c9de560de
|
3071 3072 3073 3074 3075 3076 |
ext4_fsblk_t start; ext4_fsblk_t end; int len; /* found preallocated blocks, use them */ start = pa->pa_pstart + (ac->ac_o_ex.fe_logical - pa->pa_lstart); |
53accfa9f
|
3077 3078 3079 |
end = min(pa->pa_pstart + EXT4_C2B(sbi, pa->pa_len), start + EXT4_C2B(sbi, ac->ac_o_ex.fe_len)); len = EXT4_NUM_B2C(sbi, end - start); |
c9de560de
|
3080 3081 3082 3083 3084 3085 3086 |
ext4_get_group_no_and_offset(ac->ac_sb, start, &ac->ac_b_ex.fe_group, &ac->ac_b_ex.fe_start); ac->ac_b_ex.fe_len = len; ac->ac_status = AC_STATUS_FOUND; ac->ac_pa = pa; BUG_ON(start < pa->pa_pstart); |
53accfa9f
|
3087 |
BUG_ON(end > pa->pa_pstart + EXT4_C2B(sbi, pa->pa_len)); |
c9de560de
|
3088 3089 |
BUG_ON(pa->pa_free < len); pa->pa_free -= len; |
6ba495e92
|
3090 3091 |
mb_debug(1, "use %llu/%u from inode pa %p ", start, len, pa); |
c9de560de
|
3092 3093 3094 3095 3096 3097 3098 3099 |
} /* * use blocks preallocated to locality group */ static void ext4_mb_use_group_pa(struct ext4_allocation_context *ac, struct ext4_prealloc_space *pa) { |
03cddb80e
|
3100 |
unsigned int len = ac->ac_o_ex.fe_len; |
6be2ded1d
|
3101 |
|
c9de560de
|
3102 3103 3104 3105 3106 3107 3108 3109 |
ext4_get_group_no_and_offset(ac->ac_sb, pa->pa_pstart, &ac->ac_b_ex.fe_group, &ac->ac_b_ex.fe_start); ac->ac_b_ex.fe_len = len; ac->ac_status = AC_STATUS_FOUND; ac->ac_pa = pa; /* we don't correct pa_pstart or pa_plen here to avoid |
26346ff68
|
3110 |
* possible race when the group is being loaded concurrently |
c9de560de
|
3111 |
* instead we correct pa later, after blocks are marked |
26346ff68
|
3112 3113 |
* in on-disk bitmap -- see ext4_mb_release_context() * Other CPUs are prevented from allocating from this pa by lg_mutex |
c9de560de
|
3114 |
*/ |
6ba495e92
|
3115 3116 |
mb_debug(1, "use %u/%u from group pa %p ", pa->pa_lstart-len, len, pa); |
c9de560de
|
3117 3118 3119 |
} /* |
5e745b041
|
3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 |
* Return the prealloc space that have minimal distance * from the goal block. @cpa is the prealloc * space that is having currently known minimal distance * from the goal block. */ static struct ext4_prealloc_space * ext4_mb_check_group_pa(ext4_fsblk_t goal_block, struct ext4_prealloc_space *pa, struct ext4_prealloc_space *cpa) { ext4_fsblk_t cur_distance, new_distance; if (cpa == NULL) { atomic_inc(&pa->pa_count); return pa; } cur_distance = abs(goal_block - cpa->pa_pstart); new_distance = abs(goal_block - pa->pa_pstart); |
5a54b2f19
|
3138 |
if (cur_distance <= new_distance) |
5e745b041
|
3139 3140 3141 3142 3143 3144 3145 3146 3147 |
return cpa; /* drop the previous reference */ atomic_dec(&cpa->pa_count); atomic_inc(&pa->pa_count); return pa; } /* |
c9de560de
|
3148 3149 |
* search goal blocks in preallocated space */ |
4ddfef7b4
|
3150 3151 |
static noinline_for_stack int ext4_mb_use_preallocated(struct ext4_allocation_context *ac) |
c9de560de
|
3152 |
{ |
53accfa9f
|
3153 |
struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb); |
6be2ded1d
|
3154 |
int order, i; |
c9de560de
|
3155 3156 |
struct ext4_inode_info *ei = EXT4_I(ac->ac_inode); struct ext4_locality_group *lg; |
5e745b041
|
3157 3158 |
struct ext4_prealloc_space *pa, *cpa = NULL; ext4_fsblk_t goal_block; |
c9de560de
|
3159 3160 3161 3162 3163 3164 3165 |
/* only data can be preallocated */ if (!(ac->ac_flags & EXT4_MB_HINT_DATA)) return 0; /* first, try per-file preallocation */ rcu_read_lock(); |
9a0762c5a
|
3166 |
list_for_each_entry_rcu(pa, &ei->i_prealloc_list, pa_inode_list) { |
c9de560de
|
3167 3168 3169 3170 |
/* all fields in this condition don't change, * so we can skip locking for them */ if (ac->ac_o_ex.fe_logical < pa->pa_lstart || |
53accfa9f
|
3171 3172 |
ac->ac_o_ex.fe_logical >= (pa->pa_lstart + EXT4_C2B(sbi, pa->pa_len))) |
c9de560de
|
3173 |
continue; |
fb0a387dc
|
3174 |
/* non-extent files can't have physical blocks past 2^32 */ |
12e9b8920
|
3175 |
if (!(ext4_test_inode_flag(ac->ac_inode, EXT4_INODE_EXTENTS)) && |
53accfa9f
|
3176 3177 |
(pa->pa_pstart + EXT4_C2B(sbi, pa->pa_len) > EXT4_MAX_BLOCK_FILE_PHYS)) |
fb0a387dc
|
3178 |
continue; |
c9de560de
|
3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 |
/* found preallocated blocks, use them */ spin_lock(&pa->pa_lock); if (pa->pa_deleted == 0 && pa->pa_free) { atomic_inc(&pa->pa_count); ext4_mb_use_inode_pa(ac, pa); spin_unlock(&pa->pa_lock); ac->ac_criteria = 10; rcu_read_unlock(); return 1; } spin_unlock(&pa->pa_lock); } rcu_read_unlock(); /* can we use group allocation? */ if (!(ac->ac_flags & EXT4_MB_HINT_GROUP_ALLOC)) return 0; /* inode may have no locality group for some reason */ lg = ac->ac_lg; if (lg == NULL) return 0; |
6be2ded1d
|
3201 3202 3203 3204 |
order = fls(ac->ac_o_ex.fe_len) - 1; if (order > PREALLOC_TB_SIZE - 1) /* The max size of hash table is PREALLOC_TB_SIZE */ order = PREALLOC_TB_SIZE - 1; |
bda00de7e
|
3205 |
goal_block = ext4_grp_offs_to_block(ac->ac_sb, &ac->ac_g_ex); |
5e745b041
|
3206 3207 3208 3209 |
/* * search for the prealloc space that is having * minimal distance from the goal block. */ |
6be2ded1d
|
3210 3211 3212 3213 3214 3215 3216 |
for (i = order; i < PREALLOC_TB_SIZE; i++) { rcu_read_lock(); list_for_each_entry_rcu(pa, &lg->lg_prealloc_list[i], pa_inode_list) { spin_lock(&pa->pa_lock); if (pa->pa_deleted == 0 && pa->pa_free >= ac->ac_o_ex.fe_len) { |
5e745b041
|
3217 3218 3219 |
cpa = ext4_mb_check_group_pa(goal_block, pa, cpa); |
6be2ded1d
|
3220 |
} |
c9de560de
|
3221 |
spin_unlock(&pa->pa_lock); |
c9de560de
|
3222 |
} |
6be2ded1d
|
3223 |
rcu_read_unlock(); |
c9de560de
|
3224 |
} |
5e745b041
|
3225 3226 3227 3228 3229 |
if (cpa) { ext4_mb_use_group_pa(ac, cpa); ac->ac_criteria = 20; return 1; } |
c9de560de
|
3230 3231 3232 3233 |
return 0; } /* |
7a2fcbf7f
|
3234 3235 3236 |
* the function goes through all block freed in the group * but not yet committed and marks them used in in-core bitmap. * buddy must be generated from this bitmap |
955ce5f5b
|
3237 |
* Need to be called with the ext4 group lock held |
7a2fcbf7f
|
3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 |
*/ static void ext4_mb_generate_from_freelist(struct super_block *sb, void *bitmap, ext4_group_t group) { struct rb_node *n; struct ext4_group_info *grp; struct ext4_free_data *entry; grp = ext4_get_group_info(sb, group); n = rb_first(&(grp->bb_free_root)); while (n) { entry = rb_entry(n, struct ext4_free_data, node); |
84130193e
|
3251 |
ext4_set_bits(bitmap, entry->start_cluster, entry->count); |
7a2fcbf7f
|
3252 3253 3254 3255 3256 3257 |
n = rb_next(n); } return; } /* |
c9de560de
|
3258 3259 |
* the function goes through all preallocation in this group and marks them * used in in-core bitmap. buddy must be generated from this bitmap |
955ce5f5b
|
3260 |
* Need to be called with ext4 group lock held |
c9de560de
|
3261 |
*/ |
089ceecc1
|
3262 3263 |
static noinline_for_stack void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap, |
c9de560de
|
3264 3265 3266 3267 3268 3269 3270 3271 |
ext4_group_t group) { struct ext4_group_info *grp = ext4_get_group_info(sb, group); struct ext4_prealloc_space *pa; struct list_head *cur; ext4_group_t groupnr; ext4_grpblk_t start; int preallocated = 0; |
c9de560de
|
3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 |
int len; /* all form of preallocation discards first load group, * so the only competing code is preallocation use. * we don't need any locking here * notice we do NOT ignore preallocations with pa_deleted * otherwise we could leave used blocks available for * allocation in buddy when concurrent ext4_mb_put_pa() * is dropping preallocation */ list_for_each(cur, &grp->bb_prealloc_list) { pa = list_entry(cur, struct ext4_prealloc_space, pa_group_list); spin_lock(&pa->pa_lock); ext4_get_group_no_and_offset(sb, pa->pa_pstart, &groupnr, &start); len = pa->pa_len; spin_unlock(&pa->pa_lock); if (unlikely(len == 0)) continue; BUG_ON(groupnr != group); |
c3e94d1df
|
3292 |
ext4_set_bits(bitmap, start, len); |
c9de560de
|
3293 |
preallocated += len; |
c9de560de
|
3294 |
} |
6ba495e92
|
3295 3296 |
mb_debug(1, "prellocated %u for group %u ", preallocated, group); |
c9de560de
|
3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 |
} static void ext4_mb_pa_callback(struct rcu_head *head) { struct ext4_prealloc_space *pa; pa = container_of(head, struct ext4_prealloc_space, u.pa_rcu); kmem_cache_free(ext4_pspace_cachep, pa); } /* * drops a reference to preallocated space descriptor * if this was the last reference and the space is consumed */ static void ext4_mb_put_pa(struct ext4_allocation_context *ac, struct super_block *sb, struct ext4_prealloc_space *pa) { |
a9df9a491
|
3313 |
ext4_group_t grp; |
d33a1976f
|
3314 |
ext4_fsblk_t grp_blk; |
c9de560de
|
3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 |
if (!atomic_dec_and_test(&pa->pa_count) || pa->pa_free != 0) return; /* in this short window concurrent discard can set pa_deleted */ spin_lock(&pa->pa_lock); if (pa->pa_deleted == 1) { spin_unlock(&pa->pa_lock); return; } pa->pa_deleted = 1; spin_unlock(&pa->pa_lock); |
d33a1976f
|
3328 |
grp_blk = pa->pa_pstart; |
60e6679e2
|
3329 |
/* |
cc0fb9ad7
|
3330 3331 3332 3333 |
* If doing group-based preallocation, pa_pstart may be in the * next group when pa is used up */ if (pa->pa_type == MB_GROUP_PA) |
d33a1976f
|
3334 3335 3336 |
grp_blk--; ext4_get_group_no_and_offset(sb, grp_blk, &grp, NULL); |
c9de560de
|
3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 |
/* * possible race: * * P1 (buddy init) P2 (regular allocation) * find block B in PA * copy on-disk bitmap to buddy * mark B in on-disk bitmap * drop PA from group * mark all PAs in buddy * * thus, P1 initializes buddy with B available. to prevent this * we make "copy" and "mark all PAs" atomic and serialize "drop PA" * against that pair */ ext4_lock_group(sb, grp); list_del(&pa->pa_group_list); ext4_unlock_group(sb, grp); spin_lock(pa->pa_obj_lock); list_del_rcu(&pa->pa_inode_list); spin_unlock(pa->pa_obj_lock); call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback); } /* * creates new preallocated space for given inode */ |
4ddfef7b4
|
3366 3367 |
static noinline_for_stack int ext4_mb_new_inode_pa(struct ext4_allocation_context *ac) |
c9de560de
|
3368 3369 |
{ struct super_block *sb = ac->ac_sb; |
53accfa9f
|
3370 |
struct ext4_sb_info *sbi = EXT4_SB(sb); |
c9de560de
|
3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 |
struct ext4_prealloc_space *pa; struct ext4_group_info *grp; struct ext4_inode_info *ei; /* preallocate only when found space is larger then requested */ BUG_ON(ac->ac_o_ex.fe_len >= ac->ac_b_ex.fe_len); BUG_ON(ac->ac_status != AC_STATUS_FOUND); BUG_ON(!S_ISREG(ac->ac_inode->i_mode)); pa = kmem_cache_alloc(ext4_pspace_cachep, GFP_NOFS); if (pa == NULL) return -ENOMEM; if (ac->ac_b_ex.fe_len < ac->ac_g_ex.fe_len) { int winl; int wins; int win; int offs; /* we can't allocate as much as normalizer wants. * so, found space must get proper lstart * to cover original request */ BUG_ON(ac->ac_g_ex.fe_logical > ac->ac_o_ex.fe_logical); BUG_ON(ac->ac_g_ex.fe_len < ac->ac_o_ex.fe_len); /* we're limited by original request in that * logical block must be covered any way * winl is window we can move our chunk within */ winl = ac->ac_o_ex.fe_logical - ac->ac_g_ex.fe_logical; /* also, we should cover whole original request */ |
53accfa9f
|
3402 |
wins = EXT4_C2B(sbi, ac->ac_b_ex.fe_len - ac->ac_o_ex.fe_len); |
c9de560de
|
3403 3404 3405 |
/* the smallest one defines real window */ win = min(winl, wins); |
53accfa9f
|
3406 3407 |
offs = ac->ac_o_ex.fe_logical % EXT4_C2B(sbi, ac->ac_b_ex.fe_len); |
c9de560de
|
3408 3409 |
if (offs && offs < win) win = offs; |
53accfa9f
|
3410 3411 |
ac->ac_b_ex.fe_logical = ac->ac_o_ex.fe_logical - EXT4_B2C(sbi, win); |
c9de560de
|
3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 |
BUG_ON(ac->ac_o_ex.fe_logical < ac->ac_b_ex.fe_logical); BUG_ON(ac->ac_o_ex.fe_len > ac->ac_b_ex.fe_len); } /* preallocation can change ac_b_ex, thus we store actually * allocated blocks for history */ ac->ac_f_ex = ac->ac_b_ex; pa->pa_lstart = ac->ac_b_ex.fe_logical; pa->pa_pstart = ext4_grp_offs_to_block(sb, &ac->ac_b_ex); pa->pa_len = ac->ac_b_ex.fe_len; pa->pa_free = pa->pa_len; atomic_set(&pa->pa_count, 1); spin_lock_init(&pa->pa_lock); |
d794bf8e0
|
3426 3427 |
INIT_LIST_HEAD(&pa->pa_inode_list); INIT_LIST_HEAD(&pa->pa_group_list); |
c9de560de
|
3428 |
pa->pa_deleted = 0; |
cc0fb9ad7
|
3429 |
pa->pa_type = MB_INODE_PA; |
c9de560de
|
3430 |
|
6ba495e92
|
3431 3432 |
mb_debug(1, "new inode pa %p: %llu/%u for %u ", pa, |
c9de560de
|
3433 |
pa->pa_pstart, pa->pa_len, pa->pa_lstart); |
9bffad1ed
|
3434 |
trace_ext4_mb_new_inode_pa(ac, pa); |
c9de560de
|
3435 3436 |
ext4_mb_use_inode_pa(ac, pa); |
53accfa9f
|
3437 |
atomic_add(pa->pa_free, &sbi->s_mb_preallocated); |
c9de560de
|
3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 |
ei = EXT4_I(ac->ac_inode); grp = ext4_get_group_info(sb, ac->ac_b_ex.fe_group); pa->pa_obj_lock = &ei->i_prealloc_lock; pa->pa_inode = ac->ac_inode; ext4_lock_group(sb, ac->ac_b_ex.fe_group); list_add(&pa->pa_group_list, &grp->bb_prealloc_list); ext4_unlock_group(sb, ac->ac_b_ex.fe_group); spin_lock(pa->pa_obj_lock); list_add_rcu(&pa->pa_inode_list, &ei->i_prealloc_list); spin_unlock(pa->pa_obj_lock); return 0; } /* * creates new preallocated space for locality group inodes belongs to */ |
4ddfef7b4
|
3459 3460 |
static noinline_for_stack int ext4_mb_new_group_pa(struct ext4_allocation_context *ac) |
c9de560de
|
3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 |
{ struct super_block *sb = ac->ac_sb; struct ext4_locality_group *lg; struct ext4_prealloc_space *pa; struct ext4_group_info *grp; /* preallocate only when found space is larger then requested */ BUG_ON(ac->ac_o_ex.fe_len >= ac->ac_b_ex.fe_len); BUG_ON(ac->ac_status != AC_STATUS_FOUND); BUG_ON(!S_ISREG(ac->ac_inode->i_mode)); BUG_ON(ext4_pspace_cachep == NULL); pa = kmem_cache_alloc(ext4_pspace_cachep, GFP_NOFS); if (pa == NULL) return -ENOMEM; /* preallocation can change ac_b_ex, thus we store actually * allocated blocks for history */ ac->ac_f_ex = ac->ac_b_ex; pa->pa_pstart = ext4_grp_offs_to_block(sb, &ac->ac_b_ex); pa->pa_lstart = pa->pa_pstart; pa->pa_len = ac->ac_b_ex.fe_len; pa->pa_free = pa->pa_len; atomic_set(&pa->pa_count, 1); spin_lock_init(&pa->pa_lock); |
6be2ded1d
|
3487 |
INIT_LIST_HEAD(&pa->pa_inode_list); |
d794bf8e0
|
3488 |
INIT_LIST_HEAD(&pa->pa_group_list); |
c9de560de
|
3489 |
pa->pa_deleted = 0; |
cc0fb9ad7
|
3490 |
pa->pa_type = MB_GROUP_PA; |
c9de560de
|
3491 |
|
6ba495e92
|
3492 3493 |
mb_debug(1, "new group pa %p: %llu/%u for %u ", pa, |
9bffad1ed
|
3494 3495 |
pa->pa_pstart, pa->pa_len, pa->pa_lstart); trace_ext4_mb_new_group_pa(ac, pa); |
c9de560de
|
3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 |
ext4_mb_use_group_pa(ac, pa); atomic_add(pa->pa_free, &EXT4_SB(sb)->s_mb_preallocated); grp = ext4_get_group_info(sb, ac->ac_b_ex.fe_group); lg = ac->ac_lg; BUG_ON(lg == NULL); pa->pa_obj_lock = &lg->lg_prealloc_lock; pa->pa_inode = NULL; ext4_lock_group(sb, ac->ac_b_ex.fe_group); list_add(&pa->pa_group_list, &grp->bb_prealloc_list); ext4_unlock_group(sb, ac->ac_b_ex.fe_group); |
6be2ded1d
|
3510 3511 3512 3513 |
/* * We will later add the new pa to the right bucket * after updating the pa_free in ext4_mb_release_context */ |
c9de560de
|
3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 |
return 0; } static int ext4_mb_new_preallocation(struct ext4_allocation_context *ac) { int err; if (ac->ac_flags & EXT4_MB_HINT_GROUP_ALLOC) err = ext4_mb_new_group_pa(ac); else err = ext4_mb_new_inode_pa(ac); return err; } /* * finds all unused blocks in on-disk bitmap, frees them in * in-core bitmap and buddy. * @pa must be unlinked from inode and group lists, so that * nobody else can find/use it. * the caller MUST hold group/inode locks. * TODO: optimize the case when there are no in-core structures yet */ |
4ddfef7b4
|
3536 3537 |
static noinline_for_stack int ext4_mb_release_inode_pa(struct ext4_buddy *e4b, struct buffer_head *bitmap_bh, |
3e1e5f501
|
3538 |
struct ext4_prealloc_space *pa) |
c9de560de
|
3539 |
{ |
c9de560de
|
3540 3541 |
struct super_block *sb = e4b->bd_sb; struct ext4_sb_info *sbi = EXT4_SB(sb); |
498e5f241
|
3542 3543 |
unsigned int end; unsigned int next; |
c9de560de
|
3544 3545 |
ext4_group_t group; ext4_grpblk_t bit; |
ba80b1019
|
3546 |
unsigned long long grp_blk_start; |
c9de560de
|
3547 3548 3549 3550 3551 |
int err = 0; int free = 0; BUG_ON(pa->pa_deleted == 0); ext4_get_group_no_and_offset(sb, pa->pa_pstart, &group, &bit); |
53accfa9f
|
3552 |
grp_blk_start = pa->pa_pstart - EXT4_C2B(sbi, bit); |
c9de560de
|
3553 3554 |
BUG_ON(group != e4b->bd_group && pa->pa_len != 0); end = bit + pa->pa_len; |
c9de560de
|
3555 |
while (bit < end) { |
ffad0a44b
|
3556 |
bit = mb_find_next_zero_bit(bitmap_bh->b_data, end, bit); |
c9de560de
|
3557 3558 |
if (bit >= end) break; |
ffad0a44b
|
3559 |
next = mb_find_next_bit(bitmap_bh->b_data, end, bit); |
6ba495e92
|
3560 3561 |
mb_debug(1, " free preallocated %u/%u in group %u ", |
5a0790c2c
|
3562 3563 |
(unsigned) ext4_group_first_block_no(sb, group) + bit, (unsigned) next - bit, (unsigned) group); |
c9de560de
|
3564 |
free += next - bit; |
3e1e5f501
|
3565 |
trace_ext4_mballoc_discard(sb, NULL, group, bit, next - bit); |
53accfa9f
|
3566 3567 |
trace_ext4_mb_release_inode_pa(pa, (grp_blk_start + EXT4_C2B(sbi, bit)), |
a9c667f8f
|
3568 |
next - bit); |
c9de560de
|
3569 3570 3571 3572 |
mb_free_blocks(pa->pa_inode, e4b, bit, next - bit); bit = next + 1; } if (free != pa->pa_free) { |
9d8b9ec44
|
3573 3574 3575 3576 3577 |
ext4_msg(e4b->bd_sb, KERN_CRIT, "pa %p: logic %lu, phys. %lu, len %lu", pa, (unsigned long) pa->pa_lstart, (unsigned long) pa->pa_pstart, (unsigned long) pa->pa_len); |
e29136f80
|
3578 |
ext4_grp_locked_error(sb, group, 0, 0, "free %u, pa_free %u", |
5d1b1b3f4
|
3579 |
free, pa->pa_free); |
e56eb6590
|
3580 3581 3582 3583 |
/* * pa is already deleted so we use the value obtained * from the bitmap and continue. */ |
c9de560de
|
3584 |
} |
c9de560de
|
3585 3586 3587 3588 |
atomic_add(free, &sbi->s_mb_discarded); return err; } |
4ddfef7b4
|
3589 3590 |
static noinline_for_stack int ext4_mb_release_group_pa(struct ext4_buddy *e4b, |
3e1e5f501
|
3591 |
struct ext4_prealloc_space *pa) |
c9de560de
|
3592 |
{ |
c9de560de
|
3593 3594 3595 |
struct super_block *sb = e4b->bd_sb; ext4_group_t group; ext4_grpblk_t bit; |
60e07cf51
|
3596 |
trace_ext4_mb_release_group_pa(sb, pa); |
c9de560de
|
3597 3598 3599 3600 3601 |
BUG_ON(pa->pa_deleted == 0); ext4_get_group_no_and_offset(sb, pa->pa_pstart, &group, &bit); BUG_ON(group != e4b->bd_group && pa->pa_len != 0); mb_free_blocks(pa->pa_inode, e4b, bit, pa->pa_len); atomic_add(pa->pa_len, &EXT4_SB(sb)->s_mb_discarded); |
3e1e5f501
|
3602 |
trace_ext4_mballoc_discard(sb, NULL, group, bit, pa->pa_len); |
c9de560de
|
3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 |
return 0; } /* * releases all preallocations in given group * * first, we need to decide discard policy: * - when do we discard * 1) ENOSPC * - how many do we discard * 1) how many requested */ |
4ddfef7b4
|
3616 3617 |
static noinline_for_stack int ext4_mb_discard_group_preallocations(struct super_block *sb, |
c9de560de
|
3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 |
ext4_group_t group, int needed) { struct ext4_group_info *grp = ext4_get_group_info(sb, group); struct buffer_head *bitmap_bh = NULL; struct ext4_prealloc_space *pa, *tmp; struct list_head list; struct ext4_buddy e4b; int err; int busy = 0; int free = 0; |
6ba495e92
|
3628 3629 |
mb_debug(1, "discard preallocation for group %u ", group); |
c9de560de
|
3630 3631 3632 |
if (list_empty(&grp->bb_prealloc_list)) return 0; |
574ca174c
|
3633 |
bitmap_bh = ext4_read_block_bitmap(sb, group); |
c9de560de
|
3634 |
if (bitmap_bh == NULL) { |
12062dddd
|
3635 |
ext4_error(sb, "Error reading block bitmap for %u", group); |
ce89f46cb
|
3636 |
return 0; |
c9de560de
|
3637 3638 3639 |
} err = ext4_mb_load_buddy(sb, group, &e4b); |
ce89f46cb
|
3640 |
if (err) { |
12062dddd
|
3641 |
ext4_error(sb, "Error loading buddy information for %u", group); |
ce89f46cb
|
3642 3643 3644 |
put_bh(bitmap_bh); return 0; } |
c9de560de
|
3645 3646 |
if (needed == 0) |
7137d7a48
|
3647 |
needed = EXT4_CLUSTERS_PER_GROUP(sb) + 1; |
c9de560de
|
3648 |
|
c9de560de
|
3649 |
INIT_LIST_HEAD(&list); |
c9de560de
|
3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 |
repeat: ext4_lock_group(sb, group); list_for_each_entry_safe(pa, tmp, &grp->bb_prealloc_list, pa_group_list) { spin_lock(&pa->pa_lock); if (atomic_read(&pa->pa_count)) { spin_unlock(&pa->pa_lock); busy = 1; continue; } if (pa->pa_deleted) { spin_unlock(&pa->pa_lock); continue; } /* seems this one can be freed ... */ pa->pa_deleted = 1; /* we can trust pa_free ... */ free += pa->pa_free; spin_unlock(&pa->pa_lock); list_del(&pa->pa_group_list); list_add(&pa->u.pa_tmp_list, &list); } /* if we still need more blocks and some PAs were used, try again */ if (free < needed && busy) { busy = 0; ext4_unlock_group(sb, group); /* * Yield the CPU here so that we don't get soft lockup * in non preempt case. */ yield(); goto repeat; } /* found anything to free? */ if (list_empty(&list)) { BUG_ON(free != 0); goto out; } /* now free all selected PAs */ list_for_each_entry_safe(pa, tmp, &list, u.pa_tmp_list) { /* remove from object (inode or locality group) */ spin_lock(pa->pa_obj_lock); list_del_rcu(&pa->pa_inode_list); spin_unlock(pa->pa_obj_lock); |
cc0fb9ad7
|
3702 |
if (pa->pa_type == MB_GROUP_PA) |
3e1e5f501
|
3703 |
ext4_mb_release_group_pa(&e4b, pa); |
c9de560de
|
3704 |
else |
3e1e5f501
|
3705 |
ext4_mb_release_inode_pa(&e4b, bitmap_bh, pa); |
c9de560de
|
3706 3707 3708 3709 3710 3711 3712 |
list_del(&pa->u.pa_tmp_list); call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback); } out: ext4_unlock_group(sb, group); |
e39e07fdf
|
3713 |
ext4_mb_unload_buddy(&e4b); |
c9de560de
|
3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 |
put_bh(bitmap_bh); return free; } /* * releases all non-used preallocated blocks for given inode * * It's important to discard preallocations under i_data_sem * We don't want another block to be served from the prealloc * space when we are discarding the inode prealloc space. * * FIXME!! Make sure it is valid at all the call sites */ |
c2ea3fde6
|
3727 |
void ext4_discard_preallocations(struct inode *inode) |
c9de560de
|
3728 3729 3730 3731 3732 3733 3734 3735 3736 |
{ struct ext4_inode_info *ei = EXT4_I(inode); struct super_block *sb = inode->i_sb; struct buffer_head *bitmap_bh = NULL; struct ext4_prealloc_space *pa, *tmp; ext4_group_t group = 0; struct list_head list; struct ext4_buddy e4b; int err; |
c2ea3fde6
|
3737 |
if (!S_ISREG(inode->i_mode)) { |
c9de560de
|
3738 3739 3740 |
/*BUG_ON(!list_empty(&ei->i_prealloc_list));*/ return; } |
6ba495e92
|
3741 3742 |
mb_debug(1, "discard preallocation for inode %lu ", inode->i_ino); |
9bffad1ed
|
3743 |
trace_ext4_discard_preallocations(inode); |
c9de560de
|
3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 |
INIT_LIST_HEAD(&list); repeat: /* first, collect all pa's in the inode */ spin_lock(&ei->i_prealloc_lock); while (!list_empty(&ei->i_prealloc_list)) { pa = list_entry(ei->i_prealloc_list.next, struct ext4_prealloc_space, pa_inode_list); BUG_ON(pa->pa_obj_lock != &ei->i_prealloc_lock); spin_lock(&pa->pa_lock); if (atomic_read(&pa->pa_count)) { /* this shouldn't happen often - nobody should * use preallocation while we're discarding it */ spin_unlock(&pa->pa_lock); spin_unlock(&ei->i_prealloc_lock); |
9d8b9ec44
|
3760 3761 |
ext4_msg(sb, KERN_ERR, "uh-oh! used pa while discarding"); |
c9de560de
|
3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 |
WARN_ON(1); schedule_timeout_uninterruptible(HZ); goto repeat; } if (pa->pa_deleted == 0) { pa->pa_deleted = 1; spin_unlock(&pa->pa_lock); list_del_rcu(&pa->pa_inode_list); list_add(&pa->u.pa_tmp_list, &list); continue; } /* someone is deleting pa right now */ spin_unlock(&pa->pa_lock); spin_unlock(&ei->i_prealloc_lock); /* we have to wait here because pa_deleted * doesn't mean pa is already unlinked from * the list. as we might be called from * ->clear_inode() the inode will get freed * and concurrent thread which is unlinking * pa from inode's list may access already * freed memory, bad-bad-bad */ /* XXX: if this happens too often, we can * add a flag to force wait only in case * of ->clear_inode(), but not in case of * regular truncate */ schedule_timeout_uninterruptible(HZ); goto repeat; } spin_unlock(&ei->i_prealloc_lock); list_for_each_entry_safe(pa, tmp, &list, u.pa_tmp_list) { |
cc0fb9ad7
|
3797 |
BUG_ON(pa->pa_type != MB_INODE_PA); |
c9de560de
|
3798 3799 3800 |
ext4_get_group_no_and_offset(sb, pa->pa_pstart, &group, NULL); err = ext4_mb_load_buddy(sb, group, &e4b); |
ce89f46cb
|
3801 |
if (err) { |
12062dddd
|
3802 3803 |
ext4_error(sb, "Error loading buddy information for %u", group); |
ce89f46cb
|
3804 3805 |
continue; } |
c9de560de
|
3806 |
|
574ca174c
|
3807 |
bitmap_bh = ext4_read_block_bitmap(sb, group); |
c9de560de
|
3808 |
if (bitmap_bh == NULL) { |
12062dddd
|
3809 3810 |
ext4_error(sb, "Error reading block bitmap for %u", group); |
e39e07fdf
|
3811 |
ext4_mb_unload_buddy(&e4b); |
ce89f46cb
|
3812 |
continue; |
c9de560de
|
3813 3814 3815 3816 |
} ext4_lock_group(sb, group); list_del(&pa->pa_group_list); |
3e1e5f501
|
3817 |
ext4_mb_release_inode_pa(&e4b, bitmap_bh, pa); |
c9de560de
|
3818 |
ext4_unlock_group(sb, group); |
e39e07fdf
|
3819 |
ext4_mb_unload_buddy(&e4b); |
c9de560de
|
3820 3821 3822 3823 3824 3825 |
put_bh(bitmap_bh); list_del(&pa->u.pa_tmp_list); call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback); } } |
6ba495e92
|
3826 |
#ifdef CONFIG_EXT4_DEBUG |
c9de560de
|
3827 3828 3829 |
static void ext4_mb_show_ac(struct ext4_allocation_context *ac) { struct super_block *sb = ac->ac_sb; |
8df9675f8
|
3830 |
ext4_group_t ngroups, i; |
c9de560de
|
3831 |
|
4dd89fc62
|
3832 3833 |
if (!mb_enable_debug || (EXT4_SB(sb)->s_mount_flags & EXT4_MF_FS_ABORTED)) |
e3570639c
|
3834 |
return; |
9d8b9ec44
|
3835 3836 3837 |
ext4_msg(ac->ac_sb, KERN_ERR, "EXT4-fs: Can't allocate:" " Allocation context details:"); ext4_msg(ac->ac_sb, KERN_ERR, "EXT4-fs: status %d flags %d", |
c9de560de
|
3838 |
ac->ac_status, ac->ac_flags); |
9d8b9ec44
|
3839 3840 3841 |
ext4_msg(ac->ac_sb, KERN_ERR, "EXT4-fs: orig %lu/%lu/%lu@%lu, " "goal %lu/%lu/%lu@%lu, " "best %lu/%lu/%lu@%lu cr %d", |
c9de560de
|
3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 |
(unsigned long)ac->ac_o_ex.fe_group, (unsigned long)ac->ac_o_ex.fe_start, (unsigned long)ac->ac_o_ex.fe_len, (unsigned long)ac->ac_o_ex.fe_logical, (unsigned long)ac->ac_g_ex.fe_group, (unsigned long)ac->ac_g_ex.fe_start, (unsigned long)ac->ac_g_ex.fe_len, (unsigned long)ac->ac_g_ex.fe_logical, (unsigned long)ac->ac_b_ex.fe_group, (unsigned long)ac->ac_b_ex.fe_start, (unsigned long)ac->ac_b_ex.fe_len, (unsigned long)ac->ac_b_ex.fe_logical, (int)ac->ac_criteria); |
9d8b9ec44
|
3855 3856 3857 |
ext4_msg(ac->ac_sb, KERN_ERR, "EXT4-fs: %lu scanned, %d found", ac->ac_ex_scanned, ac->ac_found); ext4_msg(ac->ac_sb, KERN_ERR, "EXT4-fs: groups: "); |
8df9675f8
|
3858 3859 |
ngroups = ext4_get_groups_count(sb); for (i = 0; i < ngroups; i++) { |
c9de560de
|
3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 |
struct ext4_group_info *grp = ext4_get_group_info(sb, i); struct ext4_prealloc_space *pa; ext4_grpblk_t start; struct list_head *cur; ext4_lock_group(sb, i); list_for_each(cur, &grp->bb_prealloc_list) { pa = list_entry(cur, struct ext4_prealloc_space, pa_group_list); spin_lock(&pa->pa_lock); ext4_get_group_no_and_offset(sb, pa->pa_pstart, NULL, &start); spin_unlock(&pa->pa_lock); |
1c7185051
|
3872 3873 3874 |
printk(KERN_ERR "PA:%u:%d:%u ", i, start, pa->pa_len); |
c9de560de
|
3875 |
} |
60bd63d19
|
3876 |
ext4_unlock_group(sb, i); |
c9de560de
|
3877 3878 3879 |
if (grp->bb_free == 0) continue; |
1c7185051
|
3880 3881 |
printk(KERN_ERR "%u: %d/%d ", |
c9de560de
|
3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 |
i, grp->bb_free, grp->bb_fragments); } printk(KERN_ERR " "); } #else static inline void ext4_mb_show_ac(struct ext4_allocation_context *ac) { return; } #endif /* * We use locality group preallocation for small size file. The size of the * file is determined by the current size or the resulting size after * allocation which ever is larger * |
b713a5ec5
|
3899 |
* One can tune this size via /sys/fs/ext4/<partition>/mb_stream_req |
c9de560de
|
3900 3901 3902 3903 3904 3905 3906 3907 3908 |
*/ static void ext4_mb_group_or_file(struct ext4_allocation_context *ac) { struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb); int bsbits = ac->ac_sb->s_blocksize_bits; loff_t size, isize; if (!(ac->ac_flags & EXT4_MB_HINT_DATA)) return; |
4ba74d00a
|
3909 3910 |
if (unlikely(ac->ac_flags & EXT4_MB_HINT_GOAL_ONLY)) return; |
53accfa9f
|
3911 |
size = ac->ac_o_ex.fe_logical + EXT4_C2B(sbi, ac->ac_o_ex.fe_len); |
50797481a
|
3912 3913 |
isize = (i_size_read(ac->ac_inode) + ac->ac_sb->s_blocksize - 1) >> bsbits; |
c9de560de
|
3914 |
|
50797481a
|
3915 3916 3917 3918 3919 3920 |
if ((size == isize) && !ext4_fs_is_busy(sbi) && (atomic_read(&ac->ac_inode->i_writecount) == 0)) { ac->ac_flags |= EXT4_MB_HINT_NOPREALLOC; return; } |
ebbe02779
|
3921 3922 3923 3924 |
if (sbi->s_mb_group_prealloc <= 0) { ac->ac_flags |= EXT4_MB_STREAM_ALLOC; return; } |
c9de560de
|
3925 |
/* don't use group allocation for large files */ |
717805773
|
3926 |
size = max(size, isize); |
cc483f102
|
3927 |
if (size > sbi->s_mb_stream_request) { |
4ba74d00a
|
3928 |
ac->ac_flags |= EXT4_MB_STREAM_ALLOC; |
c9de560de
|
3929 |
return; |
4ba74d00a
|
3930 |
} |
c9de560de
|
3931 3932 3933 3934 3935 3936 3937 |
BUG_ON(ac->ac_lg != NULL); /* * locality group prealloc space are per cpu. The reason for having * per cpu locality group is to reduce the contention between block * request from multiple CPUs. */ |
ca0c9584b
|
3938 |
ac->ac_lg = __this_cpu_ptr(sbi->s_locality_groups); |
c9de560de
|
3939 3940 3941 3942 3943 3944 3945 |
/* we're going to use group allocation */ ac->ac_flags |= EXT4_MB_HINT_GROUP_ALLOC; /* serialize all allocations in the group */ mutex_lock(&ac->ac_lg->lg_mutex); } |
4ddfef7b4
|
3946 3947 |
static noinline_for_stack int ext4_mb_initialize_context(struct ext4_allocation_context *ac, |
c9de560de
|
3948 3949 3950 3951 3952 3953 |
struct ext4_allocation_request *ar) { struct super_block *sb = ar->inode->i_sb; struct ext4_sb_info *sbi = EXT4_SB(sb); struct ext4_super_block *es = sbi->s_es; ext4_group_t group; |
498e5f241
|
3954 3955 |
unsigned int len; ext4_fsblk_t goal; |
c9de560de
|
3956 3957 3958 3959 3960 3961 |
ext4_grpblk_t block; /* we can't allocate > group size */ len = ar->len; /* just a dirty hack to filter too big requests */ |
7137d7a48
|
3962 3963 |
if (len >= EXT4_CLUSTERS_PER_GROUP(sb) - 10) len = EXT4_CLUSTERS_PER_GROUP(sb) - 10; |
c9de560de
|
3964 3965 3966 3967 3968 3969 3970 3971 3972 |
/* start searching from the goal */ goal = ar->goal; if (goal < le32_to_cpu(es->s_first_data_block) || goal >= ext4_blocks_count(es)) goal = le32_to_cpu(es->s_first_data_block); ext4_get_group_no_and_offset(sb, goal, &group, &block); /* set up allocation goals */ |
833576b36
|
3973 |
memset(ac, 0, sizeof(struct ext4_allocation_context)); |
53accfa9f
|
3974 |
ac->ac_b_ex.fe_logical = ar->logical & ~(sbi->s_cluster_ratio - 1); |
c9de560de
|
3975 |
ac->ac_status = AC_STATUS_CONTINUE; |
c9de560de
|
3976 3977 |
ac->ac_sb = sb; ac->ac_inode = ar->inode; |
53accfa9f
|
3978 |
ac->ac_o_ex.fe_logical = ac->ac_b_ex.fe_logical; |
c9de560de
|
3979 3980 3981 |
ac->ac_o_ex.fe_group = group; ac->ac_o_ex.fe_start = block; ac->ac_o_ex.fe_len = len; |
53accfa9f
|
3982 |
ac->ac_g_ex = ac->ac_o_ex; |
c9de560de
|
3983 |
ac->ac_flags = ar->flags; |
c9de560de
|
3984 3985 3986 3987 |
/* we have to define context: we'll we work with a file or * locality group. this is a policy, actually */ ext4_mb_group_or_file(ac); |
6ba495e92
|
3988 |
mb_debug(1, "init ac: %u blocks @ %u, goal %u, flags %x, 2^%d, " |
c9de560de
|
3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 |
"left: %u/%u, right %u/%u to %swritable ", (unsigned) ar->len, (unsigned) ar->logical, (unsigned) ar->goal, ac->ac_flags, ac->ac_2order, (unsigned) ar->lleft, (unsigned) ar->pleft, (unsigned) ar->lright, (unsigned) ar->pright, atomic_read(&ar->inode->i_writecount) ? "" : "non-"); return 0; } |
6be2ded1d
|
3999 4000 4001 4002 4003 4004 4005 4006 4007 |
static noinline_for_stack void ext4_mb_discard_lg_preallocations(struct super_block *sb, struct ext4_locality_group *lg, int order, int total_entries) { ext4_group_t group = 0; struct ext4_buddy e4b; struct list_head discard_list; struct ext4_prealloc_space *pa, *tmp; |
6be2ded1d
|
4008 |
|
6ba495e92
|
4009 4010 |
mb_debug(1, "discard locality group preallocation "); |
6be2ded1d
|
4011 4012 |
INIT_LIST_HEAD(&discard_list); |
6be2ded1d
|
4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 |
spin_lock(&lg->lg_prealloc_lock); list_for_each_entry_rcu(pa, &lg->lg_prealloc_list[order], pa_inode_list) { spin_lock(&pa->pa_lock); if (atomic_read(&pa->pa_count)) { /* * This is the pa that we just used * for block allocation. So don't * free that */ spin_unlock(&pa->pa_lock); continue; } if (pa->pa_deleted) { spin_unlock(&pa->pa_lock); continue; } /* only lg prealloc space */ |
cc0fb9ad7
|
4032 |
BUG_ON(pa->pa_type != MB_GROUP_PA); |
6be2ded1d
|
4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 |
/* seems this one can be freed ... */ pa->pa_deleted = 1; spin_unlock(&pa->pa_lock); list_del_rcu(&pa->pa_inode_list); list_add(&pa->u.pa_tmp_list, &discard_list); total_entries--; if (total_entries <= 5) { /* * we want to keep only 5 entries * allowing it to grow to 8. This * mak sure we don't call discard * soon for this list. */ break; } } spin_unlock(&lg->lg_prealloc_lock); list_for_each_entry_safe(pa, tmp, &discard_list, u.pa_tmp_list) { ext4_get_group_no_and_offset(sb, pa->pa_pstart, &group, NULL); if (ext4_mb_load_buddy(sb, group, &e4b)) { |
12062dddd
|
4058 4059 |
ext4_error(sb, "Error loading buddy information for %u", group); |
6be2ded1d
|
4060 4061 4062 4063 |
continue; } ext4_lock_group(sb, group); list_del(&pa->pa_group_list); |
3e1e5f501
|
4064 |
ext4_mb_release_group_pa(&e4b, pa); |
6be2ded1d
|
4065 |
ext4_unlock_group(sb, group); |
e39e07fdf
|
4066 |
ext4_mb_unload_buddy(&e4b); |
6be2ded1d
|
4067 4068 4069 |
list_del(&pa->u.pa_tmp_list); call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback); } |
6be2ded1d
|
4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 |
} /* * We have incremented pa_count. So it cannot be freed at this * point. Also we hold lg_mutex. So no parallel allocation is * possible from this lg. That means pa_free cannot be updated. * * A parallel ext4_mb_discard_group_preallocations is possible. * which can cause the lg_prealloc_list to be updated. */ static void ext4_mb_add_n_trim(struct ext4_allocation_context *ac) { int order, added = 0, lg_prealloc_count = 1; struct super_block *sb = ac->ac_sb; struct ext4_locality_group *lg = ac->ac_lg; struct ext4_prealloc_space *tmp_pa, *pa = ac->ac_pa; order = fls(pa->pa_free) - 1; if (order > PREALLOC_TB_SIZE - 1) /* The max size of hash table is PREALLOC_TB_SIZE */ order = PREALLOC_TB_SIZE - 1; /* Add the prealloc space to lg */ rcu_read_lock(); list_for_each_entry_rcu(tmp_pa, &lg->lg_prealloc_list[order], pa_inode_list) { spin_lock(&tmp_pa->pa_lock); if (tmp_pa->pa_deleted) { |
e7c9e3e99
|
4098 |
spin_unlock(&tmp_pa->pa_lock); |
6be2ded1d
|
4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 |
continue; } if (!added && pa->pa_free < tmp_pa->pa_free) { /* Add to the tail of the previous entry */ list_add_tail_rcu(&pa->pa_inode_list, &tmp_pa->pa_inode_list); added = 1; /* * we want to count the total * number of entries in the list */ } spin_unlock(&tmp_pa->pa_lock); lg_prealloc_count++; } if (!added) list_add_tail_rcu(&pa->pa_inode_list, &lg->lg_prealloc_list[order]); rcu_read_unlock(); /* Now trim the list to be not more than 8 elements */ if (lg_prealloc_count > 8) { ext4_mb_discard_lg_preallocations(sb, lg, order, lg_prealloc_count); return; } return ; } |
c9de560de
|
4127 4128 4129 4130 4131 |
/* * release all resource we used in allocation */ static int ext4_mb_release_context(struct ext4_allocation_context *ac) { |
53accfa9f
|
4132 |
struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb); |
6be2ded1d
|
4133 4134 |
struct ext4_prealloc_space *pa = ac->ac_pa; if (pa) { |
cc0fb9ad7
|
4135 |
if (pa->pa_type == MB_GROUP_PA) { |
c9de560de
|
4136 |
/* see comment in ext4_mb_use_group_pa() */ |
6be2ded1d
|
4137 |
spin_lock(&pa->pa_lock); |
53accfa9f
|
4138 4139 |
pa->pa_pstart += EXT4_C2B(sbi, ac->ac_b_ex.fe_len); pa->pa_lstart += EXT4_C2B(sbi, ac->ac_b_ex.fe_len); |
6be2ded1d
|
4140 4141 4142 |
pa->pa_free -= ac->ac_b_ex.fe_len; pa->pa_len -= ac->ac_b_ex.fe_len; spin_unlock(&pa->pa_lock); |
c9de560de
|
4143 |
} |
c9de560de
|
4144 |
} |
ba4439165
|
4145 4146 4147 4148 4149 |
if (pa) { /* * We want to add the pa to the right bucket. * Remove it from the list and while adding * make sure the list to which we are adding |
44183d423
|
4150 |
* doesn't grow big. |
ba4439165
|
4151 |
*/ |
cc0fb9ad7
|
4152 |
if ((pa->pa_type == MB_GROUP_PA) && likely(pa->pa_free)) { |
ba4439165
|
4153 4154 4155 4156 4157 4158 4159 |
spin_lock(pa->pa_obj_lock); list_del_rcu(&pa->pa_inode_list); spin_unlock(pa->pa_obj_lock); ext4_mb_add_n_trim(ac); } ext4_mb_put_pa(ac, ac->ac_sb, pa); } |
c9de560de
|
4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 |
if (ac->ac_bitmap_page) page_cache_release(ac->ac_bitmap_page); if (ac->ac_buddy_page) page_cache_release(ac->ac_buddy_page); if (ac->ac_flags & EXT4_MB_HINT_GROUP_ALLOC) mutex_unlock(&ac->ac_lg->lg_mutex); ext4_mb_collect_stats(ac); return 0; } static int ext4_mb_discard_preallocations(struct super_block *sb, int needed) { |
8df9675f8
|
4172 |
ext4_group_t i, ngroups = ext4_get_groups_count(sb); |
c9de560de
|
4173 4174 |
int ret; int freed = 0; |
9bffad1ed
|
4175 |
trace_ext4_mb_discard_preallocations(sb, needed); |
8df9675f8
|
4176 |
for (i = 0; i < ngroups && needed > 0; i++) { |
c9de560de
|
4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 |
ret = ext4_mb_discard_group_preallocations(sb, i, needed); freed += ret; needed -= ret; } return freed; } /* * Main entry point into mballoc to allocate blocks * it tries to use preallocation first, then falls back * to usual allocation */ ext4_fsblk_t ext4_mb_new_blocks(handle_t *handle, |
6c7a120ac
|
4191 |
struct ext4_allocation_request *ar, int *errp) |
c9de560de
|
4192 |
{ |
6bc6e63fc
|
4193 |
int freed; |
256bdb497
|
4194 |
struct ext4_allocation_context *ac = NULL; |
c9de560de
|
4195 4196 4197 |
struct ext4_sb_info *sbi; struct super_block *sb; ext4_fsblk_t block = 0; |
60e58e0f3
|
4198 |
unsigned int inquota = 0; |
53accfa9f
|
4199 |
unsigned int reserv_clstrs = 0; |
c9de560de
|
4200 4201 4202 |
sb = ar->inode->i_sb; sbi = EXT4_SB(sb); |
9bffad1ed
|
4203 |
trace_ext4_request_blocks(ar); |
ba80b1019
|
4204 |
|
45dc63e7d
|
4205 4206 4207 |
/* Allow to use superuser reservation for quota file */ if (IS_NOQUOTA(ar->inode)) ar->flags |= EXT4_MB_USE_ROOT_BLOCKS; |
60e58e0f3
|
4208 4209 4210 4211 4212 |
/* * For delayed allocation, we could skip the ENOSPC and * EDQUOT check, as blocks and quotas have been already * reserved when data being copied into pagecache. */ |
f23210977
|
4213 |
if (ext4_test_inode_state(ar->inode, EXT4_STATE_DELALLOC_RESERVED)) |
60e58e0f3
|
4214 4215 4216 4217 4218 |
ar->flags |= EXT4_MB_DELALLOC_RESERVED; else { /* Without delayed allocation we need to verify * there is enough free blocks to do block allocation * and verify allocation doesn't exceed the quota limits. |
d2a176379
|
4219 |
*/ |
55f020db6
|
4220 |
while (ar->len && |
e7d5f3156
|
4221 |
ext4_claim_free_clusters(sbi, ar->len, ar->flags)) { |
55f020db6
|
4222 |
|
030ba6bc6
|
4223 4224 4225 4226 4227 |
/* let others to free the space */ yield(); ar->len = ar->len >> 1; } if (!ar->len) { |
a30d542a0
|
4228 4229 4230 |
*errp = -ENOSPC; return 0; } |
53accfa9f
|
4231 |
reserv_clstrs = ar->len; |
55f020db6
|
4232 |
if (ar->flags & EXT4_MB_USE_ROOT_BLOCKS) { |
53accfa9f
|
4233 4234 |
dquot_alloc_block_nofail(ar->inode, EXT4_C2B(sbi, ar->len)); |
55f020db6
|
4235 4236 |
} else { while (ar->len && |
53accfa9f
|
4237 4238 |
dquot_alloc_block(ar->inode, EXT4_C2B(sbi, ar->len))) { |
55f020db6
|
4239 4240 4241 4242 |
ar->flags |= EXT4_MB_HINT_NOPREALLOC; ar->len--; } |
60e58e0f3
|
4243 4244 4245 4246 |
} inquota = ar->len; if (ar->len == 0) { *errp = -EDQUOT; |
6c7a120ac
|
4247 |
goto out; |
60e58e0f3
|
4248 |
} |
070314310
|
4249 |
} |
d2a176379
|
4250 |
|
256bdb497
|
4251 |
ac = kmem_cache_alloc(ext4_ac_cachep, GFP_NOFS); |
833576b36
|
4252 |
if (!ac) { |
363d4251d
|
4253 |
ar->len = 0; |
256bdb497
|
4254 |
*errp = -ENOMEM; |
6c7a120ac
|
4255 |
goto out; |
256bdb497
|
4256 |
} |
256bdb497
|
4257 |
*errp = ext4_mb_initialize_context(ac, ar); |
c9de560de
|
4258 4259 |
if (*errp) { ar->len = 0; |
6c7a120ac
|
4260 |
goto out; |
c9de560de
|
4261 |
} |
256bdb497
|
4262 4263 |
ac->ac_op = EXT4_MB_HISTORY_PREALLOC; if (!ext4_mb_use_preallocated(ac)) { |
256bdb497
|
4264 4265 |
ac->ac_op = EXT4_MB_HISTORY_ALLOC; ext4_mb_normalize_request(ac, ar); |
c9de560de
|
4266 4267 |
repeat: /* allocate space in core */ |
6c7a120ac
|
4268 4269 4270 |
*errp = ext4_mb_regular_allocator(ac); if (*errp) goto errout; |
c9de560de
|
4271 4272 4273 4274 |
/* as we've just preallocated more space than * user requested orinally, we store allocated * space in a special descriptor */ |
256bdb497
|
4275 4276 4277 |
if (ac->ac_status == AC_STATUS_FOUND && ac->ac_o_ex.fe_len < ac->ac_b_ex.fe_len) ext4_mb_new_preallocation(ac); |
c9de560de
|
4278 |
} |
256bdb497
|
4279 |
if (likely(ac->ac_status == AC_STATUS_FOUND)) { |
53accfa9f
|
4280 |
*errp = ext4_mb_mark_diskspace_used(ac, handle, reserv_clstrs); |
6c7a120ac
|
4281 |
if (*errp == -EAGAIN) { |
8556e8f3b
|
4282 4283 4284 4285 4286 |
/* * drop the reference that we took * in ext4_mb_use_best_found */ ext4_mb_release_context(ac); |
519deca04
|
4287 4288 4289 4290 4291 |
ac->ac_b_ex.fe_group = 0; ac->ac_b_ex.fe_start = 0; ac->ac_b_ex.fe_len = 0; ac->ac_status = AC_STATUS_CONTINUE; goto repeat; |
6c7a120ac
|
4292 4293 |
} else if (*errp) errout: |
b844167ed
|
4294 |
ext4_discard_allocated_blocks(ac); |
6c7a120ac
|
4295 |
else { |
519deca04
|
4296 4297 4298 |
block = ext4_grp_offs_to_block(sb, &ac->ac_b_ex); ar->len = ac->ac_b_ex.fe_len; } |
c9de560de
|
4299 |
} else { |
256bdb497
|
4300 |
freed = ext4_mb_discard_preallocations(sb, ac->ac_o_ex.fe_len); |
c9de560de
|
4301 4302 4303 |
if (freed) goto repeat; *errp = -ENOSPC; |
6c7a120ac
|
4304 4305 4306 |
} if (*errp) { |
256bdb497
|
4307 |
ac->ac_b_ex.fe_len = 0; |
c9de560de
|
4308 |
ar->len = 0; |
256bdb497
|
4309 |
ext4_mb_show_ac(ac); |
c9de560de
|
4310 |
} |
256bdb497
|
4311 |
ext4_mb_release_context(ac); |
6c7a120ac
|
4312 4313 4314 |
out: if (ac) kmem_cache_free(ext4_ac_cachep, ac); |
60e58e0f3
|
4315 |
if (inquota && ar->len < inquota) |
53accfa9f
|
4316 |
dquot_free_block(ar->inode, EXT4_C2B(sbi, inquota - ar->len)); |
0087d9fb3
|
4317 |
if (!ar->len) { |
f23210977
|
4318 4319 |
if (!ext4_test_inode_state(ar->inode, EXT4_STATE_DELALLOC_RESERVED)) |
0087d9fb3
|
4320 |
/* release all the reserved blocks if non delalloc */ |
570426518
|
4321 |
percpu_counter_sub(&sbi->s_dirtyclusters_counter, |
53accfa9f
|
4322 |
reserv_clstrs); |
0087d9fb3
|
4323 |
} |
c9de560de
|
4324 |
|
9bffad1ed
|
4325 |
trace_ext4_allocate_blocks(ar, (unsigned long long)block); |
ba80b1019
|
4326 |
|
c9de560de
|
4327 4328 |
return block; } |
c9de560de
|
4329 |
|
c894058d6
|
4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 |
/* * We can merge two free data extents only if the physical blocks * are contiguous, AND the extents were freed by the same transaction, * AND the blocks are associated with the same group. */ static int can_merge(struct ext4_free_data *entry1, struct ext4_free_data *entry2) { if ((entry1->t_tid == entry2->t_tid) && (entry1->group == entry2->group) && |
84130193e
|
4340 |
((entry1->start_cluster + entry1->count) == entry2->start_cluster)) |
c894058d6
|
4341 4342 4343 |
return 1; return 0; } |
4ddfef7b4
|
4344 4345 |
static noinline_for_stack int ext4_mb_free_metadata(handle_t *handle, struct ext4_buddy *e4b, |
7a2fcbf7f
|
4346 |
struct ext4_free_data *new_entry) |
c9de560de
|
4347 |
{ |
e29136f80
|
4348 |
ext4_group_t group = e4b->bd_group; |
84130193e
|
4349 |
ext4_grpblk_t cluster; |
7a2fcbf7f
|
4350 |
struct ext4_free_data *entry; |
c9de560de
|
4351 4352 4353 |
struct ext4_group_info *db = e4b->bd_info; struct super_block *sb = e4b->bd_sb; struct ext4_sb_info *sbi = EXT4_SB(sb); |
c894058d6
|
4354 4355 |
struct rb_node **n = &db->bb_free_root.rb_node, *node; struct rb_node *parent = NULL, *new_node; |
0390131ba
|
4356 |
BUG_ON(!ext4_handle_valid(handle)); |
c9de560de
|
4357 4358 |
BUG_ON(e4b->bd_bitmap_page == NULL); BUG_ON(e4b->bd_buddy_page == NULL); |
c894058d6
|
4359 |
new_node = &new_entry->node; |
84130193e
|
4360 |
cluster = new_entry->start_cluster; |
c894058d6
|
4361 |
|
c894058d6
|
4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 |
if (!*n) { /* first free block exent. We need to protect buddy cache from being freed, * otherwise we'll refresh it from * on-disk bitmap and lose not-yet-available * blocks */ page_cache_get(e4b->bd_buddy_page); page_cache_get(e4b->bd_bitmap_page); } while (*n) { parent = *n; entry = rb_entry(parent, struct ext4_free_data, node); |
84130193e
|
4374 |
if (cluster < entry->start_cluster) |
c894058d6
|
4375 |
n = &(*n)->rb_left; |
84130193e
|
4376 |
else if (cluster >= (entry->start_cluster + entry->count)) |
c894058d6
|
4377 4378 |
n = &(*n)->rb_right; else { |
e29136f80
|
4379 |
ext4_grp_locked_error(sb, group, 0, |
84130193e
|
4380 4381 |
ext4_group_first_block_no(sb, group) + EXT4_C2B(sbi, cluster), |
e29136f80
|
4382 |
"Block already on to-be-freed list"); |
c894058d6
|
4383 |
return 0; |
c9de560de
|
4384 |
} |
c894058d6
|
4385 |
} |
c9de560de
|
4386 |
|
c894058d6
|
4387 4388 4389 4390 4391 4392 4393 4394 |
rb_link_node(new_node, parent, n); rb_insert_color(new_node, &db->bb_free_root); /* Now try to see the extent can be merged to left and right */ node = rb_prev(new_node); if (node) { entry = rb_entry(node, struct ext4_free_data, node); if (can_merge(entry, new_entry)) { |
84130193e
|
4395 |
new_entry->start_cluster = entry->start_cluster; |
c894058d6
|
4396 4397 4398 4399 4400 4401 |
new_entry->count += entry->count; rb_erase(node, &(db->bb_free_root)); spin_lock(&sbi->s_md_lock); list_del(&entry->list); spin_unlock(&sbi->s_md_lock); kmem_cache_free(ext4_free_ext_cachep, entry); |
c9de560de
|
4402 |
} |
c894058d6
|
4403 |
} |
c9de560de
|
4404 |
|
c894058d6
|
4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 |
node = rb_next(new_node); if (node) { entry = rb_entry(node, struct ext4_free_data, node); if (can_merge(new_entry, entry)) { new_entry->count += entry->count; rb_erase(node, &(db->bb_free_root)); spin_lock(&sbi->s_md_lock); list_del(&entry->list); spin_unlock(&sbi->s_md_lock); kmem_cache_free(ext4_free_ext_cachep, entry); |
c9de560de
|
4415 4416 |
} } |
3e624fc72
|
4417 |
/* Add the extent to transaction's private list */ |
c894058d6
|
4418 |
spin_lock(&sbi->s_md_lock); |
3e624fc72
|
4419 |
list_add(&new_entry->list, &handle->h_transaction->t_private_list); |
c894058d6
|
4420 |
spin_unlock(&sbi->s_md_lock); |
c9de560de
|
4421 4422 |
return 0; } |
443387113
|
4423 4424 4425 4426 4427 4428 |
/** * ext4_free_blocks() -- Free given blocks and update quota * @handle: handle for this transaction * @inode: inode * @block: start physical block to free * @count: number of blocks to count |
5def13602
|
4429 |
* @flags: flags used by ext4_free_blocks |
c9de560de
|
4430 |
*/ |
443387113
|
4431 |
void ext4_free_blocks(handle_t *handle, struct inode *inode, |
e6362609b
|
4432 4433 |
struct buffer_head *bh, ext4_fsblk_t block, unsigned long count, int flags) |
c9de560de
|
4434 |
{ |
26346ff68
|
4435 |
struct buffer_head *bitmap_bh = NULL; |
c9de560de
|
4436 |
struct super_block *sb = inode->i_sb; |
c9de560de
|
4437 |
struct ext4_group_desc *gdp; |
443387113
|
4438 |
unsigned long freed = 0; |
498e5f241
|
4439 |
unsigned int overflow; |
c9de560de
|
4440 4441 4442 4443 4444 |
ext4_grpblk_t bit; struct buffer_head *gd_bh; ext4_group_t block_group; struct ext4_sb_info *sbi; struct ext4_buddy e4b; |
84130193e
|
4445 |
unsigned int count_clusters; |
c9de560de
|
4446 4447 |
int err = 0; int ret; |
e6362609b
|
4448 4449 4450 4451 4452 4453 |
if (bh) { if (block) BUG_ON(block != bh->b_blocknr); else block = bh->b_blocknr; } |
c9de560de
|
4454 |
|
c9de560de
|
4455 |
sbi = EXT4_SB(sb); |
1f2acb601
|
4456 4457 |
if (!(flags & EXT4_FREE_BLOCKS_VALIDATED) && !ext4_data_block_valid(sbi, block, count)) { |
12062dddd
|
4458 |
ext4_error(sb, "Freeing blocks not in datazone - " |
1f2acb601
|
4459 |
"block = %llu, count = %lu", block, count); |
c9de560de
|
4460 4461 |
goto error_return; } |
0610b6e99
|
4462 4463 |
ext4_debug("freeing block %llu ", block); |
e6362609b
|
4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 |
trace_ext4_free_blocks(inode, block, count, flags); if (flags & EXT4_FREE_BLOCKS_FORGET) { struct buffer_head *tbh = bh; int i; BUG_ON(bh && (count > 1)); for (i = 0; i < count; i++) { if (!bh) tbh = sb_find_get_block(inode->i_sb, block + i); |
877836905
|
4476 4477 |
if (unlikely(!tbh)) continue; |
60e6679e2
|
4478 |
ext4_forget(handle, flags & EXT4_FREE_BLOCKS_METADATA, |
e6362609b
|
4479 4480 4481 |
inode, tbh, block + i); } } |
60e6679e2
|
4482 |
/* |
e6362609b
|
4483 4484 4485 4486 4487 4488 4489 4490 |
* We need to make sure we don't reuse the freed block until * after the transaction is committed, which we can do by * treating the block as metadata, below. We make an * exception if the inode is to be written in writeback mode * since writeback mode has weak data consistency guarantees. */ if (!ext4_should_writeback_data(inode)) flags |= EXT4_FREE_BLOCKS_METADATA; |
c9de560de
|
4491 |
|
84130193e
|
4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 |
/* * If the extent to be freed does not begin on a cluster * boundary, we need to deal with partial clusters at the * beginning and end of the extent. Normally we will free * blocks at the beginning or the end unless we are explicitly * requested to avoid doing so. */ overflow = block & (sbi->s_cluster_ratio - 1); if (overflow) { if (flags & EXT4_FREE_BLOCKS_NOFREE_FIRST_CLUSTER) { overflow = sbi->s_cluster_ratio - overflow; block += overflow; if (count > overflow) count -= overflow; else return; } else { block -= overflow; count += overflow; } } overflow = count & (sbi->s_cluster_ratio - 1); if (overflow) { if (flags & EXT4_FREE_BLOCKS_NOFREE_LAST_CLUSTER) { if (count > overflow) count -= overflow; else return; } else count += sbi->s_cluster_ratio - overflow; } |
c9de560de
|
4523 4524 4525 4526 4527 4528 4529 4530 |
do_more: overflow = 0; ext4_get_group_no_and_offset(sb, block, &block_group, &bit); /* * Check to see if we are freeing blocks across a group * boundary. */ |
84130193e
|
4531 4532 4533 |
if (EXT4_C2B(sbi, bit) + count > EXT4_BLOCKS_PER_GROUP(sb)) { overflow = EXT4_C2B(sbi, bit) + count - EXT4_BLOCKS_PER_GROUP(sb); |
c9de560de
|
4534 4535 |
count -= overflow; } |
84130193e
|
4536 |
count_clusters = EXT4_B2C(sbi, count); |
574ca174c
|
4537 |
bitmap_bh = ext4_read_block_bitmap(sb, block_group); |
ce89f46cb
|
4538 4539 |
if (!bitmap_bh) { err = -EIO; |
c9de560de
|
4540 |
goto error_return; |
ce89f46cb
|
4541 |
} |
c9de560de
|
4542 |
gdp = ext4_get_group_desc(sb, block_group, &gd_bh); |
ce89f46cb
|
4543 4544 |
if (!gdp) { err = -EIO; |
c9de560de
|
4545 |
goto error_return; |
ce89f46cb
|
4546 |
} |
c9de560de
|
4547 4548 4549 4550 |
if (in_range(ext4_block_bitmap(sb, gdp), block, count) || in_range(ext4_inode_bitmap(sb, gdp), block, count) || in_range(block, ext4_inode_table(sb, gdp), |
84130193e
|
4551 |
EXT4_SB(sb)->s_itb_per_group) || |
c9de560de
|
4552 |
in_range(block + count - 1, ext4_inode_table(sb, gdp), |
84130193e
|
4553 |
EXT4_SB(sb)->s_itb_per_group)) { |
c9de560de
|
4554 |
|
12062dddd
|
4555 |
ext4_error(sb, "Freeing blocks in system zone - " |
0610b6e99
|
4556 |
"Block = %llu, count = %lu", block, count); |
519deca04
|
4557 4558 |
/* err = 0. ext4_std_error should be a no op */ goto error_return; |
c9de560de
|
4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 |
} BUFFER_TRACE(bitmap_bh, "getting write access"); err = ext4_journal_get_write_access(handle, bitmap_bh); if (err) goto error_return; /* * We are about to modify some metadata. Call the journal APIs * to unshare ->b_data if a currently-committing transaction is * using it */ BUFFER_TRACE(gd_bh, "get_write_access"); err = ext4_journal_get_write_access(handle, gd_bh); if (err) goto error_return; |
c9de560de
|
4575 4576 4577 |
#ifdef AGGRESSIVE_CHECK { int i; |
84130193e
|
4578 |
for (i = 0; i < count_clusters; i++) |
c9de560de
|
4579 4580 4581 |
BUG_ON(!mb_test_bit(bit + i, bitmap_bh->b_data)); } #endif |
84130193e
|
4582 |
trace_ext4_mballoc_free(sb, inode, block_group, bit, count_clusters); |
c9de560de
|
4583 |
|
920313a72
|
4584 4585 4586 |
err = ext4_mb_load_buddy(sb, block_group, &e4b); if (err) goto error_return; |
e6362609b
|
4587 4588 |
if ((flags & EXT4_FREE_BLOCKS_METADATA) && ext4_handle_valid(handle)) { |
7a2fcbf7f
|
4589 4590 4591 4592 4593 |
struct ext4_free_data *new_entry; /* * blocks being freed are metadata. these blocks shouldn't * be used until this transaction is committed */ |
b72143ab3
|
4594 4595 4596 4597 4598 |
new_entry = kmem_cache_alloc(ext4_free_ext_cachep, GFP_NOFS); if (!new_entry) { err = -ENOMEM; goto error_return; } |
84130193e
|
4599 |
new_entry->start_cluster = bit; |
7a2fcbf7f
|
4600 |
new_entry->group = block_group; |
84130193e
|
4601 |
new_entry->count = count_clusters; |
7a2fcbf7f
|
4602 |
new_entry->t_tid = handle->h_transaction->t_tid; |
955ce5f5b
|
4603 |
|
7a2fcbf7f
|
4604 |
ext4_lock_group(sb, block_group); |
84130193e
|
4605 |
mb_clear_bits(bitmap_bh->b_data, bit, count_clusters); |
7a2fcbf7f
|
4606 |
ext4_mb_free_metadata(handle, &e4b, new_entry); |
c9de560de
|
4607 |
} else { |
7a2fcbf7f
|
4608 4609 4610 4611 |
/* need to update group_info->bb_free and bitmap * with group lock held. generate_buddy look at * them with group lock_held */ |
955ce5f5b
|
4612 |
ext4_lock_group(sb, block_group); |
84130193e
|
4613 4614 |
mb_clear_bits(bitmap_bh->b_data, bit, count_clusters); mb_free_blocks(inode, &e4b, bit, count_clusters); |
c9de560de
|
4615 |
} |
021b65bb1
|
4616 4617 |
ret = ext4_free_group_clusters(sb, gdp) + count_clusters; ext4_free_group_clusters_set(sb, gdp, ret); |
c9de560de
|
4618 |
gdp->bg_checksum = ext4_group_desc_csum(sbi, block_group, gdp); |
955ce5f5b
|
4619 |
ext4_unlock_group(sb, block_group); |
570426518
|
4620 |
percpu_counter_add(&sbi->s_freeclusters_counter, count_clusters); |
c9de560de
|
4621 |
|
772cb7c83
|
4622 4623 |
if (sbi->s_log_groups_per_flex) { ext4_group_t flex_group = ext4_flex_group(sbi, block_group); |
24aaa8ef4
|
4624 4625 |
atomic_add(count_clusters, &sbi->s_flex_groups[flex_group].free_clusters); |
772cb7c83
|
4626 |
} |
e39e07fdf
|
4627 |
ext4_mb_unload_buddy(&e4b); |
c9de560de
|
4628 |
|
443387113
|
4629 |
freed += count; |
c9de560de
|
4630 |
|
7b415bf60
|
4631 4632 |
if (!(flags & EXT4_FREE_BLOCKS_NO_QUOT_UPDATE)) dquot_free_block(inode, EXT4_C2B(sbi, count_clusters)); |
7a2fcbf7f
|
4633 4634 4635 |
/* We dirtied the bitmap block */ BUFFER_TRACE(bitmap_bh, "dirtied bitmap block"); err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh); |
c9de560de
|
4636 4637 |
/* And the group descriptor block */ BUFFER_TRACE(gd_bh, "dirtied group descriptor block"); |
0390131ba
|
4638 |
ret = ext4_handle_dirty_metadata(handle, NULL, gd_bh); |
c9de560de
|
4639 4640 4641 4642 4643 4644 4645 4646 4647 |
if (!err) err = ret; if (overflow && !err) { block += count; count = overflow; put_bh(bitmap_bh); goto do_more; } |
a0375156c
|
4648 |
ext4_mark_super_dirty(sb); |
c9de560de
|
4649 4650 4651 4652 4653 |
error_return: brelse(bitmap_bh); ext4_std_error(sb, err); return; } |
7360d1731
|
4654 4655 |
/** |
0529155e8
|
4656 |
* ext4_group_add_blocks() -- Add given blocks to an existing group |
2846e8200
|
4657 4658 4659 4660 4661 |
* @handle: handle to this transaction * @sb: super block * @block: start physcial block to add to the block group * @count: number of blocks to free * |
e73a347b7
|
4662 |
* This marks the blocks as free in the bitmap and buddy. |
2846e8200
|
4663 |
*/ |
cc7365dfe
|
4664 |
int ext4_group_add_blocks(handle_t *handle, struct super_block *sb, |
2846e8200
|
4665 4666 4667 4668 4669 4670 4671 4672 4673 |
ext4_fsblk_t block, unsigned long count) { struct buffer_head *bitmap_bh = NULL; struct buffer_head *gd_bh; ext4_group_t block_group; ext4_grpblk_t bit; unsigned int i; struct ext4_group_desc *desc; struct ext4_sb_info *sbi = EXT4_SB(sb); |
e73a347b7
|
4674 |
struct ext4_buddy e4b; |
2846e8200
|
4675 4676 |
int err = 0, ret, blk_free_count; ext4_grpblk_t blocks_freed; |
2846e8200
|
4677 4678 4679 |
ext4_debug("Adding block(s) %llu-%llu ", block, block + count - 1); |
4740b830e
|
4680 4681 |
if (count == 0) return 0; |
2846e8200
|
4682 |
ext4_get_group_no_and_offset(sb, block, &block_group, &bit); |
2846e8200
|
4683 4684 4685 4686 |
/* * Check to see if we are freeing blocks across a group * boundary. */ |
cc7365dfe
|
4687 4688 4689 4690 4691 |
if (bit + count > EXT4_BLOCKS_PER_GROUP(sb)) { ext4_warning(sb, "too much blocks added to group %u ", block_group); err = -EINVAL; |
2846e8200
|
4692 |
goto error_return; |
cc7365dfe
|
4693 |
} |
2cd05cc39
|
4694 |
|
2846e8200
|
4695 |
bitmap_bh = ext4_read_block_bitmap(sb, block_group); |
cc7365dfe
|
4696 4697 |
if (!bitmap_bh) { err = -EIO; |
2846e8200
|
4698 |
goto error_return; |
cc7365dfe
|
4699 |
} |
2846e8200
|
4700 |
desc = ext4_get_group_desc(sb, block_group, &gd_bh); |
cc7365dfe
|
4701 4702 |
if (!desc) { err = -EIO; |
2846e8200
|
4703 |
goto error_return; |
cc7365dfe
|
4704 |
} |
2846e8200
|
4705 4706 4707 4708 4709 4710 4711 4712 4713 |
if (in_range(ext4_block_bitmap(sb, desc), block, count) || in_range(ext4_inode_bitmap(sb, desc), block, count) || in_range(block, ext4_inode_table(sb, desc), sbi->s_itb_per_group) || in_range(block + count - 1, ext4_inode_table(sb, desc), sbi->s_itb_per_group)) { ext4_error(sb, "Adding blocks in system zones - " "Block = %llu, count = %lu", block, count); |
cc7365dfe
|
4714 |
err = -EINVAL; |
2846e8200
|
4715 4716 |
goto error_return; } |
2cd05cc39
|
4717 4718 |
BUFFER_TRACE(bitmap_bh, "getting write access"); err = ext4_journal_get_write_access(handle, bitmap_bh); |
2846e8200
|
4719 4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 |
if (err) goto error_return; /* * We are about to modify some metadata. Call the journal APIs * to unshare ->b_data if a currently-committing transaction is * using it */ BUFFER_TRACE(gd_bh, "get_write_access"); err = ext4_journal_get_write_access(handle, gd_bh); if (err) goto error_return; |
e73a347b7
|
4731 |
|
2846e8200
|
4732 4733 |
for (i = 0, blocks_freed = 0; i < count; i++) { BUFFER_TRACE(bitmap_bh, "clear bit"); |
e73a347b7
|
4734 |
if (!mb_test_bit(bit + i, bitmap_bh->b_data)) { |
2846e8200
|
4735 4736 4737 4738 4739 4740 4741 |
ext4_error(sb, "bit already cleared for block %llu", (ext4_fsblk_t)(block + i)); BUFFER_TRACE(bitmap_bh, "bit already cleared"); } else { blocks_freed++; } } |
e73a347b7
|
4742 4743 4744 4745 4746 4747 4748 4749 4750 4751 |
err = ext4_mb_load_buddy(sb, block_group, &e4b); if (err) goto error_return; /* * need to update group_info->bb_free and bitmap * with group lock held. generate_buddy look at * them with group lock_held */ |
2846e8200
|
4752 |
ext4_lock_group(sb, block_group); |
e73a347b7
|
4753 4754 |
mb_clear_bits(bitmap_bh->b_data, bit, count); mb_free_blocks(NULL, &e4b, bit, count); |
021b65bb1
|
4755 4756 |
blk_free_count = blocks_freed + ext4_free_group_clusters(sb, desc); ext4_free_group_clusters_set(sb, desc, blk_free_count); |
2846e8200
|
4757 4758 |
desc->bg_checksum = ext4_group_desc_csum(sbi, block_group, desc); ext4_unlock_group(sb, block_group); |
570426518
|
4759 4760 |
percpu_counter_add(&sbi->s_freeclusters_counter, EXT4_B2C(sbi, blocks_freed)); |
2846e8200
|
4761 4762 4763 |
if (sbi->s_log_groups_per_flex) { ext4_group_t flex_group = ext4_flex_group(sbi, block_group); |
24aaa8ef4
|
4764 4765 |
atomic_add(EXT4_B2C(sbi, blocks_freed), &sbi->s_flex_groups[flex_group].free_clusters); |
2846e8200
|
4766 |
} |
e73a347b7
|
4767 4768 |
ext4_mb_unload_buddy(&e4b); |
2846e8200
|
4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 |
/* We dirtied the bitmap block */ BUFFER_TRACE(bitmap_bh, "dirtied bitmap block"); err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh); /* And the group descriptor block */ BUFFER_TRACE(gd_bh, "dirtied group descriptor block"); ret = ext4_handle_dirty_metadata(handle, NULL, gd_bh); if (!err) err = ret; error_return: brelse(bitmap_bh); ext4_std_error(sb, err); |
cc7365dfe
|
4783 |
return err; |
2846e8200
|
4784 4785 4786 |
} /** |
7360d1731
|
4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 |
* ext4_trim_extent -- function to TRIM one single free extent in the group * @sb: super block for the file system * @start: starting block of the free extent in the alloc. group * @count: number of blocks to TRIM * @group: alloc. group we are working with * @e4b: ext4 buddy for the group * * Trim "count" blocks starting at "start" in the "group". To assure that no * one will allocate those blocks, mark it as used in buddy bitmap. This must * be called with under the group lock. */ |
d9f34504e
|
4798 4799 |
static void ext4_trim_extent(struct super_block *sb, int start, int count, ext4_group_t group, struct ext4_buddy *e4b) |
7360d1731
|
4800 4801 |
{ struct ext4_free_extent ex; |
7360d1731
|
4802 |
|
b3d4c2b10
|
4803 |
trace_ext4_trim_extent(sb, group, start, count); |
7360d1731
|
4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 |
assert_spin_locked(ext4_group_lock_ptr(sb, group)); ex.fe_start = start; ex.fe_group = group; ex.fe_len = count; /* * Mark blocks used, so no one can reuse them while * being trimmed. */ mb_mark_used(e4b, &ex); ext4_unlock_group(sb, group); |
d9f34504e
|
4816 |
ext4_issue_discard(sb, group, start, count); |
7360d1731
|
4817 4818 |
ext4_lock_group(sb, group); mb_free_blocks(NULL, e4b, start, ex.fe_len); |
7360d1731
|
4819 4820 4821 4822 4823 |
} /** * ext4_trim_all_free -- function to trim all free space in alloc. group * @sb: super block for file system |
22612283f
|
4824 |
* @group: group to be trimmed |
7360d1731
|
4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 |
* @start: first group block to examine * @max: last group block to examine * @minblocks: minimum extent block count * * ext4_trim_all_free walks through group's buddy bitmap searching for free * extents. When the free block is found, ext4_trim_extent is called to TRIM * the extent. * * * ext4_trim_all_free walks through group's block bitmap searching for free * extents. When the free extent is found, mark it as used in group buddy * bitmap. Then issue a TRIM command on this extent and free the extent in * the group buddy bitmap. This is done until whole group is scanned. */ |
0b75a8401
|
4839 |
static ext4_grpblk_t |
789440866
|
4840 4841 4842 |
ext4_trim_all_free(struct super_block *sb, ext4_group_t group, ext4_grpblk_t start, ext4_grpblk_t max, ext4_grpblk_t minblocks) |
7360d1731
|
4843 4844 |
{ void *bitmap; |
169ddc3ec
|
4845 |
ext4_grpblk_t next, count = 0, free_count = 0; |
789440866
|
4846 4847 |
struct ext4_buddy e4b; int ret; |
7360d1731
|
4848 |
|
b3d4c2b10
|
4849 |
trace_ext4_trim_all_free(sb, group, start, max); |
789440866
|
4850 4851 4852 4853 4854 4855 |
ret = ext4_mb_load_buddy(sb, group, &e4b); if (ret) { ext4_error(sb, "Error in loading buddy " "information for %u", group); return ret; } |
789440866
|
4856 |
bitmap = e4b.bd_bitmap; |
28739eea9
|
4857 4858 |
ext4_lock_group(sb, group); |
3d56b8d2c
|
4859 4860 4861 |
if (EXT4_MB_GRP_WAS_TRIMMED(e4b.bd_info) && minblocks >= atomic_read(&EXT4_SB(sb)->s_last_trim_minblks)) goto out; |
789440866
|
4862 4863 |
start = (e4b.bd_info->bb_first_free > start) ? e4b.bd_info->bb_first_free : start; |
7360d1731
|
4864 4865 4866 4867 4868 4869 4870 4871 |
while (start < max) { start = mb_find_next_zero_bit(bitmap, max, start); if (start >= max) break; next = mb_find_next_bit(bitmap, max, start); if ((next - start) >= minblocks) { |
d9f34504e
|
4872 |
ext4_trim_extent(sb, start, |
789440866
|
4873 |
next - start, group, &e4b); |
7360d1731
|
4874 4875 |
count += next - start; } |
169ddc3ec
|
4876 |
free_count += next - start; |
7360d1731
|
4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 |
start = next + 1; if (fatal_signal_pending(current)) { count = -ERESTARTSYS; break; } if (need_resched()) { ext4_unlock_group(sb, group); cond_resched(); ext4_lock_group(sb, group); } |
169ddc3ec
|
4889 |
if ((e4b.bd_info->bb_free - free_count) < minblocks) |
7360d1731
|
4890 4891 |
break; } |
3d56b8d2c
|
4892 4893 4894 4895 |
if (!ret) EXT4_MB_GRP_SET_TRIMMED(e4b.bd_info); out: |
7360d1731
|
4896 |
ext4_unlock_group(sb, group); |
789440866
|
4897 |
ext4_mb_unload_buddy(&e4b); |
7360d1731
|
4898 4899 4900 4901 |
ext4_debug("trimmed %d blocks in the group %d ", count, group); |
7360d1731
|
4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918 |
return count; } /** * ext4_trim_fs() -- trim ioctl handle function * @sb: superblock for filesystem * @range: fstrim_range structure * * start: First Byte to trim * len: number of Bytes to trim from start * minlen: minimum extent length in Bytes * ext4_trim_fs goes through all allocation groups containing Bytes from * start to start+len. For each such a group ext4_trim_all_free function * is invoked to trim all free space. */ int ext4_trim_fs(struct super_block *sb, struct fstrim_range *range) { |
789440866
|
4919 |
struct ext4_group_info *grp; |
7360d1731
|
4920 4921 |
ext4_group_t first_group, last_group; ext4_group_t group, ngroups = ext4_get_groups_count(sb); |
7137d7a48
|
4922 |
ext4_grpblk_t cnt = 0, first_cluster, last_cluster; |
789440866
|
4923 |
uint64_t start, len, minlen, trimmed = 0; |
0f0a25bf5
|
4924 4925 |
ext4_fsblk_t first_data_blk = le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block); |
7360d1731
|
4926 4927 4928 4929 4930 |
int ret = 0; start = range->start >> sb->s_blocksize_bits; len = range->len >> sb->s_blocksize_bits; minlen = range->minlen >> sb->s_blocksize_bits; |
7360d1731
|
4931 |
|
7137d7a48
|
4932 |
if (unlikely(minlen > EXT4_CLUSTERS_PER_GROUP(sb))) |
7360d1731
|
4933 |
return -EINVAL; |
22f104574
|
4934 4935 |
if (start + len <= first_data_blk) goto out; |
0f0a25bf5
|
4936 4937 4938 4939 |
if (start < first_data_blk) { len -= first_data_blk - start; start = first_data_blk; } |
7360d1731
|
4940 4941 4942 |
/* Determine first and last group to examine based on start and len */ ext4_get_group_no_and_offset(sb, (ext4_fsblk_t) start, |
7137d7a48
|
4943 |
&first_group, &first_cluster); |
7360d1731
|
4944 |
ext4_get_group_no_and_offset(sb, (ext4_fsblk_t) (start + len), |
7137d7a48
|
4945 |
&last_group, &last_cluster); |
7360d1731
|
4946 |
last_group = (last_group > ngroups - 1) ? ngroups - 1 : last_group; |
7137d7a48
|
4947 |
last_cluster = EXT4_CLUSTERS_PER_GROUP(sb); |
7360d1731
|
4948 4949 4950 4951 4952 |
if (first_group > last_group) return -EINVAL; for (group = first_group; group <= last_group; group++) { |
789440866
|
4953 4954 4955 4956 4957 4958 |
grp = ext4_get_group_info(sb, group); /* We only do this if the grp has never been initialized */ if (unlikely(EXT4_MB_GRP_NEED_INIT(grp))) { ret = ext4_mb_init_group(sb, group); if (ret) break; |
7360d1731
|
4959 |
} |
0ba085171
|
4960 4961 4962 4963 4964 4965 |
/* * For all the groups except the last one, last block will * always be EXT4_BLOCKS_PER_GROUP(sb), so we only need to * change it for the last group in which case start + * len < EXT4_BLOCKS_PER_GROUP(sb). */ |
7137d7a48
|
4966 4967 4968 |
if (first_cluster + len < EXT4_CLUSTERS_PER_GROUP(sb)) last_cluster = first_cluster + len; len -= last_cluster - first_cluster; |
7360d1731
|
4969 |
|
789440866
|
4970 |
if (grp->bb_free >= minlen) { |
7137d7a48
|
4971 4972 |
cnt = ext4_trim_all_free(sb, group, first_cluster, last_cluster, minlen); |
7360d1731
|
4973 4974 |
if (cnt < 0) { ret = cnt; |
7360d1731
|
4975 4976 4977 |
break; } } |
7360d1731
|
4978 |
trimmed += cnt; |
7137d7a48
|
4979 |
first_cluster = 0; |
7360d1731
|
4980 4981 |
} range->len = trimmed * sb->s_blocksize; |
3d56b8d2c
|
4982 4983 |
if (!ret) atomic_set(&EXT4_SB(sb)->s_last_trim_minblks, minlen); |
22f104574
|
4984 |
out: |
7360d1731
|
4985 4986 |
return ret; } |