aboutsummaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@woody.linux-foundation.org>2007-07-18 13:32:00 -0400
committerLinus Torvalds <torvalds@woody.linux-foundation.org>2007-07-18 13:32:00 -0400
commitd756d10e246a01515d07f8161181b8a14afba7cc (patch)
treede7336f2b4b596881468bf65cb2f2f88cedcde86 /fs
parentcdf4a6482dd4c739f8c1132c5a9356912911fec5 (diff)
parente9f410b1c035b6e63f0b4c3d6cfe4298d6a04492 (diff)
Merge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4
* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: ext4: extent macros cleanup Fix compilation with EXT_DEBUG, also fix leXX_to_cpu conversions. ext4: remove extra IS_RDONLY() check ext4: Use is_power_of_2() Use zero_user_page() in ext4 where possible ext4: Remove 65000 subdirectory limit ext4: Expand extra_inodes space per the s_{want,min}_extra_isize fields ext4: Add nanosecond timestamps jbd2: Move jbd2-debug file to debugfs jbd2: Fix CONFIG_JBD_DEBUG ifdef to be CONFIG_JBD2_DEBUG ext4: Set the journal JBD2_FEATURE_INCOMPAT_64BIT on large devices ext4: Make extents code sanely handle on-disk corruption ext4: copy i_flags to inode flags on write ext4: Enable extents by default Change on-disk format to support 2^15 uninitialized extents write support for preallocated blocks fallocate support in ext4 sys_fallocate() implementation on i386, x86_64 and powerpc
Diffstat (limited to 'fs')
-rw-r--r--fs/Kconfig10
-rw-r--r--fs/ext4/balloc.c4
-rw-r--r--fs/ext4/extents.c682
-rw-r--r--fs/ext4/file.c1
-rw-r--r--fs/ext4/ialloc.c8
-rw-r--r--fs/ext4/inode.c118
-rw-r--r--fs/ext4/ioctl.c9
-rw-r--r--fs/ext4/namei.c76
-rw-r--r--fs/ext4/super.c50
-rw-r--r--fs/ext4/xattr.c276
-rw-r--r--fs/ext4/xattr.h17
-rw-r--r--fs/jbd2/journal.c79
-rw-r--r--fs/jbd2/recovery.c2
-rw-r--r--fs/open.c59
14 files changed, 1143 insertions, 248 deletions
diff --git a/fs/Kconfig b/fs/Kconfig
index 613df554728d..6a649902c5ac 100644
--- a/fs/Kconfig
+++ b/fs/Kconfig
@@ -251,7 +251,7 @@ config JBD2
251 251
252config JBD2_DEBUG 252config JBD2_DEBUG
253 bool "JBD2 (ext4dev/ext4) debugging support" 253 bool "JBD2 (ext4dev/ext4) debugging support"
254 depends on JBD2 254 depends on JBD2 && DEBUG_FS
255 help 255 help
256 If you are using the ext4dev/ext4 journaled file system (or 256 If you are using the ext4dev/ext4 journaled file system (or
257 potentially any other filesystem/device using JBD2), this option 257 potentially any other filesystem/device using JBD2), this option
@@ -260,10 +260,10 @@ config JBD2_DEBUG
260 By default, the debugging output will be turned off. 260 By default, the debugging output will be turned off.
261 261
262 If you select Y here, then you will be able to turn on debugging 262 If you select Y here, then you will be able to turn on debugging
263 with "echo N > /proc/sys/fs/jbd2-debug", where N is a number between 263 with "echo N > /sys/kernel/debug/jbd2/jbd2-debug", where N is a
264 1 and 5. The higher the number, the more debugging output is 264 number between 1 and 5. The higher the number, the more debugging
265 generated. To turn debugging off again, do 265 output is generated. To turn debugging off again, do
266 "echo 0 > /proc/sys/fs/jbd2-debug". 266 "echo 0 > /sys/kernel/debug/jbd2/jbd2-debug".
267 267
268config FS_MBCACHE 268config FS_MBCACHE
269# Meta block cache for Extended Attributes (ext2/ext3/ext4) 269# Meta block cache for Extended Attributes (ext2/ext3/ext4)
diff --git a/fs/ext4/balloc.c b/fs/ext4/balloc.c
index 9de54ae48dee..e53b4af52f11 100644
--- a/fs/ext4/balloc.c
+++ b/fs/ext4/balloc.c
@@ -517,7 +517,7 @@ do_more:
517 /* 517 /*
518 * An HJ special. This is expensive... 518 * An HJ special. This is expensive...
519 */ 519 */
520#ifdef CONFIG_JBD_DEBUG 520#ifdef CONFIG_JBD2_DEBUG
521 jbd_unlock_bh_state(bitmap_bh); 521 jbd_unlock_bh_state(bitmap_bh);
522 { 522 {
523 struct buffer_head *debug_bh; 523 struct buffer_head *debug_bh;
@@ -1597,7 +1597,7 @@ allocated:
1597 1597
1598 performed_allocation = 1; 1598 performed_allocation = 1;
1599 1599
1600#ifdef CONFIG_JBD_DEBUG 1600#ifdef CONFIG_JBD2_DEBUG
1601 { 1601 {
1602 struct buffer_head *debug_bh; 1602 struct buffer_head *debug_bh;
1603 1603
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index b9ce24129070..750c46f7d893 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -39,6 +39,7 @@
39#include <linux/quotaops.h> 39#include <linux/quotaops.h>
40#include <linux/string.h> 40#include <linux/string.h>
41#include <linux/slab.h> 41#include <linux/slab.h>
42#include <linux/falloc.h>
42#include <linux/ext4_fs_extents.h> 43#include <linux/ext4_fs_extents.h>
43#include <asm/uaccess.h> 44#include <asm/uaccess.h>
44 45
@@ -91,36 +92,6 @@ static void ext4_idx_store_pblock(struct ext4_extent_idx *ix, ext4_fsblk_t pb)
91 ix->ei_leaf_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) & 0xffff); 92 ix->ei_leaf_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) & 0xffff);
92} 93}
93 94
94static int ext4_ext_check_header(const char *function, struct inode *inode,
95 struct ext4_extent_header *eh)
96{
97 const char *error_msg = NULL;
98
99 if (unlikely(eh->eh_magic != EXT4_EXT_MAGIC)) {
100 error_msg = "invalid magic";
101 goto corrupted;
102 }
103 if (unlikely(eh->eh_max == 0)) {
104 error_msg = "invalid eh_max";
105 goto corrupted;
106 }
107 if (unlikely(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max))) {
108 error_msg = "invalid eh_entries";
109 goto corrupted;
110 }
111 return 0;
112
113corrupted:
114 ext4_error(inode->i_sb, function,
115 "bad header in inode #%lu: %s - magic %x, "
116 "entries %u, max %u, depth %u",
117 inode->i_ino, error_msg, le16_to_cpu(eh->eh_magic),
118 le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max),
119 le16_to_cpu(eh->eh_depth));
120
121 return -EIO;
122}
123
124static handle_t *ext4_ext_journal_restart(handle_t *handle, int needed) 95static handle_t *ext4_ext_journal_restart(handle_t *handle, int needed)
125{ 96{
126 int err; 97 int err;
@@ -269,6 +240,70 @@ static int ext4_ext_space_root_idx(struct inode *inode)
269 return size; 240 return size;
270} 241}
271 242
243static int
244ext4_ext_max_entries(struct inode *inode, int depth)
245{
246 int max;
247
248 if (depth == ext_depth(inode)) {
249 if (depth == 0)
250 max = ext4_ext_space_root(inode);
251 else
252 max = ext4_ext_space_root_idx(inode);
253 } else {
254 if (depth == 0)
255 max = ext4_ext_space_block(inode);
256 else
257 max = ext4_ext_space_block_idx(inode);
258 }
259
260 return max;
261}
262
263static int __ext4_ext_check_header(const char *function, struct inode *inode,
264 struct ext4_extent_header *eh,
265 int depth)
266{
267 const char *error_msg;
268 int max = 0;
269
270 if (unlikely(eh->eh_magic != EXT4_EXT_MAGIC)) {
271 error_msg = "invalid magic";
272 goto corrupted;
273 }
274 if (unlikely(le16_to_cpu(eh->eh_depth) != depth)) {
275 error_msg = "unexpected eh_depth";
276 goto corrupted;
277 }
278 if (unlikely(eh->eh_max == 0)) {
279 error_msg = "invalid eh_max";
280 goto corrupted;
281 }
282 max = ext4_ext_max_entries(inode, depth);
283 if (unlikely(le16_to_cpu(eh->eh_max) > max)) {
284 error_msg = "too large eh_max";
285 goto corrupted;
286 }
287 if (unlikely(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max))) {
288 error_msg = "invalid eh_entries";
289 goto corrupted;
290 }
291 return 0;
292
293corrupted:
294 ext4_error(inode->i_sb, function,
295 "bad header in inode #%lu: %s - magic %x, "
296 "entries %u, max %u(%u), depth %u(%u)",
297 inode->i_ino, error_msg, le16_to_cpu(eh->eh_magic),
298 le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max),
299 max, le16_to_cpu(eh->eh_depth), depth);
300
301 return -EIO;
302}
303
304#define ext4_ext_check_header(inode, eh, depth) \
305 __ext4_ext_check_header(__FUNCTION__, inode, eh, depth)
306
272#ifdef EXT_DEBUG 307#ifdef EXT_DEBUG
273static void ext4_ext_show_path(struct inode *inode, struct ext4_ext_path *path) 308static void ext4_ext_show_path(struct inode *inode, struct ext4_ext_path *path)
274{ 309{
@@ -282,7 +317,7 @@ static void ext4_ext_show_path(struct inode *inode, struct ext4_ext_path *path)
282 } else if (path->p_ext) { 317 } else if (path->p_ext) {
283 ext_debug(" %d:%d:%llu ", 318 ext_debug(" %d:%d:%llu ",
284 le32_to_cpu(path->p_ext->ee_block), 319 le32_to_cpu(path->p_ext->ee_block),
285 le16_to_cpu(path->p_ext->ee_len), 320 ext4_ext_get_actual_len(path->p_ext),
286 ext_pblock(path->p_ext)); 321 ext_pblock(path->p_ext));
287 } else 322 } else
288 ext_debug(" []"); 323 ext_debug(" []");
@@ -305,7 +340,7 @@ static void ext4_ext_show_leaf(struct inode *inode, struct ext4_ext_path *path)
305 340
306 for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ex++) { 341 for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ex++) {
307 ext_debug("%d:%d:%llu ", le32_to_cpu(ex->ee_block), 342 ext_debug("%d:%d:%llu ", le32_to_cpu(ex->ee_block),
308 le16_to_cpu(ex->ee_len), ext_pblock(ex)); 343 ext4_ext_get_actual_len(ex), ext_pblock(ex));
309 } 344 }
310 ext_debug("\n"); 345 ext_debug("\n");
311} 346}
@@ -329,6 +364,7 @@ static void ext4_ext_drop_refs(struct ext4_ext_path *path)
329/* 364/*
330 * ext4_ext_binsearch_idx: 365 * ext4_ext_binsearch_idx:
331 * binary search for the closest index of the given block 366 * binary search for the closest index of the given block
367 * the header must be checked before calling this
332 */ 368 */
333static void 369static void
334ext4_ext_binsearch_idx(struct inode *inode, struct ext4_ext_path *path, int block) 370ext4_ext_binsearch_idx(struct inode *inode, struct ext4_ext_path *path, int block)
@@ -336,27 +372,25 @@ ext4_ext_binsearch_idx(struct inode *inode, struct ext4_ext_path *path, int bloc
336 struct ext4_extent_header *eh = path->p_hdr; 372 struct ext4_extent_header *eh = path->p_hdr;
337 struct ext4_extent_idx *r, *l, *m; 373 struct ext4_extent_idx *r, *l, *m;
338 374
339 BUG_ON(eh->eh_magic != EXT4_EXT_MAGIC);
340 BUG_ON(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max));
341 BUG_ON(le16_to_cpu(eh->eh_entries) <= 0);
342 375
343 ext_debug("binsearch for %d(idx): ", block); 376 ext_debug("binsearch for %d(idx): ", block);
344 377
345 l = EXT_FIRST_INDEX(eh) + 1; 378 l = EXT_FIRST_INDEX(eh) + 1;
346 r = EXT_FIRST_INDEX(eh) + le16_to_cpu(eh->eh_entries) - 1; 379 r = EXT_LAST_INDEX(eh);
347 while (l <= r) { 380 while (l <= r) {
348 m = l + (r - l) / 2; 381 m = l + (r - l) / 2;
349 if (block < le32_to_cpu(m->ei_block)) 382 if (block < le32_to_cpu(m->ei_block))
350 r = m - 1; 383 r = m - 1;
351 else 384 else
352 l = m + 1; 385 l = m + 1;
353 ext_debug("%p(%u):%p(%u):%p(%u) ", l, l->ei_block, 386 ext_debug("%p(%u):%p(%u):%p(%u) ", l, le32_to_cpu(l->ei_block),
354 m, m->ei_block, r, r->ei_block); 387 m, le32_to_cpu(m->ei_block),
388 r, le32_to_cpu(r->ei_block));
355 } 389 }
356 390
357 path->p_idx = l - 1; 391 path->p_idx = l - 1;
358 ext_debug(" -> %d->%lld ", le32_to_cpu(path->p_idx->ei_block), 392 ext_debug(" -> %d->%lld ", le32_to_cpu(path->p_idx->ei_block),
359 idx_block(path->p_idx)); 393 idx_pblock(path->p_idx));
360 394
361#ifdef CHECK_BINSEARCH 395#ifdef CHECK_BINSEARCH
362 { 396 {
@@ -388,6 +422,7 @@ ext4_ext_binsearch_idx(struct inode *inode, struct ext4_ext_path *path, int bloc
388/* 422/*
389 * ext4_ext_binsearch: 423 * ext4_ext_binsearch:
390 * binary search for closest extent of the given block 424 * binary search for closest extent of the given block
425 * the header must be checked before calling this
391 */ 426 */
392static void 427static void
393ext4_ext_binsearch(struct inode *inode, struct ext4_ext_path *path, int block) 428ext4_ext_binsearch(struct inode *inode, struct ext4_ext_path *path, int block)
@@ -395,9 +430,6 @@ ext4_ext_binsearch(struct inode *inode, struct ext4_ext_path *path, int block)
395 struct ext4_extent_header *eh = path->p_hdr; 430 struct ext4_extent_header *eh = path->p_hdr;
396 struct ext4_extent *r, *l, *m; 431 struct ext4_extent *r, *l, *m;
397 432
398 BUG_ON(eh->eh_magic != EXT4_EXT_MAGIC);
399 BUG_ON(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max));
400
401 if (eh->eh_entries == 0) { 433 if (eh->eh_entries == 0) {
402 /* 434 /*
403 * this leaf is empty: 435 * this leaf is empty:
@@ -409,7 +441,7 @@ ext4_ext_binsearch(struct inode *inode, struct ext4_ext_path *path, int block)
409 ext_debug("binsearch for %d: ", block); 441 ext_debug("binsearch for %d: ", block);
410 442
411 l = EXT_FIRST_EXTENT(eh) + 1; 443 l = EXT_FIRST_EXTENT(eh) + 1;
412 r = EXT_FIRST_EXTENT(eh) + le16_to_cpu(eh->eh_entries) - 1; 444 r = EXT_LAST_EXTENT(eh);
413 445
414 while (l <= r) { 446 while (l <= r) {
415 m = l + (r - l) / 2; 447 m = l + (r - l) / 2;
@@ -417,15 +449,16 @@ ext4_ext_binsearch(struct inode *inode, struct ext4_ext_path *path, int block)
417 r = m - 1; 449 r = m - 1;
418 else 450 else
419 l = m + 1; 451 l = m + 1;
420 ext_debug("%p(%u):%p(%u):%p(%u) ", l, l->ee_block, 452 ext_debug("%p(%u):%p(%u):%p(%u) ", l, le32_to_cpu(l->ee_block),
421 m, m->ee_block, r, r->ee_block); 453 m, le32_to_cpu(m->ee_block),
454 r, le32_to_cpu(r->ee_block));
422 } 455 }
423 456
424 path->p_ext = l - 1; 457 path->p_ext = l - 1;
425 ext_debug(" -> %d:%llu:%d ", 458 ext_debug(" -> %d:%llu:%d ",
426 le32_to_cpu(path->p_ext->ee_block), 459 le32_to_cpu(path->p_ext->ee_block),
427 ext_pblock(path->p_ext), 460 ext_pblock(path->p_ext),
428 le16_to_cpu(path->p_ext->ee_len)); 461 ext4_ext_get_actual_len(path->p_ext));
429 462
430#ifdef CHECK_BINSEARCH 463#ifdef CHECK_BINSEARCH
431 { 464 {
@@ -468,11 +501,10 @@ ext4_ext_find_extent(struct inode *inode, int block, struct ext4_ext_path *path)
468 short int depth, i, ppos = 0, alloc = 0; 501 short int depth, i, ppos = 0, alloc = 0;
469 502
470 eh = ext_inode_hdr(inode); 503 eh = ext_inode_hdr(inode);
471 BUG_ON(eh == NULL); 504 depth = ext_depth(inode);
472 if (ext4_ext_check_header(__FUNCTION__, inode, eh)) 505 if (ext4_ext_check_header(inode, eh, depth))
473 return ERR_PTR(-EIO); 506 return ERR_PTR(-EIO);
474 507
475 i = depth = ext_depth(inode);
476 508
477 /* account possible depth increase */ 509 /* account possible depth increase */
478 if (!path) { 510 if (!path) {
@@ -484,10 +516,12 @@ ext4_ext_find_extent(struct inode *inode, int block, struct ext4_ext_path *path)
484 } 516 }
485 path[0].p_hdr = eh; 517 path[0].p_hdr = eh;
486 518
519 i = depth;
487 /* walk through the tree */ 520 /* walk through the tree */
488 while (i) { 521 while (i) {
489 ext_debug("depth %d: num %d, max %d\n", 522 ext_debug("depth %d: num %d, max %d\n",
490 ppos, le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max)); 523 ppos, le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max));
524
491 ext4_ext_binsearch_idx(inode, path + ppos, block); 525 ext4_ext_binsearch_idx(inode, path + ppos, block);
492 path[ppos].p_block = idx_pblock(path[ppos].p_idx); 526 path[ppos].p_block = idx_pblock(path[ppos].p_idx);
493 path[ppos].p_depth = i; 527 path[ppos].p_depth = i;
@@ -504,7 +538,7 @@ ext4_ext_find_extent(struct inode *inode, int block, struct ext4_ext_path *path)
504 path[ppos].p_hdr = eh; 538 path[ppos].p_hdr = eh;
505 i--; 539 i--;
506 540
507 if (ext4_ext_check_header(__FUNCTION__, inode, eh)) 541 if (ext4_ext_check_header(inode, eh, i))
508 goto err; 542 goto err;
509 } 543 }
510 544
@@ -513,9 +547,6 @@ ext4_ext_find_extent(struct inode *inode, int block, struct ext4_ext_path *path)
513 path[ppos].p_ext = NULL; 547 path[ppos].p_ext = NULL;
514 path[ppos].p_idx = NULL; 548 path[ppos].p_idx = NULL;
515 549
516 if (ext4_ext_check_header(__FUNCTION__, inode, eh))
517 goto err;
518
519 /* find extent */ 550 /* find extent */
520 ext4_ext_binsearch(inode, path + ppos, block); 551 ext4_ext_binsearch(inode, path + ppos, block);
521 552
@@ -553,7 +584,7 @@ static int ext4_ext_insert_index(handle_t *handle, struct inode *inode,
553 if (curp->p_idx != EXT_LAST_INDEX(curp->p_hdr)) { 584 if (curp->p_idx != EXT_LAST_INDEX(curp->p_hdr)) {
554 len = (len - 1) * sizeof(struct ext4_extent_idx); 585 len = (len - 1) * sizeof(struct ext4_extent_idx);
555 len = len < 0 ? 0 : len; 586 len = len < 0 ? 0 : len;
556 ext_debug("insert new index %d after: %d. " 587 ext_debug("insert new index %d after: %llu. "
557 "move %d from 0x%p to 0x%p\n", 588 "move %d from 0x%p to 0x%p\n",
558 logical, ptr, len, 589 logical, ptr, len,
559 (curp->p_idx + 1), (curp->p_idx + 2)); 590 (curp->p_idx + 1), (curp->p_idx + 2));
@@ -564,7 +595,7 @@ static int ext4_ext_insert_index(handle_t *handle, struct inode *inode,
564 /* insert before */ 595 /* insert before */
565 len = len * sizeof(struct ext4_extent_idx); 596 len = len * sizeof(struct ext4_extent_idx);
566 len = len < 0 ? 0 : len; 597 len = len < 0 ? 0 : len;
567 ext_debug("insert new index %d before: %d. " 598 ext_debug("insert new index %d before: %llu. "
568 "move %d from 0x%p to 0x%p\n", 599 "move %d from 0x%p to 0x%p\n",
569 logical, ptr, len, 600 logical, ptr, len,
570 curp->p_idx, (curp->p_idx + 1)); 601 curp->p_idx, (curp->p_idx + 1));
@@ -686,7 +717,7 @@ static int ext4_ext_split(handle_t *handle, struct inode *inode,
686 ext_debug("move %d:%llu:%d in new leaf %llu\n", 717 ext_debug("move %d:%llu:%d in new leaf %llu\n",
687 le32_to_cpu(path[depth].p_ext->ee_block), 718 le32_to_cpu(path[depth].p_ext->ee_block),
688 ext_pblock(path[depth].p_ext), 719 ext_pblock(path[depth].p_ext),
689 le16_to_cpu(path[depth].p_ext->ee_len), 720 ext4_ext_get_actual_len(path[depth].p_ext),
690 newblock); 721 newblock);
691 /*memmove(ex++, path[depth].p_ext++, 722 /*memmove(ex++, path[depth].p_ext++,
692 sizeof(struct ext4_extent)); 723 sizeof(struct ext4_extent));
@@ -764,7 +795,7 @@ static int ext4_ext_split(handle_t *handle, struct inode *inode,
764 BUG_ON(EXT_MAX_INDEX(path[i].p_hdr) != 795 BUG_ON(EXT_MAX_INDEX(path[i].p_hdr) !=
765 EXT_LAST_INDEX(path[i].p_hdr)); 796 EXT_LAST_INDEX(path[i].p_hdr));
766 while (path[i].p_idx <= EXT_MAX_INDEX(path[i].p_hdr)) { 797 while (path[i].p_idx <= EXT_MAX_INDEX(path[i].p_hdr)) {
767 ext_debug("%d: move %d:%d in new index %llu\n", i, 798 ext_debug("%d: move %d:%llu in new index %llu\n", i,
768 le32_to_cpu(path[i].p_idx->ei_block), 799 le32_to_cpu(path[i].p_idx->ei_block),
769 idx_pblock(path[i].p_idx), 800 idx_pblock(path[i].p_idx),
770 newblock); 801 newblock);
@@ -893,8 +924,13 @@ static int ext4_ext_grow_indepth(handle_t *handle, struct inode *inode,
893 curp->p_hdr->eh_max = cpu_to_le16(ext4_ext_space_root_idx(inode)); 924 curp->p_hdr->eh_max = cpu_to_le16(ext4_ext_space_root_idx(inode));
894 curp->p_hdr->eh_entries = cpu_to_le16(1); 925 curp->p_hdr->eh_entries = cpu_to_le16(1);
895 curp->p_idx = EXT_FIRST_INDEX(curp->p_hdr); 926 curp->p_idx = EXT_FIRST_INDEX(curp->p_hdr);
896 /* FIXME: it works, but actually path[0] can be index */ 927
897 curp->p_idx->ei_block = EXT_FIRST_EXTENT(path[0].p_hdr)->ee_block; 928 if (path[0].p_hdr->eh_depth)
929 curp->p_idx->ei_block =
930 EXT_FIRST_INDEX(path[0].p_hdr)->ei_block;
931 else
932 curp->p_idx->ei_block =
933 EXT_FIRST_EXTENT(path[0].p_hdr)->ee_block;
898 ext4_idx_store_pblock(curp->p_idx, newblock); 934 ext4_idx_store_pblock(curp->p_idx, newblock);
899 935
900 neh = ext_inode_hdr(inode); 936 neh = ext_inode_hdr(inode);
@@ -1106,7 +1142,24 @@ static int
1106ext4_can_extents_be_merged(struct inode *inode, struct ext4_extent *ex1, 1142ext4_can_extents_be_merged(struct inode *inode, struct ext4_extent *ex1,
1107 struct ext4_extent *ex2) 1143 struct ext4_extent *ex2)
1108{ 1144{
1109 if (le32_to_cpu(ex1->ee_block) + le16_to_cpu(ex1->ee_len) != 1145 unsigned short ext1_ee_len, ext2_ee_len, max_len;
1146
1147 /*
1148 * Make sure that either both extents are uninitialized, or
1149 * both are _not_.
1150 */
1151 if (ext4_ext_is_uninitialized(ex1) ^ ext4_ext_is_uninitialized(ex2))
1152 return 0;
1153
1154 if (ext4_ext_is_uninitialized(ex1))
1155 max_len = EXT_UNINIT_MAX_LEN;
1156 else
1157 max_len = EXT_INIT_MAX_LEN;
1158
1159 ext1_ee_len = ext4_ext_get_actual_len(ex1);
1160 ext2_ee_len = ext4_ext_get_actual_len(ex2);
1161
1162 if (le32_to_cpu(ex1->ee_block) + ext1_ee_len !=
1110 le32_to_cpu(ex2->ee_block)) 1163 le32_to_cpu(ex2->ee_block))
1111 return 0; 1164 return 0;
1112 1165
@@ -1115,19 +1168,66 @@ ext4_can_extents_be_merged(struct inode *inode, struct ext4_extent *ex1,
1115 * as an RO_COMPAT feature, refuse to merge to extents if 1168 * as an RO_COMPAT feature, refuse to merge to extents if
1116 * this can result in the top bit of ee_len being set. 1169 * this can result in the top bit of ee_len being set.
1117 */ 1170 */
1118 if (le16_to_cpu(ex1->ee_len) + le16_to_cpu(ex2->ee_len) > EXT_MAX_LEN) 1171 if (ext1_ee_len + ext2_ee_len > max_len)
1119 return 0; 1172 return 0;
1120#ifdef AGGRESSIVE_TEST 1173#ifdef AGGRESSIVE_TEST
1121 if (le16_to_cpu(ex1->ee_len) >= 4) 1174 if (le16_to_cpu(ex1->ee_len) >= 4)
1122 return 0; 1175 return 0;
1123#endif 1176#endif
1124 1177
1125 if (ext_pblock(ex1) + le16_to_cpu(ex1->ee_len) == ext_pblock(ex2)) 1178 if (ext_pblock(ex1) + ext1_ee_len == ext_pblock(ex2))
1126 return 1; 1179 return 1;
1127 return 0; 1180 return 0;
1128} 1181}
1129 1182
1130/* 1183/*
1184 * This function tries to merge the "ex" extent to the next extent in the tree.
1185 * It always tries to merge towards right. If you want to merge towards
1186 * left, pass "ex - 1" as argument instead of "ex".
1187 * Returns 0 if the extents (ex and ex+1) were _not_ merged and returns
1188 * 1 if they got merged.
1189 */
1190int ext4_ext_try_to_merge(struct inode *inode,
1191 struct ext4_ext_path *path,
1192 struct ext4_extent *ex)
1193{
1194 struct ext4_extent_header *eh;
1195 unsigned int depth, len;
1196 int merge_done = 0;
1197 int uninitialized = 0;
1198
1199 depth = ext_depth(inode);
1200 BUG_ON(path[depth].p_hdr == NULL);
1201 eh = path[depth].p_hdr;
1202
1203 while (ex < EXT_LAST_EXTENT(eh)) {
1204 if (!ext4_can_extents_be_merged(inode, ex, ex + 1))
1205 break;
1206 /* merge with next extent! */
1207 if (ext4_ext_is_uninitialized(ex))
1208 uninitialized = 1;
1209 ex->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex)
1210 + ext4_ext_get_actual_len(ex + 1));
1211 if (uninitialized)
1212 ext4_ext_mark_uninitialized(ex);
1213
1214 if (ex + 1 < EXT_LAST_EXTENT(eh)) {
1215 len = (EXT_LAST_EXTENT(eh) - ex - 1)
1216 * sizeof(struct ext4_extent);
1217 memmove(ex + 1, ex + 2, len);
1218 }
1219 eh->eh_entries = cpu_to_le16(le16_to_cpu(eh->eh_entries) - 1);
1220 merge_done = 1;
1221 WARN_ON(eh->eh_entries == 0);
1222 if (!eh->eh_entries)
1223 ext4_error(inode->i_sb, "ext4_ext_try_to_merge",
1224 "inode#%lu, eh->eh_entries = 0!", inode->i_ino);
1225 }
1226
1227 return merge_done;
1228}
1229
1230/*
1131 * check if a portion of the "newext" extent overlaps with an 1231 * check if a portion of the "newext" extent overlaps with an
1132 * existing extent. 1232 * existing extent.
1133 * 1233 *
@@ -1144,7 +1244,7 @@ unsigned int ext4_ext_check_overlap(struct inode *inode,
1144 unsigned int ret = 0; 1244 unsigned int ret = 0;
1145 1245
1146 b1 = le32_to_cpu(newext->ee_block); 1246 b1 = le32_to_cpu(newext->ee_block);
1147 len1 = le16_to_cpu(newext->ee_len); 1247 len1 = ext4_ext_get_actual_len(newext);
1148 depth = ext_depth(inode); 1248 depth = ext_depth(inode);
1149 if (!path[depth].p_ext) 1249 if (!path[depth].p_ext)
1150 goto out; 1250 goto out;
@@ -1191,8 +1291,9 @@ int ext4_ext_insert_extent(handle_t *handle, struct inode *inode,
1191 struct ext4_extent *nearex; /* nearest extent */ 1291 struct ext4_extent *nearex; /* nearest extent */
1192 struct ext4_ext_path *npath = NULL; 1292 struct ext4_ext_path *npath = NULL;
1193 int depth, len, err, next; 1293 int depth, len, err, next;
1294 unsigned uninitialized = 0;
1194 1295
1195 BUG_ON(newext->ee_len == 0); 1296 BUG_ON(ext4_ext_get_actual_len(newext) == 0);
1196 depth = ext_depth(inode); 1297 depth = ext_depth(inode);
1197 ex = path[depth].p_ext; 1298 ex = path[depth].p_ext;
1198 BUG_ON(path[depth].p_hdr == NULL); 1299 BUG_ON(path[depth].p_hdr == NULL);
@@ -1200,14 +1301,24 @@ int ext4_ext_insert_extent(handle_t *handle, struct inode *inode,
1200 /* try to insert block into found extent and return */ 1301 /* try to insert block into found extent and return */
1201 if (ex && ext4_can_extents_be_merged(inode, ex, newext)) { 1302 if (ex && ext4_can_extents_be_merged(inode, ex, newext)) {
1202 ext_debug("append %d block to %d:%d (from %llu)\n", 1303 ext_debug("append %d block to %d:%d (from %llu)\n",
1203 le16_to_cpu(newext->ee_len), 1304 ext4_ext_get_actual_len(newext),
1204 le32_to_cpu(ex->ee_block), 1305 le32_to_cpu(ex->ee_block),
1205 le16_to_cpu(ex->ee_len), ext_pblock(ex)); 1306 ext4_ext_get_actual_len(ex), ext_pblock(ex));
1206 err = ext4_ext_get_access(handle, inode, path + depth); 1307 err = ext4_ext_get_access(handle, inode, path + depth);
1207 if (err) 1308 if (err)
1208 return err; 1309 return err;
1209 ex->ee_len = cpu_to_le16(le16_to_cpu(ex->ee_len) 1310
1210 + le16_to_cpu(newext->ee_len)); 1311 /*
1312 * ext4_can_extents_be_merged should have checked that either
1313 * both extents are uninitialized, or both aren't. Thus we
1314 * need to check only one of them here.
1315 */
1316 if (ext4_ext_is_uninitialized(ex))
1317 uninitialized = 1;
1318 ex->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex)
1319 + ext4_ext_get_actual_len(newext));
1320 if (uninitialized)
1321 ext4_ext_mark_uninitialized(ex);
1211 eh = path[depth].p_hdr; 1322 eh = path[depth].p_hdr;
1212 nearex = ex; 1323 nearex = ex;
1213 goto merge; 1324 goto merge;
@@ -1263,7 +1374,7 @@ has_space:
1263 ext_debug("first extent in the leaf: %d:%llu:%d\n", 1374 ext_debug("first extent in the leaf: %d:%llu:%d\n",
1264 le32_to_cpu(newext->ee_block), 1375 le32_to_cpu(newext->ee_block),
1265 ext_pblock(newext), 1376 ext_pblock(newext),
1266 le16_to_cpu(newext->ee_len)); 1377 ext4_ext_get_actual_len(newext));
1267 path[depth].p_ext = EXT_FIRST_EXTENT(eh); 1378 path[depth].p_ext = EXT_FIRST_EXTENT(eh);
1268 } else if (le32_to_cpu(newext->ee_block) 1379 } else if (le32_to_cpu(newext->ee_block)
1269 > le32_to_cpu(nearex->ee_block)) { 1380 > le32_to_cpu(nearex->ee_block)) {
@@ -1276,7 +1387,7 @@ has_space:
1276 "move %d from 0x%p to 0x%p\n", 1387 "move %d from 0x%p to 0x%p\n",
1277 le32_to_cpu(newext->ee_block), 1388 le32_to_cpu(newext->ee_block),
1278 ext_pblock(newext), 1389 ext_pblock(newext),
1279 le16_to_cpu(newext->ee_len), 1390 ext4_ext_get_actual_len(newext),
1280 nearex, len, nearex + 1, nearex + 2); 1391 nearex, len, nearex + 1, nearex + 2);
1281 memmove(nearex + 2, nearex + 1, len); 1392 memmove(nearex + 2, nearex + 1, len);
1282 } 1393 }
@@ -1289,7 +1400,7 @@ has_space:
1289 "move %d from 0x%p to 0x%p\n", 1400 "move %d from 0x%p to 0x%p\n",
1290 le32_to_cpu(newext->ee_block), 1401 le32_to_cpu(newext->ee_block),
1291 ext_pblock(newext), 1402 ext_pblock(newext),
1292 le16_to_cpu(newext->ee_len), 1403 ext4_ext_get_actual_len(newext),
1293 nearex, len, nearex + 1, nearex + 2); 1404 nearex, len, nearex + 1, nearex + 2);
1294 memmove(nearex + 1, nearex, len); 1405 memmove(nearex + 1, nearex, len);
1295 path[depth].p_ext = nearex; 1406 path[depth].p_ext = nearex;
@@ -1304,20 +1415,7 @@ has_space:
1304 1415
1305merge: 1416merge:
1306 /* try to merge extents to the right */ 1417 /* try to merge extents to the right */
1307 while (nearex < EXT_LAST_EXTENT(eh)) { 1418 ext4_ext_try_to_merge(inode, path, nearex);
1308 if (!ext4_can_extents_be_merged(inode, nearex, nearex + 1))
1309 break;
1310 /* merge with next extent! */
1311 nearex->ee_len = cpu_to_le16(le16_to_cpu(nearex->ee_len)
1312 + le16_to_cpu(nearex[1].ee_len));
1313 if (nearex + 1 < EXT_LAST_EXTENT(eh)) {
1314 len = (EXT_LAST_EXTENT(eh) - nearex - 1)
1315 * sizeof(struct ext4_extent);
1316 memmove(nearex + 1, nearex + 2, len);
1317 }
1318 eh->eh_entries = cpu_to_le16(le16_to_cpu(eh->eh_entries)-1);
1319 BUG_ON(eh->eh_entries == 0);
1320 }
1321 1419
1322 /* try to merge extents to the left */ 1420 /* try to merge extents to the left */
1323 1421
@@ -1379,8 +1477,8 @@ int ext4_ext_walk_space(struct inode *inode, unsigned long block,
1379 end = le32_to_cpu(ex->ee_block); 1477 end = le32_to_cpu(ex->ee_block);
1380 if (block + num < end) 1478 if (block + num < end)
1381 end = block + num; 1479 end = block + num;
1382 } else if (block >= 1480 } else if (block >= le32_to_cpu(ex->ee_block)
1383 le32_to_cpu(ex->ee_block) + le16_to_cpu(ex->ee_len)) { 1481 + ext4_ext_get_actual_len(ex)) {
1384 /* need to allocate space after found extent */ 1482 /* need to allocate space after found extent */
1385 start = block; 1483 start = block;
1386 end = block + num; 1484 end = block + num;
@@ -1392,7 +1490,8 @@ int ext4_ext_walk_space(struct inode *inode, unsigned long block,
1392 * by found extent 1490 * by found extent
1393 */ 1491 */
1394 start = block; 1492 start = block;
1395 end = le32_to_cpu(ex->ee_block) + le16_to_cpu(ex->ee_len); 1493 end = le32_to_cpu(ex->ee_block)
1494 + ext4_ext_get_actual_len(ex);
1396 if (block + num < end) 1495 if (block + num < end)
1397 end = block + num; 1496 end = block + num;
1398 exists = 1; 1497 exists = 1;
@@ -1408,7 +1507,7 @@ int ext4_ext_walk_space(struct inode *inode, unsigned long block,
1408 cbex.ec_type = EXT4_EXT_CACHE_GAP; 1507 cbex.ec_type = EXT4_EXT_CACHE_GAP;
1409 } else { 1508 } else {
1410 cbex.ec_block = le32_to_cpu(ex->ee_block); 1509 cbex.ec_block = le32_to_cpu(ex->ee_block);
1411 cbex.ec_len = le16_to_cpu(ex->ee_len); 1510 cbex.ec_len = ext4_ext_get_actual_len(ex);
1412 cbex.ec_start = ext_pblock(ex); 1511 cbex.ec_start = ext_pblock(ex);
1413 cbex.ec_type = EXT4_EXT_CACHE_EXTENT; 1512 cbex.ec_type = EXT4_EXT_CACHE_EXTENT;
1414 } 1513 }
@@ -1481,15 +1580,15 @@ ext4_ext_put_gap_in_cache(struct inode *inode, struct ext4_ext_path *path,
1481 ext_debug("cache gap(before): %lu [%lu:%lu]", 1580 ext_debug("cache gap(before): %lu [%lu:%lu]",
1482 (unsigned long) block, 1581 (unsigned long) block,
1483 (unsigned long) le32_to_cpu(ex->ee_block), 1582 (unsigned long) le32_to_cpu(ex->ee_block),
1484 (unsigned long) le16_to_cpu(ex->ee_len)); 1583 (unsigned long) ext4_ext_get_actual_len(ex));
1485 } else if (block >= le32_to_cpu(ex->ee_block) 1584 } else if (block >= le32_to_cpu(ex->ee_block)
1486 + le16_to_cpu(ex->ee_len)) { 1585 + ext4_ext_get_actual_len(ex)) {
1487 lblock = le32_to_cpu(ex->ee_block) 1586 lblock = le32_to_cpu(ex->ee_block)
1488 + le16_to_cpu(ex->ee_len); 1587 + ext4_ext_get_actual_len(ex);
1489 len = ext4_ext_next_allocated_block(path); 1588 len = ext4_ext_next_allocated_block(path);
1490 ext_debug("cache gap(after): [%lu:%lu] %lu", 1589 ext_debug("cache gap(after): [%lu:%lu] %lu",
1491 (unsigned long) le32_to_cpu(ex->ee_block), 1590 (unsigned long) le32_to_cpu(ex->ee_block),
1492 (unsigned long) le16_to_cpu(ex->ee_len), 1591 (unsigned long) ext4_ext_get_actual_len(ex),
1493 (unsigned long) block); 1592 (unsigned long) block);
1494 BUG_ON(len == lblock); 1593 BUG_ON(len == lblock);
1495 len = len - lblock; 1594 len = len - lblock;
@@ -1619,12 +1718,12 @@ static int ext4_remove_blocks(handle_t *handle, struct inode *inode,
1619 unsigned long from, unsigned long to) 1718 unsigned long from, unsigned long to)
1620{ 1719{
1621 struct buffer_head *bh; 1720 struct buffer_head *bh;
1721 unsigned short ee_len = ext4_ext_get_actual_len(ex);
1622 int i; 1722 int i;
1623 1723
1624#ifdef EXTENTS_STATS 1724#ifdef EXTENTS_STATS
1625 { 1725 {
1626 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 1726 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1627 unsigned short ee_len = le16_to_cpu(ex->ee_len);
1628 spin_lock(&sbi->s_ext_stats_lock); 1727 spin_lock(&sbi->s_ext_stats_lock);
1629 sbi->s_ext_blocks += ee_len; 1728 sbi->s_ext_blocks += ee_len;
1630 sbi->s_ext_extents++; 1729 sbi->s_ext_extents++;
@@ -1638,12 +1737,12 @@ static int ext4_remove_blocks(handle_t *handle, struct inode *inode,
1638 } 1737 }
1639#endif 1738#endif
1640 if (from >= le32_to_cpu(ex->ee_block) 1739 if (from >= le32_to_cpu(ex->ee_block)
1641 && to == le32_to_cpu(ex->ee_block) + le16_to_cpu(ex->ee_len) - 1) { 1740 && to == le32_to_cpu(ex->ee_block) + ee_len - 1) {
1642 /* tail removal */ 1741 /* tail removal */
1643 unsigned long num; 1742 unsigned long num;
1644 ext4_fsblk_t start; 1743 ext4_fsblk_t start;
1645 num = le32_to_cpu(ex->ee_block) + le16_to_cpu(ex->ee_len) - from; 1744 num = le32_to_cpu(ex->ee_block) + ee_len - from;
1646 start = ext_pblock(ex) + le16_to_cpu(ex->ee_len) - num; 1745 start = ext_pblock(ex) + ee_len - num;
1647 ext_debug("free last %lu blocks starting %llu\n", num, start); 1746 ext_debug("free last %lu blocks starting %llu\n", num, start);
1648 for (i = 0; i < num; i++) { 1747 for (i = 0; i < num; i++) {
1649 bh = sb_find_get_block(inode->i_sb, start + i); 1748 bh = sb_find_get_block(inode->i_sb, start + i);
@@ -1651,12 +1750,12 @@ static int ext4_remove_blocks(handle_t *handle, struct inode *inode,
1651 } 1750 }
1652 ext4_free_blocks(handle, inode, start, num); 1751 ext4_free_blocks(handle, inode, start, num);
1653 } else if (from == le32_to_cpu(ex->ee_block) 1752 } else if (from == le32_to_cpu(ex->ee_block)
1654 && to <= le32_to_cpu(ex->ee_block) + le16_to_cpu(ex->ee_len) - 1) { 1753 && to <= le32_to_cpu(ex->ee_block) + ee_len - 1) {
1655 printk("strange request: removal %lu-%lu from %u:%u\n", 1754 printk("strange request: removal %lu-%lu from %u:%u\n",
1656 from, to, le32_to_cpu(ex->ee_block), le16_to_cpu(ex->ee_len)); 1755 from, to, le32_to_cpu(ex->ee_block), ee_len);
1657 } else { 1756 } else {
1658 printk("strange request: removal(2) %lu-%lu from %u:%u\n", 1757 printk("strange request: removal(2) %lu-%lu from %u:%u\n",
1659 from, to, le32_to_cpu(ex->ee_block), le16_to_cpu(ex->ee_len)); 1758 from, to, le32_to_cpu(ex->ee_block), ee_len);
1660 } 1759 }
1661 return 0; 1760 return 0;
1662} 1761}
@@ -1671,21 +1770,23 @@ ext4_ext_rm_leaf(handle_t *handle, struct inode *inode,
1671 unsigned a, b, block, num; 1770 unsigned a, b, block, num;
1672 unsigned long ex_ee_block; 1771 unsigned long ex_ee_block;
1673 unsigned short ex_ee_len; 1772 unsigned short ex_ee_len;
1773 unsigned uninitialized = 0;
1674 struct ext4_extent *ex; 1774 struct ext4_extent *ex;
1675 1775
1776 /* the header must be checked already in ext4_ext_remove_space() */
1676 ext_debug("truncate since %lu in leaf\n", start); 1777 ext_debug("truncate since %lu in leaf\n", start);
1677 if (!path[depth].p_hdr) 1778 if (!path[depth].p_hdr)
1678 path[depth].p_hdr = ext_block_hdr(path[depth].p_bh); 1779 path[depth].p_hdr = ext_block_hdr(path[depth].p_bh);
1679 eh = path[depth].p_hdr; 1780 eh = path[depth].p_hdr;
1680 BUG_ON(eh == NULL); 1781 BUG_ON(eh == NULL);
1681 BUG_ON(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max));
1682 BUG_ON(eh->eh_magic != EXT4_EXT_MAGIC);
1683 1782
1684 /* find where to start removing */ 1783 /* find where to start removing */
1685 ex = EXT_LAST_EXTENT(eh); 1784 ex = EXT_LAST_EXTENT(eh);
1686 1785
1687 ex_ee_block = le32_to_cpu(ex->ee_block); 1786 ex_ee_block = le32_to_cpu(ex->ee_block);
1688 ex_ee_len = le16_to_cpu(ex->ee_len); 1787 if (ext4_ext_is_uninitialized(ex))
1788 uninitialized = 1;
1789 ex_ee_len = ext4_ext_get_actual_len(ex);
1689 1790
1690 while (ex >= EXT_FIRST_EXTENT(eh) && 1791 while (ex >= EXT_FIRST_EXTENT(eh) &&
1691 ex_ee_block + ex_ee_len > start) { 1792 ex_ee_block + ex_ee_len > start) {
@@ -1753,6 +1854,12 @@ ext4_ext_rm_leaf(handle_t *handle, struct inode *inode,
1753 1854
1754 ex->ee_block = cpu_to_le32(block); 1855 ex->ee_block = cpu_to_le32(block);
1755 ex->ee_len = cpu_to_le16(num); 1856 ex->ee_len = cpu_to_le16(num);
1857 /*
1858 * Do not mark uninitialized if all the blocks in the
1859 * extent have been removed.
1860 */
1861 if (uninitialized && num)
1862 ext4_ext_mark_uninitialized(ex);
1756 1863
1757 err = ext4_ext_dirty(handle, inode, path + depth); 1864 err = ext4_ext_dirty(handle, inode, path + depth);
1758 if (err) 1865 if (err)
@@ -1762,7 +1869,7 @@ ext4_ext_rm_leaf(handle_t *handle, struct inode *inode,
1762 ext_pblock(ex)); 1869 ext_pblock(ex));
1763 ex--; 1870 ex--;
1764 ex_ee_block = le32_to_cpu(ex->ee_block); 1871 ex_ee_block = le32_to_cpu(ex->ee_block);
1765 ex_ee_len = le16_to_cpu(ex->ee_len); 1872 ex_ee_len = ext4_ext_get_actual_len(ex);
1766 } 1873 }
1767 1874
1768 if (correct_index && eh->eh_entries) 1875 if (correct_index && eh->eh_entries)
@@ -1825,7 +1932,7 @@ int ext4_ext_remove_space(struct inode *inode, unsigned long start)
1825 return -ENOMEM; 1932 return -ENOMEM;
1826 } 1933 }
1827 path[0].p_hdr = ext_inode_hdr(inode); 1934 path[0].p_hdr = ext_inode_hdr(inode);
1828 if (ext4_ext_check_header(__FUNCTION__, inode, path[0].p_hdr)) { 1935 if (ext4_ext_check_header(inode, path[0].p_hdr, depth)) {
1829 err = -EIO; 1936 err = -EIO;
1830 goto out; 1937 goto out;
1831 } 1938 }
@@ -1846,17 +1953,8 @@ int ext4_ext_remove_space(struct inode *inode, unsigned long start)
1846 if (!path[i].p_hdr) { 1953 if (!path[i].p_hdr) {
1847 ext_debug("initialize header\n"); 1954 ext_debug("initialize header\n");
1848 path[i].p_hdr = ext_block_hdr(path[i].p_bh); 1955 path[i].p_hdr = ext_block_hdr(path[i].p_bh);
1849 if (ext4_ext_check_header(__FUNCTION__, inode,
1850 path[i].p_hdr)) {
1851 err = -EIO;
1852 goto out;
1853 }
1854 } 1956 }
1855 1957
1856 BUG_ON(le16_to_cpu(path[i].p_hdr->eh_entries)
1857 > le16_to_cpu(path[i].p_hdr->eh_max));
1858 BUG_ON(path[i].p_hdr->eh_magic != EXT4_EXT_MAGIC);
1859
1860 if (!path[i].p_idx) { 1958 if (!path[i].p_idx) {
1861 /* this level hasn't been touched yet */ 1959 /* this level hasn't been touched yet */
1862 path[i].p_idx = EXT_LAST_INDEX(path[i].p_hdr); 1960 path[i].p_idx = EXT_LAST_INDEX(path[i].p_hdr);
@@ -1873,17 +1971,27 @@ int ext4_ext_remove_space(struct inode *inode, unsigned long start)
1873 i, EXT_FIRST_INDEX(path[i].p_hdr), 1971 i, EXT_FIRST_INDEX(path[i].p_hdr),
1874 path[i].p_idx); 1972 path[i].p_idx);
1875 if (ext4_ext_more_to_rm(path + i)) { 1973 if (ext4_ext_more_to_rm(path + i)) {
1974 struct buffer_head *bh;
1876 /* go to the next level */ 1975 /* go to the next level */
1877 ext_debug("move to level %d (block %llu)\n", 1976 ext_debug("move to level %d (block %llu)\n",
1878 i + 1, idx_pblock(path[i].p_idx)); 1977 i + 1, idx_pblock(path[i].p_idx));
1879 memset(path + i + 1, 0, sizeof(*path)); 1978 memset(path + i + 1, 0, sizeof(*path));
1880 path[i+1].p_bh = 1979 bh = sb_bread(sb, idx_pblock(path[i].p_idx));
1881 sb_bread(sb, idx_pblock(path[i].p_idx)); 1980 if (!bh) {
1882 if (!path[i+1].p_bh) {
1883 /* should we reset i_size? */ 1981 /* should we reset i_size? */
1884 err = -EIO; 1982 err = -EIO;
1885 break; 1983 break;
1886 } 1984 }
1985 if (WARN_ON(i + 1 > depth)) {
1986 err = -EIO;
1987 break;
1988 }
1989 if (ext4_ext_check_header(inode, ext_block_hdr(bh),
1990 depth - i - 1)) {
1991 err = -EIO;
1992 break;
1993 }
1994 path[i + 1].p_bh = bh;
1887 1995
1888 /* save actual number of indexes since this 1996 /* save actual number of indexes since this
1889 * number is changed at the next iteration */ 1997 * number is changed at the next iteration */
@@ -1977,15 +2085,158 @@ void ext4_ext_release(struct super_block *sb)
1977#endif 2085#endif
1978} 2086}
1979 2087
2088/*
2089 * This function is called by ext4_ext_get_blocks() if someone tries to write
2090 * to an uninitialized extent. It may result in splitting the uninitialized
2091 * extent into multiple extents (upto three - one initialized and two
2092 * uninitialized).
2093 * There are three possibilities:
2094 * a> There is no split required: Entire extent should be initialized
2095 * b> Splits in two extents: Write is happening at either end of the extent
2096 * c> Splits in three extents: Somone is writing in middle of the extent
2097 */
2098int ext4_ext_convert_to_initialized(handle_t *handle, struct inode *inode,
2099 struct ext4_ext_path *path,
2100 ext4_fsblk_t iblock,
2101 unsigned long max_blocks)
2102{
2103 struct ext4_extent *ex, newex;
2104 struct ext4_extent *ex1 = NULL;
2105 struct ext4_extent *ex2 = NULL;
2106 struct ext4_extent *ex3 = NULL;
2107 struct ext4_extent_header *eh;
2108 unsigned int allocated, ee_block, ee_len, depth;
2109 ext4_fsblk_t newblock;
2110 int err = 0;
2111 int ret = 0;
2112
2113 depth = ext_depth(inode);
2114 eh = path[depth].p_hdr;
2115 ex = path[depth].p_ext;
2116 ee_block = le32_to_cpu(ex->ee_block);
2117 ee_len = ext4_ext_get_actual_len(ex);
2118 allocated = ee_len - (iblock - ee_block);
2119 newblock = iblock - ee_block + ext_pblock(ex);
2120 ex2 = ex;
2121
2122 /* ex1: ee_block to iblock - 1 : uninitialized */
2123 if (iblock > ee_block) {
2124 ex1 = ex;
2125 ex1->ee_len = cpu_to_le16(iblock - ee_block);
2126 ext4_ext_mark_uninitialized(ex1);
2127 ex2 = &newex;
2128 }
2129 /*
2130 * for sanity, update the length of the ex2 extent before
2131 * we insert ex3, if ex1 is NULL. This is to avoid temporary
2132 * overlap of blocks.
2133 */
2134 if (!ex1 && allocated > max_blocks)
2135 ex2->ee_len = cpu_to_le16(max_blocks);
2136 /* ex3: to ee_block + ee_len : uninitialised */
2137 if (allocated > max_blocks) {
2138 unsigned int newdepth;
2139 ex3 = &newex;
2140 ex3->ee_block = cpu_to_le32(iblock + max_blocks);
2141 ext4_ext_store_pblock(ex3, newblock + max_blocks);
2142 ex3->ee_len = cpu_to_le16(allocated - max_blocks);
2143 ext4_ext_mark_uninitialized(ex3);
2144 err = ext4_ext_insert_extent(handle, inode, path, ex3);
2145 if (err)
2146 goto out;
2147 /*
2148 * The depth, and hence eh & ex might change
2149 * as part of the insert above.
2150 */
2151 newdepth = ext_depth(inode);
2152 if (newdepth != depth) {
2153 depth = newdepth;
2154 path = ext4_ext_find_extent(inode, iblock, NULL);
2155 if (IS_ERR(path)) {
2156 err = PTR_ERR(path);
2157 path = NULL;
2158 goto out;
2159 }
2160 eh = path[depth].p_hdr;
2161 ex = path[depth].p_ext;
2162 if (ex2 != &newex)
2163 ex2 = ex;
2164 }
2165 allocated = max_blocks;
2166 }
2167 /*
2168 * If there was a change of depth as part of the
2169 * insertion of ex3 above, we need to update the length
2170 * of the ex1 extent again here
2171 */
2172 if (ex1 && ex1 != ex) {
2173 ex1 = ex;
2174 ex1->ee_len = cpu_to_le16(iblock - ee_block);
2175 ext4_ext_mark_uninitialized(ex1);
2176 ex2 = &newex;
2177 }
2178 /* ex2: iblock to iblock + maxblocks-1 : initialised */
2179 ex2->ee_block = cpu_to_le32(iblock);
2180 ex2->ee_start = cpu_to_le32(newblock);
2181 ext4_ext_store_pblock(ex2, newblock);
2182 ex2->ee_len = cpu_to_le16(allocated);
2183 if (ex2 != ex)
2184 goto insert;
2185 err = ext4_ext_get_access(handle, inode, path + depth);
2186 if (err)
2187 goto out;
2188 /*
2189 * New (initialized) extent starts from the first block
2190 * in the current extent. i.e., ex2 == ex
2191 * We have to see if it can be merged with the extent
2192 * on the left.
2193 */
2194 if (ex2 > EXT_FIRST_EXTENT(eh)) {
2195 /*
2196 * To merge left, pass "ex2 - 1" to try_to_merge(),
2197 * since it merges towards right _only_.
2198 */
2199 ret = ext4_ext_try_to_merge(inode, path, ex2 - 1);
2200 if (ret) {
2201 err = ext4_ext_correct_indexes(handle, inode, path);
2202 if (err)
2203 goto out;
2204 depth = ext_depth(inode);
2205 ex2--;
2206 }
2207 }
2208 /*
2209 * Try to Merge towards right. This might be required
2210 * only when the whole extent is being written to.
2211 * i.e. ex2 == ex and ex3 == NULL.
2212 */
2213 if (!ex3) {
2214 ret = ext4_ext_try_to_merge(inode, path, ex2);
2215 if (ret) {
2216 err = ext4_ext_correct_indexes(handle, inode, path);
2217 if (err)
2218 goto out;
2219 }
2220 }
2221 /* Mark modified extent as dirty */
2222 err = ext4_ext_dirty(handle, inode, path + depth);
2223 goto out;
2224insert:
2225 err = ext4_ext_insert_extent(handle, inode, path, &newex);
2226out:
2227 return err ? err : allocated;
2228}
2229
1980int ext4_ext_get_blocks(handle_t *handle, struct inode *inode, 2230int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
1981 ext4_fsblk_t iblock, 2231 ext4_fsblk_t iblock,
1982 unsigned long max_blocks, struct buffer_head *bh_result, 2232 unsigned long max_blocks, struct buffer_head *bh_result,
1983 int create, int extend_disksize) 2233 int create, int extend_disksize)
1984{ 2234{
1985 struct ext4_ext_path *path = NULL; 2235 struct ext4_ext_path *path = NULL;
2236 struct ext4_extent_header *eh;
1986 struct ext4_extent newex, *ex; 2237 struct ext4_extent newex, *ex;
1987 ext4_fsblk_t goal, newblock; 2238 ext4_fsblk_t goal, newblock;
1988 int err = 0, depth; 2239 int err = 0, depth, ret;
1989 unsigned long allocated = 0; 2240 unsigned long allocated = 0;
1990 2241
1991 __clear_bit(BH_New, &bh_result->b_state); 2242 __clear_bit(BH_New, &bh_result->b_state);
@@ -1998,8 +2249,10 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
1998 if (goal) { 2249 if (goal) {
1999 if (goal == EXT4_EXT_CACHE_GAP) { 2250 if (goal == EXT4_EXT_CACHE_GAP) {
2000 if (!create) { 2251 if (!create) {
2001 /* block isn't allocated yet and 2252 /*
2002 * user doesn't want to allocate it */ 2253 * block isn't allocated yet and
2254 * user doesn't want to allocate it
2255 */
2003 goto out2; 2256 goto out2;
2004 } 2257 }
2005 /* we should allocate requested block */ 2258 /* we should allocate requested block */
@@ -2033,21 +2286,19 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
2033 * this is why assert can't be put in ext4_ext_find_extent() 2286 * this is why assert can't be put in ext4_ext_find_extent()
2034 */ 2287 */
2035 BUG_ON(path[depth].p_ext == NULL && depth != 0); 2288 BUG_ON(path[depth].p_ext == NULL && depth != 0);
2289 eh = path[depth].p_hdr;
2036 2290
2037 ex = path[depth].p_ext; 2291 ex = path[depth].p_ext;
2038 if (ex) { 2292 if (ex) {
2039 unsigned long ee_block = le32_to_cpu(ex->ee_block); 2293 unsigned long ee_block = le32_to_cpu(ex->ee_block);
2040 ext4_fsblk_t ee_start = ext_pblock(ex); 2294 ext4_fsblk_t ee_start = ext_pblock(ex);
2041 unsigned short ee_len = le16_to_cpu(ex->ee_len); 2295 unsigned short ee_len;
2042 2296
2043 /* 2297 /*
2044 * Allow future support for preallocated extents to be added
2045 * as an RO_COMPAT feature:
2046 * Uninitialized extents are treated as holes, except that 2298 * Uninitialized extents are treated as holes, except that
2047 * we avoid (fail) allocating new blocks during a write. 2299 * we split out initialized portions during a write.
2048 */ 2300 */
2049 if (ee_len > EXT_MAX_LEN) 2301 ee_len = ext4_ext_get_actual_len(ex);
2050 goto out2;
2051 /* if found extent covers block, simply return it */ 2302 /* if found extent covers block, simply return it */
2052 if (iblock >= ee_block && iblock < ee_block + ee_len) { 2303 if (iblock >= ee_block && iblock < ee_block + ee_len) {
2053 newblock = iblock - ee_block + ee_start; 2304 newblock = iblock - ee_block + ee_start;
@@ -2055,9 +2306,27 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
2055 allocated = ee_len - (iblock - ee_block); 2306 allocated = ee_len - (iblock - ee_block);
2056 ext_debug("%d fit into %lu:%d -> %llu\n", (int) iblock, 2307 ext_debug("%d fit into %lu:%d -> %llu\n", (int) iblock,
2057 ee_block, ee_len, newblock); 2308 ee_block, ee_len, newblock);
2058 ext4_ext_put_in_cache(inode, ee_block, ee_len, 2309
2059 ee_start, EXT4_EXT_CACHE_EXTENT); 2310 /* Do not put uninitialized extent in the cache */
2060 goto out; 2311 if (!ext4_ext_is_uninitialized(ex)) {
2312 ext4_ext_put_in_cache(inode, ee_block,
2313 ee_len, ee_start,
2314 EXT4_EXT_CACHE_EXTENT);
2315 goto out;
2316 }
2317 if (create == EXT4_CREATE_UNINITIALIZED_EXT)
2318 goto out;
2319 if (!create)
2320 goto out2;
2321
2322 ret = ext4_ext_convert_to_initialized(handle, inode,
2323 path, iblock,
2324 max_blocks);
2325 if (ret <= 0)
2326 goto out2;
2327 else
2328 allocated = ret;
2329 goto outnew;
2061 } 2330 }
2062 } 2331 }
2063 2332
@@ -2066,8 +2335,10 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
2066 * we couldn't try to create block if create flag is zero 2335 * we couldn't try to create block if create flag is zero
2067 */ 2336 */
2068 if (!create) { 2337 if (!create) {
2069 /* put just found gap into cache to speed up 2338 /*
2070 * subsequent requests */ 2339 * put just found gap into cache to speed up
2340 * subsequent requests
2341 */
2071 ext4_ext_put_gap_in_cache(inode, path, iblock); 2342 ext4_ext_put_gap_in_cache(inode, path, iblock);
2072 goto out2; 2343 goto out2;
2073 } 2344 }
@@ -2081,6 +2352,19 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
2081 /* allocate new block */ 2352 /* allocate new block */
2082 goal = ext4_ext_find_goal(inode, path, iblock); 2353 goal = ext4_ext_find_goal(inode, path, iblock);
2083 2354
2355 /*
2356 * See if request is beyond maximum number of blocks we can have in
2357 * a single extent. For an initialized extent this limit is
2358 * EXT_INIT_MAX_LEN and for an uninitialized extent this limit is
2359 * EXT_UNINIT_MAX_LEN.
2360 */
2361 if (max_blocks > EXT_INIT_MAX_LEN &&
2362 create != EXT4_CREATE_UNINITIALIZED_EXT)
2363 max_blocks = EXT_INIT_MAX_LEN;
2364 else if (max_blocks > EXT_UNINIT_MAX_LEN &&
2365 create == EXT4_CREATE_UNINITIALIZED_EXT)
2366 max_blocks = EXT_UNINIT_MAX_LEN;
2367
2084 /* Check if we can really insert (iblock)::(iblock+max_blocks) extent */ 2368 /* Check if we can really insert (iblock)::(iblock+max_blocks) extent */
2085 newex.ee_block = cpu_to_le32(iblock); 2369 newex.ee_block = cpu_to_le32(iblock);
2086 newex.ee_len = cpu_to_le16(max_blocks); 2370 newex.ee_len = cpu_to_le16(max_blocks);
@@ -2098,6 +2382,8 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
2098 /* try to insert new extent into found leaf and return */ 2382 /* try to insert new extent into found leaf and return */
2099 ext4_ext_store_pblock(&newex, newblock); 2383 ext4_ext_store_pblock(&newex, newblock);
2100 newex.ee_len = cpu_to_le16(allocated); 2384 newex.ee_len = cpu_to_le16(allocated);
2385 if (create == EXT4_CREATE_UNINITIALIZED_EXT) /* Mark uninitialized */
2386 ext4_ext_mark_uninitialized(&newex);
2101 err = ext4_ext_insert_extent(handle, inode, path, &newex); 2387 err = ext4_ext_insert_extent(handle, inode, path, &newex);
2102 if (err) { 2388 if (err) {
2103 /* free data blocks we just allocated */ 2389 /* free data blocks we just allocated */
@@ -2111,10 +2397,13 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
2111 2397
2112 /* previous routine could use block we allocated */ 2398 /* previous routine could use block we allocated */
2113 newblock = ext_pblock(&newex); 2399 newblock = ext_pblock(&newex);
2400outnew:
2114 __set_bit(BH_New, &bh_result->b_state); 2401 __set_bit(BH_New, &bh_result->b_state);
2115 2402
2116 ext4_ext_put_in_cache(inode, iblock, allocated, newblock, 2403 /* Cache only when it is _not_ an uninitialized extent */
2117 EXT4_EXT_CACHE_EXTENT); 2404 if (create != EXT4_CREATE_UNINITIALIZED_EXT)
2405 ext4_ext_put_in_cache(inode, iblock, allocated, newblock,
2406 EXT4_EXT_CACHE_EXTENT);
2118out: 2407out:
2119 if (allocated > max_blocks) 2408 if (allocated > max_blocks)
2120 allocated = max_blocks; 2409 allocated = max_blocks;
@@ -2178,7 +2467,8 @@ void ext4_ext_truncate(struct inode * inode, struct page *page)
2178 err = ext4_ext_remove_space(inode, last_block); 2467 err = ext4_ext_remove_space(inode, last_block);
2179 2468
2180 /* In a multi-transaction truncate, we only make the final 2469 /* In a multi-transaction truncate, we only make the final
2181 * transaction synchronous. */ 2470 * transaction synchronous.
2471 */
2182 if (IS_SYNC(inode)) 2472 if (IS_SYNC(inode))
2183 handle->h_sync = 1; 2473 handle->h_sync = 1;
2184 2474
@@ -2217,3 +2507,127 @@ int ext4_ext_writepage_trans_blocks(struct inode *inode, int num)
2217 2507
2218 return needed; 2508 return needed;
2219} 2509}
2510
2511/*
2512 * preallocate space for a file. This implements ext4's fallocate inode
2513 * operation, which gets called from sys_fallocate system call.
2514 * For block-mapped files, posix_fallocate should fall back to the method
2515 * of writing zeroes to the required new blocks (the same behavior which is
2516 * expected for file systems which do not support fallocate() system call).
2517 */
2518long ext4_fallocate(struct inode *inode, int mode, loff_t offset, loff_t len)
2519{
2520 handle_t *handle;
2521 ext4_fsblk_t block, max_blocks;
2522 ext4_fsblk_t nblocks = 0;
2523 int ret = 0;
2524 int ret2 = 0;
2525 int retries = 0;
2526 struct buffer_head map_bh;
2527 unsigned int credits, blkbits = inode->i_blkbits;
2528
2529 /*
2530 * currently supporting (pre)allocate mode for extent-based
2531 * files _only_
2532 */
2533 if (!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL))
2534 return -EOPNOTSUPP;
2535
2536 /* preallocation to directories is currently not supported */
2537 if (S_ISDIR(inode->i_mode))
2538 return -ENODEV;
2539
2540 block = offset >> blkbits;
2541 max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits)
2542 - block;
2543
2544 /*
2545 * credits to insert 1 extent into extent tree + buffers to be able to
2546 * modify 1 super block, 1 block bitmap and 1 group descriptor.
2547 */
2548 credits = EXT4_DATA_TRANS_BLOCKS(inode->i_sb) + 3;
2549retry:
2550 while (ret >= 0 && ret < max_blocks) {
2551 block = block + ret;
2552 max_blocks = max_blocks - ret;
2553 handle = ext4_journal_start(inode, credits);
2554 if (IS_ERR(handle)) {
2555 ret = PTR_ERR(handle);
2556 break;
2557 }
2558
2559 ret = ext4_ext_get_blocks(handle, inode, block,
2560 max_blocks, &map_bh,
2561 EXT4_CREATE_UNINITIALIZED_EXT, 0);
2562 WARN_ON(!ret);
2563 if (!ret) {
2564 ext4_error(inode->i_sb, "ext4_fallocate",
2565 "ext4_ext_get_blocks returned 0! inode#%lu"
2566 ", block=%llu, max_blocks=%llu",
2567 inode->i_ino, block, max_blocks);
2568 ret = -EIO;
2569 ext4_mark_inode_dirty(handle, inode);
2570 ret2 = ext4_journal_stop(handle);
2571 break;
2572 }
2573 if (ret > 0) {
2574 /* check wrap through sign-bit/zero here */
2575 if ((block + ret) < 0 || (block + ret) < block) {
2576 ret = -EIO;
2577 ext4_mark_inode_dirty(handle, inode);
2578 ret2 = ext4_journal_stop(handle);
2579 break;
2580 }
2581 if (buffer_new(&map_bh) && ((block + ret) >
2582 (EXT4_BLOCK_ALIGN(i_size_read(inode), blkbits)
2583 >> blkbits)))
2584 nblocks = nblocks + ret;
2585 }
2586
2587 /* Update ctime if new blocks get allocated */
2588 if (nblocks) {
2589 struct timespec now;
2590
2591 now = current_fs_time(inode->i_sb);
2592 if (!timespec_equal(&inode->i_ctime, &now))
2593 inode->i_ctime = now;
2594 }
2595
2596 ext4_mark_inode_dirty(handle, inode);
2597 ret2 = ext4_journal_stop(handle);
2598 if (ret2)
2599 break;
2600 }
2601
2602 if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
2603 goto retry;
2604
2605 /*
2606 * Time to update the file size.
2607 * Update only when preallocation was requested beyond the file size.
2608 */
2609 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
2610 (offset + len) > i_size_read(inode)) {
2611 if (ret > 0) {
2612 /*
2613 * if no error, we assume preallocation succeeded
2614 * completely
2615 */
2616 mutex_lock(&inode->i_mutex);
2617 i_size_write(inode, offset + len);
2618 EXT4_I(inode)->i_disksize = i_size_read(inode);
2619 mutex_unlock(&inode->i_mutex);
2620 } else if (ret < 0 && nblocks) {
2621 /* Handle partial allocation scenario */
2622 loff_t newsize;
2623
2624 mutex_lock(&inode->i_mutex);
2625 newsize = (nblocks << blkbits) + i_size_read(inode);
2626 i_size_write(inode, EXT4_BLOCK_ALIGN(newsize, blkbits));
2627 EXT4_I(inode)->i_disksize = i_size_read(inode);
2628 mutex_unlock(&inode->i_mutex);
2629 }
2630 }
2631
2632 return ret > 0 ? ret2 : ret;
2633}
diff --git a/fs/ext4/file.c b/fs/ext4/file.c
index d4c8186aed64..1a81cd66d63b 100644
--- a/fs/ext4/file.c
+++ b/fs/ext4/file.c
@@ -134,5 +134,6 @@ const struct inode_operations ext4_file_inode_operations = {
134 .removexattr = generic_removexattr, 134 .removexattr = generic_removexattr,
135#endif 135#endif
136 .permission = ext4_permission, 136 .permission = ext4_permission,
137 .fallocate = ext4_fallocate,
137}; 138};
138 139
diff --git a/fs/ext4/ialloc.c b/fs/ext4/ialloc.c
index c88b439ba5cd..427f83066a0d 100644
--- a/fs/ext4/ialloc.c
+++ b/fs/ext4/ialloc.c
@@ -563,7 +563,8 @@ got:
563 inode->i_ino = ino; 563 inode->i_ino = ino;
564 /* This is the optimal IO size (for stat), not the fs block size */ 564 /* This is the optimal IO size (for stat), not the fs block size */
565 inode->i_blocks = 0; 565 inode->i_blocks = 0;
566 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC; 566 inode->i_mtime = inode->i_atime = inode->i_ctime = ei->i_crtime =
567 ext4_current_time(inode);
567 568
568 memset(ei->i_data, 0, sizeof(ei->i_data)); 569 memset(ei->i_data, 0, sizeof(ei->i_data));
569 ei->i_dir_start_lookup = 0; 570 ei->i_dir_start_lookup = 0;
@@ -595,9 +596,8 @@ got:
595 spin_unlock(&sbi->s_next_gen_lock); 596 spin_unlock(&sbi->s_next_gen_lock);
596 597
597 ei->i_state = EXT4_STATE_NEW; 598 ei->i_state = EXT4_STATE_NEW;
598 ei->i_extra_isize = 599
599 (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE) ? 600 ei->i_extra_isize = EXT4_SB(sb)->s_want_extra_isize;
600 sizeof(struct ext4_inode) - EXT4_GOOD_OLD_INODE_SIZE : 0;
601 601
602 ret = inode; 602 ret = inode;
603 if(DQUOT_ALLOC_INODE(inode)) { 603 if(DQUOT_ALLOC_INODE(inode)) {
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 8416fa28c422..de26c25d6a18 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -726,7 +726,7 @@ static int ext4_splice_branch(handle_t *handle, struct inode *inode,
726 726
727 /* We are done with atomic stuff, now do the rest of housekeeping */ 727 /* We are done with atomic stuff, now do the rest of housekeeping */
728 728
729 inode->i_ctime = CURRENT_TIME_SEC; 729 inode->i_ctime = ext4_current_time(inode);
730 ext4_mark_inode_dirty(handle, inode); 730 ext4_mark_inode_dirty(handle, inode);
731 731
732 /* had we spliced it onto indirect block? */ 732 /* had we spliced it onto indirect block? */
@@ -1766,7 +1766,6 @@ int ext4_block_truncate_page(handle_t *handle, struct page *page,
1766 struct inode *inode = mapping->host; 1766 struct inode *inode = mapping->host;
1767 struct buffer_head *bh; 1767 struct buffer_head *bh;
1768 int err = 0; 1768 int err = 0;
1769 void *kaddr;
1770 1769
1771 blocksize = inode->i_sb->s_blocksize; 1770 blocksize = inode->i_sb->s_blocksize;
1772 length = blocksize - (offset & (blocksize - 1)); 1771 length = blocksize - (offset & (blocksize - 1));
@@ -1778,10 +1777,7 @@ int ext4_block_truncate_page(handle_t *handle, struct page *page,
1778 */ 1777 */
1779 if (!page_has_buffers(page) && test_opt(inode->i_sb, NOBH) && 1778 if (!page_has_buffers(page) && test_opt(inode->i_sb, NOBH) &&
1780 ext4_should_writeback_data(inode) && PageUptodate(page)) { 1779 ext4_should_writeback_data(inode) && PageUptodate(page)) {
1781 kaddr = kmap_atomic(page, KM_USER0); 1780 zero_user_page(page, offset, length, KM_USER0);
1782 memset(kaddr + offset, 0, length);
1783 flush_dcache_page(page);
1784 kunmap_atomic(kaddr, KM_USER0);
1785 set_page_dirty(page); 1781 set_page_dirty(page);
1786 goto unlock; 1782 goto unlock;
1787 } 1783 }
@@ -1834,10 +1830,7 @@ int ext4_block_truncate_page(handle_t *handle, struct page *page,
1834 goto unlock; 1830 goto unlock;
1835 } 1831 }
1836 1832
1837 kaddr = kmap_atomic(page, KM_USER0); 1833 zero_user_page(page, offset, length, KM_USER0);
1838 memset(kaddr + offset, 0, length);
1839 flush_dcache_page(page);
1840 kunmap_atomic(kaddr, KM_USER0);
1841 1834
1842 BUFFER_TRACE(bh, "zeroed end of block"); 1835 BUFFER_TRACE(bh, "zeroed end of block");
1843 1836
@@ -2375,7 +2368,7 @@ do_indirects:
2375 ext4_discard_reservation(inode); 2368 ext4_discard_reservation(inode);
2376 2369
2377 mutex_unlock(&ei->truncate_mutex); 2370 mutex_unlock(&ei->truncate_mutex);
2378 inode->i_mtime = inode->i_ctime = CURRENT_TIME_SEC; 2371 inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
2379 ext4_mark_inode_dirty(handle, inode); 2372 ext4_mark_inode_dirty(handle, inode);
2380 2373
2381 /* 2374 /*
@@ -2583,6 +2576,25 @@ void ext4_set_inode_flags(struct inode *inode)
2583 inode->i_flags |= S_DIRSYNC; 2576 inode->i_flags |= S_DIRSYNC;
2584} 2577}
2585 2578
2579/* Propagate flags from i_flags to EXT4_I(inode)->i_flags */
2580void ext4_get_inode_flags(struct ext4_inode_info *ei)
2581{
2582 unsigned int flags = ei->vfs_inode.i_flags;
2583
2584 ei->i_flags &= ~(EXT4_SYNC_FL|EXT4_APPEND_FL|
2585 EXT4_IMMUTABLE_FL|EXT4_NOATIME_FL|EXT4_DIRSYNC_FL);
2586 if (flags & S_SYNC)
2587 ei->i_flags |= EXT4_SYNC_FL;
2588 if (flags & S_APPEND)
2589 ei->i_flags |= EXT4_APPEND_FL;
2590 if (flags & S_IMMUTABLE)
2591 ei->i_flags |= EXT4_IMMUTABLE_FL;
2592 if (flags & S_NOATIME)
2593 ei->i_flags |= EXT4_NOATIME_FL;
2594 if (flags & S_DIRSYNC)
2595 ei->i_flags |= EXT4_DIRSYNC_FL;
2596}
2597
2586void ext4_read_inode(struct inode * inode) 2598void ext4_read_inode(struct inode * inode)
2587{ 2599{
2588 struct ext4_iloc iloc; 2600 struct ext4_iloc iloc;
@@ -2610,10 +2622,6 @@ void ext4_read_inode(struct inode * inode)
2610 } 2622 }
2611 inode->i_nlink = le16_to_cpu(raw_inode->i_links_count); 2623 inode->i_nlink = le16_to_cpu(raw_inode->i_links_count);
2612 inode->i_size = le32_to_cpu(raw_inode->i_size); 2624 inode->i_size = le32_to_cpu(raw_inode->i_size);
2613 inode->i_atime.tv_sec = (signed)le32_to_cpu(raw_inode->i_atime);
2614 inode->i_ctime.tv_sec = (signed)le32_to_cpu(raw_inode->i_ctime);
2615 inode->i_mtime.tv_sec = (signed)le32_to_cpu(raw_inode->i_mtime);
2616 inode->i_atime.tv_nsec = inode->i_ctime.tv_nsec = inode->i_mtime.tv_nsec = 0;
2617 2625
2618 ei->i_state = 0; 2626 ei->i_state = 0;
2619 ei->i_dir_start_lookup = 0; 2627 ei->i_dir_start_lookup = 0;
@@ -2691,6 +2699,11 @@ void ext4_read_inode(struct inode * inode)
2691 } else 2699 } else
2692 ei->i_extra_isize = 0; 2700 ei->i_extra_isize = 0;
2693 2701
2702 EXT4_INODE_GET_XTIME(i_ctime, inode, raw_inode);
2703 EXT4_INODE_GET_XTIME(i_mtime, inode, raw_inode);
2704 EXT4_INODE_GET_XTIME(i_atime, inode, raw_inode);
2705 EXT4_EINODE_GET_XTIME(i_crtime, ei, raw_inode);
2706
2694 if (S_ISREG(inode->i_mode)) { 2707 if (S_ISREG(inode->i_mode)) {
2695 inode->i_op = &ext4_file_inode_operations; 2708 inode->i_op = &ext4_file_inode_operations;
2696 inode->i_fop = &ext4_file_operations; 2709 inode->i_fop = &ext4_file_operations;
@@ -2744,6 +2757,7 @@ static int ext4_do_update_inode(handle_t *handle,
2744 if (ei->i_state & EXT4_STATE_NEW) 2757 if (ei->i_state & EXT4_STATE_NEW)
2745 memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size); 2758 memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
2746 2759
2760 ext4_get_inode_flags(ei);
2747 raw_inode->i_mode = cpu_to_le16(inode->i_mode); 2761 raw_inode->i_mode = cpu_to_le16(inode->i_mode);
2748 if(!(test_opt(inode->i_sb, NO_UID32))) { 2762 if(!(test_opt(inode->i_sb, NO_UID32))) {
2749 raw_inode->i_uid_low = cpu_to_le16(low_16_bits(inode->i_uid)); 2763 raw_inode->i_uid_low = cpu_to_le16(low_16_bits(inode->i_uid));
@@ -2771,9 +2785,12 @@ static int ext4_do_update_inode(handle_t *handle,
2771 } 2785 }
2772 raw_inode->i_links_count = cpu_to_le16(inode->i_nlink); 2786 raw_inode->i_links_count = cpu_to_le16(inode->i_nlink);
2773 raw_inode->i_size = cpu_to_le32(ei->i_disksize); 2787 raw_inode->i_size = cpu_to_le32(ei->i_disksize);
2774 raw_inode->i_atime = cpu_to_le32(inode->i_atime.tv_sec); 2788
2775 raw_inode->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec); 2789 EXT4_INODE_SET_XTIME(i_ctime, inode, raw_inode);
2776 raw_inode->i_mtime = cpu_to_le32(inode->i_mtime.tv_sec); 2790 EXT4_INODE_SET_XTIME(i_mtime, inode, raw_inode);
2791 EXT4_INODE_SET_XTIME(i_atime, inode, raw_inode);
2792 EXT4_EINODE_SET_XTIME(i_crtime, ei, raw_inode);
2793
2777 raw_inode->i_blocks = cpu_to_le32(inode->i_blocks); 2794 raw_inode->i_blocks = cpu_to_le32(inode->i_blocks);
2778 raw_inode->i_dtime = cpu_to_le32(ei->i_dtime); 2795 raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
2779 raw_inode->i_flags = cpu_to_le32(ei->i_flags); 2796 raw_inode->i_flags = cpu_to_le32(ei->i_flags);
@@ -3082,6 +3099,39 @@ ext4_reserve_inode_write(handle_t *handle, struct inode *inode,
3082} 3099}
3083 3100
3084/* 3101/*
3102 * Expand an inode by new_extra_isize bytes.
3103 * Returns 0 on success or negative error number on failure.
3104 */
3105int ext4_expand_extra_isize(struct inode *inode, unsigned int new_extra_isize,
3106 struct ext4_iloc iloc, handle_t *handle)
3107{
3108 struct ext4_inode *raw_inode;
3109 struct ext4_xattr_ibody_header *header;
3110 struct ext4_xattr_entry *entry;
3111
3112 if (EXT4_I(inode)->i_extra_isize >= new_extra_isize)
3113 return 0;
3114
3115 raw_inode = ext4_raw_inode(&iloc);
3116
3117 header = IHDR(inode, raw_inode);
3118 entry = IFIRST(header);
3119
3120 /* No extended attributes present */
3121 if (!(EXT4_I(inode)->i_state & EXT4_STATE_XATTR) ||
3122 header->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC)) {
3123 memset((void *)raw_inode + EXT4_GOOD_OLD_INODE_SIZE, 0,
3124 new_extra_isize);
3125 EXT4_I(inode)->i_extra_isize = new_extra_isize;
3126 return 0;
3127 }
3128
3129 /* try to expand with EAs present */
3130 return ext4_expand_extra_isize_ea(inode, new_extra_isize,
3131 raw_inode, handle);
3132}
3133
3134/*
3085 * What we do here is to mark the in-core inode as clean with respect to inode 3135 * What we do here is to mark the in-core inode as clean with respect to inode
3086 * dirtiness (it may still be data-dirty). 3136 * dirtiness (it may still be data-dirty).
3087 * This means that the in-core inode may be reaped by prune_icache 3137 * This means that the in-core inode may be reaped by prune_icache
@@ -3105,10 +3155,38 @@ ext4_reserve_inode_write(handle_t *handle, struct inode *inode,
3105int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode) 3155int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode)
3106{ 3156{
3107 struct ext4_iloc iloc; 3157 struct ext4_iloc iloc;
3108 int err; 3158 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
3159 static unsigned int mnt_count;
3160 int err, ret;
3109 3161
3110 might_sleep(); 3162 might_sleep();
3111 err = ext4_reserve_inode_write(handle, inode, &iloc); 3163 err = ext4_reserve_inode_write(handle, inode, &iloc);
3164 if (EXT4_I(inode)->i_extra_isize < sbi->s_want_extra_isize &&
3165 !(EXT4_I(inode)->i_state & EXT4_STATE_NO_EXPAND)) {
3166 /*
3167 * We need extra buffer credits since we may write into EA block
3168 * with this same handle. If journal_extend fails, then it will
3169 * only result in a minor loss of functionality for that inode.
3170 * If this is felt to be critical, then e2fsck should be run to
3171 * force a large enough s_min_extra_isize.
3172 */
3173 if ((jbd2_journal_extend(handle,
3174 EXT4_DATA_TRANS_BLOCKS(inode->i_sb))) == 0) {
3175 ret = ext4_expand_extra_isize(inode,
3176 sbi->s_want_extra_isize,
3177 iloc, handle);
3178 if (ret) {
3179 EXT4_I(inode)->i_state |= EXT4_STATE_NO_EXPAND;
3180 if (mnt_count != sbi->s_es->s_mnt_count) {
3181 ext4_warning(inode->i_sb, __FUNCTION__,
3182 "Unable to expand inode %lu. Delete"
3183 " some EAs or run e2fsck.",
3184 inode->i_ino);
3185 mnt_count = sbi->s_es->s_mnt_count;
3186 }
3187 }
3188 }
3189 }
3112 if (!err) 3190 if (!err)
3113 err = ext4_mark_iloc_dirty(handle, inode, &iloc); 3191 err = ext4_mark_iloc_dirty(handle, inode, &iloc);
3114 return err; 3192 return err;
@@ -3197,7 +3275,7 @@ int ext4_change_inode_journal_flag(struct inode *inode, int val)
3197 */ 3275 */
3198 3276
3199 journal = EXT4_JOURNAL(inode); 3277 journal = EXT4_JOURNAL(inode);
3200 if (is_journal_aborted(journal) || IS_RDONLY(inode)) 3278 if (is_journal_aborted(journal))
3201 return -EROFS; 3279 return -EROFS;
3202 3280
3203 jbd2_journal_lock_updates(journal); 3281 jbd2_journal_lock_updates(journal);
diff --git a/fs/ext4/ioctl.c b/fs/ext4/ioctl.c
index 7b4aa4543c83..c04c7ccba9e3 100644
--- a/fs/ext4/ioctl.c
+++ b/fs/ext4/ioctl.c
@@ -28,6 +28,7 @@ int ext4_ioctl (struct inode * inode, struct file * filp, unsigned int cmd,
28 28
29 switch (cmd) { 29 switch (cmd) {
30 case EXT4_IOC_GETFLAGS: 30 case EXT4_IOC_GETFLAGS:
31 ext4_get_inode_flags(ei);
31 flags = ei->i_flags & EXT4_FL_USER_VISIBLE; 32 flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
32 return put_user(flags, (int __user *) arg); 33 return put_user(flags, (int __user *) arg);
33 case EXT4_IOC_SETFLAGS: { 34 case EXT4_IOC_SETFLAGS: {
@@ -96,7 +97,7 @@ int ext4_ioctl (struct inode * inode, struct file * filp, unsigned int cmd,
96 ei->i_flags = flags; 97 ei->i_flags = flags;
97 98
98 ext4_set_inode_flags(inode); 99 ext4_set_inode_flags(inode);
99 inode->i_ctime = CURRENT_TIME_SEC; 100 inode->i_ctime = ext4_current_time(inode);
100 101
101 err = ext4_mark_iloc_dirty(handle, inode, &iloc); 102 err = ext4_mark_iloc_dirty(handle, inode, &iloc);
102flags_err: 103flags_err:
@@ -133,14 +134,14 @@ flags_err:
133 return PTR_ERR(handle); 134 return PTR_ERR(handle);
134 err = ext4_reserve_inode_write(handle, inode, &iloc); 135 err = ext4_reserve_inode_write(handle, inode, &iloc);
135 if (err == 0) { 136 if (err == 0) {
136 inode->i_ctime = CURRENT_TIME_SEC; 137 inode->i_ctime = ext4_current_time(inode);
137 inode->i_generation = generation; 138 inode->i_generation = generation;
138 err = ext4_mark_iloc_dirty(handle, inode, &iloc); 139 err = ext4_mark_iloc_dirty(handle, inode, &iloc);
139 } 140 }
140 ext4_journal_stop(handle); 141 ext4_journal_stop(handle);
141 return err; 142 return err;
142 } 143 }
143#ifdef CONFIG_JBD_DEBUG 144#ifdef CONFIG_JBD2_DEBUG
144 case EXT4_IOC_WAIT_FOR_READONLY: 145 case EXT4_IOC_WAIT_FOR_READONLY:
145 /* 146 /*
146 * This is racy - by the time we're woken up and running, 147 * This is racy - by the time we're woken up and running,
@@ -282,7 +283,7 @@ long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
282 case EXT4_IOC32_SETVERSION_OLD: 283 case EXT4_IOC32_SETVERSION_OLD:
283 cmd = EXT4_IOC_SETVERSION_OLD; 284 cmd = EXT4_IOC_SETVERSION_OLD;
284 break; 285 break;
285#ifdef CONFIG_JBD_DEBUG 286#ifdef CONFIG_JBD2_DEBUG
286 case EXT4_IOC32_WAIT_FOR_READONLY: 287 case EXT4_IOC32_WAIT_FOR_READONLY:
287 cmd = EXT4_IOC_WAIT_FOR_READONLY; 288 cmd = EXT4_IOC_WAIT_FOR_READONLY;
288 break; 289 break;
diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
index 2de339dd7554..da224974af78 100644
--- a/fs/ext4/namei.c
+++ b/fs/ext4/namei.c
@@ -1295,7 +1295,7 @@ static int add_dirent_to_buf(handle_t *handle, struct dentry *dentry,
1295 * happen is that the times are slightly out of date 1295 * happen is that the times are slightly out of date
1296 * and/or different from the directory change time. 1296 * and/or different from the directory change time.
1297 */ 1297 */
1298 dir->i_mtime = dir->i_ctime = CURRENT_TIME_SEC; 1298 dir->i_mtime = dir->i_ctime = ext4_current_time(dir);
1299 ext4_update_dx_flag(dir); 1299 ext4_update_dx_flag(dir);
1300 dir->i_version++; 1300 dir->i_version++;
1301 ext4_mark_inode_dirty(handle, dir); 1301 ext4_mark_inode_dirty(handle, dir);
@@ -1629,6 +1629,35 @@ static int ext4_delete_entry (handle_t *handle,
1629 return -ENOENT; 1629 return -ENOENT;
1630} 1630}
1631 1631
1632/*
1633 * DIR_NLINK feature is set if 1) nlinks > EXT4_LINK_MAX or 2) nlinks == 2,
1634 * since this indicates that nlinks count was previously 1.
1635 */
1636static void ext4_inc_count(handle_t *handle, struct inode *inode)
1637{
1638 inc_nlink(inode);
1639 if (is_dx(inode) && inode->i_nlink > 1) {
1640 /* limit is 16-bit i_links_count */
1641 if (inode->i_nlink >= EXT4_LINK_MAX || inode->i_nlink == 2) {
1642 inode->i_nlink = 1;
1643 EXT4_SET_RO_COMPAT_FEATURE(inode->i_sb,
1644 EXT4_FEATURE_RO_COMPAT_DIR_NLINK);
1645 }
1646 }
1647}
1648
1649/*
1650 * If a directory had nlink == 1, then we should let it be 1. This indicates
1651 * directory has >EXT4_LINK_MAX subdirs.
1652 */
1653static void ext4_dec_count(handle_t *handle, struct inode *inode)
1654{
1655 drop_nlink(inode);
1656 if (S_ISDIR(inode->i_mode) && inode->i_nlink == 0)
1657 inc_nlink(inode);
1658}
1659
1660
1632static int ext4_add_nondir(handle_t *handle, 1661static int ext4_add_nondir(handle_t *handle,
1633 struct dentry *dentry, struct inode *inode) 1662 struct dentry *dentry, struct inode *inode)
1634{ 1663{
@@ -1725,7 +1754,7 @@ static int ext4_mkdir(struct inode * dir, struct dentry * dentry, int mode)
1725 struct ext4_dir_entry_2 * de; 1754 struct ext4_dir_entry_2 * de;
1726 int err, retries = 0; 1755 int err, retries = 0;
1727 1756
1728 if (dir->i_nlink >= EXT4_LINK_MAX) 1757 if (EXT4_DIR_LINK_MAX(dir))
1729 return -EMLINK; 1758 return -EMLINK;
1730 1759
1731retry: 1760retry:
@@ -1748,7 +1777,7 @@ retry:
1748 inode->i_size = EXT4_I(inode)->i_disksize = inode->i_sb->s_blocksize; 1777 inode->i_size = EXT4_I(inode)->i_disksize = inode->i_sb->s_blocksize;
1749 dir_block = ext4_bread (handle, inode, 0, 1, &err); 1778 dir_block = ext4_bread (handle, inode, 0, 1, &err);
1750 if (!dir_block) { 1779 if (!dir_block) {
1751 drop_nlink(inode); /* is this nlink == 0? */ 1780 ext4_dec_count(handle, inode); /* is this nlink == 0? */
1752 ext4_mark_inode_dirty(handle, inode); 1781 ext4_mark_inode_dirty(handle, inode);
1753 iput (inode); 1782 iput (inode);
1754 goto out_stop; 1783 goto out_stop;
@@ -1780,7 +1809,7 @@ retry:
1780 iput (inode); 1809 iput (inode);
1781 goto out_stop; 1810 goto out_stop;
1782 } 1811 }
1783 inc_nlink(dir); 1812 ext4_inc_count(handle, dir);
1784 ext4_update_dx_flag(dir); 1813 ext4_update_dx_flag(dir);
1785 ext4_mark_inode_dirty(handle, dir); 1814 ext4_mark_inode_dirty(handle, dir);
1786 d_instantiate(dentry, inode); 1815 d_instantiate(dentry, inode);
@@ -2045,9 +2074,9 @@ static int ext4_rmdir (struct inode * dir, struct dentry *dentry)
2045 retval = ext4_delete_entry(handle, dir, de, bh); 2074 retval = ext4_delete_entry(handle, dir, de, bh);
2046 if (retval) 2075 if (retval)
2047 goto end_rmdir; 2076 goto end_rmdir;
2048 if (inode->i_nlink != 2) 2077 if (!EXT4_DIR_LINK_EMPTY(inode))
2049 ext4_warning (inode->i_sb, "ext4_rmdir", 2078 ext4_warning (inode->i_sb, "ext4_rmdir",
2050 "empty directory has nlink!=2 (%d)", 2079 "empty directory has too many links (%d)",
2051 inode->i_nlink); 2080 inode->i_nlink);
2052 inode->i_version++; 2081 inode->i_version++;
2053 clear_nlink(inode); 2082 clear_nlink(inode);
@@ -2056,9 +2085,9 @@ static int ext4_rmdir (struct inode * dir, struct dentry *dentry)
2056 * recovery. */ 2085 * recovery. */
2057 inode->i_size = 0; 2086 inode->i_size = 0;
2058 ext4_orphan_add(handle, inode); 2087 ext4_orphan_add(handle, inode);
2059 inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME_SEC; 2088 inode->i_ctime = dir->i_ctime = dir->i_mtime = ext4_current_time(inode);
2060 ext4_mark_inode_dirty(handle, inode); 2089 ext4_mark_inode_dirty(handle, inode);
2061 drop_nlink(dir); 2090 ext4_dec_count(handle, dir);
2062 ext4_update_dx_flag(dir); 2091 ext4_update_dx_flag(dir);
2063 ext4_mark_inode_dirty(handle, dir); 2092 ext4_mark_inode_dirty(handle, dir);
2064 2093
@@ -2106,13 +2135,13 @@ static int ext4_unlink(struct inode * dir, struct dentry *dentry)
2106 retval = ext4_delete_entry(handle, dir, de, bh); 2135 retval = ext4_delete_entry(handle, dir, de, bh);
2107 if (retval) 2136 if (retval)
2108 goto end_unlink; 2137 goto end_unlink;
2109 dir->i_ctime = dir->i_mtime = CURRENT_TIME_SEC; 2138 dir->i_ctime = dir->i_mtime = ext4_current_time(dir);
2110 ext4_update_dx_flag(dir); 2139 ext4_update_dx_flag(dir);
2111 ext4_mark_inode_dirty(handle, dir); 2140 ext4_mark_inode_dirty(handle, dir);
2112 drop_nlink(inode); 2141 ext4_dec_count(handle, inode);
2113 if (!inode->i_nlink) 2142 if (!inode->i_nlink)
2114 ext4_orphan_add(handle, inode); 2143 ext4_orphan_add(handle, inode);
2115 inode->i_ctime = dir->i_ctime; 2144 inode->i_ctime = ext4_current_time(inode);
2116 ext4_mark_inode_dirty(handle, inode); 2145 ext4_mark_inode_dirty(handle, inode);
2117 retval = 0; 2146 retval = 0;
2118 2147
@@ -2159,7 +2188,7 @@ retry:
2159 err = __page_symlink(inode, symname, l, 2188 err = __page_symlink(inode, symname, l,
2160 mapping_gfp_mask(inode->i_mapping) & ~__GFP_FS); 2189 mapping_gfp_mask(inode->i_mapping) & ~__GFP_FS);
2161 if (err) { 2190 if (err) {
2162 drop_nlink(inode); 2191 ext4_dec_count(handle, inode);
2163 ext4_mark_inode_dirty(handle, inode); 2192 ext4_mark_inode_dirty(handle, inode);
2164 iput (inode); 2193 iput (inode);
2165 goto out_stop; 2194 goto out_stop;
@@ -2185,8 +2214,9 @@ static int ext4_link (struct dentry * old_dentry,
2185 struct inode *inode = old_dentry->d_inode; 2214 struct inode *inode = old_dentry->d_inode;
2186 int err, retries = 0; 2215 int err, retries = 0;
2187 2216
2188 if (inode->i_nlink >= EXT4_LINK_MAX) 2217 if (EXT4_DIR_LINK_MAX(inode))
2189 return -EMLINK; 2218 return -EMLINK;
2219
2190 /* 2220 /*
2191 * Return -ENOENT if we've raced with unlink and i_nlink is 0. Doing 2221 * Return -ENOENT if we've raced with unlink and i_nlink is 0. Doing
2192 * otherwise has the potential to corrupt the orphan inode list. 2222 * otherwise has the potential to corrupt the orphan inode list.
@@ -2203,8 +2233,8 @@ retry:
2203 if (IS_DIRSYNC(dir)) 2233 if (IS_DIRSYNC(dir))
2204 handle->h_sync = 1; 2234 handle->h_sync = 1;
2205 2235
2206 inode->i_ctime = CURRENT_TIME_SEC; 2236 inode->i_ctime = ext4_current_time(inode);
2207 inc_nlink(inode); 2237 ext4_inc_count(handle, inode);
2208 atomic_inc(&inode->i_count); 2238 atomic_inc(&inode->i_count);
2209 2239
2210 err = ext4_add_nondir(handle, dentry, inode); 2240 err = ext4_add_nondir(handle, dentry, inode);
@@ -2305,7 +2335,7 @@ static int ext4_rename (struct inode * old_dir, struct dentry *old_dentry,
2305 * Like most other Unix systems, set the ctime for inodes on a 2335 * Like most other Unix systems, set the ctime for inodes on a
2306 * rename. 2336 * rename.
2307 */ 2337 */
2308 old_inode->i_ctime = CURRENT_TIME_SEC; 2338 old_inode->i_ctime = ext4_current_time(old_inode);
2309 ext4_mark_inode_dirty(handle, old_inode); 2339 ext4_mark_inode_dirty(handle, old_inode);
2310 2340
2311 /* 2341 /*
@@ -2337,10 +2367,10 @@ static int ext4_rename (struct inode * old_dir, struct dentry *old_dentry,
2337 } 2367 }
2338 2368
2339 if (new_inode) { 2369 if (new_inode) {
2340 drop_nlink(new_inode); 2370 ext4_dec_count(handle, new_inode);
2341 new_inode->i_ctime = CURRENT_TIME_SEC; 2371 new_inode->i_ctime = ext4_current_time(new_inode);
2342 } 2372 }
2343 old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME_SEC; 2373 old_dir->i_ctime = old_dir->i_mtime = ext4_current_time(old_dir);
2344 ext4_update_dx_flag(old_dir); 2374 ext4_update_dx_flag(old_dir);
2345 if (dir_bh) { 2375 if (dir_bh) {
2346 BUFFER_TRACE(dir_bh, "get_write_access"); 2376 BUFFER_TRACE(dir_bh, "get_write_access");
@@ -2348,11 +2378,13 @@ static int ext4_rename (struct inode * old_dir, struct dentry *old_dentry,
2348 PARENT_INO(dir_bh->b_data) = cpu_to_le32(new_dir->i_ino); 2378 PARENT_INO(dir_bh->b_data) = cpu_to_le32(new_dir->i_ino);
2349 BUFFER_TRACE(dir_bh, "call ext4_journal_dirty_metadata"); 2379 BUFFER_TRACE(dir_bh, "call ext4_journal_dirty_metadata");
2350 ext4_journal_dirty_metadata(handle, dir_bh); 2380 ext4_journal_dirty_metadata(handle, dir_bh);
2351 drop_nlink(old_dir); 2381 ext4_dec_count(handle, old_dir);
2352 if (new_inode) { 2382 if (new_inode) {
2353 drop_nlink(new_inode); 2383 /* checked empty_dir above, can't have another parent,
2384 * ext3_dec_count() won't work for many-linked dirs */
2385 new_inode->i_nlink = 0;
2354 } else { 2386 } else {
2355 inc_nlink(new_dir); 2387 ext4_inc_count(handle, new_dir);
2356 ext4_update_dx_flag(new_dir); 2388 ext4_update_dx_flag(new_dir);
2357 ext4_mark_inode_dirty(handle, new_dir); 2389 ext4_mark_inode_dirty(handle, new_dir);
2358 } 2390 }
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index b806e689c4aa..6dcbb28dc06d 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -36,6 +36,7 @@
36#include <linux/namei.h> 36#include <linux/namei.h>
37#include <linux/quotaops.h> 37#include <linux/quotaops.h>
38#include <linux/seq_file.h> 38#include <linux/seq_file.h>
39#include <linux/log2.h>
39 40
40#include <asm/uaccess.h> 41#include <asm/uaccess.h>
41 42
@@ -734,7 +735,7 @@ enum {
734 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 735 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
735 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota, 736 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
736 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota, 737 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
737 Opt_grpquota, Opt_extents, 738 Opt_grpquota, Opt_extents, Opt_noextents,
738}; 739};
739 740
740static match_table_t tokens = { 741static match_table_t tokens = {
@@ -785,6 +786,7 @@ static match_table_t tokens = {
785 {Opt_usrquota, "usrquota"}, 786 {Opt_usrquota, "usrquota"},
786 {Opt_barrier, "barrier=%u"}, 787 {Opt_barrier, "barrier=%u"},
787 {Opt_extents, "extents"}, 788 {Opt_extents, "extents"},
789 {Opt_noextents, "noextents"},
788 {Opt_err, NULL}, 790 {Opt_err, NULL},
789 {Opt_resize, "resize"}, 791 {Opt_resize, "resize"},
790}; 792};
@@ -1120,6 +1122,9 @@ clear_qf_name:
1120 case Opt_extents: 1122 case Opt_extents:
1121 set_opt (sbi->s_mount_opt, EXTENTS); 1123 set_opt (sbi->s_mount_opt, EXTENTS);
1122 break; 1124 break;
1125 case Opt_noextents:
1126 clear_opt (sbi->s_mount_opt, EXTENTS);
1127 break;
1123 default: 1128 default:
1124 printk (KERN_ERR 1129 printk (KERN_ERR
1125 "EXT4-fs: Unrecognized mount option \"%s\" " 1130 "EXT4-fs: Unrecognized mount option \"%s\" "
@@ -1551,6 +1556,12 @@ static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1551 1556
1552 set_opt(sbi->s_mount_opt, RESERVATION); 1557 set_opt(sbi->s_mount_opt, RESERVATION);
1553 1558
1559 /*
1560 * turn on extents feature by default in ext4 filesystem
1561 * User -o noextents to turn it off
1562 */
1563 set_opt(sbi->s_mount_opt, EXTENTS);
1564
1554 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum, 1565 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1555 NULL, 0)) 1566 NULL, 0))
1556 goto failed_mount; 1567 goto failed_mount;
@@ -1634,13 +1645,15 @@ static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1634 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 1645 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1635 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 1646 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1636 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 1647 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1637 (sbi->s_inode_size & (sbi->s_inode_size - 1)) || 1648 (!is_power_of_2(sbi->s_inode_size)) ||
1638 (sbi->s_inode_size > blocksize)) { 1649 (sbi->s_inode_size > blocksize)) {
1639 printk (KERN_ERR 1650 printk (KERN_ERR
1640 "EXT4-fs: unsupported inode size: %d\n", 1651 "EXT4-fs: unsupported inode size: %d\n",
1641 sbi->s_inode_size); 1652 sbi->s_inode_size);
1642 goto failed_mount; 1653 goto failed_mount;
1643 } 1654 }
1655 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
1656 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
1644 } 1657 }
1645 sbi->s_frag_size = EXT4_MIN_FRAG_SIZE << 1658 sbi->s_frag_size = EXT4_MIN_FRAG_SIZE <<
1646 le32_to_cpu(es->s_log_frag_size); 1659 le32_to_cpu(es->s_log_frag_size);
@@ -1803,6 +1816,13 @@ static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1803 goto failed_mount3; 1816 goto failed_mount3;
1804 } 1817 }
1805 1818
1819 if (ext4_blocks_count(es) > 0xffffffffULL &&
1820 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
1821 JBD2_FEATURE_INCOMPAT_64BIT)) {
1822 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
1823 goto failed_mount4;
1824 }
1825
1806 /* We have now updated the journal if required, so we can 1826 /* We have now updated the journal if required, so we can
1807 * validate the data journaling mode. */ 1827 * validate the data journaling mode. */
1808 switch (test_opt(sb, DATA_FLAGS)) { 1828 switch (test_opt(sb, DATA_FLAGS)) {
@@ -1857,6 +1877,32 @@ static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1857 } 1877 }
1858 1878
1859 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY); 1879 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1880
1881 /* determine the minimum size of new large inodes, if present */
1882 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
1883 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
1884 EXT4_GOOD_OLD_INODE_SIZE;
1885 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1886 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
1887 if (sbi->s_want_extra_isize <
1888 le16_to_cpu(es->s_want_extra_isize))
1889 sbi->s_want_extra_isize =
1890 le16_to_cpu(es->s_want_extra_isize);
1891 if (sbi->s_want_extra_isize <
1892 le16_to_cpu(es->s_min_extra_isize))
1893 sbi->s_want_extra_isize =
1894 le16_to_cpu(es->s_min_extra_isize);
1895 }
1896 }
1897 /* Check if enough inode space is available */
1898 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
1899 sbi->s_inode_size) {
1900 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
1901 EXT4_GOOD_OLD_INODE_SIZE;
1902 printk(KERN_INFO "EXT4-fs: required extra inode space not"
1903 "available.\n");
1904 }
1905
1860 /* 1906 /*
1861 * akpm: core read_super() calls in here with the superblock locked. 1907 * akpm: core read_super() calls in here with the superblock locked.
1862 * That deadlocks, because orphan cleanup needs to lock the superblock 1908 * That deadlocks, because orphan cleanup needs to lock the superblock
diff --git a/fs/ext4/xattr.c b/fs/ext4/xattr.c
index e832e96095b3..b10d68fffb55 100644
--- a/fs/ext4/xattr.c
+++ b/fs/ext4/xattr.c
@@ -66,13 +66,6 @@
66#define BFIRST(bh) ENTRY(BHDR(bh)+1) 66#define BFIRST(bh) ENTRY(BHDR(bh)+1)
67#define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0) 67#define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0)
68 68
69#define IHDR(inode, raw_inode) \
70 ((struct ext4_xattr_ibody_header *) \
71 ((void *)raw_inode + \
72 EXT4_GOOD_OLD_INODE_SIZE + \
73 EXT4_I(inode)->i_extra_isize))
74#define IFIRST(hdr) ((struct ext4_xattr_entry *)((hdr)+1))
75
76#ifdef EXT4_XATTR_DEBUG 69#ifdef EXT4_XATTR_DEBUG
77# define ea_idebug(inode, f...) do { \ 70# define ea_idebug(inode, f...) do { \
78 printk(KERN_DEBUG "inode %s:%lu: ", \ 71 printk(KERN_DEBUG "inode %s:%lu: ", \
@@ -508,6 +501,24 @@ out:
508 return; 501 return;
509} 502}
510 503
504/*
505 * Find the available free space for EAs. This also returns the total number of
506 * bytes used by EA entries.
507 */
508static size_t ext4_xattr_free_space(struct ext4_xattr_entry *last,
509 size_t *min_offs, void *base, int *total)
510{
511 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
512 *total += EXT4_XATTR_LEN(last->e_name_len);
513 if (!last->e_value_block && last->e_value_size) {
514 size_t offs = le16_to_cpu(last->e_value_offs);
515 if (offs < *min_offs)
516 *min_offs = offs;
517 }
518 }
519 return (*min_offs - ((void *)last - base) - sizeof(__u32));
520}
521
511struct ext4_xattr_info { 522struct ext4_xattr_info {
512 int name_index; 523 int name_index;
513 const char *name; 524 const char *name;
@@ -1013,7 +1024,9 @@ ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
1013 } 1024 }
1014 if (!error) { 1025 if (!error) {
1015 ext4_xattr_update_super_block(handle, inode->i_sb); 1026 ext4_xattr_update_super_block(handle, inode->i_sb);
1016 inode->i_ctime = CURRENT_TIME_SEC; 1027 inode->i_ctime = ext4_current_time(inode);
1028 if (!value)
1029 EXT4_I(inode)->i_state &= ~EXT4_STATE_NO_EXPAND;
1017 error = ext4_mark_iloc_dirty(handle, inode, &is.iloc); 1030 error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1018 /* 1031 /*
1019 * The bh is consumed by ext4_mark_iloc_dirty, even with 1032 * The bh is consumed by ext4_mark_iloc_dirty, even with
@@ -1067,6 +1080,253 @@ retry:
1067} 1080}
1068 1081
1069/* 1082/*
1083 * Shift the EA entries in the inode to create space for the increased
1084 * i_extra_isize.
1085 */
1086static void ext4_xattr_shift_entries(struct ext4_xattr_entry *entry,
1087 int value_offs_shift, void *to,
1088 void *from, size_t n, int blocksize)
1089{
1090 struct ext4_xattr_entry *last = entry;
1091 int new_offs;
1092
1093 /* Adjust the value offsets of the entries */
1094 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1095 if (!last->e_value_block && last->e_value_size) {
1096 new_offs = le16_to_cpu(last->e_value_offs) +
1097 value_offs_shift;
1098 BUG_ON(new_offs + le32_to_cpu(last->e_value_size)
1099 > blocksize);
1100 last->e_value_offs = cpu_to_le16(new_offs);
1101 }
1102 }
1103 /* Shift the entries by n bytes */
1104 memmove(to, from, n);
1105}
1106
1107/*
1108 * Expand an inode by new_extra_isize bytes when EAs are present.
1109 * Returns 0 on success or negative error number on failure.
1110 */
1111int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
1112 struct ext4_inode *raw_inode, handle_t *handle)
1113{
1114 struct ext4_xattr_ibody_header *header;
1115 struct ext4_xattr_entry *entry, *last, *first;
1116 struct buffer_head *bh = NULL;
1117 struct ext4_xattr_ibody_find *is = NULL;
1118 struct ext4_xattr_block_find *bs = NULL;
1119 char *buffer = NULL, *b_entry_name = NULL;
1120 size_t min_offs, free;
1121 int total_ino, total_blk;
1122 void *base, *start, *end;
1123 int extra_isize = 0, error = 0, tried_min_extra_isize = 0;
1124 int s_min_extra_isize = EXT4_SB(inode->i_sb)->s_es->s_min_extra_isize;
1125
1126 down_write(&EXT4_I(inode)->xattr_sem);
1127retry:
1128 if (EXT4_I(inode)->i_extra_isize >= new_extra_isize) {
1129 up_write(&EXT4_I(inode)->xattr_sem);
1130 return 0;
1131 }
1132
1133 header = IHDR(inode, raw_inode);
1134 entry = IFIRST(header);
1135
1136 /*
1137 * Check if enough free space is available in the inode to shift the
1138 * entries ahead by new_extra_isize.
1139 */
1140
1141 base = start = entry;
1142 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
1143 min_offs = end - base;
1144 last = entry;
1145 total_ino = sizeof(struct ext4_xattr_ibody_header);
1146
1147 free = ext4_xattr_free_space(last, &min_offs, base, &total_ino);
1148 if (free >= new_extra_isize) {
1149 entry = IFIRST(header);
1150 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize
1151 - new_extra_isize, (void *)raw_inode +
1152 EXT4_GOOD_OLD_INODE_SIZE + new_extra_isize,
1153 (void *)header, total_ino,
1154 inode->i_sb->s_blocksize);
1155 EXT4_I(inode)->i_extra_isize = new_extra_isize;
1156 error = 0;
1157 goto cleanup;
1158 }
1159
1160 /*
1161 * Enough free space isn't available in the inode, check if
1162 * EA block can hold new_extra_isize bytes.
1163 */
1164 if (EXT4_I(inode)->i_file_acl) {
1165 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1166 error = -EIO;
1167 if (!bh)
1168 goto cleanup;
1169 if (ext4_xattr_check_block(bh)) {
1170 ext4_error(inode->i_sb, __FUNCTION__,
1171 "inode %lu: bad block %llu", inode->i_ino,
1172 EXT4_I(inode)->i_file_acl);
1173 error = -EIO;
1174 goto cleanup;
1175 }
1176 base = BHDR(bh);
1177 first = BFIRST(bh);
1178 end = bh->b_data + bh->b_size;
1179 min_offs = end - base;
1180 free = ext4_xattr_free_space(first, &min_offs, base,
1181 &total_blk);
1182 if (free < new_extra_isize) {
1183 if (!tried_min_extra_isize && s_min_extra_isize) {
1184 tried_min_extra_isize++;
1185 new_extra_isize = s_min_extra_isize;
1186 brelse(bh);
1187 goto retry;
1188 }
1189 error = -1;
1190 goto cleanup;
1191 }
1192 } else {
1193 free = inode->i_sb->s_blocksize;
1194 }
1195
1196 while (new_extra_isize > 0) {
1197 size_t offs, size, entry_size;
1198 struct ext4_xattr_entry *small_entry = NULL;
1199 struct ext4_xattr_info i = {
1200 .value = NULL,
1201 .value_len = 0,
1202 };
1203 unsigned int total_size; /* EA entry size + value size */
1204 unsigned int shift_bytes; /* No. of bytes to shift EAs by? */
1205 unsigned int min_total_size = ~0U;
1206
1207 is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS);
1208 bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS);
1209 if (!is || !bs) {
1210 error = -ENOMEM;
1211 goto cleanup;
1212 }
1213
1214 is->s.not_found = -ENODATA;
1215 bs->s.not_found = -ENODATA;
1216 is->iloc.bh = NULL;
1217 bs->bh = NULL;
1218
1219 last = IFIRST(header);
1220 /* Find the entry best suited to be pushed into EA block */
1221 entry = NULL;
1222 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1223 total_size =
1224 EXT4_XATTR_SIZE(le32_to_cpu(last->e_value_size)) +
1225 EXT4_XATTR_LEN(last->e_name_len);
1226 if (total_size <= free && total_size < min_total_size) {
1227 if (total_size < new_extra_isize) {
1228 small_entry = last;
1229 } else {
1230 entry = last;
1231 min_total_size = total_size;
1232 }
1233 }
1234 }
1235
1236 if (entry == NULL) {
1237 if (small_entry) {
1238 entry = small_entry;
1239 } else {
1240 if (!tried_min_extra_isize &&
1241 s_min_extra_isize) {
1242 tried_min_extra_isize++;
1243 new_extra_isize = s_min_extra_isize;
1244 goto retry;
1245 }
1246 error = -1;
1247 goto cleanup;
1248 }
1249 }
1250 offs = le16_to_cpu(entry->e_value_offs);
1251 size = le32_to_cpu(entry->e_value_size);
1252 entry_size = EXT4_XATTR_LEN(entry->e_name_len);
1253 i.name_index = entry->e_name_index,
1254 buffer = kmalloc(EXT4_XATTR_SIZE(size), GFP_NOFS);
1255 b_entry_name = kmalloc(entry->e_name_len + 1, GFP_NOFS);
1256 if (!buffer || !b_entry_name) {
1257 error = -ENOMEM;
1258 goto cleanup;
1259 }
1260 /* Save the entry name and the entry value */
1261 memcpy(buffer, (void *)IFIRST(header) + offs,
1262 EXT4_XATTR_SIZE(size));
1263 memcpy(b_entry_name, entry->e_name, entry->e_name_len);
1264 b_entry_name[entry->e_name_len] = '\0';
1265 i.name = b_entry_name;
1266
1267 error = ext4_get_inode_loc(inode, &is->iloc);
1268 if (error)
1269 goto cleanup;
1270
1271 error = ext4_xattr_ibody_find(inode, &i, is);
1272 if (error)
1273 goto cleanup;
1274
1275 /* Remove the chosen entry from the inode */
1276 error = ext4_xattr_ibody_set(handle, inode, &i, is);
1277
1278 entry = IFIRST(header);
1279 if (entry_size + EXT4_XATTR_SIZE(size) >= new_extra_isize)
1280 shift_bytes = new_extra_isize;
1281 else
1282 shift_bytes = entry_size + size;
1283 /* Adjust the offsets and shift the remaining entries ahead */
1284 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize -
1285 shift_bytes, (void *)raw_inode +
1286 EXT4_GOOD_OLD_INODE_SIZE + extra_isize + shift_bytes,
1287 (void *)header, total_ino - entry_size,
1288 inode->i_sb->s_blocksize);
1289
1290 extra_isize += shift_bytes;
1291 new_extra_isize -= shift_bytes;
1292 EXT4_I(inode)->i_extra_isize = extra_isize;
1293
1294 i.name = b_entry_name;
1295 i.value = buffer;
1296 i.value_len = cpu_to_le32(size);
1297 error = ext4_xattr_block_find(inode, &i, bs);
1298 if (error)
1299 goto cleanup;
1300
1301 /* Add entry which was removed from the inode into the block */
1302 error = ext4_xattr_block_set(handle, inode, &i, bs);
1303 if (error)
1304 goto cleanup;
1305 kfree(b_entry_name);
1306 kfree(buffer);
1307 brelse(is->iloc.bh);
1308 kfree(is);
1309 kfree(bs);
1310 }
1311 brelse(bh);
1312 up_write(&EXT4_I(inode)->xattr_sem);
1313 return 0;
1314
1315cleanup:
1316 kfree(b_entry_name);
1317 kfree(buffer);
1318 if (is)
1319 brelse(is->iloc.bh);
1320 kfree(is);
1321 kfree(bs);
1322 brelse(bh);
1323 up_write(&EXT4_I(inode)->xattr_sem);
1324 return error;
1325}
1326
1327
1328
1329/*
1070 * ext4_xattr_delete_inode() 1330 * ext4_xattr_delete_inode()
1071 * 1331 *
1072 * Free extended attribute resources associated with this inode. This 1332 * Free extended attribute resources associated with this inode. This
diff --git a/fs/ext4/xattr.h b/fs/ext4/xattr.h
index 79432b35398f..d7f5d6a12651 100644
--- a/fs/ext4/xattr.h
+++ b/fs/ext4/xattr.h
@@ -56,6 +56,13 @@ struct ext4_xattr_entry {
56#define EXT4_XATTR_SIZE(size) \ 56#define EXT4_XATTR_SIZE(size) \
57 (((size) + EXT4_XATTR_ROUND) & ~EXT4_XATTR_ROUND) 57 (((size) + EXT4_XATTR_ROUND) & ~EXT4_XATTR_ROUND)
58 58
59#define IHDR(inode, raw_inode) \
60 ((struct ext4_xattr_ibody_header *) \
61 ((void *)raw_inode + \
62 EXT4_GOOD_OLD_INODE_SIZE + \
63 EXT4_I(inode)->i_extra_isize))
64#define IFIRST(hdr) ((struct ext4_xattr_entry *)((hdr)+1))
65
59# ifdef CONFIG_EXT4DEV_FS_XATTR 66# ifdef CONFIG_EXT4DEV_FS_XATTR
60 67
61extern struct xattr_handler ext4_xattr_user_handler; 68extern struct xattr_handler ext4_xattr_user_handler;
@@ -74,6 +81,9 @@ extern int ext4_xattr_set_handle(handle_t *, struct inode *, int, const char *,
74extern void ext4_xattr_delete_inode(handle_t *, struct inode *); 81extern void ext4_xattr_delete_inode(handle_t *, struct inode *);
75extern void ext4_xattr_put_super(struct super_block *); 82extern void ext4_xattr_put_super(struct super_block *);
76 83
84extern int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
85 struct ext4_inode *raw_inode, handle_t *handle);
86
77extern int init_ext4_xattr(void); 87extern int init_ext4_xattr(void);
78extern void exit_ext4_xattr(void); 88extern void exit_ext4_xattr(void);
79 89
@@ -129,6 +139,13 @@ exit_ext4_xattr(void)
129{ 139{
130} 140}
131 141
142static inline int
143ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
144 struct ext4_inode *raw_inode, handle_t *handle)
145{
146 return -EOPNOTSUPP;
147}
148
132#define ext4_xattr_handlers NULL 149#define ext4_xattr_handlers NULL
133 150
134# endif /* CONFIG_EXT4DEV_FS_XATTR */ 151# endif /* CONFIG_EXT4DEV_FS_XATTR */
diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
index 78d63b818f0b..f290cb7cb834 100644
--- a/fs/jbd2/journal.c
+++ b/fs/jbd2/journal.c
@@ -35,6 +35,7 @@
35#include <linux/kthread.h> 35#include <linux/kthread.h>
36#include <linux/poison.h> 36#include <linux/poison.h>
37#include <linux/proc_fs.h> 37#include <linux/proc_fs.h>
38#include <linux/debugfs.h>
38 39
39#include <asm/uaccess.h> 40#include <asm/uaccess.h>
40#include <asm/page.h> 41#include <asm/page.h>
@@ -528,7 +529,7 @@ int jbd2_log_wait_commit(journal_t *journal, tid_t tid)
528{ 529{
529 int err = 0; 530 int err = 0;
530 531
531#ifdef CONFIG_JBD_DEBUG 532#ifdef CONFIG_JBD2_DEBUG
532 spin_lock(&journal->j_state_lock); 533 spin_lock(&journal->j_state_lock);
533 if (!tid_geq(journal->j_commit_request, tid)) { 534 if (!tid_geq(journal->j_commit_request, tid)) {
534 printk(KERN_EMERG 535 printk(KERN_EMERG
@@ -1709,7 +1710,7 @@ void jbd2_slab_free(void *ptr, size_t size)
1709 * Journal_head storage management 1710 * Journal_head storage management
1710 */ 1711 */
1711static struct kmem_cache *jbd2_journal_head_cache; 1712static struct kmem_cache *jbd2_journal_head_cache;
1712#ifdef CONFIG_JBD_DEBUG 1713#ifdef CONFIG_JBD2_DEBUG
1713static atomic_t nr_journal_heads = ATOMIC_INIT(0); 1714static atomic_t nr_journal_heads = ATOMIC_INIT(0);
1714#endif 1715#endif
1715 1716
@@ -1747,7 +1748,7 @@ static struct journal_head *journal_alloc_journal_head(void)
1747 struct journal_head *ret; 1748 struct journal_head *ret;
1748 static unsigned long last_warning; 1749 static unsigned long last_warning;
1749 1750
1750#ifdef CONFIG_JBD_DEBUG 1751#ifdef CONFIG_JBD2_DEBUG
1751 atomic_inc(&nr_journal_heads); 1752 atomic_inc(&nr_journal_heads);
1752#endif 1753#endif
1753 ret = kmem_cache_alloc(jbd2_journal_head_cache, GFP_NOFS); 1754 ret = kmem_cache_alloc(jbd2_journal_head_cache, GFP_NOFS);
@@ -1768,7 +1769,7 @@ static struct journal_head *journal_alloc_journal_head(void)
1768 1769
1769static void journal_free_journal_head(struct journal_head *jh) 1770static void journal_free_journal_head(struct journal_head *jh)
1770{ 1771{
1771#ifdef CONFIG_JBD_DEBUG 1772#ifdef CONFIG_JBD2_DEBUG
1772 atomic_dec(&nr_journal_heads); 1773 atomic_dec(&nr_journal_heads);
1773 memset(jh, JBD_POISON_FREE, sizeof(*jh)); 1774 memset(jh, JBD_POISON_FREE, sizeof(*jh));
1774#endif 1775#endif
@@ -1951,64 +1952,50 @@ void jbd2_journal_put_journal_head(struct journal_head *jh)
1951} 1952}
1952 1953
1953/* 1954/*
1954 * /proc tunables 1955 * debugfs tunables
1955 */ 1956 */
1956#if defined(CONFIG_JBD_DEBUG) 1957#if defined(CONFIG_JBD2_DEBUG)
1957int jbd2_journal_enable_debug; 1958u8 jbd2_journal_enable_debug;
1958EXPORT_SYMBOL(jbd2_journal_enable_debug); 1959EXPORT_SYMBOL(jbd2_journal_enable_debug);
1959#endif 1960#endif
1960 1961
1961#if defined(CONFIG_JBD_DEBUG) && defined(CONFIG_PROC_FS) 1962#if defined(CONFIG_JBD2_DEBUG) && defined(CONFIG_DEBUG_FS)
1962 1963
1963static struct proc_dir_entry *proc_jbd_debug; 1964#define JBD2_DEBUG_NAME "jbd2-debug"
1964 1965
1965static int read_jbd_debug(char *page, char **start, off_t off, 1966struct dentry *jbd2_debugfs_dir, *jbd2_debug;
1966 int count, int *eof, void *data)
1967{
1968 int ret;
1969 1967
1970 ret = sprintf(page + off, "%d\n", jbd2_journal_enable_debug); 1968static void __init jbd2_create_debugfs_entry(void)
1971 *eof = 1; 1969{
1972 return ret; 1970 jbd2_debugfs_dir = debugfs_create_dir("jbd2", NULL);
1971 if (jbd2_debugfs_dir)
1972 jbd2_debug = debugfs_create_u8(JBD2_DEBUG_NAME, S_IRUGO,
1973 jbd2_debugfs_dir,
1974 &jbd2_journal_enable_debug);
1973} 1975}
1974 1976
1975static int write_jbd_debug(struct file *file, const char __user *buffer, 1977static void __exit jbd2_remove_debugfs_entry(void)
1976 unsigned long count, void *data)
1977{ 1978{
1978 char buf[32]; 1979 if (jbd2_debug)
1979 1980 debugfs_remove(jbd2_debug);
1980 if (count > ARRAY_SIZE(buf) - 1) 1981 if (jbd2_debugfs_dir)
1981 count = ARRAY_SIZE(buf) - 1; 1982 debugfs_remove(jbd2_debugfs_dir);
1982 if (copy_from_user(buf, buffer, count))
1983 return -EFAULT;
1984 buf[ARRAY_SIZE(buf) - 1] = '\0';
1985 jbd2_journal_enable_debug = simple_strtoul(buf, NULL, 10);
1986 return count;
1987} 1983}
1988 1984
1989#define JBD_PROC_NAME "sys/fs/jbd2-debug" 1985#else
1990 1986
1991static void __init create_jbd_proc_entry(void) 1987static void __init jbd2_create_debugfs_entry(void)
1992{ 1988{
1993 proc_jbd_debug = create_proc_entry(JBD_PROC_NAME, 0644, NULL); 1989 do {
1994 if (proc_jbd_debug) { 1990 } while (0);
1995 /* Why is this so hard? */
1996 proc_jbd_debug->read_proc = read_jbd_debug;
1997 proc_jbd_debug->write_proc = write_jbd_debug;
1998 }
1999} 1991}
2000 1992
2001static void __exit jbd2_remove_jbd_proc_entry(void) 1993static void __exit jbd2_remove_debugfs_entry(void)
2002{ 1994{
2003 if (proc_jbd_debug) 1995 do {
2004 remove_proc_entry(JBD_PROC_NAME, NULL); 1996 } while (0);
2005} 1997}
2006 1998
2007#else
2008
2009#define create_jbd_proc_entry() do {} while (0)
2010#define jbd2_remove_jbd_proc_entry() do {} while (0)
2011
2012#endif 1999#endif
2013 2000
2014struct kmem_cache *jbd2_handle_cache; 2001struct kmem_cache *jbd2_handle_cache;
@@ -2067,18 +2054,18 @@ static int __init journal_init(void)
2067 ret = journal_init_caches(); 2054 ret = journal_init_caches();
2068 if (ret != 0) 2055 if (ret != 0)
2069 jbd2_journal_destroy_caches(); 2056 jbd2_journal_destroy_caches();
2070 create_jbd_proc_entry(); 2057 jbd2_create_debugfs_entry();
2071 return ret; 2058 return ret;
2072} 2059}
2073 2060
2074static void __exit journal_exit(void) 2061static void __exit journal_exit(void)
2075{ 2062{
2076#ifdef CONFIG_JBD_DEBUG 2063#ifdef CONFIG_JBD2_DEBUG
2077 int n = atomic_read(&nr_journal_heads); 2064 int n = atomic_read(&nr_journal_heads);
2078 if (n) 2065 if (n)
2079 printk(KERN_EMERG "JBD: leaked %d journal_heads!\n", n); 2066 printk(KERN_EMERG "JBD: leaked %d journal_heads!\n", n);
2080#endif 2067#endif
2081 jbd2_remove_jbd_proc_entry(); 2068 jbd2_remove_debugfs_entry();
2082 jbd2_journal_destroy_caches(); 2069 jbd2_journal_destroy_caches();
2083} 2070}
2084 2071
diff --git a/fs/jbd2/recovery.c b/fs/jbd2/recovery.c
index 395c92a04ac9..e7730a045b93 100644
--- a/fs/jbd2/recovery.c
+++ b/fs/jbd2/recovery.c
@@ -295,7 +295,7 @@ int jbd2_journal_skip_recovery(journal_t *journal)
295 printk(KERN_ERR "JBD: error %d scanning journal\n", err); 295 printk(KERN_ERR "JBD: error %d scanning journal\n", err);
296 ++journal->j_transaction_sequence; 296 ++journal->j_transaction_sequence;
297 } else { 297 } else {
298#ifdef CONFIG_JBD_DEBUG 298#ifdef CONFIG_JBD2_DEBUG
299 int dropped = info.end_transaction - be32_to_cpu(sb->s_sequence); 299 int dropped = info.end_transaction - be32_to_cpu(sb->s_sequence);
300#endif 300#endif
301 jbd_debug(0, 301 jbd_debug(0,
diff --git a/fs/open.c b/fs/open.c
index be6a457f4226..a6b054edacba 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -26,6 +26,7 @@
26#include <linux/syscalls.h> 26#include <linux/syscalls.h>
27#include <linux/rcupdate.h> 27#include <linux/rcupdate.h>
28#include <linux/audit.h> 28#include <linux/audit.h>
29#include <linux/falloc.h>
29 30
30int vfs_statfs(struct dentry *dentry, struct kstatfs *buf) 31int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
31{ 32{
@@ -352,6 +353,64 @@ asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
352} 353}
353#endif 354#endif
354 355
356asmlinkage long sys_fallocate(int fd, int mode, loff_t offset, loff_t len)
357{
358 struct file *file;
359 struct inode *inode;
360 long ret = -EINVAL;
361
362 if (offset < 0 || len <= 0)
363 goto out;
364
365 /* Return error if mode is not supported */
366 ret = -EOPNOTSUPP;
367 if (mode && !(mode & FALLOC_FL_KEEP_SIZE))
368 goto out;
369
370 ret = -EBADF;
371 file = fget(fd);
372 if (!file)
373 goto out;
374 if (!(file->f_mode & FMODE_WRITE))
375 goto out_fput;
376 /*
377 * Revalidate the write permissions, in case security policy has
378 * changed since the files were opened.
379 */
380 ret = security_file_permission(file, MAY_WRITE);
381 if (ret)
382 goto out_fput;
383
384 inode = file->f_path.dentry->d_inode;
385
386 ret = -ESPIPE;
387 if (S_ISFIFO(inode->i_mode))
388 goto out_fput;
389
390 ret = -ENODEV;
391 /*
392 * Let individual file system decide if it supports preallocation
393 * for directories or not.
394 */
395 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
396 goto out_fput;
397
398 ret = -EFBIG;
399 /* Check for wrap through zero too */
400 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
401 goto out_fput;
402
403 if (inode->i_op && inode->i_op->fallocate)
404 ret = inode->i_op->fallocate(inode, mode, offset, len);
405 else
406 ret = -ENOSYS;
407
408out_fput:
409 fput(file);
410out:
411 return ret;
412}
413
355/* 414/*
356 * access() needs to use the real uid/gid, not the effective uid/gid. 415 * access() needs to use the real uid/gid, not the effective uid/gid.
357 * We do this by temporarily clearing all FS-related capabilities and 416 * We do this by temporarily clearing all FS-related capabilities and
generated by cgit v1.2.2 (git 2.25.0) at 2024-10-06 22:14:58 -0400