aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2013-11-13 01:24:40 -0500
committerLinus Torvalds <torvalds@linux-foundation.org>2013-11-13 01:24:40 -0500
commitdd1d1399f2884102172f761816c32aa311bceafb (patch)
treef3c350bd0a1c43bc61d6aefde32d5d5a8aa22011
parenta9986464564609dd0962e6023513f7d3d313dc80 (diff)
parent29e59c14ae5c21d25db1580d9651b5855d656a30 (diff)
Merge tag 'for-f2fs-3.13' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs
Pull f2fs updates from Jaegeuk Kim: "This patch-set includes the following major enhancement patches. - add a sysfs to control reclaiming free segments - enhance the f2fs global lock procedures - enhance the victim selection flow - wait for selected node blocks during fsync - add some tracepoints - add a config to remove abundant BUG_ONs The other bug fixes are as follows. - fix deadlock on acl operations - fix some bugs with respect to orphan inodes And, there are a bunch of cleanups" * tag 'for-f2fs-3.13' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs: (42 commits) f2fs: issue more large discard command f2fs: fix memory leak after kobject init failed in fill_super f2fs: cleanup waiting routine for writeback pages in cp f2fs: avoid to use a NULL point in destroy_segment_manager f2fs: remove unnecessary TestClearPageError when wait pages writeback f2fs: update f2fs document f2fs: avoid to wait all the node blocks during fsync f2fs: check all ones or zeros bitmap with bitops for better mount performance f2fs: change the method of calculating the number summary blocks f2fs: fix calculating incorrect free size when update xattr in __f2fs_setxattr f2fs: add an option to avoid unnecessary BUG_ONs f2fs: introduce CONFIG_F2FS_CHECK_FS for BUG_ON control f2fs: fix a deadlock during init_acl procedure f2fs: clean up acl flow for better readability f2fs: remove unnecessary segment bitmap updates f2fs: add tracepoint for vm_page_mkwrite f2fs: add tracepoint for set_page_dirty f2fs: remove redundant set_page_dirty from write_compacted_summaries f2fs: add reclaiming control by sysfs f2fs: introduce f2fs_balance_fs_bg for some background jobs ...
Diffstat
-rw-r--r--Documentation/filesystems/f2fs.txt7
-rw-r--r--fs/f2fs/Kconfig8
-rw-r--r--fs/f2fs/acl.c36
-rw-r--r--fs/f2fs/acl.h9
-rw-r--r--fs/f2fs/checkpoint.c75
-rw-r--r--fs/f2fs/data.c29
-rw-r--r--fs/f2fs/dir.c4
-rw-r--r--fs/f2fs/f2fs.h117
-rw-r--r--fs/f2fs/file.c45
-rw-r--r--fs/f2fs/gc.c31
-rw-r--r--fs/f2fs/inode.c62
-rw-r--r--fs/f2fs/namei.c52
-rw-r--r--fs/f2fs/node.c142
-rw-r--r--fs/f2fs/recovery.c45
-rw-r--r--fs/f2fs/segment.c133
-rw-r--r--fs/f2fs/segment.h38
-rw-r--r--fs/f2fs/super.c143
-rw-r--r--fs/f2fs/xattr.c36
-rw-r--r--include/trace/events/f2fs.h51
19 files changed, 651 insertions, 412 deletions
diff --git a/Documentation/filesystems/f2fs.txt b/Documentation/filesystems/f2fs.txt
index 3cd27bed6349..a3fe811bbdbc 100644
--- a/Documentation/filesystems/f2fs.txt
+++ b/Documentation/filesystems/f2fs.txt
@@ -119,6 +119,7 @@ active_logs=%u Support configuring the number of active logs. In the
119 Default number is 6. 119 Default number is 6.
120disable_ext_identify Disable the extension list configured by mkfs, so f2fs 120disable_ext_identify Disable the extension list configured by mkfs, so f2fs
121 does not aware of cold files such as media files. 121 does not aware of cold files such as media files.
122inline_xattr Enable the inline xattrs feature.
122 123
123================================================================================ 124================================================================================
124DEBUGFS ENTRIES 125DEBUGFS ENTRIES
@@ -164,6 +165,12 @@ Files in /sys/fs/f2fs/<devname>
164 gc_idle = 1 will select the Cost Benefit approach 165 gc_idle = 1 will select the Cost Benefit approach
165 & setting gc_idle = 2 will select the greedy aproach. 166 & setting gc_idle = 2 will select the greedy aproach.
166 167
168 reclaim_segments This parameter controls the number of prefree
169 segments to be reclaimed. If the number of prefree
170 segments is larger than this number, f2fs tries to
171 conduct checkpoint to reclaim the prefree segments
172 to free segments. By default, 100 segments, 200MB.
173
167================================================================================ 174================================================================================
168USAGE 175USAGE
169================================================================================ 176================================================================================
diff --git a/fs/f2fs/Kconfig b/fs/f2fs/Kconfig
index e06e0995e00f..214fe1054fce 100644
--- a/fs/f2fs/Kconfig
+++ b/fs/f2fs/Kconfig
@@ -63,3 +63,11 @@ config F2FS_FS_SECURITY
63 the extended attribute support in advance. 63 the extended attribute support in advance.
64 64
65 If you are not using a security module, say N. 65 If you are not using a security module, say N.
66
67config F2FS_CHECK_FS
68 bool "F2FS consistency checking feature"
69 depends on F2FS_FS
70 help
71 Enables BUG_ONs which check the file system consistency in runtime.
72
73 If you want to improve the performance, say N.
diff --git a/fs/f2fs/acl.c b/fs/f2fs/acl.c
index b7826ec1b470..d0fc287efeff 100644
--- a/fs/f2fs/acl.c
+++ b/fs/f2fs/acl.c
@@ -205,7 +205,8 @@ struct posix_acl *f2fs_get_acl(struct inode *inode, int type)
205 return acl; 205 return acl;
206} 206}
207 207
208static int f2fs_set_acl(struct inode *inode, int type, struct posix_acl *acl) 208static int f2fs_set_acl(struct inode *inode, int type,
209 struct posix_acl *acl, struct page *ipage)
209{ 210{
210 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 211 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
211 struct f2fs_inode_info *fi = F2FS_I(inode); 212 struct f2fs_inode_info *fi = F2FS_I(inode);
@@ -250,7 +251,7 @@ static int f2fs_set_acl(struct inode *inode, int type, struct posix_acl *acl)
250 } 251 }
251 } 252 }
252 253
253 error = f2fs_setxattr(inode, name_index, "", value, size, NULL); 254 error = f2fs_setxattr(inode, name_index, "", value, size, ipage);
254 255
255 kfree(value); 256 kfree(value);
256 if (!error) 257 if (!error)
@@ -260,10 +261,10 @@ static int f2fs_set_acl(struct inode *inode, int type, struct posix_acl *acl)
260 return error; 261 return error;
261} 262}
262 263
263int f2fs_init_acl(struct inode *inode, struct inode *dir) 264int f2fs_init_acl(struct inode *inode, struct inode *dir, struct page *ipage)
264{ 265{
265 struct posix_acl *acl = NULL;
266 struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb); 266 struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
267 struct posix_acl *acl = NULL;
267 int error = 0; 268 int error = 0;
268 269
269 if (!S_ISLNK(inode->i_mode)) { 270 if (!S_ISLNK(inode->i_mode)) {
@@ -276,19 +277,19 @@ int f2fs_init_acl(struct inode *inode, struct inode *dir)
276 inode->i_mode &= ~current_umask(); 277 inode->i_mode &= ~current_umask();
277 } 278 }
278 279
279 if (test_opt(sbi, POSIX_ACL) && acl) { 280 if (!test_opt(sbi, POSIX_ACL) || !acl)
281 goto cleanup;
280 282
281 if (S_ISDIR(inode->i_mode)) { 283 if (S_ISDIR(inode->i_mode)) {
282 error = f2fs_set_acl(inode, ACL_TYPE_DEFAULT, acl); 284 error = f2fs_set_acl(inode, ACL_TYPE_DEFAULT, acl, ipage);
283 if (error) 285 if (error)
284 goto cleanup; 286 goto cleanup;
285 }
286 error = posix_acl_create(&acl, GFP_KERNEL, &inode->i_mode);
287 if (error < 0)
288 return error;
289 if (error > 0)
290 error = f2fs_set_acl(inode, ACL_TYPE_ACCESS, acl);
291 } 287 }
288 error = posix_acl_create(&acl, GFP_KERNEL, &inode->i_mode);
289 if (error < 0)
290 return error;
291 if (error > 0)
292 error = f2fs_set_acl(inode, ACL_TYPE_ACCESS, acl, ipage);
292cleanup: 293cleanup:
293 posix_acl_release(acl); 294 posix_acl_release(acl);
294 return error; 295 return error;
@@ -313,7 +314,8 @@ int f2fs_acl_chmod(struct inode *inode)
313 error = posix_acl_chmod(&acl, GFP_KERNEL, mode); 314 error = posix_acl_chmod(&acl, GFP_KERNEL, mode);
314 if (error) 315 if (error)
315 return error; 316 return error;
316 error = f2fs_set_acl(inode, ACL_TYPE_ACCESS, acl); 317
318 error = f2fs_set_acl(inode, ACL_TYPE_ACCESS, acl, NULL);
317 posix_acl_release(acl); 319 posix_acl_release(acl);
318 return error; 320 return error;
319} 321}
@@ -388,7 +390,7 @@ static int f2fs_xattr_set_acl(struct dentry *dentry, const char *name,
388 acl = NULL; 390 acl = NULL;
389 } 391 }
390 392
391 error = f2fs_set_acl(inode, type, acl); 393 error = f2fs_set_acl(inode, type, acl, NULL);
392 394
393release_and_out: 395release_and_out:
394 posix_acl_release(acl); 396 posix_acl_release(acl);
diff --git a/fs/f2fs/acl.h b/fs/f2fs/acl.h
index 80f430674417..49633131e038 100644
--- a/fs/f2fs/acl.h
+++ b/fs/f2fs/acl.h
@@ -36,9 +36,9 @@ struct f2fs_acl_header {
36 36
37#ifdef CONFIG_F2FS_FS_POSIX_ACL 37#ifdef CONFIG_F2FS_FS_POSIX_ACL
38 38
39extern struct posix_acl *f2fs_get_acl(struct inode *inode, int type); 39extern struct posix_acl *f2fs_get_acl(struct inode *, int);
40extern int f2fs_acl_chmod(struct inode *inode); 40extern int f2fs_acl_chmod(struct inode *);
41extern int f2fs_init_acl(struct inode *inode, struct inode *dir); 41extern int f2fs_init_acl(struct inode *, struct inode *, struct page *);
42#else 42#else
43#define f2fs_check_acl NULL 43#define f2fs_check_acl NULL
44#define f2fs_get_acl NULL 44#define f2fs_get_acl NULL
@@ -49,7 +49,8 @@ static inline int f2fs_acl_chmod(struct inode *inode)
49 return 0; 49 return 0;
50} 50}
51 51
52static inline int f2fs_init_acl(struct inode *inode, struct inode *dir) 52static inline int f2fs_init_acl(struct inode *inode, struct inode *dir,
53 struct page *page)
53{ 54{
54 return 0; 55 return 0;
55} 56}
diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
index bb312201ca95..5716e5eb4e8e 100644
--- a/fs/f2fs/checkpoint.c
+++ b/fs/f2fs/checkpoint.c
@@ -81,7 +81,7 @@ static int f2fs_write_meta_page(struct page *page,
81 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 81 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
82 82
83 /* Should not write any meta pages, if any IO error was occurred */ 83 /* Should not write any meta pages, if any IO error was occurred */
84 if (wbc->for_reclaim || 84 if (wbc->for_reclaim || sbi->por_doing ||
85 is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ERROR_FLAG)) { 85 is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ERROR_FLAG)) {
86 dec_page_count(sbi, F2FS_DIRTY_META); 86 dec_page_count(sbi, F2FS_DIRTY_META);
87 wbc->pages_skipped++; 87 wbc->pages_skipped++;
@@ -142,8 +142,8 @@ long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type,
142 for (i = 0; i < nr_pages; i++) { 142 for (i = 0; i < nr_pages; i++) {
143 struct page *page = pvec.pages[i]; 143 struct page *page = pvec.pages[i];
144 lock_page(page); 144 lock_page(page);
145 BUG_ON(page->mapping != mapping); 145 f2fs_bug_on(page->mapping != mapping);
146 BUG_ON(!PageDirty(page)); 146 f2fs_bug_on(!PageDirty(page));
147 clear_page_dirty_for_io(page); 147 clear_page_dirty_for_io(page);
148 if (f2fs_write_meta_page(page, &wbc)) { 148 if (f2fs_write_meta_page(page, &wbc)) {
149 unlock_page(page); 149 unlock_page(page);
@@ -167,6 +167,8 @@ static int f2fs_set_meta_page_dirty(struct page *page)
167 struct address_space *mapping = page->mapping; 167 struct address_space *mapping = page->mapping;
168 struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb); 168 struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb);
169 169
170 trace_f2fs_set_page_dirty(page, META);
171
170 SetPageUptodate(page); 172 SetPageUptodate(page);
171 if (!PageDirty(page)) { 173 if (!PageDirty(page)) {
172 __set_page_dirty_nobuffers(page); 174 __set_page_dirty_nobuffers(page);
@@ -206,6 +208,7 @@ int acquire_orphan_inode(struct f2fs_sb_info *sbi)
206void release_orphan_inode(struct f2fs_sb_info *sbi) 208void release_orphan_inode(struct f2fs_sb_info *sbi)
207{ 209{
208 mutex_lock(&sbi->orphan_inode_mutex); 210 mutex_lock(&sbi->orphan_inode_mutex);
211 f2fs_bug_on(sbi->n_orphans == 0);
209 sbi->n_orphans--; 212 sbi->n_orphans--;
210 mutex_unlock(&sbi->orphan_inode_mutex); 213 mutex_unlock(&sbi->orphan_inode_mutex);
211} 214}
@@ -225,12 +228,8 @@ void add_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino)
225 break; 228 break;
226 orphan = NULL; 229 orphan = NULL;
227 } 230 }
228retry: 231
229 new = kmem_cache_alloc(orphan_entry_slab, GFP_ATOMIC); 232 new = f2fs_kmem_cache_alloc(orphan_entry_slab, GFP_ATOMIC);
230 if (!new) {
231 cond_resched();
232 goto retry;
233 }
234 new->ino = ino; 233 new->ino = ino;
235 234
236 /* add new_oentry into list which is sorted by inode number */ 235 /* add new_oentry into list which is sorted by inode number */
@@ -253,6 +252,7 @@ void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino)
253 if (orphan->ino == ino) { 252 if (orphan->ino == ino) {
254 list_del(&orphan->list); 253 list_del(&orphan->list);
255 kmem_cache_free(orphan_entry_slab, orphan); 254 kmem_cache_free(orphan_entry_slab, orphan);
255 f2fs_bug_on(sbi->n_orphans == 0);
256 sbi->n_orphans--; 256 sbi->n_orphans--;
257 break; 257 break;
258 } 258 }
@@ -263,7 +263,7 @@ void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino)
263static void recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino) 263static void recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino)
264{ 264{
265 struct inode *inode = f2fs_iget(sbi->sb, ino); 265 struct inode *inode = f2fs_iget(sbi->sb, ino);
266 BUG_ON(IS_ERR(inode)); 266 f2fs_bug_on(IS_ERR(inode));
267 clear_nlink(inode); 267 clear_nlink(inode);
268 268
269 /* truncate all the data during iput */ 269 /* truncate all the data during iput */
@@ -277,7 +277,7 @@ int recover_orphan_inodes(struct f2fs_sb_info *sbi)
277 if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG)) 277 if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG))
278 return 0; 278 return 0;
279 279
280 sbi->por_doing = 1; 280 sbi->por_doing = true;
281 start_blk = __start_cp_addr(sbi) + 1; 281 start_blk = __start_cp_addr(sbi) + 1;
282 orphan_blkaddr = __start_sum_addr(sbi) - 1; 282 orphan_blkaddr = __start_sum_addr(sbi) - 1;
283 283
@@ -294,7 +294,7 @@ int recover_orphan_inodes(struct f2fs_sb_info *sbi)
294 } 294 }
295 /* clear Orphan Flag */ 295 /* clear Orphan Flag */
296 clear_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG); 296 clear_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG);
297 sbi->por_doing = 0; 297 sbi->por_doing = false;
298 return 0; 298 return 0;
299} 299}
300 300
@@ -469,9 +469,7 @@ static int __add_dirty_inode(struct inode *inode, struct dir_inode_entry *new)
469 return -EEXIST; 469 return -EEXIST;
470 } 470 }
471 list_add_tail(&new->list, head); 471 list_add_tail(&new->list, head);
472#ifdef CONFIG_F2FS_STAT_FS 472 stat_inc_dirty_dir(sbi);
473 sbi->n_dirty_dirs++;
474#endif
475 return 0; 473 return 0;
476} 474}
477 475
@@ -482,12 +480,8 @@ void set_dirty_dir_page(struct inode *inode, struct page *page)
482 480
483 if (!S_ISDIR(inode->i_mode)) 481 if (!S_ISDIR(inode->i_mode))
484 return; 482 return;
485retry: 483
486 new = kmem_cache_alloc(inode_entry_slab, GFP_NOFS); 484 new = f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS);
487 if (!new) {
488 cond_resched();
489 goto retry;
490 }
491 new->inode = inode; 485 new->inode = inode;
492 INIT_LIST_HEAD(&new->list); 486 INIT_LIST_HEAD(&new->list);
493 487
@@ -504,13 +498,9 @@ retry:
504void add_dirty_dir_inode(struct inode *inode) 498void add_dirty_dir_inode(struct inode *inode)
505{ 499{
506 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 500 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
507 struct dir_inode_entry *new; 501 struct dir_inode_entry *new =
508retry: 502 f2fs_kmem_cache_alloc(inode_entry_slab, GFP_NOFS);
509 new = kmem_cache_alloc(inode_entry_slab, GFP_NOFS); 503
510 if (!new) {
511 cond_resched();
512 goto retry;
513 }
514 new->inode = inode; 504 new->inode = inode;
515 INIT_LIST_HEAD(&new->list); 505 INIT_LIST_HEAD(&new->list);
516 506
@@ -541,9 +531,7 @@ void remove_dirty_dir_inode(struct inode *inode)
541 if (entry->inode == inode) { 531 if (entry->inode == inode) {
542 list_del(&entry->list); 532 list_del(&entry->list);
543 kmem_cache_free(inode_entry_slab, entry); 533 kmem_cache_free(inode_entry_slab, entry);
544#ifdef CONFIG_F2FS_STAT_FS 534 stat_dec_dirty_dir(sbi);
545 sbi->n_dirty_dirs--;
546#endif
547 break; 535 break;
548 } 536 }
549 } 537 }
@@ -617,11 +605,10 @@ static void block_operations(struct f2fs_sb_info *sbi)
617 blk_start_plug(&plug); 605 blk_start_plug(&plug);
618 606
619retry_flush_dents: 607retry_flush_dents:
620 mutex_lock_all(sbi); 608 f2fs_lock_all(sbi);
621
622 /* write all the dirty dentry pages */ 609 /* write all the dirty dentry pages */
623 if (get_pages(sbi, F2FS_DIRTY_DENTS)) { 610 if (get_pages(sbi, F2FS_DIRTY_DENTS)) {
624 mutex_unlock_all(sbi); 611 f2fs_unlock_all(sbi);
625 sync_dirty_dir_inodes(sbi); 612 sync_dirty_dir_inodes(sbi);
626 goto retry_flush_dents; 613 goto retry_flush_dents;
627 } 614 }
@@ -644,7 +631,22 @@ retry_flush_nodes:
644static void unblock_operations(struct f2fs_sb_info *sbi) 631static void unblock_operations(struct f2fs_sb_info *sbi)
645{ 632{
646 mutex_unlock(&sbi->node_write); 633 mutex_unlock(&sbi->node_write);
647 mutex_unlock_all(sbi); 634 f2fs_unlock_all(sbi);
635}
636
637static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi)
638{
639 DEFINE_WAIT(wait);
640
641 for (;;) {
642 prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE);
643
644 if (!get_pages(sbi, F2FS_WRITEBACK))
645 break;
646
647 io_schedule();
648 }
649 finish_wait(&sbi->cp_wait, &wait);
648} 650}
649 651
650static void do_checkpoint(struct f2fs_sb_info *sbi, bool is_umount) 652static void do_checkpoint(struct f2fs_sb_info *sbi, bool is_umount)
@@ -756,8 +758,7 @@ static void do_checkpoint(struct f2fs_sb_info *sbi, bool is_umount)
756 f2fs_put_page(cp_page, 1); 758 f2fs_put_page(cp_page, 1);
757 759
758 /* wait for previous submitted node/meta pages writeback */ 760 /* wait for previous submitted node/meta pages writeback */
759 while (get_pages(sbi, F2FS_WRITEBACK)) 761 wait_on_all_pages_writeback(sbi);
760 congestion_wait(BLK_RW_ASYNC, HZ / 50);
761 762
762 filemap_fdatawait_range(sbi->node_inode->i_mapping, 0, LONG_MAX); 763 filemap_fdatawait_range(sbi->node_inode->i_mapping, 0, LONG_MAX);
763 filemap_fdatawait_range(sbi->meta_inode->i_mapping, 0, LONG_MAX); 764 filemap_fdatawait_range(sbi->meta_inode->i_mapping, 0, LONG_MAX);
diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 941f9b9ca3a5..aa3438c571fa 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -68,9 +68,6 @@ static int check_extent_cache(struct inode *inode, pgoff_t pgofs,
68 struct buffer_head *bh_result) 68 struct buffer_head *bh_result)
69{ 69{
70 struct f2fs_inode_info *fi = F2FS_I(inode); 70 struct f2fs_inode_info *fi = F2FS_I(inode);
71#ifdef CONFIG_F2FS_STAT_FS
72 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
73#endif
74 pgoff_t start_fofs, end_fofs; 71 pgoff_t start_fofs, end_fofs;
75 block_t start_blkaddr; 72 block_t start_blkaddr;
76 73
@@ -80,9 +77,8 @@ static int check_extent_cache(struct inode *inode, pgoff_t pgofs,
80 return 0; 77 return 0;
81 } 78 }
82 79
83#ifdef CONFIG_F2FS_STAT_FS 80 stat_inc_total_hit(inode->i_sb);
84 sbi->total_hit_ext++; 81
85#endif
86 start_fofs = fi->ext.fofs; 82 start_fofs = fi->ext.fofs;
87 end_fofs = fi->ext.fofs + fi->ext.len - 1; 83 end_fofs = fi->ext.fofs + fi->ext.len - 1;
88 start_blkaddr = fi->ext.blk_addr; 84 start_blkaddr = fi->ext.blk_addr;
@@ -100,9 +96,7 @@ static int check_extent_cache(struct inode *inode, pgoff_t pgofs,
100 else 96 else
101 bh_result->b_size = UINT_MAX; 97 bh_result->b_size = UINT_MAX;
102 98
103#ifdef CONFIG_F2FS_STAT_FS 99 stat_inc_read_hit(inode->i_sb);
104 sbi->read_hit_ext++;
105#endif
106 read_unlock(&fi->ext.ext_lock); 100 read_unlock(&fi->ext.ext_lock);
107 return 1; 101 return 1;
108 } 102 }
@@ -116,7 +110,7 @@ void update_extent_cache(block_t blk_addr, struct dnode_of_data *dn)
116 pgoff_t fofs, start_fofs, end_fofs; 110 pgoff_t fofs, start_fofs, end_fofs;
117 block_t start_blkaddr, end_blkaddr; 111 block_t start_blkaddr, end_blkaddr;
118 112
119 BUG_ON(blk_addr == NEW_ADDR); 113 f2fs_bug_on(blk_addr == NEW_ADDR);
120 fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) + 114 fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) +
121 dn->ofs_in_node; 115 dn->ofs_in_node;
122 116
@@ -442,7 +436,7 @@ static int get_data_block_ro(struct inode *inode, sector_t iblock,
442 } 436 }
443 437
444 /* It does not support data allocation */ 438 /* It does not support data allocation */
445 BUG_ON(create); 439 f2fs_bug_on(create);
446 440
447 if (dn.data_blkaddr != NEW_ADDR && dn.data_blkaddr != NULL_ADDR) { 441 if (dn.data_blkaddr != NEW_ADDR && dn.data_blkaddr != NULL_ADDR) {
448 int i; 442 int i;
@@ -560,9 +554,9 @@ write:
560 inode_dec_dirty_dents(inode); 554 inode_dec_dirty_dents(inode);
561 err = do_write_data_page(page); 555 err = do_write_data_page(page);
562 } else { 556 } else {
563 int ilock = mutex_lock_op(sbi); 557 f2fs_lock_op(sbi);
564 err = do_write_data_page(page); 558 err = do_write_data_page(page);
565 mutex_unlock_op(sbi, ilock); 559 f2fs_unlock_op(sbi);
566 need_balance_fs = true; 560 need_balance_fs = true;
567 } 561 }
568 if (err == -ENOENT) 562 if (err == -ENOENT)
@@ -641,7 +635,6 @@ static int f2fs_write_begin(struct file *file, struct address_space *mapping,
641 pgoff_t index = ((unsigned long long) pos) >> PAGE_CACHE_SHIFT; 635 pgoff_t index = ((unsigned long long) pos) >> PAGE_CACHE_SHIFT;
642 struct dnode_of_data dn; 636 struct dnode_of_data dn;
643 int err = 0; 637 int err = 0;
644 int ilock;
645 638
646 f2fs_balance_fs(sbi); 639 f2fs_balance_fs(sbi);
647repeat: 640repeat:
@@ -650,7 +643,7 @@ repeat:
650 return -ENOMEM; 643 return -ENOMEM;
651 *pagep = page; 644 *pagep = page;
652 645
653 ilock = mutex_lock_op(sbi); 646 f2fs_lock_op(sbi);
654 647
655 set_new_dnode(&dn, inode, NULL, NULL, 0); 648 set_new_dnode(&dn, inode, NULL, NULL, 0);
656 err = get_dnode_of_data(&dn, index, ALLOC_NODE); 649 err = get_dnode_of_data(&dn, index, ALLOC_NODE);
@@ -664,7 +657,7 @@ repeat:
664 if (err) 657 if (err)
665 goto err; 658 goto err;
666 659
667 mutex_unlock_op(sbi, ilock); 660 f2fs_unlock_op(sbi);
668 661
669 if ((len == PAGE_CACHE_SIZE) || PageUptodate(page)) 662 if ((len == PAGE_CACHE_SIZE) || PageUptodate(page))
670 return 0; 663 return 0;
@@ -700,7 +693,7 @@ out:
700 return 0; 693 return 0;
701 694
702err: 695err:
703 mutex_unlock_op(sbi, ilock); 696 f2fs_unlock_op(sbi);
704 f2fs_put_page(page, 1); 697 f2fs_put_page(page, 1);
705 return err; 698 return err;
706} 699}
@@ -763,6 +756,8 @@ static int f2fs_set_data_page_dirty(struct page *page)
763 struct address_space *mapping = page->mapping; 756 struct address_space *mapping = page->mapping;
764 struct inode *inode = mapping->host; 757 struct inode *inode = mapping->host;
765 758
759 trace_f2fs_set_page_dirty(page, DATA);
760
766 SetPageUptodate(page); 761 SetPageUptodate(page);
767 if (!PageDirty(page)) { 762 if (!PageDirty(page)) {
768 __set_page_dirty_nobuffers(page); 763 __set_page_dirty_nobuffers(page);
diff --git a/fs/f2fs/dir.c b/fs/f2fs/dir.c
index 384c6daf9a89..594fc1bb64ef 100644
--- a/fs/f2fs/dir.c
+++ b/fs/f2fs/dir.c
@@ -139,7 +139,7 @@ static struct f2fs_dir_entry *find_in_level(struct inode *dir,
139 bool room = false; 139 bool room = false;
140 int max_slots = 0; 140 int max_slots = 0;
141 141
142 BUG_ON(level > MAX_DIR_HASH_DEPTH); 142 f2fs_bug_on(level > MAX_DIR_HASH_DEPTH);
143 143
144 nbucket = dir_buckets(level); 144 nbucket = dir_buckets(level);
145 nblock = bucket_blocks(level); 145 nblock = bucket_blocks(level);
@@ -346,7 +346,7 @@ static struct page *init_inode_metadata(struct inode *inode,
346 goto error; 346 goto error;
347 } 347 }
348 348
349 err = f2fs_init_acl(inode, dir); 349 err = f2fs_init_acl(inode, dir, page);
350 if (err) 350 if (err)
351 goto error; 351 goto error;
352 352
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 608f0df5b919..89dc7508faf2 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -18,6 +18,13 @@
18#include <linux/crc32.h> 18#include <linux/crc32.h>
19#include <linux/magic.h> 19#include <linux/magic.h>
20#include <linux/kobject.h> 20#include <linux/kobject.h>
21#include <linux/sched.h>
22
23#ifdef CONFIG_F2FS_CHECK_FS
24#define f2fs_bug_on(condition) BUG_ON(condition)
25#else
26#define f2fs_bug_on(condition)
27#endif
21 28
22/* 29/*
23 * For mount options 30 * For mount options
@@ -298,6 +305,9 @@ struct f2fs_sm_info {
298 unsigned int main_segments; /* # of segments in main area */ 305 unsigned int main_segments; /* # of segments in main area */
299 unsigned int reserved_segments; /* # of reserved segments */ 306 unsigned int reserved_segments; /* # of reserved segments */
300 unsigned int ovp_segments; /* # of overprovision segments */ 307 unsigned int ovp_segments; /* # of overprovision segments */
308
309 /* a threshold to reclaim prefree segments */
310 unsigned int rec_prefree_segments;
301}; 311};
302 312
303/* 313/*
@@ -318,14 +328,6 @@ enum count_type {
318}; 328};
319 329
320/* 330/*
321 * Uses as sbi->fs_lock[NR_GLOBAL_LOCKS].
322 * The checkpoint procedure blocks all the locks in this fs_lock array.
323 * Some FS operations grab free locks, and if there is no free lock,
324 * then wait to grab a lock in a round-robin manner.
325 */
326#define NR_GLOBAL_LOCKS 8
327
328/*
329 * The below are the page types of bios used in submti_bio(). 331 * The below are the page types of bios used in submti_bio().
330 * The available types are: 332 * The available types are:
331 * DATA User data pages. It operates as async mode. 333 * DATA User data pages. It operates as async mode.
@@ -365,12 +367,12 @@ struct f2fs_sb_info {
365 struct f2fs_checkpoint *ckpt; /* raw checkpoint pointer */ 367 struct f2fs_checkpoint *ckpt; /* raw checkpoint pointer */
366 struct inode *meta_inode; /* cache meta blocks */ 368 struct inode *meta_inode; /* cache meta blocks */
367 struct mutex cp_mutex; /* checkpoint procedure lock */ 369 struct mutex cp_mutex; /* checkpoint procedure lock */
368 struct mutex fs_lock[NR_GLOBAL_LOCKS]; /* blocking FS operations */ 370 struct rw_semaphore cp_rwsem; /* blocking FS operations */
369 struct mutex node_write; /* locking node writes */ 371 struct mutex node_write; /* locking node writes */
370 struct mutex writepages; /* mutex for writepages() */ 372 struct mutex writepages; /* mutex for writepages() */
371 unsigned char next_lock_num; /* round-robin global locks */ 373 bool por_doing; /* recovery is doing or not */
372 int por_doing; /* recovery is doing or not */ 374 bool on_build_free_nids; /* build_free_nids is doing */
373 int on_build_free_nids; /* build_free_nids is doing */ 375 wait_queue_head_t cp_wait;
374 376
375 /* for orphan inode management */ 377 /* for orphan inode management */
376 struct list_head orphan_inode_list; /* orphan inode list */ 378 struct list_head orphan_inode_list; /* orphan inode list */
@@ -520,48 +522,24 @@ static inline void clear_ckpt_flags(struct f2fs_checkpoint *cp, unsigned int f)
520 cp->ckpt_flags = cpu_to_le32(ckpt_flags); 522 cp->ckpt_flags = cpu_to_le32(ckpt_flags);
521} 523}
522 524
523static inline void mutex_lock_all(struct f2fs_sb_info *sbi) 525static inline void f2fs_lock_op(struct f2fs_sb_info *sbi)
524{ 526{
525 int i; 527 down_read(&sbi->cp_rwsem);
526
527 for (i = 0; i < NR_GLOBAL_LOCKS; i++) {
528 /*
529 * This is the only time we take multiple fs_lock[]
530 * instances; the order is immaterial since we
531 * always hold cp_mutex, which serializes multiple
532 * such operations.
533 */
534 mutex_lock_nest_lock(&sbi->fs_lock[i], &sbi->cp_mutex);
535 }
536} 528}
537 529
538static inline void mutex_unlock_all(struct f2fs_sb_info *sbi) 530static inline void f2fs_unlock_op(struct f2fs_sb_info *sbi)
539{ 531{
540 int i = 0; 532 up_read(&sbi->cp_rwsem);
541 for (; i < NR_GLOBAL_LOCKS; i++)
542 mutex_unlock(&sbi->fs_lock[i]);
543} 533}
544 534
545static inline int mutex_lock_op(struct f2fs_sb_info *sbi) 535static inline void f2fs_lock_all(struct f2fs_sb_info *sbi)
546{ 536{
547 unsigned char next_lock = sbi->next_lock_num % NR_GLOBAL_LOCKS; 537 down_write_nest_lock(&sbi->cp_rwsem, &sbi->cp_mutex);
548 int i = 0;
549
550 for (; i < NR_GLOBAL_LOCKS; i++)
551 if (mutex_trylock(&sbi->fs_lock[i]))
552 return i;
553
554 mutex_lock(&sbi->fs_lock[next_lock]);
555 sbi->next_lock_num++;
556 return next_lock;
557} 538}
558 539
559static inline void mutex_unlock_op(struct f2fs_sb_info *sbi, int ilock) 540static inline void f2fs_unlock_all(struct f2fs_sb_info *sbi)
560{ 541{
561 if (ilock < 0) 542 up_write(&sbi->cp_rwsem);
562 return;
563 BUG_ON(ilock >= NR_GLOBAL_LOCKS);
564 mutex_unlock(&sbi->fs_lock[ilock]);
565} 543}
566 544
567/* 545/*
@@ -612,8 +590,8 @@ static inline int dec_valid_block_count(struct f2fs_sb_info *sbi,
612 blkcnt_t count) 590 blkcnt_t count)
613{ 591{
614 spin_lock(&sbi->stat_lock); 592 spin_lock(&sbi->stat_lock);
615 BUG_ON(sbi->total_valid_block_count < (block_t) count); 593 f2fs_bug_on(sbi->total_valid_block_count < (block_t) count);
616 BUG_ON(inode->i_blocks < count); 594 f2fs_bug_on(inode->i_blocks < count);
617 inode->i_blocks -= count; 595 inode->i_blocks -= count;
618 sbi->total_valid_block_count -= (block_t)count; 596 sbi->total_valid_block_count -= (block_t)count;
619 spin_unlock(&sbi->stat_lock); 597 spin_unlock(&sbi->stat_lock);
@@ -745,9 +723,9 @@ static inline void dec_valid_node_count(struct f2fs_sb_info *sbi,
745{ 723{
746 spin_lock(&sbi->stat_lock); 724 spin_lock(&sbi->stat_lock);
747 725
748 BUG_ON(sbi->total_valid_block_count < count); 726 f2fs_bug_on(sbi->total_valid_block_count < count);
749 BUG_ON(sbi->total_valid_node_count < count); 727 f2fs_bug_on(sbi->total_valid_node_count < count);
750 BUG_ON(inode->i_blocks < count); 728 f2fs_bug_on(inode->i_blocks < count);
751 729
752 inode->i_blocks -= count; 730 inode->i_blocks -= count;
753 sbi->total_valid_node_count -= count; 731 sbi->total_valid_node_count -= count;
@@ -768,7 +746,7 @@ static inline unsigned int valid_node_count(struct f2fs_sb_info *sbi)
768static inline void inc_valid_inode_count(struct f2fs_sb_info *sbi) 746static inline void inc_valid_inode_count(struct f2fs_sb_info *sbi)
769{ 747{
770 spin_lock(&sbi->stat_lock); 748 spin_lock(&sbi->stat_lock);
771 BUG_ON(sbi->total_valid_inode_count == sbi->total_node_count); 749 f2fs_bug_on(sbi->total_valid_inode_count == sbi->total_node_count);
772 sbi->total_valid_inode_count++; 750 sbi->total_valid_inode_count++;
773 spin_unlock(&sbi->stat_lock); 751 spin_unlock(&sbi->stat_lock);
774} 752}
@@ -776,7 +754,7 @@ static inline void inc_valid_inode_count(struct f2fs_sb_info *sbi)
776static inline int dec_valid_inode_count(struct f2fs_sb_info *sbi) 754static inline int dec_valid_inode_count(struct f2fs_sb_info *sbi)
777{ 755{
778 spin_lock(&sbi->stat_lock); 756 spin_lock(&sbi->stat_lock);
779 BUG_ON(!sbi->total_valid_inode_count); 757 f2fs_bug_on(!sbi->total_valid_inode_count);
780 sbi->total_valid_inode_count--; 758 sbi->total_valid_inode_count--;
781 spin_unlock(&sbi->stat_lock); 759 spin_unlock(&sbi->stat_lock);
782 return 0; 760 return 0;
@@ -797,7 +775,7 @@ static inline void f2fs_put_page(struct page *page, int unlock)
797 return; 775 return;
798 776
799 if (unlock) { 777 if (unlock) {
800 BUG_ON(!PageLocked(page)); 778 f2fs_bug_on(!PageLocked(page));
801 unlock_page(page); 779 unlock_page(page);
802 } 780 }
803 page_cache_release(page); 781 page_cache_release(page);
@@ -819,6 +797,20 @@ static inline struct kmem_cache *f2fs_kmem_cache_create(const char *name,
819 return kmem_cache_create(name, size, 0, SLAB_RECLAIM_ACCOUNT, ctor); 797 return kmem_cache_create(name, size, 0, SLAB_RECLAIM_ACCOUNT, ctor);
820} 798}
821 799
800static inline void *f2fs_kmem_cache_alloc(struct kmem_cache *cachep,
801 gfp_t flags)
802{
803 void *entry;
804retry:
805 entry = kmem_cache_alloc(cachep, flags);
806 if (!entry) {
807 cond_resched();
808 goto retry;
809 }
810
811 return entry;
812}
813
822#define RAW_IS_INODE(p) ((p)->footer.nid == (p)->footer.ino) 814#define RAW_IS_INODE(p) ((p)->footer.nid == (p)->footer.ino)
823 815
824static inline bool IS_INODE(struct page *page) 816static inline bool IS_INODE(struct page *page)
@@ -979,6 +971,7 @@ long f2fs_compat_ioctl(struct file *, unsigned int, unsigned long);
979 */ 971 */
980void f2fs_set_inode_flags(struct inode *); 972void f2fs_set_inode_flags(struct inode *);
981struct inode *f2fs_iget(struct super_block *, unsigned long); 973struct inode *f2fs_iget(struct super_block *, unsigned long);
974int try_to_free_nats(struct f2fs_sb_info *, int);
982void update_inode(struct inode *, struct page *); 975void update_inode(struct inode *, struct page *);
983int update_inode_page(struct inode *); 976int update_inode_page(struct inode *);
984int f2fs_write_inode(struct inode *, struct writeback_control *); 977int f2fs_write_inode(struct inode *, struct writeback_control *);
@@ -1033,6 +1026,7 @@ void get_node_info(struct f2fs_sb_info *, nid_t, struct node_info *);
1033int get_dnode_of_data(struct dnode_of_data *, pgoff_t, int); 1026int get_dnode_of_data(struct dnode_of_data *, pgoff_t, int);
1034int truncate_inode_blocks(struct inode *, pgoff_t); 1027int truncate_inode_blocks(struct inode *, pgoff_t);
1035int truncate_xattr_node(struct inode *, struct page *); 1028int truncate_xattr_node(struct inode *, struct page *);
1029int wait_on_node_pages_writeback(struct f2fs_sb_info *, nid_t);
1036int remove_inode_page(struct inode *); 1030int remove_inode_page(struct inode *);
1037struct page *new_inode_page(struct inode *, const struct qstr *); 1031struct page *new_inode_page(struct inode *, const struct qstr *);
1038struct page *new_node_page(struct dnode_of_data *, unsigned int, struct page *); 1032struct page *new_node_page(struct dnode_of_data *, unsigned int, struct page *);
@@ -1059,6 +1053,7 @@ void destroy_node_manager_caches(void);
1059 * segment.c 1053 * segment.c
1060 */ 1054 */
1061void f2fs_balance_fs(struct f2fs_sb_info *); 1055void f2fs_balance_fs(struct f2fs_sb_info *);
1056void f2fs_balance_fs_bg(struct f2fs_sb_info *);
1062void invalidate_blocks(struct f2fs_sb_info *, block_t); 1057void invalidate_blocks(struct f2fs_sb_info *, block_t);
1063void clear_prefree_segments(struct f2fs_sb_info *); 1058void clear_prefree_segments(struct f2fs_sb_info *);
1064int npages_for_summary_flush(struct f2fs_sb_info *); 1059int npages_for_summary_flush(struct f2fs_sb_info *);
@@ -1172,7 +1167,16 @@ static inline struct f2fs_stat_info *F2FS_STAT(struct f2fs_sb_info *sbi)
1172 return (struct f2fs_stat_info*)sbi->stat_info; 1167 return (struct f2fs_stat_info*)sbi->stat_info;
1173} 1168}
1174 1169
1175#define stat_inc_call_count(si) ((si)->call_count++) 1170#define stat_inc_call_count(si) ((si)->call_count++)
1171#define stat_inc_bggc_count(sbi) ((sbi)->bg_gc++)
1172#define stat_inc_dirty_dir(sbi) ((sbi)->n_dirty_dirs++)
1173#define stat_dec_dirty_dir(sbi) ((sbi)->n_dirty_dirs--)
1174#define stat_inc_total_hit(sb) ((F2FS_SB(sb))->total_hit_ext++)
1175#define stat_inc_read_hit(sb) ((F2FS_SB(sb))->read_hit_ext++)
1176#define stat_inc_seg_type(sbi, curseg) \
1177 ((sbi)->segment_count[(curseg)->alloc_type]++)
1178#define stat_inc_block_count(sbi, curseg) \
1179 ((sbi)->block_count[(curseg)->alloc_type]++)
1176 1180
1177#define stat_inc_seg_count(sbi, type) \ 1181#define stat_inc_seg_count(sbi, type) \
1178 do { \ 1182 do { \
@@ -1207,6 +1211,13 @@ void __init f2fs_create_root_stats(void);
1207void f2fs_destroy_root_stats(void); 1211void f2fs_destroy_root_stats(void);
1208#else 1212#else
1209#define stat_inc_call_count(si) 1213#define stat_inc_call_count(si)
1214#define stat_inc_bggc_count(si)
1215#define stat_inc_dirty_dir(sbi)
1216#define stat_dec_dirty_dir(sbi)
1217#define stat_inc_total_hit(sb)
1218#define stat_inc_read_hit(sb)
1219#define stat_inc_seg_type(sbi, curseg)
1220#define stat_inc_block_count(sbi, curseg)
1210#define stat_inc_seg_count(si, type) 1221#define stat_inc_seg_count(si, type)
1211#define stat_inc_tot_blk_count(si, blks) 1222#define stat_inc_tot_blk_count(si, blks)
1212#define stat_inc_data_blk_count(si, blks) 1223#define stat_inc_data_blk_count(si, blks)
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index 02c906971cc6..7d714f4972d5 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -35,18 +35,18 @@ static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
35 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 35 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
36 block_t old_blk_addr; 36 block_t old_blk_addr;
37 struct dnode_of_data dn; 37 struct dnode_of_data dn;
38 int err, ilock; 38 int err;
39 39
40 f2fs_balance_fs(sbi); 40 f2fs_balance_fs(sbi);
41 41
42 sb_start_pagefault(inode->i_sb); 42 sb_start_pagefault(inode->i_sb);
43 43
44 /* block allocation */ 44 /* block allocation */
45 ilock = mutex_lock_op(sbi); 45 f2fs_lock_op(sbi);
46 set_new_dnode(&dn, inode, NULL, NULL, 0); 46 set_new_dnode(&dn, inode, NULL, NULL, 0);
47 err = get_dnode_of_data(&dn, page->index, ALLOC_NODE); 47 err = get_dnode_of_data(&dn, page->index, ALLOC_NODE);
48 if (err) { 48 if (err) {
49 mutex_unlock_op(sbi, ilock); 49 f2fs_unlock_op(sbi);
50 goto out; 50 goto out;
51 } 51 }
52 52
@@ -56,12 +56,12 @@ static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
56 err = reserve_new_block(&dn); 56 err = reserve_new_block(&dn);
57 if (err) { 57 if (err) {
58 f2fs_put_dnode(&dn); 58 f2fs_put_dnode(&dn);
59 mutex_unlock_op(sbi, ilock); 59 f2fs_unlock_op(sbi);
60 goto out; 60 goto out;
61 } 61 }
62 } 62 }
63 f2fs_put_dnode(&dn); 63 f2fs_put_dnode(&dn);
64 mutex_unlock_op(sbi, ilock); 64 f2fs_unlock_op(sbi);
65 65
66 file_update_time(vma->vm_file); 66 file_update_time(vma->vm_file);
67 lock_page(page); 67 lock_page(page);
@@ -88,6 +88,7 @@ static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
88 set_page_dirty(page); 88 set_page_dirty(page);
89 SetPageUptodate(page); 89 SetPageUptodate(page);
90 90
91 trace_f2fs_vm_page_mkwrite(page, DATA);
91mapped: 92mapped:
92 /* fill the page */ 93 /* fill the page */
93 wait_on_page_writeback(page); 94 wait_on_page_writeback(page);
@@ -188,8 +189,9 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
188 if (ret) 189 if (ret)
189 goto out; 190 goto out;
190 } 191 }
191 filemap_fdatawait_range(sbi->node_inode->i_mapping, 192 ret = wait_on_node_pages_writeback(sbi, inode->i_ino);
192 0, LONG_MAX); 193 if (ret)
194 goto out;
193 ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL); 195 ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
194 } 196 }
195out: 197out:
@@ -270,7 +272,7 @@ static int truncate_blocks(struct inode *inode, u64 from)
270 unsigned int blocksize = inode->i_sb->s_blocksize; 272 unsigned int blocksize = inode->i_sb->s_blocksize;
271 struct dnode_of_data dn; 273 struct dnode_of_data dn;
272 pgoff_t free_from; 274 pgoff_t free_from;
273 int count = 0, ilock = -1; 275 int count = 0;
274 int err; 276 int err;
275 277
276 trace_f2fs_truncate_blocks_enter(inode, from); 278 trace_f2fs_truncate_blocks_enter(inode, from);
@@ -278,13 +280,13 @@ static int truncate_blocks(struct inode *inode, u64 from)
278 free_from = (pgoff_t) 280 free_from = (pgoff_t)
279 ((from + blocksize - 1) >> (sbi->log_blocksize)); 281 ((from + blocksize - 1) >> (sbi->log_blocksize));
280 282
281 ilock = mutex_lock_op(sbi); 283 f2fs_lock_op(sbi);
282 set_new_dnode(&dn, inode, NULL, NULL, 0); 284 set_new_dnode(&dn, inode, NULL, NULL, 0);
283 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE); 285 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
284 if (err) { 286 if (err) {
285 if (err == -ENOENT) 287 if (err == -ENOENT)
286 goto free_next; 288 goto free_next;
287 mutex_unlock_op(sbi, ilock); 289 f2fs_unlock_op(sbi);
288 trace_f2fs_truncate_blocks_exit(inode, err); 290 trace_f2fs_truncate_blocks_exit(inode, err);
289 return err; 291 return err;
290 } 292 }
@@ -295,7 +297,7 @@ static int truncate_blocks(struct inode *inode, u64 from)
295 count = ADDRS_PER_BLOCK; 297 count = ADDRS_PER_BLOCK;
296 298
297 count -= dn.ofs_in_node; 299 count -= dn.ofs_in_node;
298 BUG_ON(count < 0); 300 f2fs_bug_on(count < 0);
299 301
300 if (dn.ofs_in_node || IS_INODE(dn.node_page)) { 302 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
301 truncate_data_blocks_range(&dn, count); 303 truncate_data_blocks_range(&dn, count);
@@ -305,7 +307,7 @@ static int truncate_blocks(struct inode *inode, u64 from)
305 f2fs_put_dnode(&dn); 307 f2fs_put_dnode(&dn);
306free_next: 308free_next:
307 err = truncate_inode_blocks(inode, free_from); 309 err = truncate_inode_blocks(inode, free_from);
308 mutex_unlock_op(sbi, ilock); 310 f2fs_unlock_op(sbi);
309 311
310 /* lastly zero out the first data page */ 312 /* lastly zero out the first data page */
311 truncate_partial_data_page(inode, from); 313 truncate_partial_data_page(inode, from);
@@ -416,16 +418,15 @@ static void fill_zero(struct inode *inode, pgoff_t index,
416{ 418{
417 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 419 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
418 struct page *page; 420 struct page *page;
419 int ilock;
420 421
421 if (!len) 422 if (!len)
422 return; 423 return;
423 424
424 f2fs_balance_fs(sbi); 425 f2fs_balance_fs(sbi);
425 426
426 ilock = mutex_lock_op(sbi); 427 f2fs_lock_op(sbi);
427 page = get_new_data_page(inode, NULL, index, false); 428 page = get_new_data_page(inode, NULL, index, false);
428 mutex_unlock_op(sbi, ilock); 429 f2fs_unlock_op(sbi);
429 430
430 if (!IS_ERR(page)) { 431 if (!IS_ERR(page)) {
431 wait_on_page_writeback(page); 432 wait_on_page_writeback(page);
@@ -484,7 +485,6 @@ static int punch_hole(struct inode *inode, loff_t offset, loff_t len, int mode)
484 struct address_space *mapping = inode->i_mapping; 485 struct address_space *mapping = inode->i_mapping;
485 loff_t blk_start, blk_end; 486 loff_t blk_start, blk_end;
486 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 487 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
487 int ilock;
488 488
489 f2fs_balance_fs(sbi); 489 f2fs_balance_fs(sbi);
490 490
@@ -493,9 +493,9 @@ static int punch_hole(struct inode *inode, loff_t offset, loff_t len, int mode)
493 truncate_inode_pages_range(mapping, blk_start, 493 truncate_inode_pages_range(mapping, blk_start,
494 blk_end - 1); 494 blk_end - 1);
495 495
496 ilock = mutex_lock_op(sbi); 496 f2fs_lock_op(sbi);
497 ret = truncate_hole(inode, pg_start, pg_end); 497 ret = truncate_hole(inode, pg_start, pg_end);
498 mutex_unlock_op(sbi, ilock); 498 f2fs_unlock_op(sbi);
499 } 499 }
500 } 500 }
501 501
@@ -529,13 +529,12 @@ static int expand_inode_data(struct inode *inode, loff_t offset,
529 529
530 for (index = pg_start; index <= pg_end; index++) { 530 for (index = pg_start; index <= pg_end; index++) {
531 struct dnode_of_data dn; 531 struct dnode_of_data dn;
532 int ilock;
533 532
534 ilock = mutex_lock_op(sbi); 533 f2fs_lock_op(sbi);
535 set_new_dnode(&dn, inode, NULL, NULL, 0); 534 set_new_dnode(&dn, inode, NULL, NULL, 0);
536 ret = get_dnode_of_data(&dn, index, ALLOC_NODE); 535 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
537 if (ret) { 536 if (ret) {
538 mutex_unlock_op(sbi, ilock); 537 f2fs_unlock_op(sbi);
539 break; 538 break;
540 } 539 }
541 540
@@ -543,12 +542,12 @@ static int expand_inode_data(struct inode *inode, loff_t offset,
543 ret = reserve_new_block(&dn); 542 ret = reserve_new_block(&dn);
544 if (ret) { 543 if (ret) {
545 f2fs_put_dnode(&dn); 544 f2fs_put_dnode(&dn);
546 mutex_unlock_op(sbi, ilock); 545 f2fs_unlock_op(sbi);
547 break; 546 break;
548 } 547 }
549 } 548 }
550 f2fs_put_dnode(&dn); 549 f2fs_put_dnode(&dn);
551 mutex_unlock_op(sbi, ilock); 550 f2fs_unlock_op(sbi);
552 551
553 if (pg_start == pg_end) 552 if (pg_start == pg_end)
554 new_size = offset + len; 553 new_size = offset + len;
diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
index 2f157e883687..b7ad1ec7e4cc 100644
--- a/fs/f2fs/gc.c
+++ b/fs/f2fs/gc.c
@@ -77,13 +77,15 @@ static int gc_thread_func(void *data)
77 else 77 else
78 wait_ms = increase_sleep_time(gc_th, wait_ms); 78 wait_ms = increase_sleep_time(gc_th, wait_ms);
79 79
80#ifdef CONFIG_F2FS_STAT_FS 80 stat_inc_bggc_count(sbi);
81 sbi->bg_gc++;
82#endif
83 81
84 /* if return value is not zero, no victim was selected */ 82 /* if return value is not zero, no victim was selected */
85 if (f2fs_gc(sbi)) 83 if (f2fs_gc(sbi))
86 wait_ms = gc_th->no_gc_sleep_time; 84 wait_ms = gc_th->no_gc_sleep_time;
85
86 /* balancing f2fs's metadata periodically */
87 f2fs_balance_fs_bg(sbi);
88
87 } while (!kthread_should_stop()); 89 } while (!kthread_should_stop());
88 return 0; 90 return 0;
89} 91}
@@ -236,8 +238,8 @@ static unsigned int get_cb_cost(struct f2fs_sb_info *sbi, unsigned int segno)
236 return UINT_MAX - ((100 * (100 - u) * age) / (100 + u)); 238 return UINT_MAX - ((100 * (100 - u) * age) / (100 + u));
237} 239}
238 240
239static unsigned int get_gc_cost(struct f2fs_sb_info *sbi, unsigned int segno, 241static inline unsigned int get_gc_cost(struct f2fs_sb_info *sbi,
240 struct victim_sel_policy *p) 242 unsigned int segno, struct victim_sel_policy *p)
241{ 243{
242 if (p->alloc_mode == SSR) 244 if (p->alloc_mode == SSR)
243 return get_seg_entry(sbi, segno)->ckpt_valid_blocks; 245 return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
@@ -293,7 +295,11 @@ static int get_victim_by_default(struct f2fs_sb_info *sbi,
293 } 295 }
294 break; 296 break;
295 } 297 }
296 p.offset = ((segno / p.ofs_unit) * p.ofs_unit) + p.ofs_unit; 298
299 p.offset = segno + p.ofs_unit;
300 if (p.ofs_unit > 1)
301 p.offset -= segno % p.ofs_unit;
302
297 secno = GET_SECNO(sbi, segno); 303 secno = GET_SECNO(sbi, segno);
298 304
299 if (sec_usage_check(sbi, secno)) 305 if (sec_usage_check(sbi, secno))
@@ -306,10 +312,9 @@ static int get_victim_by_default(struct f2fs_sb_info *sbi,
306 if (p.min_cost > cost) { 312 if (p.min_cost > cost) {
307 p.min_segno = segno; 313 p.min_segno = segno;
308 p.min_cost = cost; 314 p.min_cost = cost;
309 } 315 } else if (unlikely(cost == max_cost)) {
310
311 if (cost == max_cost)
312 continue; 316 continue;
317 }
313 318
314 if (nsearched++ >= p.max_search) { 319 if (nsearched++ >= p.max_search) {
315 sbi->last_victim[p.gc_mode] = segno; 320 sbi->last_victim[p.gc_mode] = segno;
@@ -358,12 +363,8 @@ static void add_gc_inode(struct inode *inode, struct list_head *ilist)
358 iput(inode); 363 iput(inode);
359 return; 364 return;
360 } 365 }
361repeat: 366
362 new_ie = kmem_cache_alloc(winode_slab, GFP_NOFS); 367 new_ie = f2fs_kmem_cache_alloc(winode_slab, GFP_NOFS);
363 if (!new_ie) {
364 cond_resched();
365 goto repeat;
366 }
367 new_ie->inode = inode; 368 new_ie->inode = inode;
368 list_add_tail(&new_ie->list, ilist); 369 list_add_tail(&new_ie->list, ilist);
369} 370}
diff --git a/fs/f2fs/inode.c b/fs/f2fs/inode.c
index 9339cd292047..d0eaa9faeca0 100644
--- a/fs/f2fs/inode.c
+++ b/fs/f2fs/inode.c
@@ -37,6 +37,31 @@ void f2fs_set_inode_flags(struct inode *inode)
37 inode->i_flags |= S_DIRSYNC; 37 inode->i_flags |= S_DIRSYNC;
38} 38}
39 39
40static void __get_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
41{
42 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
43 S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
44 if (ri->i_addr[0])
45 inode->i_rdev = old_decode_dev(le32_to_cpu(ri->i_addr[0]));
46 else
47 inode->i_rdev = new_decode_dev(le32_to_cpu(ri->i_addr[1]));
48 }
49}
50
51static void __set_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
52{
53 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
54 if (old_valid_dev(inode->i_rdev)) {
55 ri->i_addr[0] = cpu_to_le32(old_encode_dev(inode->i_rdev));
56 ri->i_addr[1] = 0;
57 } else {
58 ri->i_addr[0] = 0;
59 ri->i_addr[1] = cpu_to_le32(new_encode_dev(inode->i_rdev));
60 ri->i_addr[2] = 0;
61 }
62 }
63}
64
40static int do_read_inode(struct inode *inode) 65static int do_read_inode(struct inode *inode)
41{ 66{
42 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 67 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
@@ -73,10 +98,6 @@ static int do_read_inode(struct inode *inode)
73 inode->i_ctime.tv_nsec = le32_to_cpu(ri->i_ctime_nsec); 98 inode->i_ctime.tv_nsec = le32_to_cpu(ri->i_ctime_nsec);
74 inode->i_mtime.tv_nsec = le32_to_cpu(ri->i_mtime_nsec); 99 inode->i_mtime.tv_nsec = le32_to_cpu(ri->i_mtime_nsec);
75 inode->i_generation = le32_to_cpu(ri->i_generation); 100 inode->i_generation = le32_to_cpu(ri->i_generation);
76 if (ri->i_addr[0])
77 inode->i_rdev = old_decode_dev(le32_to_cpu(ri->i_addr[0]));
78 else
79 inode->i_rdev = new_decode_dev(le32_to_cpu(ri->i_addr[1]));
80 101
81 fi->i_current_depth = le32_to_cpu(ri->i_current_depth); 102 fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
82 fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid); 103 fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
@@ -84,8 +105,13 @@ static int do_read_inode(struct inode *inode)
84 fi->flags = 0; 105 fi->flags = 0;
85 fi->i_advise = ri->i_advise; 106 fi->i_advise = ri->i_advise;
86 fi->i_pino = le32_to_cpu(ri->i_pino); 107 fi->i_pino = le32_to_cpu(ri->i_pino);
108
87 get_extent_info(&fi->ext, ri->i_ext); 109 get_extent_info(&fi->ext, ri->i_ext);
88 get_inline_info(fi, ri); 110 get_inline_info(fi, ri);
111
112 /* get rdev by using inline_info */
113 __get_inode_rdev(inode, ri);
114
89 f2fs_put_page(node_page, 1); 115 f2fs_put_page(node_page, 1);
90 return 0; 116 return 0;
91} 117}
@@ -179,21 +205,10 @@ void update_inode(struct inode *inode, struct page *node_page)
179 ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino); 205 ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino);
180 ri->i_generation = cpu_to_le32(inode->i_generation); 206 ri->i_generation = cpu_to_le32(inode->i_generation);
181 207
182 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) { 208 __set_inode_rdev(inode, ri);
183 if (old_valid_dev(inode->i_rdev)) {
184 ri->i_addr[0] =
185 cpu_to_le32(old_encode_dev(inode->i_rdev));
186 ri->i_addr[1] = 0;
187 } else {
188 ri->i_addr[0] = 0;
189 ri->i_addr[1] =
190 cpu_to_le32(new_encode_dev(inode->i_rdev));
191 ri->i_addr[2] = 0;
192 }
193 }
194
195 set_cold_node(inode, node_page); 209 set_cold_node(inode, node_page);
196 set_page_dirty(node_page); 210 set_page_dirty(node_page);
211
197 clear_inode_flag(F2FS_I(inode), FI_DIRTY_INODE); 212 clear_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
198} 213}
199 214
@@ -214,7 +229,7 @@ int update_inode_page(struct inode *inode)
214int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc) 229int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
215{ 230{
216 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 231 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
217 int ret, ilock; 232 int ret;
218 233
219 if (inode->i_ino == F2FS_NODE_INO(sbi) || 234 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
220 inode->i_ino == F2FS_META_INO(sbi)) 235 inode->i_ino == F2FS_META_INO(sbi))
@@ -227,9 +242,9 @@ int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
227 * We need to lock here to prevent from producing dirty node pages 242 * We need to lock here to prevent from producing dirty node pages
228 * during the urgent cleaning time when runing out of free sections. 243 * during the urgent cleaning time when runing out of free sections.
229 */ 244 */
230 ilock = mutex_lock_op(sbi); 245 f2fs_lock_op(sbi);
231 ret = update_inode_page(inode); 246 ret = update_inode_page(inode);
232 mutex_unlock_op(sbi, ilock); 247 f2fs_unlock_op(sbi);
233 248
234 if (wbc) 249 if (wbc)
235 f2fs_balance_fs(sbi); 250 f2fs_balance_fs(sbi);
@@ -243,7 +258,6 @@ int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
243void f2fs_evict_inode(struct inode *inode) 258void f2fs_evict_inode(struct inode *inode)
244{ 259{
245 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 260 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
246 int ilock;
247 261
248 trace_f2fs_evict_inode(inode); 262 trace_f2fs_evict_inode(inode);
249 truncate_inode_pages(&inode->i_data, 0); 263 truncate_inode_pages(&inode->i_data, 0);
@@ -252,7 +266,7 @@ void f2fs_evict_inode(struct inode *inode)
252 inode->i_ino == F2FS_META_INO(sbi)) 266 inode->i_ino == F2FS_META_INO(sbi))
253 goto no_delete; 267 goto no_delete;
254 268
255 BUG_ON(atomic_read(&F2FS_I(inode)->dirty_dents)); 269 f2fs_bug_on(atomic_read(&F2FS_I(inode)->dirty_dents));
256 remove_dirty_dir_inode(inode); 270 remove_dirty_dir_inode(inode);
257 271
258 if (inode->i_nlink || is_bad_inode(inode)) 272 if (inode->i_nlink || is_bad_inode(inode))
@@ -265,9 +279,9 @@ void f2fs_evict_inode(struct inode *inode)
265 if (F2FS_HAS_BLOCKS(inode)) 279 if (F2FS_HAS_BLOCKS(inode))
266 f2fs_truncate(inode); 280 f2fs_truncate(inode);
267 281
268 ilock = mutex_lock_op(sbi); 282 f2fs_lock_op(sbi);
269 remove_inode_page(inode); 283 remove_inode_page(inode);
270 mutex_unlock_op(sbi, ilock); 284 f2fs_unlock_op(sbi);
271 285
272 sb_end_intwrite(inode->i_sb); 286 sb_end_intwrite(inode->i_sb);
273no_delete: 287no_delete:
diff --git a/fs/f2fs/namei.c b/fs/f2fs/namei.c
index 2a5359c990fc..575adac17f8b 100644
--- a/fs/f2fs/namei.c
+++ b/fs/f2fs/namei.c
@@ -27,19 +27,19 @@ static struct inode *f2fs_new_inode(struct inode *dir, umode_t mode)
27 nid_t ino; 27 nid_t ino;
28 struct inode *inode; 28 struct inode *inode;
29 bool nid_free = false; 29 bool nid_free = false;
30 int err, ilock; 30 int err;
31 31
32 inode = new_inode(sb); 32 inode = new_inode(sb);
33 if (!inode) 33 if (!inode)
34 return ERR_PTR(-ENOMEM); 34 return ERR_PTR(-ENOMEM);
35 35
36 ilock = mutex_lock_op(sbi); 36 f2fs_lock_op(sbi);
37 if (!alloc_nid(sbi, &ino)) { 37 if (!alloc_nid(sbi, &ino)) {
38 mutex_unlock_op(sbi, ilock); 38 f2fs_unlock_op(sbi);
39 err = -ENOSPC; 39 err = -ENOSPC;
40 goto fail; 40 goto fail;
41 } 41 }
42 mutex_unlock_op(sbi, ilock); 42 f2fs_unlock_op(sbi);
43 43
44 inode->i_uid = current_fsuid(); 44 inode->i_uid = current_fsuid();
45 45
@@ -115,7 +115,7 @@ static int f2fs_create(struct inode *dir, struct dentry *dentry, umode_t mode,
115 struct f2fs_sb_info *sbi = F2FS_SB(sb); 115 struct f2fs_sb_info *sbi = F2FS_SB(sb);
116 struct inode *inode; 116 struct inode *inode;
117 nid_t ino = 0; 117 nid_t ino = 0;
118 int err, ilock; 118 int err;
119 119
120 f2fs_balance_fs(sbi); 120 f2fs_balance_fs(sbi);
121 121
@@ -131,9 +131,9 @@ static int f2fs_create(struct inode *dir, struct dentry *dentry, umode_t mode,
131 inode->i_mapping->a_ops = &f2fs_dblock_aops; 131 inode->i_mapping->a_ops = &f2fs_dblock_aops;
132 ino = inode->i_ino; 132 ino = inode->i_ino;
133 133
134 ilock = mutex_lock_op(sbi); 134 f2fs_lock_op(sbi);
135 err = f2fs_add_link(dentry, inode); 135 err = f2fs_add_link(dentry, inode);
136 mutex_unlock_op(sbi, ilock); 136 f2fs_unlock_op(sbi);
137 if (err) 137 if (err)
138 goto out; 138 goto out;
139 139
@@ -157,7 +157,7 @@ static int f2fs_link(struct dentry *old_dentry, struct inode *dir,
157 struct inode *inode = old_dentry->d_inode; 157 struct inode *inode = old_dentry->d_inode;
158 struct super_block *sb = dir->i_sb; 158 struct super_block *sb = dir->i_sb;
159 struct f2fs_sb_info *sbi = F2FS_SB(sb); 159 struct f2fs_sb_info *sbi = F2FS_SB(sb);
160 int err, ilock; 160 int err;
161 161
162 f2fs_balance_fs(sbi); 162 f2fs_balance_fs(sbi);
163 163
@@ -165,9 +165,9 @@ static int f2fs_link(struct dentry *old_dentry, struct inode *dir,
165 ihold(inode); 165 ihold(inode);
166 166
167 set_inode_flag(F2FS_I(inode), FI_INC_LINK); 167 set_inode_flag(F2FS_I(inode), FI_INC_LINK);
168 ilock = mutex_lock_op(sbi); 168 f2fs_lock_op(sbi);
169 err = f2fs_add_link(dentry, inode); 169 err = f2fs_add_link(dentry, inode);
170 mutex_unlock_op(sbi, ilock); 170 f2fs_unlock_op(sbi);
171 if (err) 171 if (err)
172 goto out; 172 goto out;
173 173
@@ -220,7 +220,6 @@ static int f2fs_unlink(struct inode *dir, struct dentry *dentry)
220 struct f2fs_dir_entry *de; 220 struct f2fs_dir_entry *de;
221 struct page *page; 221 struct page *page;
222 int err = -ENOENT; 222 int err = -ENOENT;
223 int ilock;
224 223
225 trace_f2fs_unlink_enter(dir, dentry); 224 trace_f2fs_unlink_enter(dir, dentry);
226 f2fs_balance_fs(sbi); 225 f2fs_balance_fs(sbi);
@@ -229,16 +228,16 @@ static int f2fs_unlink(struct inode *dir, struct dentry *dentry)
229 if (!de) 228 if (!de)
230 goto fail; 229 goto fail;
231 230
231 f2fs_lock_op(sbi);
232 err = acquire_orphan_inode(sbi); 232 err = acquire_orphan_inode(sbi);
233 if (err) { 233 if (err) {
234 f2fs_unlock_op(sbi);
234 kunmap(page); 235 kunmap(page);
235 f2fs_put_page(page, 0); 236 f2fs_put_page(page, 0);
236 goto fail; 237 goto fail;
237 } 238 }
238
239 ilock = mutex_lock_op(sbi);
240 f2fs_delete_entry(de, page, inode); 239 f2fs_delete_entry(de, page, inode);
241 mutex_unlock_op(sbi, ilock); 240 f2fs_unlock_op(sbi);
242 241
243 /* In order to evict this inode, we set it dirty */ 242 /* In order to evict this inode, we set it dirty */
244 mark_inode_dirty(inode); 243 mark_inode_dirty(inode);
@@ -254,7 +253,7 @@ static int f2fs_symlink(struct inode *dir, struct dentry *dentry,
254 struct f2fs_sb_info *sbi = F2FS_SB(sb); 253 struct f2fs_sb_info *sbi = F2FS_SB(sb);
255 struct inode *inode; 254 struct inode *inode;
256 size_t symlen = strlen(symname) + 1; 255 size_t symlen = strlen(symname) + 1;
257 int err, ilock; 256 int err;
258 257
259 f2fs_balance_fs(sbi); 258 f2fs_balance_fs(sbi);
260 259
@@ -265,9 +264,9 @@ static int f2fs_symlink(struct inode *dir, struct dentry *dentry,
265 inode->i_op = &f2fs_symlink_inode_operations; 264 inode->i_op = &f2fs_symlink_inode_operations;
266 inode->i_mapping->a_ops = &f2fs_dblock_aops; 265 inode->i_mapping->a_ops = &f2fs_dblock_aops;
267 266
268 ilock = mutex_lock_op(sbi); 267 f2fs_lock_op(sbi);
269 err = f2fs_add_link(dentry, inode); 268 err = f2fs_add_link(dentry, inode);
270 mutex_unlock_op(sbi, ilock); 269 f2fs_unlock_op(sbi);
271 if (err) 270 if (err)
272 goto out; 271 goto out;
273 272
@@ -290,7 +289,7 @@ static int f2fs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
290{ 289{
291 struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb); 290 struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
292 struct inode *inode; 291 struct inode *inode;
293 int err, ilock; 292 int err;
294 293
295 f2fs_balance_fs(sbi); 294 f2fs_balance_fs(sbi);
296 295
@@ -304,9 +303,9 @@ static int f2fs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
304 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO); 303 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO);
305 304
306 set_inode_flag(F2FS_I(inode), FI_INC_LINK); 305 set_inode_flag(F2FS_I(inode), FI_INC_LINK);
307 ilock = mutex_lock_op(sbi); 306 f2fs_lock_op(sbi);
308 err = f2fs_add_link(dentry, inode); 307 err = f2fs_add_link(dentry, inode);
309 mutex_unlock_op(sbi, ilock); 308 f2fs_unlock_op(sbi);
310 if (err) 309 if (err)
311 goto out_fail; 310 goto out_fail;
312 311
@@ -342,7 +341,6 @@ static int f2fs_mknod(struct inode *dir, struct dentry *dentry,
342 struct f2fs_sb_info *sbi = F2FS_SB(sb); 341 struct f2fs_sb_info *sbi = F2FS_SB(sb);
343 struct inode *inode; 342 struct inode *inode;
344 int err = 0; 343 int err = 0;
345 int ilock;
346 344
347 if (!new_valid_dev(rdev)) 345 if (!new_valid_dev(rdev))
348 return -EINVAL; 346 return -EINVAL;
@@ -356,9 +354,9 @@ static int f2fs_mknod(struct inode *dir, struct dentry *dentry,
356 init_special_inode(inode, inode->i_mode, rdev); 354 init_special_inode(inode, inode->i_mode, rdev);
357 inode->i_op = &f2fs_special_inode_operations; 355 inode->i_op = &f2fs_special_inode_operations;
358 356
359 ilock = mutex_lock_op(sbi); 357 f2fs_lock_op(sbi);
360 err = f2fs_add_link(dentry, inode); 358 err = f2fs_add_link(dentry, inode);
361 mutex_unlock_op(sbi, ilock); 359 f2fs_unlock_op(sbi);
362 if (err) 360 if (err)
363 goto out; 361 goto out;
364 362
@@ -387,7 +385,7 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
387 struct f2fs_dir_entry *old_dir_entry = NULL; 385 struct f2fs_dir_entry *old_dir_entry = NULL;
388 struct f2fs_dir_entry *old_entry; 386 struct f2fs_dir_entry *old_entry;
389 struct f2fs_dir_entry *new_entry; 387 struct f2fs_dir_entry *new_entry;
390 int err = -ENOENT, ilock = -1; 388 int err = -ENOENT;
391 389
392 f2fs_balance_fs(sbi); 390 f2fs_balance_fs(sbi);
393 391
@@ -402,7 +400,7 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
402 goto out_old; 400 goto out_old;
403 } 401 }
404 402
405 ilock = mutex_lock_op(sbi); 403 f2fs_lock_op(sbi);
406 404
407 if (new_inode) { 405 if (new_inode) {
408 406
@@ -467,7 +465,7 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
467 update_inode_page(old_dir); 465 update_inode_page(old_dir);
468 } 466 }
469 467
470 mutex_unlock_op(sbi, ilock); 468 f2fs_unlock_op(sbi);
471 return 0; 469 return 0;
472 470
473put_out_dir: 471put_out_dir:
@@ -477,7 +475,7 @@ out_dir:
477 kunmap(old_dir_page); 475 kunmap(old_dir_page);
478 f2fs_put_page(old_dir_page, 0); 476 f2fs_put_page(old_dir_page, 0);
479 } 477 }
480 mutex_unlock_op(sbi, ilock); 478 f2fs_unlock_op(sbi);
481out_old: 479out_old:
482 kunmap(old_page); 480 kunmap(old_page);
483 f2fs_put_page(old_page, 0); 481 f2fs_put_page(old_page, 0);
diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
index 51ef27894433..4ac4150d421d 100644
--- a/fs/f2fs/node.c
+++ b/fs/f2fs/node.c
@@ -204,7 +204,7 @@ retry:
204 } 204 }
205 e->ni = *ni; 205 e->ni = *ni;
206 e->checkpointed = true; 206 e->checkpointed = true;
207 BUG_ON(ni->blk_addr == NEW_ADDR); 207 f2fs_bug_on(ni->blk_addr == NEW_ADDR);
208 } else if (new_blkaddr == NEW_ADDR) { 208 } else if (new_blkaddr == NEW_ADDR) {
209 /* 209 /*
210 * when nid is reallocated, 210 * when nid is reallocated,
@@ -212,19 +212,19 @@ retry:
212 * So, reinitialize it with new information. 212 * So, reinitialize it with new information.
213 */ 213 */
214 e->ni = *ni; 214 e->ni = *ni;
215 BUG_ON(ni->blk_addr != NULL_ADDR); 215 f2fs_bug_on(ni->blk_addr != NULL_ADDR);
216 } 216 }
217 217
218 if (new_blkaddr == NEW_ADDR) 218 if (new_blkaddr == NEW_ADDR)
219 e->checkpointed = false; 219 e->checkpointed = false;
220 220
221 /* sanity check */ 221 /* sanity check */
222 BUG_ON(nat_get_blkaddr(e) != ni->blk_addr); 222 f2fs_bug_on(nat_get_blkaddr(e) != ni->blk_addr);
223 BUG_ON(nat_get_blkaddr(e) == NULL_ADDR && 223 f2fs_bug_on(nat_get_blkaddr(e) == NULL_ADDR &&
224 new_blkaddr == NULL_ADDR); 224 new_blkaddr == NULL_ADDR);
225 BUG_ON(nat_get_blkaddr(e) == NEW_ADDR && 225 f2fs_bug_on(nat_get_blkaddr(e) == NEW_ADDR &&
226 new_blkaddr == NEW_ADDR); 226 new_blkaddr == NEW_ADDR);
227 BUG_ON(nat_get_blkaddr(e) != NEW_ADDR && 227 f2fs_bug_on(nat_get_blkaddr(e) != NEW_ADDR &&
228 nat_get_blkaddr(e) != NULL_ADDR && 228 nat_get_blkaddr(e) != NULL_ADDR &&
229 new_blkaddr == NEW_ADDR); 229 new_blkaddr == NEW_ADDR);
230 230
@@ -240,7 +240,7 @@ retry:
240 write_unlock(&nm_i->nat_tree_lock); 240 write_unlock(&nm_i->nat_tree_lock);
241} 241}
242 242
243static int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink) 243int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink)
244{ 244{
245 struct f2fs_nm_info *nm_i = NM_I(sbi); 245 struct f2fs_nm_info *nm_i = NM_I(sbi);
246 246
@@ -495,10 +495,10 @@ static void truncate_node(struct dnode_of_data *dn)
495 495
496 get_node_info(sbi, dn->nid, &ni); 496 get_node_info(sbi, dn->nid, &ni);
497 if (dn->inode->i_blocks == 0) { 497 if (dn->inode->i_blocks == 0) {
498 BUG_ON(ni.blk_addr != NULL_ADDR); 498 f2fs_bug_on(ni.blk_addr != NULL_ADDR);
499 goto invalidate; 499 goto invalidate;
500 } 500 }
501 BUG_ON(ni.blk_addr == NULL_ADDR); 501 f2fs_bug_on(ni.blk_addr == NULL_ADDR);
502 502
503 /* Deallocate node address */ 503 /* Deallocate node address */
504 invalidate_blocks(sbi, ni.blk_addr); 504 invalidate_blocks(sbi, ni.blk_addr);
@@ -822,7 +822,7 @@ int remove_inode_page(struct inode *inode)
822 } 822 }
823 823
824 /* 0 is possible, after f2fs_new_inode() is failed */ 824 /* 0 is possible, after f2fs_new_inode() is failed */
825 BUG_ON(inode->i_blocks != 0 && inode->i_blocks != 1); 825 f2fs_bug_on(inode->i_blocks != 0 && inode->i_blocks != 1);
826 set_new_dnode(&dn, inode, page, page, ino); 826 set_new_dnode(&dn, inode, page, page, ino);
827 truncate_node(&dn); 827 truncate_node(&dn);
828 return 0; 828 return 0;
@@ -863,7 +863,7 @@ struct page *new_node_page(struct dnode_of_data *dn,
863 get_node_info(sbi, dn->nid, &old_ni); 863 get_node_info(sbi, dn->nid, &old_ni);
864 864
865 /* Reinitialize old_ni with new node page */ 865 /* Reinitialize old_ni with new node page */
866 BUG_ON(old_ni.blk_addr != NULL_ADDR); 866 f2fs_bug_on(old_ni.blk_addr != NULL_ADDR);
867 new_ni = old_ni; 867 new_ni = old_ni;
868 new_ni.ino = dn->inode->i_ino; 868 new_ni.ino = dn->inode->i_ino;
869 set_node_addr(sbi, &new_ni, NEW_ADDR); 869 set_node_addr(sbi, &new_ni, NEW_ADDR);
@@ -969,7 +969,7 @@ repeat:
969 goto repeat; 969 goto repeat;
970 } 970 }
971got_it: 971got_it:
972 BUG_ON(nid != nid_of_node(page)); 972 f2fs_bug_on(nid != nid_of_node(page));
973 mark_page_accessed(page); 973 mark_page_accessed(page);
974 return page; 974 return page;
975} 975}
@@ -1148,6 +1148,47 @@ continue_unlock:
1148 return nwritten; 1148 return nwritten;
1149} 1149}
1150 1150
1151int wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, nid_t ino)
1152{
1153 struct address_space *mapping = sbi->node_inode->i_mapping;
1154 pgoff_t index = 0, end = LONG_MAX;
1155 struct pagevec pvec;
1156 int nr_pages;
1157 int ret2 = 0, ret = 0;
1158
1159 pagevec_init(&pvec, 0);
1160 while ((index <= end) &&
1161 (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
1162 PAGECACHE_TAG_WRITEBACK,
1163 min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1)) != 0) {
1164 unsigned i;
1165
1166 for (i = 0; i < nr_pages; i++) {
1167 struct page *page = pvec.pages[i];
1168
1169 /* until radix tree lookup accepts end_index */
1170 if (page->index > end)
1171 continue;
1172
1173 if (ino && ino_of_node(page) == ino) {
1174 wait_on_page_writeback(page);
1175 if (TestClearPageError(page))
1176 ret = -EIO;
1177 }
1178 }
1179 pagevec_release(&pvec);
1180 cond_resched();
1181 }
1182
1183 if (test_and_clear_bit(AS_ENOSPC, &mapping->flags))
1184 ret2 = -ENOSPC;
1185 if (test_and_clear_bit(AS_EIO, &mapping->flags))
1186 ret2 = -EIO;
1187 if (!ret)
1188 ret = ret2;
1189 return ret;
1190}
1191
1151static int f2fs_write_node_page(struct page *page, 1192static int f2fs_write_node_page(struct page *page,
1152 struct writeback_control *wbc) 1193 struct writeback_control *wbc)
1153{ 1194{
@@ -1156,11 +1197,14 @@ static int f2fs_write_node_page(struct page *page,
1156 block_t new_addr; 1197 block_t new_addr;
1157 struct node_info ni; 1198 struct node_info ni;
1158 1199
1200 if (sbi->por_doing)
1201 goto redirty_out;
1202
1159 wait_on_page_writeback(page); 1203 wait_on_page_writeback(page);
1160 1204
1161 /* get old block addr of this node page */ 1205 /* get old block addr of this node page */
1162 nid = nid_of_node(page); 1206 nid = nid_of_node(page);
1163 BUG_ON(page->index != nid); 1207 f2fs_bug_on(page->index != nid);
1164 1208
1165 get_node_info(sbi, nid, &ni); 1209 get_node_info(sbi, nid, &ni);
1166 1210
@@ -1171,12 +1215,8 @@ static int f2fs_write_node_page(struct page *page,
1171 return 0; 1215 return 0;
1172 } 1216 }
1173 1217
1174 if (wbc->for_reclaim) { 1218 if (wbc->for_reclaim)
1175 dec_page_count(sbi, F2FS_DIRTY_NODES); 1219 goto redirty_out;
1176 wbc->pages_skipped++;
1177 set_page_dirty(page);
1178 return AOP_WRITEPAGE_ACTIVATE;
1179 }
1180 1220
1181 mutex_lock(&sbi->node_write); 1221 mutex_lock(&sbi->node_write);
1182 set_page_writeback(page); 1222 set_page_writeback(page);
@@ -1186,6 +1226,12 @@ static int f2fs_write_node_page(struct page *page,
1186 mutex_unlock(&sbi->node_write); 1226 mutex_unlock(&sbi->node_write);
1187 unlock_page(page); 1227 unlock_page(page);
1188 return 0; 1228 return 0;
1229
1230redirty_out:
1231 dec_page_count(sbi, F2FS_DIRTY_NODES);
1232 wbc->pages_skipped++;
1233 set_page_dirty(page);
1234 return AOP_WRITEPAGE_ACTIVATE;
1189} 1235}
1190 1236
1191/* 1237/*
@@ -1200,11 +1246,8 @@ static int f2fs_write_node_pages(struct address_space *mapping,
1200 struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb); 1246 struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb);
1201 long nr_to_write = wbc->nr_to_write; 1247 long nr_to_write = wbc->nr_to_write;
1202 1248
1203 /* First check balancing cached NAT entries */ 1249 /* balancing f2fs's metadata in background */
1204 if (try_to_free_nats(sbi, NAT_ENTRY_PER_BLOCK)) { 1250 f2fs_balance_fs_bg(sbi);
1205 f2fs_sync_fs(sbi->sb, true);
1206 return 0;
1207 }
1208 1251
1209 /* collect a number of dirty node pages and write together */ 1252 /* collect a number of dirty node pages and write together */
1210 if (get_pages(sbi, F2FS_DIRTY_NODES) < COLLECT_DIRTY_NODES) 1253 if (get_pages(sbi, F2FS_DIRTY_NODES) < COLLECT_DIRTY_NODES)
@@ -1223,6 +1266,8 @@ static int f2fs_set_node_page_dirty(struct page *page)
1223 struct address_space *mapping = page->mapping; 1266 struct address_space *mapping = page->mapping;
1224 struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb); 1267 struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb);
1225 1268
1269 trace_f2fs_set_page_dirty(page, NODE);
1270
1226 SetPageUptodate(page); 1271 SetPageUptodate(page);
1227 if (!PageDirty(page)) { 1272 if (!PageDirty(page)) {
1228 __set_page_dirty_nobuffers(page); 1273 __set_page_dirty_nobuffers(page);
@@ -1291,23 +1336,18 @@ static int add_free_nid(struct f2fs_nm_info *nm_i, nid_t nid, bool build)
1291 if (nid == 0) 1336 if (nid == 0)
1292 return 0; 1337 return 0;
1293 1338
1294 if (!build) 1339 if (build) {
1295 goto retry; 1340 /* do not add allocated nids */
1296 1341 read_lock(&nm_i->nat_tree_lock);
1297 /* do not add allocated nids */ 1342 ne = __lookup_nat_cache(nm_i, nid);
1298 read_lock(&nm_i->nat_tree_lock); 1343 if (ne && nat_get_blkaddr(ne) != NULL_ADDR)
1299 ne = __lookup_nat_cache(nm_i, nid); 1344 allocated = true;
1300 if (ne && nat_get_blkaddr(ne) != NULL_ADDR) 1345 read_unlock(&nm_i->nat_tree_lock);
1301 allocated = true; 1346 if (allocated)
1302 read_unlock(&nm_i->nat_tree_lock); 1347 return 0;
1303 if (allocated)
1304 return 0;
1305retry:
1306 i = kmem_cache_alloc(free_nid_slab, GFP_NOFS);
1307 if (!i) {
1308 cond_resched();
1309 goto retry;
1310 } 1348 }
1349
1350 i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS);
1311 i->nid = nid; 1351 i->nid = nid;
1312 i->state = NID_NEW; 1352 i->state = NID_NEW;
1313 1353
@@ -1350,7 +1390,7 @@ static void scan_nat_page(struct f2fs_nm_info *nm_i,
1350 break; 1390 break;
1351 1391
1352 blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr); 1392 blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr);
1353 BUG_ON(blk_addr == NEW_ADDR); 1393 f2fs_bug_on(blk_addr == NEW_ADDR);
1354 if (blk_addr == NULL_ADDR) { 1394 if (blk_addr == NULL_ADDR) {
1355 if (add_free_nid(nm_i, start_nid, true) < 0) 1395 if (add_free_nid(nm_i, start_nid, true) < 0)
1356 break; 1396 break;
@@ -1421,14 +1461,14 @@ retry:
1421 1461
1422 /* We should not use stale free nids created by build_free_nids */ 1462 /* We should not use stale free nids created by build_free_nids */
1423 if (nm_i->fcnt && !sbi->on_build_free_nids) { 1463 if (nm_i->fcnt && !sbi->on_build_free_nids) {
1424 BUG_ON(list_empty(&nm_i->free_nid_list)); 1464 f2fs_bug_on(list_empty(&nm_i->free_nid_list));
1425 list_for_each(this, &nm_i->free_nid_list) { 1465 list_for_each(this, &nm_i->free_nid_list) {
1426 i = list_entry(this, struct free_nid, list); 1466 i = list_entry(this, struct free_nid, list);
1427 if (i->state == NID_NEW) 1467 if (i->state == NID_NEW)
1428 break; 1468 break;
1429 } 1469 }
1430 1470
1431 BUG_ON(i->state != NID_NEW); 1471 f2fs_bug_on(i->state != NID_NEW);
1432 *nid = i->nid; 1472 *nid = i->nid;
1433 i->state = NID_ALLOC; 1473 i->state = NID_ALLOC;
1434 nm_i->fcnt--; 1474 nm_i->fcnt--;
@@ -1439,9 +1479,9 @@ retry:
1439 1479
1440 /* Let's scan nat pages and its caches to get free nids */ 1480 /* Let's scan nat pages and its caches to get free nids */
1441 mutex_lock(&nm_i->build_lock); 1481 mutex_lock(&nm_i->build_lock);
1442 sbi->on_build_free_nids = 1; 1482 sbi->on_build_free_nids = true;
1443 build_free_nids(sbi); 1483 build_free_nids(sbi);
1444 sbi->on_build_free_nids = 0; 1484 sbi->on_build_free_nids = false;
1445 mutex_unlock(&nm_i->build_lock); 1485 mutex_unlock(&nm_i->build_lock);
1446 goto retry; 1486 goto retry;
1447} 1487}
@@ -1456,7 +1496,7 @@ void alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid)
1456 1496
1457 spin_lock(&nm_i->free_nid_list_lock); 1497 spin_lock(&nm_i->free_nid_list_lock);
1458 i = __lookup_free_nid_list(nid, &nm_i->free_nid_list); 1498 i = __lookup_free_nid_list(nid, &nm_i->free_nid_list);
1459 BUG_ON(!i || i->state != NID_ALLOC); 1499 f2fs_bug_on(!i || i->state != NID_ALLOC);
1460 __del_from_free_nid_list(i); 1500 __del_from_free_nid_list(i);
1461 spin_unlock(&nm_i->free_nid_list_lock); 1501 spin_unlock(&nm_i->free_nid_list_lock);
1462} 1502}
@@ -1474,7 +1514,7 @@ void alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid)
1474 1514
1475 spin_lock(&nm_i->free_nid_list_lock); 1515 spin_lock(&nm_i->free_nid_list_lock);
1476 i = __lookup_free_nid_list(nid, &nm_i->free_nid_list); 1516 i = __lookup_free_nid_list(nid, &nm_i->free_nid_list);
1477 BUG_ON(!i || i->state != NID_ALLOC); 1517 f2fs_bug_on(!i || i->state != NID_ALLOC);
1478 if (nm_i->fcnt > 2 * MAX_FREE_NIDS) { 1518 if (nm_i->fcnt > 2 * MAX_FREE_NIDS) {
1479 __del_from_free_nid_list(i); 1519 __del_from_free_nid_list(i);
1480 } else { 1520 } else {
@@ -1677,7 +1717,7 @@ to_nat_page:
1677 nat_blk = page_address(page); 1717 nat_blk = page_address(page);
1678 } 1718 }
1679 1719
1680 BUG_ON(!nat_blk); 1720 f2fs_bug_on(!nat_blk);
1681 raw_ne = nat_blk->entries[nid - start_nid]; 1721 raw_ne = nat_blk->entries[nid - start_nid];
1682flush_now: 1722flush_now:
1683 new_blkaddr = nat_get_blkaddr(ne); 1723 new_blkaddr = nat_get_blkaddr(ne);
@@ -1781,11 +1821,11 @@ void destroy_node_manager(struct f2fs_sb_info *sbi)
1781 /* destroy free nid list */ 1821 /* destroy free nid list */
1782 spin_lock(&nm_i->free_nid_list_lock); 1822 spin_lock(&nm_i->free_nid_list_lock);
1783 list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) { 1823 list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) {
1784 BUG_ON(i->state == NID_ALLOC); 1824 f2fs_bug_on(i->state == NID_ALLOC);
1785 __del_from_free_nid_list(i); 1825 __del_from_free_nid_list(i);
1786 nm_i->fcnt--; 1826 nm_i->fcnt--;
1787 } 1827 }
1788 BUG_ON(nm_i->fcnt); 1828 f2fs_bug_on(nm_i->fcnt);
1789 spin_unlock(&nm_i->free_nid_list_lock); 1829 spin_unlock(&nm_i->free_nid_list_lock);
1790 1830
1791 /* destroy nat cache */ 1831 /* destroy nat cache */
@@ -1799,7 +1839,7 @@ void destroy_node_manager(struct f2fs_sb_info *sbi)
1799 __del_from_nat_cache(nm_i, e); 1839 __del_from_nat_cache(nm_i, e);
1800 } 1840 }
1801 } 1841 }
1802 BUG_ON(nm_i->nat_cnt); 1842 f2fs_bug_on(nm_i->nat_cnt);
1803 write_unlock(&nm_i->nat_tree_lock); 1843 write_unlock(&nm_i->nat_tree_lock);
1804 1844
1805 kfree(nm_i->nat_bitmap); 1845 kfree(nm_i->nat_bitmap);
diff --git a/fs/f2fs/recovery.c b/fs/f2fs/recovery.c
index 51ef5eec33d7..fdc81161f254 100644
--- a/fs/f2fs/recovery.c
+++ b/fs/f2fs/recovery.c
@@ -64,24 +64,31 @@ static int recover_dentry(struct page *ipage, struct inode *inode)
64 name.name = raw_inode->i_name; 64 name.name = raw_inode->i_name;
65retry: 65retry:
66 de = f2fs_find_entry(dir, &name, &page); 66 de = f2fs_find_entry(dir, &name, &page);
67 if (de && inode->i_ino == le32_to_cpu(de->ino)) { 67 if (de && inode->i_ino == le32_to_cpu(de->ino))
68 kunmap(page); 68 goto out_unmap_put;
69 f2fs_put_page(page, 0);
70 goto out;
71 }
72 if (de) { 69 if (de) {
73 einode = f2fs_iget(inode->i_sb, le32_to_cpu(de->ino)); 70 einode = f2fs_iget(inode->i_sb, le32_to_cpu(de->ino));
74 if (IS_ERR(einode)) { 71 if (IS_ERR(einode)) {
75 WARN_ON(1); 72 WARN_ON(1);
76 if (PTR_ERR(einode) == -ENOENT) 73 if (PTR_ERR(einode) == -ENOENT)
77 err = -EEXIST; 74 err = -EEXIST;
78 goto out; 75 goto out_unmap_put;
76 }
77 err = acquire_orphan_inode(F2FS_SB(inode->i_sb));
78 if (err) {
79 iput(einode);
80 goto out_unmap_put;
79 } 81 }
80 f2fs_delete_entry(de, page, einode); 82 f2fs_delete_entry(de, page, einode);
81 iput(einode); 83 iput(einode);
82 goto retry; 84 goto retry;
83 } 85 }
84 err = __f2fs_add_link(dir, &name, inode); 86 err = __f2fs_add_link(dir, &name, inode);
87 goto out;
88
89out_unmap_put:
90 kunmap(page);
91 f2fs_put_page(page, 0);
85out: 92out:
86 f2fs_msg(inode->i_sb, KERN_NOTICE, "recover_inode and its dentry: " 93 f2fs_msg(inode->i_sb, KERN_NOTICE, "recover_inode and its dentry: "
87 "ino = %x, name = %s, dir = %lx, err = %d", 94 "ino = %x, name = %s, dir = %lx, err = %d",
@@ -285,7 +292,6 @@ static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode,
285 struct f2fs_summary sum; 292 struct f2fs_summary sum;
286 struct node_info ni; 293 struct node_info ni;
287 int err = 0, recovered = 0; 294 int err = 0, recovered = 0;
288 int ilock;
289 295
290 start = start_bidx_of_node(ofs_of_node(page), fi); 296 start = start_bidx_of_node(ofs_of_node(page), fi);
291 if (IS_INODE(page)) 297 if (IS_INODE(page))
@@ -293,20 +299,20 @@ static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode,
293 else 299 else
294 end = start + ADDRS_PER_BLOCK; 300 end = start + ADDRS_PER_BLOCK;
295 301
296 ilock = mutex_lock_op(sbi); 302 f2fs_lock_op(sbi);
297 set_new_dnode(&dn, inode, NULL, NULL, 0); 303 set_new_dnode(&dn, inode, NULL, NULL, 0);
298 304
299 err = get_dnode_of_data(&dn, start, ALLOC_NODE); 305 err = get_dnode_of_data(&dn, start, ALLOC_NODE);
300 if (err) { 306 if (err) {
301 mutex_unlock_op(sbi, ilock); 307 f2fs_unlock_op(sbi);
302 return err; 308 return err;
303 } 309 }
304 310
305 wait_on_page_writeback(dn.node_page); 311 wait_on_page_writeback(dn.node_page);
306 312
307 get_node_info(sbi, dn.nid, &ni); 313 get_node_info(sbi, dn.nid, &ni);
308 BUG_ON(ni.ino != ino_of_node(page)); 314 f2fs_bug_on(ni.ino != ino_of_node(page));
309 BUG_ON(ofs_of_node(dn.node_page) != ofs_of_node(page)); 315 f2fs_bug_on(ofs_of_node(dn.node_page) != ofs_of_node(page));
310 316
311 for (; start < end; start++) { 317 for (; start < end; start++) {
312 block_t src, dest; 318 block_t src, dest;
@@ -316,9 +322,9 @@ static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode,
316 322
317 if (src != dest && dest != NEW_ADDR && dest != NULL_ADDR) { 323 if (src != dest && dest != NEW_ADDR && dest != NULL_ADDR) {
318 if (src == NULL_ADDR) { 324 if (src == NULL_ADDR) {
319 int err = reserve_new_block(&dn); 325 err = reserve_new_block(&dn);
320 /* We should not get -ENOSPC */ 326 /* We should not get -ENOSPC */
321 BUG_ON(err); 327 f2fs_bug_on(err);
322 } 328 }
323 329
324 /* Check the previous node page having this index */ 330 /* Check the previous node page having this index */
@@ -349,7 +355,7 @@ static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode,
349 recover_node_page(sbi, dn.node_page, &sum, &ni, blkaddr); 355 recover_node_page(sbi, dn.node_page, &sum, &ni, blkaddr);
350err: 356err:
351 f2fs_put_dnode(&dn); 357 f2fs_put_dnode(&dn);
352 mutex_unlock_op(sbi, ilock); 358 f2fs_unlock_op(sbi);
353 359
354 f2fs_msg(sbi->sb, KERN_NOTICE, "recover_data: ino = %lx, " 360 f2fs_msg(sbi->sb, KERN_NOTICE, "recover_data: ino = %lx, "
355 "recovered_data = %d blocks, err = %d", 361 "recovered_data = %d blocks, err = %d",
@@ -419,6 +425,7 @@ int recover_fsync_data(struct f2fs_sb_info *sbi)
419{ 425{
420 struct list_head inode_list; 426 struct list_head inode_list;
421 int err; 427 int err;
428 bool need_writecp = false;
422 429
423 fsync_entry_slab = f2fs_kmem_cache_create("f2fs_fsync_inode_entry", 430 fsync_entry_slab = f2fs_kmem_cache_create("f2fs_fsync_inode_entry",
424 sizeof(struct fsync_inode_entry), NULL); 431 sizeof(struct fsync_inode_entry), NULL);
@@ -428,7 +435,7 @@ int recover_fsync_data(struct f2fs_sb_info *sbi)
428 INIT_LIST_HEAD(&inode_list); 435 INIT_LIST_HEAD(&inode_list);
429 436
430 /* step #1: find fsynced inode numbers */ 437 /* step #1: find fsynced inode numbers */
431 sbi->por_doing = 1; 438 sbi->por_doing = true;
432 err = find_fsync_dnodes(sbi, &inode_list); 439 err = find_fsync_dnodes(sbi, &inode_list);
433 if (err) 440 if (err)
434 goto out; 441 goto out;
@@ -436,14 +443,16 @@ int recover_fsync_data(struct f2fs_sb_info *sbi)
436 if (list_empty(&inode_list)) 443 if (list_empty(&inode_list))
437 goto out; 444 goto out;
438 445
446 need_writecp = true;
447
439 /* step #2: recover data */ 448 /* step #2: recover data */
440 err = recover_data(sbi, &inode_list, CURSEG_WARM_NODE); 449 err = recover_data(sbi, &inode_list, CURSEG_WARM_NODE);
441 BUG_ON(!list_empty(&inode_list)); 450 f2fs_bug_on(!list_empty(&inode_list));
442out: 451out:
443 destroy_fsync_dnodes(&inode_list); 452 destroy_fsync_dnodes(&inode_list);
444 kmem_cache_destroy(fsync_entry_slab); 453 kmem_cache_destroy(fsync_entry_slab);
445 sbi->por_doing = 0; 454 sbi->por_doing = false;
446 if (!err) 455 if (!err && need_writecp)
447 write_checkpoint(sbi, false); 456 write_checkpoint(sbi, false);
448 return err; 457 return err;
449} 458}
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index 09af9c7b0f52..fa284d397199 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -36,6 +36,14 @@ void f2fs_balance_fs(struct f2fs_sb_info *sbi)
36 } 36 }
37} 37}
38 38
39void f2fs_balance_fs_bg(struct f2fs_sb_info *sbi)
40{
41 /* check the # of cached NAT entries and prefree segments */
42 if (try_to_free_nats(sbi, NAT_ENTRY_PER_BLOCK) ||
43 excess_prefree_segs(sbi))
44 f2fs_sync_fs(sbi->sb, true);
45}
46
39static void __locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno, 47static void __locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
40 enum dirty_type dirty_type) 48 enum dirty_type dirty_type)
41{ 49{
@@ -50,20 +58,10 @@ static void __locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
50 58
51 if (dirty_type == DIRTY) { 59 if (dirty_type == DIRTY) {
52 struct seg_entry *sentry = get_seg_entry(sbi, segno); 60 struct seg_entry *sentry = get_seg_entry(sbi, segno);
53 enum dirty_type t = DIRTY_HOT_DATA; 61 enum dirty_type t = sentry->type;
54
55 dirty_type = sentry->type;
56
57 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[dirty_type]))
58 dirty_i->nr_dirty[dirty_type]++;
59 62
60 /* Only one bitmap should be set */ 63 if (!test_and_set_bit(segno, dirty_i->dirty_segmap[t]))
61 for (; t <= DIRTY_COLD_NODE; t++) { 64 dirty_i->nr_dirty[t]++;
62 if (t == dirty_type)
63 continue;
64 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[t]))
65 dirty_i->nr_dirty[t]--;
66 }
67 } 65 }
68} 66}
69 67
@@ -76,12 +74,11 @@ static void __remove_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
76 dirty_i->nr_dirty[dirty_type]--; 74 dirty_i->nr_dirty[dirty_type]--;
77 75
78 if (dirty_type == DIRTY) { 76 if (dirty_type == DIRTY) {
79 enum dirty_type t = DIRTY_HOT_DATA; 77 struct seg_entry *sentry = get_seg_entry(sbi, segno);
78 enum dirty_type t = sentry->type;
80 79
81 /* clear all the bitmaps */ 80 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[t]))
82 for (; t <= DIRTY_COLD_NODE; t++) 81 dirty_i->nr_dirty[t]--;
83 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[t]))
84 dirty_i->nr_dirty[t]--;
85 82
86 if (get_valid_blocks(sbi, segno, sbi->segs_per_sec) == 0) 83 if (get_valid_blocks(sbi, segno, sbi->segs_per_sec) == 0)
87 clear_bit(GET_SECNO(sbi, segno), 84 clear_bit(GET_SECNO(sbi, segno),
@@ -142,27 +139,33 @@ static void set_prefree_as_free_segments(struct f2fs_sb_info *sbi)
142void clear_prefree_segments(struct f2fs_sb_info *sbi) 139void clear_prefree_segments(struct f2fs_sb_info *sbi)
143{ 140{
144 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi); 141 struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
145 unsigned int segno = -1; 142 unsigned long *prefree_map = dirty_i->dirty_segmap[PRE];
146 unsigned int total_segs = TOTAL_SEGS(sbi); 143 unsigned int total_segs = TOTAL_SEGS(sbi);
144 unsigned int start = 0, end = -1;
147 145
148 mutex_lock(&dirty_i->seglist_lock); 146 mutex_lock(&dirty_i->seglist_lock);
147
149 while (1) { 148 while (1) {
150 segno = find_next_bit(dirty_i->dirty_segmap[PRE], total_segs, 149 int i;
151 segno + 1); 150 start = find_next_bit(prefree_map, total_segs, end + 1);
152 if (segno >= total_segs) 151 if (start >= total_segs)
153 break; 152 break;
153 end = find_next_zero_bit(prefree_map, total_segs, start + 1);
154
155 for (i = start; i < end; i++)
156 clear_bit(i, prefree_map);
154 157
155 if (test_and_clear_bit(segno, dirty_i->dirty_segmap[PRE])) 158 dirty_i->nr_dirty[PRE] -= end - start;
156 dirty_i->nr_dirty[PRE]--; 159
157 160 if (!test_opt(sbi, DISCARD))
158 /* Let's use trim */ 161 continue;
159 if (test_opt(sbi, DISCARD)) 162
160 blkdev_issue_discard(sbi->sb->s_bdev, 163 blkdev_issue_discard(sbi->sb->s_bdev,
161 START_BLOCK(sbi, segno) << 164 START_BLOCK(sbi, start) <<
162 sbi->log_sectors_per_block, 165 sbi->log_sectors_per_block,
163 1 << (sbi->log_sectors_per_block + 166 (1 << (sbi->log_sectors_per_block +
164 sbi->log_blocks_per_seg), 167 sbi->log_blocks_per_seg)) * (end - start),
165 GFP_NOFS, 0); 168 GFP_NOFS, 0);
166 } 169 }
167 mutex_unlock(&dirty_i->seglist_lock); 170 mutex_unlock(&dirty_i->seglist_lock);
168} 171}
@@ -195,7 +198,7 @@ static void update_sit_entry(struct f2fs_sb_info *sbi, block_t blkaddr, int del)
195 new_vblocks = se->valid_blocks + del; 198 new_vblocks = se->valid_blocks + del;
196 offset = GET_SEGOFF_FROM_SEG0(sbi, blkaddr) & (sbi->blocks_per_seg - 1); 199 offset = GET_SEGOFF_FROM_SEG0(sbi, blkaddr) & (sbi->blocks_per_seg - 1);
197 200
198 BUG_ON((new_vblocks >> (sizeof(unsigned short) << 3) || 201 f2fs_bug_on((new_vblocks >> (sizeof(unsigned short) << 3) ||
199 (new_vblocks > sbi->blocks_per_seg))); 202 (new_vblocks > sbi->blocks_per_seg)));
200 203
201 se->valid_blocks = new_vblocks; 204 se->valid_blocks = new_vblocks;
@@ -235,7 +238,7 @@ void invalidate_blocks(struct f2fs_sb_info *sbi, block_t addr)
235 unsigned int segno = GET_SEGNO(sbi, addr); 238 unsigned int segno = GET_SEGNO(sbi, addr);
236 struct sit_info *sit_i = SIT_I(sbi); 239 struct sit_info *sit_i = SIT_I(sbi);
237 240
238 BUG_ON(addr == NULL_ADDR); 241 f2fs_bug_on(addr == NULL_ADDR);
239 if (addr == NEW_ADDR) 242 if (addr == NEW_ADDR)
240 return; 243 return;
241 244
@@ -267,9 +270,8 @@ static void __add_sum_entry(struct f2fs_sb_info *sbi, int type,
267 */ 270 */
268int npages_for_summary_flush(struct f2fs_sb_info *sbi) 271int npages_for_summary_flush(struct f2fs_sb_info *sbi)
269{ 272{
270 int total_size_bytes = 0;
271 int valid_sum_count = 0; 273 int valid_sum_count = 0;
272 int i, sum_space; 274 int i, sum_in_page;
273 275
274 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) { 276 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
275 if (sbi->ckpt->alloc_type[i] == SSR) 277 if (sbi->ckpt->alloc_type[i] == SSR)
@@ -278,13 +280,12 @@ int npages_for_summary_flush(struct f2fs_sb_info *sbi)
278 valid_sum_count += curseg_blkoff(sbi, i); 280 valid_sum_count += curseg_blkoff(sbi, i);
279 } 281 }
280 282
281 total_size_bytes = valid_sum_count * (SUMMARY_SIZE + 1) 283 sum_in_page = (PAGE_CACHE_SIZE - 2 * SUM_JOURNAL_SIZE -
282 + sizeof(struct nat_journal) + 2 284 SUM_FOOTER_SIZE) / SUMMARY_SIZE;
283 + sizeof(struct sit_journal) + 2; 285 if (valid_sum_count <= sum_in_page)
284 sum_space = PAGE_CACHE_SIZE - SUM_FOOTER_SIZE;
285 if (total_size_bytes < sum_space)
286 return 1; 286 return 1;
287 else if (total_size_bytes < 2 * sum_space) 287 else if ((valid_sum_count - sum_in_page) <=
288 (PAGE_CACHE_SIZE - SUM_FOOTER_SIZE) / SUMMARY_SIZE)
288 return 2; 289 return 2;
289 return 3; 290 return 3;
290} 291}
@@ -350,7 +351,7 @@ find_other_zone:
350 if (dir == ALLOC_RIGHT) { 351 if (dir == ALLOC_RIGHT) {
351 secno = find_next_zero_bit(free_i->free_secmap, 352 secno = find_next_zero_bit(free_i->free_secmap,
352 TOTAL_SECS(sbi), 0); 353 TOTAL_SECS(sbi), 0);
353 BUG_ON(secno >= TOTAL_SECS(sbi)); 354 f2fs_bug_on(secno >= TOTAL_SECS(sbi));
354 } else { 355 } else {
355 go_left = 1; 356 go_left = 1;
356 left_start = hint - 1; 357 left_start = hint - 1;
@@ -366,7 +367,7 @@ find_other_zone:
366 } 367 }
367 left_start = find_next_zero_bit(free_i->free_secmap, 368 left_start = find_next_zero_bit(free_i->free_secmap,
368 TOTAL_SECS(sbi), 0); 369 TOTAL_SECS(sbi), 0);
369 BUG_ON(left_start >= TOTAL_SECS(sbi)); 370 f2fs_bug_on(left_start >= TOTAL_SECS(sbi));
370 break; 371 break;
371 } 372 }
372 secno = left_start; 373 secno = left_start;
@@ -405,7 +406,7 @@ skip_left:
405 } 406 }
406got_it: 407got_it:
407 /* set it as dirty segment in free segmap */ 408 /* set it as dirty segment in free segmap */
408 BUG_ON(test_bit(segno, free_i->free_segmap)); 409 f2fs_bug_on(test_bit(segno, free_i->free_segmap));
409 __set_inuse(sbi, segno); 410 __set_inuse(sbi, segno);
410 *newseg = segno; 411 *newseg = segno;
411 write_unlock(&free_i->segmap_lock); 412 write_unlock(&free_i->segmap_lock);
@@ -550,9 +551,8 @@ static void allocate_segment_by_default(struct f2fs_sb_info *sbi,
550 change_curseg(sbi, type, true); 551 change_curseg(sbi, type, true);
551 else 552 else
552 new_curseg(sbi, type, false); 553 new_curseg(sbi, type, false);
553#ifdef CONFIG_F2FS_STAT_FS 554
554 sbi->segment_count[curseg->alloc_type]++; 555 stat_inc_seg_type(sbi, curseg);
555#endif
556} 556}
557 557
558void allocate_new_segments(struct f2fs_sb_info *sbi) 558void allocate_new_segments(struct f2fs_sb_info *sbi)
@@ -597,6 +597,11 @@ static void f2fs_end_io_write(struct bio *bio, int err)
597 597
598 if (p->is_sync) 598 if (p->is_sync)
599 complete(p->wait); 599 complete(p->wait);
600
601 if (!get_pages(p->sbi, F2FS_WRITEBACK) &&
602 !list_empty(&p->sbi->cp_wait.task_list))
603 wake_up(&p->sbi->cp_wait);
604
600 kfree(p); 605 kfree(p);
601 bio_put(bio); 606 bio_put(bio);
602} 607}
@@ -657,6 +662,7 @@ static void submit_write_page(struct f2fs_sb_info *sbi, struct page *page,
657 block_t blk_addr, enum page_type type) 662 block_t blk_addr, enum page_type type)
658{ 663{
659 struct block_device *bdev = sbi->sb->s_bdev; 664 struct block_device *bdev = sbi->sb->s_bdev;
665 int bio_blocks;
660 666
661 verify_block_addr(sbi, blk_addr); 667 verify_block_addr(sbi, blk_addr);
662 668
@@ -676,7 +682,8 @@ retry:
676 goto retry; 682 goto retry;
677 } 683 }
678 684
679 sbi->bio[type] = f2fs_bio_alloc(bdev, max_hw_blocks(sbi)); 685 bio_blocks = MAX_BIO_BLOCKS(max_hw_blocks(sbi));
686 sbi->bio[type] = f2fs_bio_alloc(bdev, bio_blocks);
680 sbi->bio[type]->bi_sector = SECTOR_FROM_BLOCK(sbi, blk_addr); 687 sbi->bio[type]->bi_sector = SECTOR_FROM_BLOCK(sbi, blk_addr);
681 sbi->bio[type]->bi_private = priv; 688 sbi->bio[type]->bi_private = priv;
682 /* 689 /*
@@ -771,7 +778,7 @@ static int __get_segment_type(struct page *page, enum page_type p_type)
771 return __get_segment_type_4(page, p_type); 778 return __get_segment_type_4(page, p_type);
772 } 779 }
773 /* NR_CURSEG_TYPE(6) logs by default */ 780 /* NR_CURSEG_TYPE(6) logs by default */
774 BUG_ON(sbi->active_logs != NR_CURSEG_TYPE); 781 f2fs_bug_on(sbi->active_logs != NR_CURSEG_TYPE);
775 return __get_segment_type_6(page, p_type); 782 return __get_segment_type_6(page, p_type);
776} 783}
777 784
@@ -801,9 +808,8 @@ static void do_write_page(struct f2fs_sb_info *sbi, struct page *page,
801 808
802 mutex_lock(&sit_i->sentry_lock); 809 mutex_lock(&sit_i->sentry_lock);
803 __refresh_next_blkoff(sbi, curseg); 810 __refresh_next_blkoff(sbi, curseg);
804#ifdef CONFIG_F2FS_STAT_FS 811
805 sbi->block_count[curseg->alloc_type]++; 812 stat_inc_block_count(sbi, curseg);
806#endif
807 813
808 /* 814 /*
809 * SIT information should be updated before segment allocation, 815 * SIT information should be updated before segment allocation,
@@ -849,7 +855,7 @@ void write_data_page(struct inode *inode, struct page *page,
849 struct f2fs_summary sum; 855 struct f2fs_summary sum;
850 struct node_info ni; 856 struct node_info ni;
851 857
852 BUG_ON(old_blkaddr == NULL_ADDR); 858 f2fs_bug_on(old_blkaddr == NULL_ADDR);
853 get_node_info(sbi, dn->nid, &ni); 859 get_node_info(sbi, dn->nid, &ni);
854 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version); 860 set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
855 861
@@ -1122,8 +1128,6 @@ static void write_compacted_summaries(struct f2fs_sb_info *sbi, block_t blkaddr)
1122 SUM_JOURNAL_SIZE); 1128 SUM_JOURNAL_SIZE);
1123 written_size += SUM_JOURNAL_SIZE; 1129 written_size += SUM_JOURNAL_SIZE;
1124 1130
1125 set_page_dirty(page);
1126
1127 /* Step 3: write summary entries */ 1131 /* Step 3: write summary entries */
1128 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) { 1132 for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
1129 unsigned short blkoff; 1133 unsigned short blkoff;
@@ -1142,18 +1146,20 @@ static void write_compacted_summaries(struct f2fs_sb_info *sbi, block_t blkaddr)
1142 summary = (struct f2fs_summary *)(kaddr + written_size); 1146 summary = (struct f2fs_summary *)(kaddr + written_size);
1143 *summary = seg_i->sum_blk->entries[j]; 1147 *summary = seg_i->sum_blk->entries[j];
1144 written_size += SUMMARY_SIZE; 1148 written_size += SUMMARY_SIZE;
1145 set_page_dirty(page);
1146 1149
1147 if (written_size + SUMMARY_SIZE <= PAGE_CACHE_SIZE - 1150 if (written_size + SUMMARY_SIZE <= PAGE_CACHE_SIZE -
1148 SUM_FOOTER_SIZE) 1151 SUM_FOOTER_SIZE)
1149 continue; 1152 continue;
1150 1153
1154 set_page_dirty(page);
1151 f2fs_put_page(page, 1); 1155 f2fs_put_page(page, 1);
1152 page = NULL; 1156 page = NULL;
1153 } 1157 }
1154 } 1158 }
1155 if (page) 1159 if (page) {
1160 set_page_dirty(page);
1156 f2fs_put_page(page, 1); 1161 f2fs_put_page(page, 1);
1162 }
1157} 1163}
1158 1164
1159static void write_normal_summaries(struct f2fs_sb_info *sbi, 1165static void write_normal_summaries(struct f2fs_sb_info *sbi,
@@ -1239,7 +1245,7 @@ static struct page *get_next_sit_page(struct f2fs_sb_info *sbi,
1239 /* get current sit block page without lock */ 1245 /* get current sit block page without lock */
1240 src_page = get_meta_page(sbi, src_off); 1246 src_page = get_meta_page(sbi, src_off);
1241 dst_page = grab_meta_page(sbi, dst_off); 1247 dst_page = grab_meta_page(sbi, dst_off);
1242 BUG_ON(PageDirty(src_page)); 1248 f2fs_bug_on(PageDirty(src_page));
1243 1249
1244 src_addr = page_address(src_page); 1250 src_addr = page_address(src_page);
1245 dst_addr = page_address(dst_page); 1251 dst_addr = page_address(dst_page);
@@ -1271,9 +1277,9 @@ static bool flush_sits_in_journal(struct f2fs_sb_info *sbi)
1271 __mark_sit_entry_dirty(sbi, segno); 1277 __mark_sit_entry_dirty(sbi, segno);
1272 } 1278 }
1273 update_sits_in_cursum(sum, -sits_in_cursum(sum)); 1279 update_sits_in_cursum(sum, -sits_in_cursum(sum));
1274 return 1; 1280 return true;
1275 } 1281 }
1276 return 0; 1282 return false;
1277} 1283}
1278 1284
1279/* 1285/*
@@ -1637,6 +1643,7 @@ int build_segment_manager(struct f2fs_sb_info *sbi)
1637 sm_info->ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 1643 sm_info->ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
1638 sm_info->main_segments = le32_to_cpu(raw_super->segment_count_main); 1644 sm_info->main_segments = le32_to_cpu(raw_super->segment_count_main);
1639 sm_info->ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 1645 sm_info->ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1646 sm_info->rec_prefree_segments = DEF_RECLAIM_PREFREE_SEGMENTS;
1640 1647
1641 err = build_sit_info(sbi); 1648 err = build_sit_info(sbi);
1642 if (err) 1649 if (err)
@@ -1744,6 +1751,8 @@ static void destroy_sit_info(struct f2fs_sb_info *sbi)
1744void destroy_segment_manager(struct f2fs_sb_info *sbi) 1751void destroy_segment_manager(struct f2fs_sb_info *sbi)
1745{ 1752{
1746 struct f2fs_sm_info *sm_info = SM_I(sbi); 1753 struct f2fs_sm_info *sm_info = SM_I(sbi);
1754 if (!sm_info)
1755 return;
1747 destroy_dirty_segmap(sbi); 1756 destroy_dirty_segmap(sbi);
1748 destroy_curseg(sbi); 1757 destroy_curseg(sbi);
1749 destroy_free_segmap(sbi); 1758 destroy_free_segmap(sbi);
diff --git a/fs/f2fs/segment.h b/fs/f2fs/segment.h
index bdd10eab8c40..269f690b4e24 100644
--- a/fs/f2fs/segment.h
+++ b/fs/f2fs/segment.h
@@ -14,6 +14,8 @@
14#define NULL_SEGNO ((unsigned int)(~0)) 14#define NULL_SEGNO ((unsigned int)(~0))
15#define NULL_SECNO ((unsigned int)(~0)) 15#define NULL_SECNO ((unsigned int)(~0))
16 16
17#define DEF_RECLAIM_PREFREE_SEGMENTS 100 /* 200MB of prefree segments */
18
17/* L: Logical segment # in volume, R: Relative segment # in main area */ 19/* L: Logical segment # in volume, R: Relative segment # in main area */
18#define GET_L2R_SEGNO(free_i, segno) (segno - free_i->start_segno) 20#define GET_L2R_SEGNO(free_i, segno) (segno - free_i->start_segno)
19#define GET_R2L_SEGNO(free_i, segno) (segno + free_i->start_segno) 21#define GET_R2L_SEGNO(free_i, segno) (segno + free_i->start_segno)
@@ -90,6 +92,8 @@
90 (blk_addr << ((sbi)->log_blocksize - F2FS_LOG_SECTOR_SIZE)) 92 (blk_addr << ((sbi)->log_blocksize - F2FS_LOG_SECTOR_SIZE))
91#define SECTOR_TO_BLOCK(sbi, sectors) \ 93#define SECTOR_TO_BLOCK(sbi, sectors) \
92 (sectors >> ((sbi)->log_blocksize - F2FS_LOG_SECTOR_SIZE)) 94 (sectors >> ((sbi)->log_blocksize - F2FS_LOG_SECTOR_SIZE))
95#define MAX_BIO_BLOCKS(max_hw_blocks) \
96 (min((int)max_hw_blocks, BIO_MAX_PAGES))
93 97
94/* during checkpoint, bio_private is used to synchronize the last bio */ 98/* during checkpoint, bio_private is used to synchronize the last bio */
95struct bio_private { 99struct bio_private {
@@ -470,6 +474,11 @@ static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi, int freed)
470 reserved_sections(sbi))); 474 reserved_sections(sbi)));
471} 475}
472 476
477static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
478{
479 return (prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments);
480}
481
473static inline int utilization(struct f2fs_sb_info *sbi) 482static inline int utilization(struct f2fs_sb_info *sbi)
474{ 483{
475 return div_u64((u64)valid_user_blocks(sbi) * 100, sbi->user_block_count); 484 return div_u64((u64)valid_user_blocks(sbi) * 100, sbi->user_block_count);
@@ -513,16 +522,13 @@ static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
513 return curseg->next_blkoff; 522 return curseg->next_blkoff;
514} 523}
515 524
525#ifdef CONFIG_F2FS_CHECK_FS
516static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno) 526static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
517{ 527{
518 unsigned int end_segno = SM_I(sbi)->segment_count - 1; 528 unsigned int end_segno = SM_I(sbi)->segment_count - 1;
519 BUG_ON(segno > end_segno); 529 BUG_ON(segno > end_segno);
520} 530}
521 531
522/*
523 * This function is used for only debugging.
524 * NOTE: In future, we have to remove this function.
525 */
526static inline void verify_block_addr(struct f2fs_sb_info *sbi, block_t blk_addr) 532static inline void verify_block_addr(struct f2fs_sb_info *sbi, block_t blk_addr)
527{ 533{
528 struct f2fs_sm_info *sm_info = SM_I(sbi); 534 struct f2fs_sm_info *sm_info = SM_I(sbi);
@@ -541,8 +547,9 @@ static inline void check_block_count(struct f2fs_sb_info *sbi,
541{ 547{
542 struct f2fs_sm_info *sm_info = SM_I(sbi); 548 struct f2fs_sm_info *sm_info = SM_I(sbi);
543 unsigned int end_segno = sm_info->segment_count - 1; 549 unsigned int end_segno = sm_info->segment_count - 1;
550 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
544 int valid_blocks = 0; 551 int valid_blocks = 0;
545 int i; 552 int cur_pos = 0, next_pos;
546 553
547 /* check segment usage */ 554 /* check segment usage */
548 BUG_ON(GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg); 555 BUG_ON(GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg);
@@ -551,11 +558,26 @@ static inline void check_block_count(struct f2fs_sb_info *sbi,
551 BUG_ON(segno > end_segno); 558 BUG_ON(segno > end_segno);
552 559
553 /* check bitmap with valid block count */ 560 /* check bitmap with valid block count */
554 for (i = 0; i < sbi->blocks_per_seg; i++) 561 do {
555 if (f2fs_test_bit(i, raw_sit->valid_map)) 562 if (is_valid) {
556 valid_blocks++; 563 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
564 sbi->blocks_per_seg,
565 cur_pos);
566 valid_blocks += next_pos - cur_pos;
567 } else
568 next_pos = find_next_bit_le(&raw_sit->valid_map,
569 sbi->blocks_per_seg,
570 cur_pos);
571 cur_pos = next_pos;
572 is_valid = !is_valid;
573 } while (cur_pos < sbi->blocks_per_seg);
557 BUG_ON(GET_SIT_VBLOCKS(raw_sit) != valid_blocks); 574 BUG_ON(GET_SIT_VBLOCKS(raw_sit) != valid_blocks);
558} 575}
576#else
577#define check_seg_range(sbi, segno)
578#define verify_block_addr(sbi, blk_addr)
579#define check_block_count(sbi, segno, raw_sit)
580#endif
559 581
560static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi, 582static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
561 unsigned int start) 583 unsigned int start)
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 13d0a0fe49dd..bafff72de8e8 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -43,7 +43,9 @@ enum {
43 Opt_disable_roll_forward, 43 Opt_disable_roll_forward,
44 Opt_discard, 44 Opt_discard,
45 Opt_noheap, 45 Opt_noheap,
46 Opt_user_xattr,
46 Opt_nouser_xattr, 47 Opt_nouser_xattr,
48 Opt_acl,
47 Opt_noacl, 49 Opt_noacl,
48 Opt_active_logs, 50 Opt_active_logs,
49 Opt_disable_ext_identify, 51 Opt_disable_ext_identify,
@@ -56,7 +58,9 @@ static match_table_t f2fs_tokens = {
56 {Opt_disable_roll_forward, "disable_roll_forward"}, 58 {Opt_disable_roll_forward, "disable_roll_forward"},
57 {Opt_discard, "discard"}, 59 {Opt_discard, "discard"},
58 {Opt_noheap, "no_heap"}, 60 {Opt_noheap, "no_heap"},
61 {Opt_user_xattr, "user_xattr"},
59 {Opt_nouser_xattr, "nouser_xattr"}, 62 {Opt_nouser_xattr, "nouser_xattr"},
63 {Opt_acl, "acl"},
60 {Opt_noacl, "noacl"}, 64 {Opt_noacl, "noacl"},
61 {Opt_active_logs, "active_logs=%u"}, 65 {Opt_active_logs, "active_logs=%u"},
62 {Opt_disable_ext_identify, "disable_ext_identify"}, 66 {Opt_disable_ext_identify, "disable_ext_identify"},
@@ -65,24 +69,40 @@ static match_table_t f2fs_tokens = {
65}; 69};
66 70
67/* Sysfs support for f2fs */ 71/* Sysfs support for f2fs */
72enum {
73 GC_THREAD, /* struct f2fs_gc_thread */
74 SM_INFO, /* struct f2fs_sm_info */
75};
76
68struct f2fs_attr { 77struct f2fs_attr {
69 struct attribute attr; 78 struct attribute attr;
70 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *); 79 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
71 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *, 80 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
72 const char *, size_t); 81 const char *, size_t);
82 int struct_type;
73 int offset; 83 int offset;
74}; 84};
75 85
86static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
87{
88 if (struct_type == GC_THREAD)
89 return (unsigned char *)sbi->gc_thread;
90 else if (struct_type == SM_INFO)
91 return (unsigned char *)SM_I(sbi);
92 return NULL;
93}
94
76static ssize_t f2fs_sbi_show(struct f2fs_attr *a, 95static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
77 struct f2fs_sb_info *sbi, char *buf) 96 struct f2fs_sb_info *sbi, char *buf)
78{ 97{
79 struct f2fs_gc_kthread *gc_kth = sbi->gc_thread; 98 unsigned char *ptr = NULL;
80 unsigned int *ui; 99 unsigned int *ui;
81 100
82 if (!gc_kth) 101 ptr = __struct_ptr(sbi, a->struct_type);
102 if (!ptr)
83 return -EINVAL; 103 return -EINVAL;
84 104
85 ui = (unsigned int *)(((char *)gc_kth) + a->offset); 105 ui = (unsigned int *)(ptr + a->offset);
86 106
87 return snprintf(buf, PAGE_SIZE, "%u\n", *ui); 107 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
88} 108}
@@ -91,15 +111,16 @@ static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
91 struct f2fs_sb_info *sbi, 111 struct f2fs_sb_info *sbi,
92 const char *buf, size_t count) 112 const char *buf, size_t count)
93{ 113{
94 struct f2fs_gc_kthread *gc_kth = sbi->gc_thread; 114 unsigned char *ptr;
95 unsigned long t; 115 unsigned long t;
96 unsigned int *ui; 116 unsigned int *ui;
97 ssize_t ret; 117 ssize_t ret;
98 118
99 if (!gc_kth) 119 ptr = __struct_ptr(sbi, a->struct_type);
120 if (!ptr)
100 return -EINVAL; 121 return -EINVAL;
101 122
102 ui = (unsigned int *)(((char *)gc_kth) + a->offset); 123 ui = (unsigned int *)(ptr + a->offset);
103 124
104 ret = kstrtoul(skip_spaces(buf), 0, &t); 125 ret = kstrtoul(skip_spaces(buf), 0, &t);
105 if (ret < 0) 126 if (ret < 0)
@@ -135,21 +156,25 @@ static void f2fs_sb_release(struct kobject *kobj)
135 complete(&sbi->s_kobj_unregister); 156 complete(&sbi->s_kobj_unregister);
136} 157}
137 158
138#define F2FS_ATTR_OFFSET(_name, _mode, _show, _store, _elname) \ 159#define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
139static struct f2fs_attr f2fs_attr_##_name = { \ 160static struct f2fs_attr f2fs_attr_##_name = { \
140 .attr = {.name = __stringify(_name), .mode = _mode }, \ 161 .attr = {.name = __stringify(_name), .mode = _mode }, \
141 .show = _show, \ 162 .show = _show, \
142 .store = _store, \ 163 .store = _store, \
143 .offset = offsetof(struct f2fs_gc_kthread, _elname), \ 164 .struct_type = _struct_type, \
165 .offset = _offset \
144} 166}
145 167
146#define F2FS_RW_ATTR(name, elname) \ 168#define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
147 F2FS_ATTR_OFFSET(name, 0644, f2fs_sbi_show, f2fs_sbi_store, elname) 169 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
170 f2fs_sbi_show, f2fs_sbi_store, \
171 offsetof(struct struct_name, elname))
148 172
149F2FS_RW_ATTR(gc_min_sleep_time, min_sleep_time); 173F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
150F2FS_RW_ATTR(gc_max_sleep_time, max_sleep_time); 174F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
151F2FS_RW_ATTR(gc_no_gc_sleep_time, no_gc_sleep_time); 175F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
152F2FS_RW_ATTR(gc_idle, gc_idle); 176F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
177F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
153 178
154#define ATTR_LIST(name) (&f2fs_attr_##name.attr) 179#define ATTR_LIST(name) (&f2fs_attr_##name.attr)
155static struct attribute *f2fs_attrs[] = { 180static struct attribute *f2fs_attrs[] = {
@@ -157,6 +182,7 @@ static struct attribute *f2fs_attrs[] = {
157 ATTR_LIST(gc_max_sleep_time), 182 ATTR_LIST(gc_max_sleep_time),
158 ATTR_LIST(gc_no_gc_sleep_time), 183 ATTR_LIST(gc_no_gc_sleep_time),
159 ATTR_LIST(gc_idle), 184 ATTR_LIST(gc_idle),
185 ATTR_LIST(reclaim_segments),
160 NULL, 186 NULL,
161}; 187};
162 188
@@ -237,6 +263,9 @@ static int parse_options(struct super_block *sb, char *options)
237 set_opt(sbi, NOHEAP); 263 set_opt(sbi, NOHEAP);
238 break; 264 break;
239#ifdef CONFIG_F2FS_FS_XATTR 265#ifdef CONFIG_F2FS_FS_XATTR
266 case Opt_user_xattr:
267 set_opt(sbi, XATTR_USER);
268 break;
240 case Opt_nouser_xattr: 269 case Opt_nouser_xattr:
241 clear_opt(sbi, XATTR_USER); 270 clear_opt(sbi, XATTR_USER);
242 break; 271 break;
@@ -244,6 +273,10 @@ static int parse_options(struct super_block *sb, char *options)
244 set_opt(sbi, INLINE_XATTR); 273 set_opt(sbi, INLINE_XATTR);
245 break; 274 break;
246#else 275#else
276 case Opt_user_xattr:
277 f2fs_msg(sb, KERN_INFO,
278 "user_xattr options not supported");
279 break;
247 case Opt_nouser_xattr: 280 case Opt_nouser_xattr:
248 f2fs_msg(sb, KERN_INFO, 281 f2fs_msg(sb, KERN_INFO,
249 "nouser_xattr options not supported"); 282 "nouser_xattr options not supported");
@@ -254,10 +287,16 @@ static int parse_options(struct super_block *sb, char *options)
254 break; 287 break;
255#endif 288#endif
256#ifdef CONFIG_F2FS_FS_POSIX_ACL 289#ifdef CONFIG_F2FS_FS_POSIX_ACL
290 case Opt_acl:
291 set_opt(sbi, POSIX_ACL);
292 break;
257 case Opt_noacl: 293 case Opt_noacl:
258 clear_opt(sbi, POSIX_ACL); 294 clear_opt(sbi, POSIX_ACL);
259 break; 295 break;
260#else 296#else
297 case Opt_acl:
298 f2fs_msg(sb, KERN_INFO, "acl options not supported");
299 break;
261 case Opt_noacl: 300 case Opt_noacl:
262 f2fs_msg(sb, KERN_INFO, "noacl options not supported"); 301 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
263 break; 302 break;
@@ -355,7 +394,9 @@ static void f2fs_put_super(struct super_block *sb)
355 f2fs_destroy_stats(sbi); 394 f2fs_destroy_stats(sbi);
356 stop_gc_thread(sbi); 395 stop_gc_thread(sbi);
357 396
358 write_checkpoint(sbi, true); 397 /* We don't need to do checkpoint when it's clean */
398 if (sbi->s_dirty && get_pages(sbi, F2FS_DIRTY_NODES))
399 write_checkpoint(sbi, true);
359 400
360 iput(sbi->node_inode); 401 iput(sbi->node_inode);
361 iput(sbi->meta_inode); 402 iput(sbi->meta_inode);
@@ -727,30 +768,47 @@ static void init_sb_info(struct f2fs_sb_info *sbi)
727 atomic_set(&sbi->nr_pages[i], 0); 768 atomic_set(&sbi->nr_pages[i], 0);
728} 769}
729 770
730static int validate_superblock(struct super_block *sb, 771/*
731 struct f2fs_super_block **raw_super, 772 * Read f2fs raw super block.
732 struct buffer_head **raw_super_buf, sector_t block) 773 * Because we have two copies of super block, so read the first one at first,
774 * if the first one is invalid, move to read the second one.
775 */
776static int read_raw_super_block(struct super_block *sb,
777 struct f2fs_super_block **raw_super,
778 struct buffer_head **raw_super_buf)
733{ 779{
734 const char *super = (block == 0 ? "first" : "second"); 780 int block = 0;
735 781
736 /* read f2fs raw super block */ 782retry:
737 *raw_super_buf = sb_bread(sb, block); 783 *raw_super_buf = sb_bread(sb, block);
738 if (!*raw_super_buf) { 784 if (!*raw_super_buf) {
739 f2fs_msg(sb, KERN_ERR, "unable to read %s superblock", 785 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
740 super); 786 block + 1);
741 return -EIO; 787 if (block == 0) {
788 block++;
789 goto retry;
790 } else {
791 return -EIO;
792 }
742 } 793 }
743 794
744 *raw_super = (struct f2fs_super_block *) 795 *raw_super = (struct f2fs_super_block *)
745 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET); 796 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
746 797
747 /* sanity checking of raw super */ 798 /* sanity checking of raw super */
748 if (!sanity_check_raw_super(sb, *raw_super)) 799 if (sanity_check_raw_super(sb, *raw_super)) {
749 return 0; 800 brelse(*raw_super_buf);
801 f2fs_msg(sb, KERN_ERR, "Can't find a valid F2FS filesystem "
802 "in %dth superblock", block + 1);
803 if(block == 0) {
804 block++;
805 goto retry;
806 } else {
807 return -EINVAL;
808 }
809 }
750 810
751 f2fs_msg(sb, KERN_ERR, "Can't find a valid F2FS filesystem " 811 return 0;
752 "in %s superblock", super);
753 return -EINVAL;
754} 812}
755 813
756static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 814static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
@@ -760,7 +818,6 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
760 struct buffer_head *raw_super_buf; 818 struct buffer_head *raw_super_buf;
761 struct inode *root; 819 struct inode *root;
762 long err = -EINVAL; 820 long err = -EINVAL;
763 int i;
764 821
765 /* allocate memory for f2fs-specific super block info */ 822 /* allocate memory for f2fs-specific super block info */
766 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 823 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
@@ -773,14 +830,10 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
773 goto free_sbi; 830 goto free_sbi;
774 } 831 }
775 832
776 err = validate_superblock(sb, &raw_super, &raw_super_buf, 0); 833 err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
777 if (err) { 834 if (err)
778 brelse(raw_super_buf); 835 goto free_sbi;
779 /* check secondary superblock when primary failed */ 836
780 err = validate_superblock(sb, &raw_super, &raw_super_buf, 1);
781 if (err)
782 goto free_sb_buf;
783 }
784 sb->s_fs_info = sbi; 837 sb->s_fs_info = sbi;
785 /* init some FS parameters */ 838 /* init some FS parameters */
786 sbi->active_logs = NR_CURSEG_TYPE; 839 sbi->active_logs = NR_CURSEG_TYPE;
@@ -818,12 +871,12 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
818 mutex_init(&sbi->gc_mutex); 871 mutex_init(&sbi->gc_mutex);
819 mutex_init(&sbi->writepages); 872 mutex_init(&sbi->writepages);
820 mutex_init(&sbi->cp_mutex); 873 mutex_init(&sbi->cp_mutex);
821 for (i = 0; i < NR_GLOBAL_LOCKS; i++)
822 mutex_init(&sbi->fs_lock[i]);
823 mutex_init(&sbi->node_write); 874 mutex_init(&sbi->node_write);
824 sbi->por_doing = 0; 875 sbi->por_doing = false;
825 spin_lock_init(&sbi->stat_lock); 876 spin_lock_init(&sbi->stat_lock);
826 init_rwsem(&sbi->bio_sem); 877 init_rwsem(&sbi->bio_sem);
878 init_rwsem(&sbi->cp_rwsem);
879 init_waitqueue_head(&sbi->cp_wait);
827 init_sb_info(sbi); 880 init_sb_info(sbi);
828 881
829 /* get an inode for meta space */ 882 /* get an inode for meta space */
@@ -922,12 +975,12 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
922 /* After POR, we can run background GC thread.*/ 975 /* After POR, we can run background GC thread.*/
923 err = start_gc_thread(sbi); 976 err = start_gc_thread(sbi);
924 if (err) 977 if (err)
925 goto fail; 978 goto free_gc;
926 } 979 }
927 980
928 err = f2fs_build_stats(sbi); 981 err = f2fs_build_stats(sbi);
929 if (err) 982 if (err)
930 goto fail; 983 goto free_gc;
931 984
932 if (f2fs_proc_root) 985 if (f2fs_proc_root)
933 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root); 986 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
@@ -953,6 +1006,12 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
953 1006
954 return 0; 1007 return 0;
955fail: 1008fail:
1009 if (sbi->s_proc) {
1010 remove_proc_entry("segment_info", sbi->s_proc);
1011 remove_proc_entry(sb->s_id, f2fs_proc_root);
1012 }
1013 f2fs_destroy_stats(sbi);
1014free_gc:
956 stop_gc_thread(sbi); 1015 stop_gc_thread(sbi);
957free_root_inode: 1016free_root_inode:
958 dput(sb->s_root); 1017 dput(sb->s_root);
diff --git a/fs/f2fs/xattr.c b/fs/f2fs/xattr.c
index 1ac8a5f6e380..aa7a3f139fe5 100644
--- a/fs/f2fs/xattr.c
+++ b/fs/f2fs/xattr.c
@@ -154,6 +154,9 @@ static int f2fs_xattr_advise_set(struct dentry *dentry, const char *name,
154} 154}
155 155
156#ifdef CONFIG_F2FS_FS_SECURITY 156#ifdef CONFIG_F2FS_FS_SECURITY
157static int __f2fs_setxattr(struct inode *inode, int name_index,
158 const char *name, const void *value, size_t value_len,
159 struct page *ipage);
157static int f2fs_initxattrs(struct inode *inode, const struct xattr *xattr_array, 160static int f2fs_initxattrs(struct inode *inode, const struct xattr *xattr_array,
158 void *page) 161 void *page)
159{ 162{
@@ -161,7 +164,7 @@ static int f2fs_initxattrs(struct inode *inode, const struct xattr *xattr_array,
161 int err = 0; 164 int err = 0;
162 165
163 for (xattr = xattr_array; xattr->name != NULL; xattr++) { 166 for (xattr = xattr_array; xattr->name != NULL; xattr++) {
164 err = f2fs_setxattr(inode, F2FS_XATTR_INDEX_SECURITY, 167 err = __f2fs_setxattr(inode, F2FS_XATTR_INDEX_SECURITY,
165 xattr->name, xattr->value, 168 xattr->name, xattr->value,
166 xattr->value_len, (struct page *)page); 169 xattr->value_len, (struct page *)page);
167 if (err < 0) 170 if (err < 0)
@@ -369,7 +372,7 @@ static inline int write_all_xattrs(struct inode *inode, __u32 hsize,
369 alloc_nid_failed(sbi, new_nid); 372 alloc_nid_failed(sbi, new_nid);
370 return PTR_ERR(xpage); 373 return PTR_ERR(xpage);
371 } 374 }
372 BUG_ON(new_nid); 375 f2fs_bug_on(new_nid);
373 } else { 376 } else {
374 struct dnode_of_data dn; 377 struct dnode_of_data dn;
375 set_new_dnode(&dn, inode, NULL, NULL, new_nid); 378 set_new_dnode(&dn, inode, NULL, NULL, new_nid);
@@ -469,16 +472,15 @@ cleanup:
469 return error; 472 return error;
470} 473}
471 474
472int f2fs_setxattr(struct inode *inode, int name_index, const char *name, 475static int __f2fs_setxattr(struct inode *inode, int name_index,
473 const void *value, size_t value_len, struct page *ipage) 476 const char *name, const void *value, size_t value_len,
477 struct page *ipage)
474{ 478{
475 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
476 struct f2fs_inode_info *fi = F2FS_I(inode); 479 struct f2fs_inode_info *fi = F2FS_I(inode);
477 struct f2fs_xattr_entry *here, *last; 480 struct f2fs_xattr_entry *here, *last;
478 void *base_addr; 481 void *base_addr;
479 int found, newsize; 482 int found, newsize;
480 size_t name_len; 483 size_t name_len;
481 int ilock;
482 __u32 new_hsize; 484 __u32 new_hsize;
483 int error = -ENOMEM; 485 int error = -ENOMEM;
484 486
@@ -493,10 +495,6 @@ int f2fs_setxattr(struct inode *inode, int name_index, const char *name,
493 if (name_len > F2FS_NAME_LEN || value_len > MAX_VALUE_LEN(inode)) 495 if (name_len > F2FS_NAME_LEN || value_len > MAX_VALUE_LEN(inode))
494 return -ERANGE; 496 return -ERANGE;
495 497
496 f2fs_balance_fs(sbi);
497
498 ilock = mutex_lock_op(sbi);
499
500 base_addr = read_all_xattrs(inode, ipage); 498 base_addr = read_all_xattrs(inode, ipage);
501 if (!base_addr) 499 if (!base_addr)
502 goto exit; 500 goto exit;
@@ -522,7 +520,7 @@ int f2fs_setxattr(struct inode *inode, int name_index, const char *name,
522 */ 520 */
523 free = MIN_OFFSET(inode) - ((char *)last - (char *)base_addr); 521 free = MIN_OFFSET(inode) - ((char *)last - (char *)base_addr);
524 if (found) 522 if (found)
525 free = free - ENTRY_SIZE(here); 523 free = free + ENTRY_SIZE(here);
526 524
527 if (free < newsize) { 525 if (free < newsize) {
528 error = -ENOSPC; 526 error = -ENOSPC;
@@ -578,7 +576,21 @@ int f2fs_setxattr(struct inode *inode, int name_index, const char *name,
578 else 576 else
579 update_inode_page(inode); 577 update_inode_page(inode);
580exit: 578exit:
581 mutex_unlock_op(sbi, ilock);
582 kzfree(base_addr); 579 kzfree(base_addr);
583 return error; 580 return error;
584} 581}
582
583int f2fs_setxattr(struct inode *inode, int name_index, const char *name,
584 const void *value, size_t value_len, struct page *ipage)
585{
586 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
587 int err;
588
589 f2fs_balance_fs(sbi);
590
591 f2fs_lock_op(sbi);
592 err = __f2fs_setxattr(inode, name_index, name, value, value_len, ipage);
593 f2fs_unlock_op(sbi);
594
595 return err;
596}
diff --git a/include/trace/events/f2fs.h b/include/trace/events/f2fs.h
index 52ae54828eda..e0dc355fa317 100644
--- a/include/trace/events/f2fs.h
+++ b/include/trace/events/f2fs.h
@@ -36,6 +36,11 @@
36 { CURSEG_COLD_NODE, "Cold NODE" }, \ 36 { CURSEG_COLD_NODE, "Cold NODE" }, \
37 { NO_CHECK_TYPE, "No TYPE" }) 37 { NO_CHECK_TYPE, "No TYPE" })
38 38
39#define show_file_type(type) \
40 __print_symbolic(type, \
41 { 0, "FILE" }, \
42 { 1, "DIR" })
43
39#define show_gc_type(type) \ 44#define show_gc_type(type) \
40 __print_symbolic(type, \ 45 __print_symbolic(type, \
41 { FG_GC, "Foreground GC" }, \ 46 { FG_GC, "Foreground GC" }, \
@@ -623,6 +628,52 @@ TRACE_EVENT(f2fs_do_submit_bio,
623 __entry->size) 628 __entry->size)
624); 629);
625 630
631DECLARE_EVENT_CLASS(f2fs__page,
632
633 TP_PROTO(struct page *page, int type),
634
635 TP_ARGS(page, type),
636
637 TP_STRUCT__entry(
638 __field(dev_t, dev)
639 __field(ino_t, ino)
640 __field(int, type)
641 __field(int, dir)
642 __field(pgoff_t, index)
643 __field(int, dirty)
644 ),
645
646 TP_fast_assign(
647 __entry->dev = page->mapping->host->i_sb->s_dev;
648 __entry->ino = page->mapping->host->i_ino;
649 __entry->type = type;
650 __entry->dir = S_ISDIR(page->mapping->host->i_mode);
651 __entry->index = page->index;
652 __entry->dirty = PageDirty(page);
653 ),
654
655 TP_printk("dev = (%d,%d), ino = %lu, %s, %s, index = %lu, dirty = %d",
656 show_dev_ino(__entry),
657 show_block_type(__entry->type),
658 show_file_type(__entry->dir),
659 (unsigned long)__entry->index,
660 __entry->dirty)
661);
662
663DEFINE_EVENT(f2fs__page, f2fs_set_page_dirty,
664
665 TP_PROTO(struct page *page, int type),
666
667 TP_ARGS(page, type)
668);
669
670DEFINE_EVENT(f2fs__page, f2fs_vm_page_mkwrite,
671
672 TP_PROTO(struct page *page, int type),
673
674 TP_ARGS(page, type)
675);
676
626TRACE_EVENT(f2fs_submit_write_page, 677TRACE_EVENT(f2fs_submit_write_page,
627 678
628 TP_PROTO(struct page *page, block_t blk_addr, int type), 679 TP_PROTO(struct page *page, block_t blk_addr, int type),
generated by cgit v1.2.2 (git 2.25.0) at 2025-08-23 17:16:37 -0400