aboutsummaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
Diffstat (limited to 'fs')
-rw-r--r--fs/binfmt_flat.c4
-rw-r--r--fs/binfmt_misc.c4
-rw-r--r--fs/cramfs/inode.c84
-rw-r--r--fs/ext4/balloc.c3
-rw-r--r--fs/ext4/dir.c20
-rw-r--r--fs/ext4/ext4.h4
-rw-r--r--fs/ext4/ext4_extents.h4
-rw-r--r--fs/ext4/ext4_jbd2.h8
-rw-r--r--fs/ext4/extents.c113
-rw-r--r--fs/ext4/ialloc.c2
-rw-r--r--fs/ext4/inode.c478
-rw-r--r--fs/ext4/mballoc.c53
-rw-r--r--fs/ext4/migrate.c3
-rw-r--r--fs/ext4/resize.c3
-rw-r--r--fs/ext4/super.c1
-rw-r--r--fs/fat/inode.c10
-rw-r--r--fs/ioprio.c8
-rw-r--r--fs/jffs2/jffs2_fs_i.h1
-rw-r--r--fs/proc/nommu.c4
-rw-r--r--fs/proc/task_mmu.c4
-rw-r--r--fs/xfs/xfs_dmapi.h1
21 files changed, 508 insertions, 304 deletions
diff --git a/fs/binfmt_flat.c b/fs/binfmt_flat.c
index 56372ecf1690..dfc0197905ca 100644
--- a/fs/binfmt_flat.c
+++ b/fs/binfmt_flat.c
@@ -914,7 +914,9 @@ static int load_flat_binary(struct linux_binprm * bprm, struct pt_regs * regs)
914 /* Stash our initial stack pointer into the mm structure */ 914 /* Stash our initial stack pointer into the mm structure */
915 current->mm->start_stack = (unsigned long )sp; 915 current->mm->start_stack = (unsigned long )sp;
916 916
917 917#ifdef FLAT_PLAT_INIT
918 FLAT_PLAT_INIT(regs);
919#endif
918 DBG_FLT("start_thread(regs=0x%x, entry=0x%x, start_stack=0x%x)\n", 920 DBG_FLT("start_thread(regs=0x%x, entry=0x%x, start_stack=0x%x)\n",
919 (int)regs, (int)start_addr, (int)current->mm->start_stack); 921 (int)regs, (int)start_addr, (int)current->mm->start_stack);
920 922
diff --git a/fs/binfmt_misc.c b/fs/binfmt_misc.c
index 756205314c24..8d7e88e02e0f 100644
--- a/fs/binfmt_misc.c
+++ b/fs/binfmt_misc.c
@@ -120,8 +120,6 @@ static int load_misc_binary(struct linux_binprm *bprm, struct pt_regs *regs)
120 if (bprm->misc_bang) 120 if (bprm->misc_bang)
121 goto _ret; 121 goto _ret;
122 122
123 bprm->misc_bang = 1;
124
125 /* to keep locking time low, we copy the interpreter string */ 123 /* to keep locking time low, we copy the interpreter string */
126 read_lock(&entries_lock); 124 read_lock(&entries_lock);
127 fmt = check_file(bprm); 125 fmt = check_file(bprm);
@@ -199,6 +197,8 @@ static int load_misc_binary(struct linux_binprm *bprm, struct pt_regs *regs)
199 if (retval < 0) 197 if (retval < 0)
200 goto _error; 198 goto _error;
201 199
200 bprm->misc_bang = 1;
201
202 retval = search_binary_handler (bprm, regs); 202 retval = search_binary_handler (bprm, regs);
203 if (retval < 0) 203 if (retval < 0)
204 goto _error; 204 goto _error;
diff --git a/fs/cramfs/inode.c b/fs/cramfs/inode.c
index 0c3b618c15b3..f40423eb1a14 100644
--- a/fs/cramfs/inode.c
+++ b/fs/cramfs/inode.c
@@ -43,58 +43,13 @@ static DEFINE_MUTEX(read_mutex);
43static int cramfs_iget5_test(struct inode *inode, void *opaque) 43static int cramfs_iget5_test(struct inode *inode, void *opaque)
44{ 44{
45 struct cramfs_inode *cramfs_inode = opaque; 45 struct cramfs_inode *cramfs_inode = opaque;
46 46 return inode->i_ino == CRAMINO(cramfs_inode) && inode->i_ino != 1;
47 if (inode->i_ino != CRAMINO(cramfs_inode))
48 return 0; /* does not match */
49
50 if (inode->i_ino != 1)
51 return 1;
52
53 /* all empty directories, char, block, pipe, and sock, share inode #1 */
54
55 if ((inode->i_mode != cramfs_inode->mode) ||
56 (inode->i_gid != cramfs_inode->gid) ||
57 (inode->i_uid != cramfs_inode->uid))
58 return 0; /* does not match */
59
60 if ((S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) &&
61 (inode->i_rdev != old_decode_dev(cramfs_inode->size)))
62 return 0; /* does not match */
63
64 return 1; /* matches */
65} 47}
66 48
67static int cramfs_iget5_set(struct inode *inode, void *opaque) 49static int cramfs_iget5_set(struct inode *inode, void *opaque)
68{ 50{
69 static struct timespec zerotime;
70 struct cramfs_inode *cramfs_inode = opaque; 51 struct cramfs_inode *cramfs_inode = opaque;
71 inode->i_mode = cramfs_inode->mode;
72 inode->i_uid = cramfs_inode->uid;
73 inode->i_size = cramfs_inode->size;
74 inode->i_blocks = (cramfs_inode->size - 1) / 512 + 1;
75 inode->i_gid = cramfs_inode->gid;
76 /* Struct copy intentional */
77 inode->i_mtime = inode->i_atime = inode->i_ctime = zerotime;
78 inode->i_ino = CRAMINO(cramfs_inode); 52 inode->i_ino = CRAMINO(cramfs_inode);
79 /* inode->i_nlink is left 1 - arguably wrong for directories,
80 but it's the best we can do without reading the directory
81 contents. 1 yields the right result in GNU find, even
82 without -noleaf option. */
83 if (S_ISREG(inode->i_mode)) {
84 inode->i_fop = &generic_ro_fops;
85 inode->i_data.a_ops = &cramfs_aops;
86 } else if (S_ISDIR(inode->i_mode)) {
87 inode->i_op = &cramfs_dir_inode_operations;
88 inode->i_fop = &cramfs_directory_operations;
89 } else if (S_ISLNK(inode->i_mode)) {
90 inode->i_op = &page_symlink_inode_operations;
91 inode->i_data.a_ops = &cramfs_aops;
92 } else {
93 inode->i_size = 0;
94 inode->i_blocks = 0;
95 init_special_inode(inode, inode->i_mode,
96 old_decode_dev(cramfs_inode->size));
97 }
98 return 0; 53 return 0;
99} 54}
100 55
@@ -104,12 +59,48 @@ static struct inode *get_cramfs_inode(struct super_block *sb,
104 struct inode *inode = iget5_locked(sb, CRAMINO(cramfs_inode), 59 struct inode *inode = iget5_locked(sb, CRAMINO(cramfs_inode),
105 cramfs_iget5_test, cramfs_iget5_set, 60 cramfs_iget5_test, cramfs_iget5_set,
106 cramfs_inode); 61 cramfs_inode);
62 static struct timespec zerotime;
63
107 if (inode && (inode->i_state & I_NEW)) { 64 if (inode && (inode->i_state & I_NEW)) {
65 inode->i_mode = cramfs_inode->mode;
66 inode->i_uid = cramfs_inode->uid;
67 inode->i_size = cramfs_inode->size;
68 inode->i_blocks = (cramfs_inode->size - 1) / 512 + 1;
69 inode->i_gid = cramfs_inode->gid;
70 /* Struct copy intentional */
71 inode->i_mtime = inode->i_atime = inode->i_ctime = zerotime;
72 /* inode->i_nlink is left 1 - arguably wrong for directories,
73 but it's the best we can do without reading the directory
74 contents. 1 yields the right result in GNU find, even
75 without -noleaf option. */
76 if (S_ISREG(inode->i_mode)) {
77 inode->i_fop = &generic_ro_fops;
78 inode->i_data.a_ops = &cramfs_aops;
79 } else if (S_ISDIR(inode->i_mode)) {
80 inode->i_op = &cramfs_dir_inode_operations;
81 inode->i_fop = &cramfs_directory_operations;
82 } else if (S_ISLNK(inode->i_mode)) {
83 inode->i_op = &page_symlink_inode_operations;
84 inode->i_data.a_ops = &cramfs_aops;
85 } else {
86 inode->i_size = 0;
87 inode->i_blocks = 0;
88 init_special_inode(inode, inode->i_mode,
89 old_decode_dev(cramfs_inode->size));
90 }
108 unlock_new_inode(inode); 91 unlock_new_inode(inode);
109 } 92 }
110 return inode; 93 return inode;
111} 94}
112 95
96static void cramfs_drop_inode(struct inode *inode)
97{
98 if (inode->i_ino == 1)
99 generic_delete_inode(inode);
100 else
101 generic_drop_inode(inode);
102}
103
113/* 104/*
114 * We have our own block cache: don't fill up the buffer cache 105 * We have our own block cache: don't fill up the buffer cache
115 * with the rom-image, because the way the filesystem is set 106 * with the rom-image, because the way the filesystem is set
@@ -534,6 +525,7 @@ static const struct super_operations cramfs_ops = {
534 .put_super = cramfs_put_super, 525 .put_super = cramfs_put_super,
535 .remount_fs = cramfs_remount, 526 .remount_fs = cramfs_remount,
536 .statfs = cramfs_statfs, 527 .statfs = cramfs_statfs,
528 .drop_inode = cramfs_drop_inode,
537}; 529};
538 530
539static int cramfs_get_sb(struct file_system_type *fs_type, 531static int cramfs_get_sb(struct file_system_type *fs_type,
diff --git a/fs/ext4/balloc.c b/fs/ext4/balloc.c
index 1ae5004e93fc..e9fa960ba6da 100644
--- a/fs/ext4/balloc.c
+++ b/fs/ext4/balloc.c
@@ -1626,6 +1626,9 @@ ext4_fsblk_t ext4_has_free_blocks(struct ext4_sb_info *sbi,
1626 free_blocks = 1626 free_blocks =
1627 percpu_counter_sum_and_set(&sbi->s_freeblocks_counter); 1627 percpu_counter_sum_and_set(&sbi->s_freeblocks_counter);
1628#endif 1628#endif
1629 if (free_blocks <= root_blocks)
1630 /* we don't have free space */
1631 return 0;
1629 if (free_blocks - root_blocks < nblocks) 1632 if (free_blocks - root_blocks < nblocks)
1630 return free_blocks - root_blocks; 1633 return free_blocks - root_blocks;
1631 return nblocks; 1634 return nblocks;
diff --git a/fs/ext4/dir.c b/fs/ext4/dir.c
index d3d23d73c08b..ec8e33b45219 100644
--- a/fs/ext4/dir.c
+++ b/fs/ext4/dir.c
@@ -411,7 +411,7 @@ static int call_filldir(struct file * filp, void * dirent,
411 get_dtype(sb, fname->file_type)); 411 get_dtype(sb, fname->file_type));
412 if (error) { 412 if (error) {
413 filp->f_pos = curr_pos; 413 filp->f_pos = curr_pos;
414 info->extra_fname = fname->next; 414 info->extra_fname = fname;
415 return error; 415 return error;
416 } 416 }
417 fname = fname->next; 417 fname = fname->next;
@@ -450,11 +450,21 @@ static int ext4_dx_readdir(struct file * filp,
450 * If there are any leftover names on the hash collision 450 * If there are any leftover names on the hash collision
451 * chain, return them first. 451 * chain, return them first.
452 */ 452 */
453 if (info->extra_fname && 453 if (info->extra_fname) {
454 call_filldir(filp, dirent, filldir, info->extra_fname)) 454 if (call_filldir(filp, dirent, filldir, info->extra_fname))
455 goto finished; 455 goto finished;
456 456
457 if (!info->curr_node) 457 info->extra_fname = NULL;
458 info->curr_node = rb_next(info->curr_node);
459 if (!info->curr_node) {
460 if (info->next_hash == ~0) {
461 filp->f_pos = EXT4_HTREE_EOF;
462 goto finished;
463 }
464 info->curr_hash = info->next_hash;
465 info->curr_minor_hash = 0;
466 }
467 } else if (!info->curr_node)
458 info->curr_node = rb_first(&info->root); 468 info->curr_node = rb_first(&info->root);
459 469
460 while (1) { 470 while (1) {
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 6c7924d9e358..295003241d3d 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1072,6 +1072,8 @@ extern void ext4_set_inode_flags(struct inode *);
1072extern void ext4_get_inode_flags(struct ext4_inode_info *); 1072extern void ext4_get_inode_flags(struct ext4_inode_info *);
1073extern void ext4_set_aops(struct inode *inode); 1073extern void ext4_set_aops(struct inode *inode);
1074extern int ext4_writepage_trans_blocks(struct inode *); 1074extern int ext4_writepage_trans_blocks(struct inode *);
1075extern int ext4_meta_trans_blocks(struct inode *, int nrblocks, int idxblocks);
1076extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1075extern int ext4_block_truncate_page(handle_t *handle, 1077extern int ext4_block_truncate_page(handle_t *handle,
1076 struct address_space *mapping, loff_t from); 1078 struct address_space *mapping, loff_t from);
1077extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct page *page); 1079extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct page *page);
@@ -1227,6 +1229,8 @@ extern const struct inode_operations ext4_fast_symlink_inode_operations;
1227/* extents.c */ 1229/* extents.c */
1228extern int ext4_ext_tree_init(handle_t *handle, struct inode *); 1230extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
1229extern int ext4_ext_writepage_trans_blocks(struct inode *, int); 1231extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
1232extern int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks,
1233 int chunk);
1230extern int ext4_ext_get_blocks(handle_t *handle, struct inode *inode, 1234extern int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
1231 ext4_lblk_t iblock, 1235 ext4_lblk_t iblock,
1232 unsigned long max_blocks, struct buffer_head *bh_result, 1236 unsigned long max_blocks, struct buffer_head *bh_result,
diff --git a/fs/ext4/ext4_extents.h b/fs/ext4/ext4_extents.h
index 6c166c0a54b7..d33dc56d6986 100644
--- a/fs/ext4/ext4_extents.h
+++ b/fs/ext4/ext4_extents.h
@@ -216,7 +216,9 @@ extern int ext4_ext_calc_metadata_amount(struct inode *inode, int blocks);
216extern ext4_fsblk_t idx_pblock(struct ext4_extent_idx *); 216extern ext4_fsblk_t idx_pblock(struct ext4_extent_idx *);
217extern void ext4_ext_store_pblock(struct ext4_extent *, ext4_fsblk_t); 217extern void ext4_ext_store_pblock(struct ext4_extent *, ext4_fsblk_t);
218extern int ext4_extent_tree_init(handle_t *, struct inode *); 218extern int ext4_extent_tree_init(handle_t *, struct inode *);
219extern int ext4_ext_calc_credits_for_insert(struct inode *, struct ext4_ext_path *); 219extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
220 int num,
221 struct ext4_ext_path *path);
220extern int ext4_ext_try_to_merge(struct inode *inode, 222extern int ext4_ext_try_to_merge(struct inode *inode,
221 struct ext4_ext_path *path, 223 struct ext4_ext_path *path,
222 struct ext4_extent *); 224 struct ext4_extent *);
diff --git a/fs/ext4/ext4_jbd2.h b/fs/ext4/ext4_jbd2.h
index eb8bc3afe6e9..b455c685a98b 100644
--- a/fs/ext4/ext4_jbd2.h
+++ b/fs/ext4/ext4_jbd2.h
@@ -51,6 +51,14 @@
51 EXT4_XATTR_TRANS_BLOCKS - 2 + \ 51 EXT4_XATTR_TRANS_BLOCKS - 2 + \
52 2*EXT4_QUOTA_TRANS_BLOCKS(sb)) 52 2*EXT4_QUOTA_TRANS_BLOCKS(sb))
53 53
54/*
55 * Define the number of metadata blocks we need to account to modify data.
56 *
57 * This include super block, inode block, quota blocks and xattr blocks
58 */
59#define EXT4_META_TRANS_BLOCKS(sb) (EXT4_XATTR_TRANS_BLOCKS + \
60 2*EXT4_QUOTA_TRANS_BLOCKS(sb))
61
54/* Delete operations potentially hit one directory's namespace plus an 62/* Delete operations potentially hit one directory's namespace plus an
55 * entire inode, plus arbitrary amounts of bitmap/indirection data. Be 63 * entire inode, plus arbitrary amounts of bitmap/indirection data. Be
56 * generous. We can grow the delete transaction later if necessary. */ 64 * generous. We can grow the delete transaction later if necessary. */
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 612c3d2c3824..b24d3c53f20c 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -1747,54 +1747,61 @@ static int ext4_ext_rm_idx(handle_t *handle, struct inode *inode,
1747} 1747}
1748 1748
1749/* 1749/*
1750 * ext4_ext_calc_credits_for_insert: 1750 * ext4_ext_calc_credits_for_single_extent:
1751 * This routine returns max. credits that the extent tree can consume. 1751 * This routine returns max. credits that needed to insert an extent
1752 * It should be OK for low-performance paths like ->writepage() 1752 * to the extent tree.
1753 * To allow many writing processes to fit into a single transaction, 1753 * When pass the actual path, the caller should calculate credits
1754 * the caller should calculate credits under i_data_sem and 1754 * under i_data_sem.
1755 * pass the actual path.
1756 */ 1755 */
1757int ext4_ext_calc_credits_for_insert(struct inode *inode, 1756int ext4_ext_calc_credits_for_single_extent(struct inode *inode, int nrblocks,
1758 struct ext4_ext_path *path) 1757 struct ext4_ext_path *path)
1759{ 1758{
1760 int depth, needed;
1761
1762 if (path) { 1759 if (path) {
1760 int depth = ext_depth(inode);
1761 int ret = 0;
1762
1763 /* probably there is space in leaf? */ 1763 /* probably there is space in leaf? */
1764 depth = ext_depth(inode);
1765 if (le16_to_cpu(path[depth].p_hdr->eh_entries) 1764 if (le16_to_cpu(path[depth].p_hdr->eh_entries)
1766 < le16_to_cpu(path[depth].p_hdr->eh_max)) 1765 < le16_to_cpu(path[depth].p_hdr->eh_max)) {
1767 return 1;
1768 }
1769 1766
1770 /* 1767 /*
1771 * given 32-bit logical block (4294967296 blocks), max. tree 1768 * There are some space in the leaf tree, no
1772 * can be 4 levels in depth -- 4 * 340^4 == 53453440000. 1769 * need to account for leaf block credit
1773 * Let's also add one more level for imbalance. 1770 *
1774 */ 1771 * bitmaps and block group descriptor blocks
1775 depth = 5; 1772 * and other metadat blocks still need to be
1776 1773 * accounted.
1777 /* allocation of new data block(s) */ 1774 */
1778 needed = 2; 1775 /* 1 bitmap, 1 block group descriptor */
1776 ret = 2 + EXT4_META_TRANS_BLOCKS(inode->i_sb);
1777 }
1778 }
1779 1779
1780 /* 1780 return ext4_chunk_trans_blocks(inode, nrblocks);
1781 * tree can be full, so it would need to grow in depth: 1781}
1782 * we need one credit to modify old root, credits for
1783 * new root will be added in split accounting
1784 */
1785 needed += 1;
1786 1782
1787 /* 1783/*
1788 * Index split can happen, we would need: 1784 * How many index/leaf blocks need to change/allocate to modify nrblocks?
1789 * allocate intermediate indexes (bitmap + group) 1785 *
1790 * + change two blocks at each level, but root (already included) 1786 * if nrblocks are fit in a single extent (chunk flag is 1), then
1791 */ 1787 * in the worse case, each tree level index/leaf need to be changed
1792 needed += (depth * 2) + (depth * 2); 1788 * if the tree split due to insert a new extent, then the old tree
1789 * index/leaf need to be updated too
1790 *
1791 * If the nrblocks are discontiguous, they could cause
1792 * the whole tree split more than once, but this is really rare.
1793 */
1794int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks, int chunk)
1795{
1796 int index;
1797 int depth = ext_depth(inode);
1793 1798
1794 /* any allocation modifies superblock */ 1799 if (chunk)
1795 needed += 1; 1800 index = depth * 2;
1801 else
1802 index = depth * 3;
1796 1803
1797 return needed; 1804 return index;
1798} 1805}
1799 1806
1800static int ext4_remove_blocks(handle_t *handle, struct inode *inode, 1807static int ext4_remove_blocks(handle_t *handle, struct inode *inode,
@@ -1921,9 +1928,7 @@ ext4_ext_rm_leaf(handle_t *handle, struct inode *inode,
1921 correct_index = 1; 1928 correct_index = 1;
1922 credits += (ext_depth(inode)) + 1; 1929 credits += (ext_depth(inode)) + 1;
1923 } 1930 }
1924#ifdef CONFIG_QUOTA
1925 credits += 2 * EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb); 1931 credits += 2 * EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb);
1926#endif
1927 1932
1928 err = ext4_ext_journal_restart(handle, credits); 1933 err = ext4_ext_journal_restart(handle, credits);
1929 if (err) 1934 if (err)
@@ -2805,7 +2810,7 @@ void ext4_ext_truncate(struct inode *inode)
2805 /* 2810 /*
2806 * probably first extent we're gonna free will be last in block 2811 * probably first extent we're gonna free will be last in block
2807 */ 2812 */
2808 err = ext4_writepage_trans_blocks(inode) + 3; 2813 err = ext4_writepage_trans_blocks(inode);
2809 handle = ext4_journal_start(inode, err); 2814 handle = ext4_journal_start(inode, err);
2810 if (IS_ERR(handle)) 2815 if (IS_ERR(handle))
2811 return; 2816 return;
@@ -2819,7 +2824,7 @@ void ext4_ext_truncate(struct inode *inode)
2819 down_write(&EXT4_I(inode)->i_data_sem); 2824 down_write(&EXT4_I(inode)->i_data_sem);
2820 ext4_ext_invalidate_cache(inode); 2825 ext4_ext_invalidate_cache(inode);
2821 2826
2822 ext4_mb_discard_inode_preallocations(inode); 2827 ext4_discard_reservation(inode);
2823 2828
2824 /* 2829 /*
2825 * TODO: optimization is possible here. 2830 * TODO: optimization is possible here.
@@ -2858,27 +2863,6 @@ out_stop:
2858 ext4_journal_stop(handle); 2863 ext4_journal_stop(handle);
2859} 2864}
2860 2865
2861/*
2862 * ext4_ext_writepage_trans_blocks:
2863 * calculate max number of blocks we could modify
2864 * in order to allocate new block for an inode
2865 */
2866int ext4_ext_writepage_trans_blocks(struct inode *inode, int num)
2867{
2868 int needed;
2869
2870 needed = ext4_ext_calc_credits_for_insert(inode, NULL);
2871
2872 /* caller wants to allocate num blocks, but note it includes sb */
2873 needed = needed * num - (num - 1);
2874
2875#ifdef CONFIG_QUOTA
2876 needed += 2 * EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb);
2877#endif
2878
2879 return needed;
2880}
2881
2882static void ext4_falloc_update_inode(struct inode *inode, 2866static void ext4_falloc_update_inode(struct inode *inode,
2883 int mode, loff_t new_size, int update_ctime) 2867 int mode, loff_t new_size, int update_ctime)
2884{ 2868{
@@ -2939,10 +2923,9 @@ long ext4_fallocate(struct inode *inode, int mode, loff_t offset, loff_t len)
2939 max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits) 2923 max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits)
2940 - block; 2924 - block;
2941 /* 2925 /*
2942 * credits to insert 1 extent into extent tree + buffers to be able to 2926 * credits to insert 1 extent into extent tree
2943 * modify 1 super block, 1 block bitmap and 1 group descriptor.
2944 */ 2927 */
2945 credits = EXT4_DATA_TRANS_BLOCKS(inode->i_sb) + 3; 2928 credits = ext4_chunk_trans_blocks(inode, max_blocks);
2946 mutex_lock(&inode->i_mutex); 2929 mutex_lock(&inode->i_mutex);
2947retry: 2930retry:
2948 while (ret >= 0 && ret < max_blocks) { 2931 while (ret >= 0 && ret < max_blocks) {
diff --git a/fs/ext4/ialloc.c b/fs/ext4/ialloc.c
index 655e760212b8..f344834bbf58 100644
--- a/fs/ext4/ialloc.c
+++ b/fs/ext4/ialloc.c
@@ -351,7 +351,7 @@ find_close_to_parent:
351 goto found_flexbg; 351 goto found_flexbg;
352 } 352 }
353 353
354 if (best_flex < 0 || 354 if (flex_group[best_flex].free_inodes == 0 ||
355 (flex_group[i].free_blocks > 355 (flex_group[i].free_blocks >
356 flex_group[best_flex].free_blocks && 356 flex_group[best_flex].free_blocks &&
357 flex_group[i].free_inodes)) 357 flex_group[i].free_inodes))
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 59fbbe899acc..7e91913e325b 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -41,6 +41,8 @@
41#include "acl.h" 41#include "acl.h"
42#include "ext4_extents.h" 42#include "ext4_extents.h"
43 43
44#define MPAGE_DA_EXTENT_TAIL 0x01
45
44static inline int ext4_begin_ordered_truncate(struct inode *inode, 46static inline int ext4_begin_ordered_truncate(struct inode *inode,
45 loff_t new_size) 47 loff_t new_size)
46{ 48{
@@ -1005,6 +1007,9 @@ static int ext4_indirect_calc_metadata_amount(struct inode *inode, int blocks)
1005 */ 1007 */
1006static int ext4_calc_metadata_amount(struct inode *inode, int blocks) 1008static int ext4_calc_metadata_amount(struct inode *inode, int blocks)
1007{ 1009{
1010 if (!blocks)
1011 return 0;
1012
1008 if (EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL) 1013 if (EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL)
1009 return ext4_ext_calc_metadata_amount(inode, blocks); 1014 return ext4_ext_calc_metadata_amount(inode, blocks);
1010 1015
@@ -1041,18 +1046,6 @@ static void ext4_da_update_reserve_space(struct inode *inode, int used)
1041 spin_unlock(&EXT4_I(inode)->i_block_reservation_lock); 1046 spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
1042} 1047}
1043 1048
1044/* Maximum number of blocks we map for direct IO at once. */
1045#define DIO_MAX_BLOCKS 4096
1046/*
1047 * Number of credits we need for writing DIO_MAX_BLOCKS:
1048 * We need sb + group descriptor + bitmap + inode -> 4
1049 * For B blocks with A block pointers per block we need:
1050 * 1 (triple ind.) + (B/A/A + 2) (doubly ind.) + (B/A + 2) (indirect).
1051 * If we plug in 4096 for B and 256 for A (for 1KB block size), we get 25.
1052 */
1053#define DIO_CREDITS 25
1054
1055
1056/* 1049/*
1057 * The ext4_get_blocks_wrap() function try to look up the requested blocks, 1050 * The ext4_get_blocks_wrap() function try to look up the requested blocks,
1058 * and returns if the blocks are already mapped. 1051 * and returns if the blocks are already mapped.
@@ -1164,19 +1157,23 @@ int ext4_get_blocks_wrap(handle_t *handle, struct inode *inode, sector_t block,
1164 return retval; 1157 return retval;
1165} 1158}
1166 1159
1160/* Maximum number of blocks we map for direct IO at once. */
1161#define DIO_MAX_BLOCKS 4096
1162
1167static int ext4_get_block(struct inode *inode, sector_t iblock, 1163static int ext4_get_block(struct inode *inode, sector_t iblock,
1168 struct buffer_head *bh_result, int create) 1164 struct buffer_head *bh_result, int create)
1169{ 1165{
1170 handle_t *handle = ext4_journal_current_handle(); 1166 handle_t *handle = ext4_journal_current_handle();
1171 int ret = 0, started = 0; 1167 int ret = 0, started = 0;
1172 unsigned max_blocks = bh_result->b_size >> inode->i_blkbits; 1168 unsigned max_blocks = bh_result->b_size >> inode->i_blkbits;
1169 int dio_credits;
1173 1170
1174 if (create && !handle) { 1171 if (create && !handle) {
1175 /* Direct IO write... */ 1172 /* Direct IO write... */
1176 if (max_blocks > DIO_MAX_BLOCKS) 1173 if (max_blocks > DIO_MAX_BLOCKS)
1177 max_blocks = DIO_MAX_BLOCKS; 1174 max_blocks = DIO_MAX_BLOCKS;
1178 handle = ext4_journal_start(inode, DIO_CREDITS + 1175 dio_credits = ext4_chunk_trans_blocks(inode, max_blocks);
1179 2 * EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb)); 1176 handle = ext4_journal_start(inode, dio_credits);
1180 if (IS_ERR(handle)) { 1177 if (IS_ERR(handle)) {
1181 ret = PTR_ERR(handle); 1178 ret = PTR_ERR(handle);
1182 goto out; 1179 goto out;
@@ -1559,7 +1556,25 @@ static void ext4_da_release_space(struct inode *inode, int to_free)
1559 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 1556 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1560 int total, mdb, mdb_free, release; 1557 int total, mdb, mdb_free, release;
1561 1558
1559 if (!to_free)
1560 return; /* Nothing to release, exit */
1561
1562 spin_lock(&EXT4_I(inode)->i_block_reservation_lock); 1562 spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
1563
1564 if (!EXT4_I(inode)->i_reserved_data_blocks) {
1565 /*
1566 * if there is no reserved blocks, but we try to free some
1567 * then the counter is messed up somewhere.
1568 * but since this function is called from invalidate
1569 * page, it's harmless to return without any action
1570 */
1571 printk(KERN_INFO "ext4 delalloc try to release %d reserved "
1572 "blocks for inode %lu, but there is no reserved "
1573 "data blocks\n", to_free, inode->i_ino);
1574 spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
1575 return;
1576 }
1577
1563 /* recalculate the number of metablocks still need to be reserved */ 1578 /* recalculate the number of metablocks still need to be reserved */
1564 total = EXT4_I(inode)->i_reserved_data_blocks - to_free; 1579 total = EXT4_I(inode)->i_reserved_data_blocks - to_free;
1565 mdb = ext4_calc_metadata_amount(inode, total); 1580 mdb = ext4_calc_metadata_amount(inode, total);
@@ -1613,11 +1628,13 @@ struct mpage_da_data {
1613 unsigned long first_page, next_page; /* extent of pages */ 1628 unsigned long first_page, next_page; /* extent of pages */
1614 get_block_t *get_block; 1629 get_block_t *get_block;
1615 struct writeback_control *wbc; 1630 struct writeback_control *wbc;
1631 int io_done;
1632 long pages_written;
1616}; 1633};
1617 1634
1618/* 1635/*
1619 * mpage_da_submit_io - walks through extent of pages and try to write 1636 * mpage_da_submit_io - walks through extent of pages and try to write
1620 * them with __mpage_writepage() 1637 * them with writepage() call back
1621 * 1638 *
1622 * @mpd->inode: inode 1639 * @mpd->inode: inode
1623 * @mpd->first_page: first page of the extent 1640 * @mpd->first_page: first page of the extent
@@ -1632,18 +1649,11 @@ struct mpage_da_data {
1632static int mpage_da_submit_io(struct mpage_da_data *mpd) 1649static int mpage_da_submit_io(struct mpage_da_data *mpd)
1633{ 1650{
1634 struct address_space *mapping = mpd->inode->i_mapping; 1651 struct address_space *mapping = mpd->inode->i_mapping;
1635 struct mpage_data mpd_pp = {
1636 .bio = NULL,
1637 .last_block_in_bio = 0,
1638 .get_block = mpd->get_block,
1639 .use_writepage = 1,
1640 };
1641 int ret = 0, err, nr_pages, i; 1652 int ret = 0, err, nr_pages, i;
1642 unsigned long index, end; 1653 unsigned long index, end;
1643 struct pagevec pvec; 1654 struct pagevec pvec;
1644 1655
1645 BUG_ON(mpd->next_page <= mpd->first_page); 1656 BUG_ON(mpd->next_page <= mpd->first_page);
1646
1647 pagevec_init(&pvec, 0); 1657 pagevec_init(&pvec, 0);
1648 index = mpd->first_page; 1658 index = mpd->first_page;
1649 end = mpd->next_page - 1; 1659 end = mpd->next_page - 1;
@@ -1661,8 +1671,9 @@ static int mpage_da_submit_io(struct mpage_da_data *mpd)
1661 break; 1671 break;
1662 index++; 1672 index++;
1663 1673
1664 err = __mpage_writepage(page, mpd->wbc, &mpd_pp); 1674 err = mapping->a_ops->writepage(page, mpd->wbc);
1665 1675 if (!err)
1676 mpd->pages_written++;
1666 /* 1677 /*
1667 * In error case, we have to continue because 1678 * In error case, we have to continue because
1668 * remaining pages are still locked 1679 * remaining pages are still locked
@@ -1673,9 +1684,6 @@ static int mpage_da_submit_io(struct mpage_da_data *mpd)
1673 } 1684 }
1674 pagevec_release(&pvec); 1685 pagevec_release(&pvec);
1675 } 1686 }
1676 if (mpd_pp.bio)
1677 mpage_bio_submit(WRITE, mpd_pp.bio);
1678
1679 return ret; 1687 return ret;
1680} 1688}
1681 1689
@@ -1698,7 +1706,7 @@ static void mpage_put_bnr_to_bhs(struct mpage_da_data *mpd, sector_t logical,
1698 int blocks = exbh->b_size >> inode->i_blkbits; 1706 int blocks = exbh->b_size >> inode->i_blkbits;
1699 sector_t pblock = exbh->b_blocknr, cur_logical; 1707 sector_t pblock = exbh->b_blocknr, cur_logical;
1700 struct buffer_head *head, *bh; 1708 struct buffer_head *head, *bh;
1701 unsigned long index, end; 1709 pgoff_t index, end;
1702 struct pagevec pvec; 1710 struct pagevec pvec;
1703 int nr_pages, i; 1711 int nr_pages, i;
1704 1712
@@ -1741,6 +1749,13 @@ static void mpage_put_bnr_to_bhs(struct mpage_da_data *mpd, sector_t logical,
1741 if (buffer_delay(bh)) { 1749 if (buffer_delay(bh)) {
1742 bh->b_blocknr = pblock; 1750 bh->b_blocknr = pblock;
1743 clear_buffer_delay(bh); 1751 clear_buffer_delay(bh);
1752 bh->b_bdev = inode->i_sb->s_bdev;
1753 } else if (buffer_unwritten(bh)) {
1754 bh->b_blocknr = pblock;
1755 clear_buffer_unwritten(bh);
1756 set_buffer_mapped(bh);
1757 set_buffer_new(bh);
1758 bh->b_bdev = inode->i_sb->s_bdev;
1744 } else if (buffer_mapped(bh)) 1759 } else if (buffer_mapped(bh))
1745 BUG_ON(bh->b_blocknr != pblock); 1760 BUG_ON(bh->b_blocknr != pblock);
1746 1761
@@ -1776,13 +1791,11 @@ static inline void __unmap_underlying_blocks(struct inode *inode,
1776 * 1791 *
1777 * The function skips space we know is already mapped to disk blocks. 1792 * The function skips space we know is already mapped to disk blocks.
1778 * 1793 *
1779 * The function ignores errors ->get_block() returns, thus real
1780 * error handling is postponed to __mpage_writepage()
1781 */ 1794 */
1782static void mpage_da_map_blocks(struct mpage_da_data *mpd) 1795static void mpage_da_map_blocks(struct mpage_da_data *mpd)
1783{ 1796{
1797 int err = 0;
1784 struct buffer_head *lbh = &mpd->lbh; 1798 struct buffer_head *lbh = &mpd->lbh;
1785 int err = 0, remain = lbh->b_size;
1786 sector_t next = lbh->b_blocknr; 1799 sector_t next = lbh->b_blocknr;
1787 struct buffer_head new; 1800 struct buffer_head new;
1788 1801
@@ -1792,38 +1805,36 @@ static void mpage_da_map_blocks(struct mpage_da_data *mpd)
1792 if (buffer_mapped(lbh) && !buffer_delay(lbh)) 1805 if (buffer_mapped(lbh) && !buffer_delay(lbh))
1793 return; 1806 return;
1794 1807
1795 while (remain) { 1808 new.b_state = lbh->b_state;
1796 new.b_state = lbh->b_state; 1809 new.b_blocknr = 0;
1797 new.b_blocknr = 0; 1810 new.b_size = lbh->b_size;
1798 new.b_size = remain;
1799 err = mpd->get_block(mpd->inode, next, &new, 1);
1800 if (err) {
1801 /*
1802 * Rather than implement own error handling
1803 * here, we just leave remaining blocks
1804 * unallocated and try again with ->writepage()
1805 */
1806 break;
1807 }
1808 BUG_ON(new.b_size == 0);
1809 1811
1810 if (buffer_new(&new)) 1812 /*
1811 __unmap_underlying_blocks(mpd->inode, &new); 1813 * If we didn't accumulate anything
1814 * to write simply return
1815 */
1816 if (!new.b_size)
1817 return;
1818 err = mpd->get_block(mpd->inode, next, &new, 1);
1819 if (err)
1820 return;
1821 BUG_ON(new.b_size == 0);
1812 1822
1813 /* 1823 if (buffer_new(&new))
1814 * If blocks are delayed marked, we need to 1824 __unmap_underlying_blocks(mpd->inode, &new);
1815 * put actual blocknr and drop delayed bit
1816 */
1817 if (buffer_delay(lbh))
1818 mpage_put_bnr_to_bhs(mpd, next, &new);
1819 1825
1820 /* go for the remaining blocks */ 1826 /*
1821 next += new.b_size >> mpd->inode->i_blkbits; 1827 * If blocks are delayed marked, we need to
1822 remain -= new.b_size; 1828 * put actual blocknr and drop delayed bit
1823 } 1829 */
1830 if (buffer_delay(lbh) || buffer_unwritten(lbh))
1831 mpage_put_bnr_to_bhs(mpd, next, &new);
1832
1833 return;
1824} 1834}
1825 1835
1826#define BH_FLAGS ((1 << BH_Uptodate) | (1 << BH_Mapped) | (1 << BH_Delay)) 1836#define BH_FLAGS ((1 << BH_Uptodate) | (1 << BH_Mapped) | \
1837 (1 << BH_Delay) | (1 << BH_Unwritten))
1827 1838
1828/* 1839/*
1829 * mpage_add_bh_to_extent - try to add one more block to extent of blocks 1840 * mpage_add_bh_to_extent - try to add one more block to extent of blocks
@@ -1837,41 +1848,61 @@ static void mpage_da_map_blocks(struct mpage_da_data *mpd)
1837static void mpage_add_bh_to_extent(struct mpage_da_data *mpd, 1848static void mpage_add_bh_to_extent(struct mpage_da_data *mpd,
1838 sector_t logical, struct buffer_head *bh) 1849 sector_t logical, struct buffer_head *bh)
1839{ 1850{
1840 struct buffer_head *lbh = &mpd->lbh;
1841 sector_t next; 1851 sector_t next;
1852 size_t b_size = bh->b_size;
1853 struct buffer_head *lbh = &mpd->lbh;
1854 int nrblocks = lbh->b_size >> mpd->inode->i_blkbits;
1842 1855
1843 next = lbh->b_blocknr + (lbh->b_size >> mpd->inode->i_blkbits); 1856 /* check if thereserved journal credits might overflow */
1844 1857 if (!(EXT4_I(mpd->inode)->i_flags & EXT4_EXTENTS_FL)) {
1858 if (nrblocks >= EXT4_MAX_TRANS_DATA) {
1859 /*
1860 * With non-extent format we are limited by the journal
1861 * credit available. Total credit needed to insert
1862 * nrblocks contiguous blocks is dependent on the
1863 * nrblocks. So limit nrblocks.
1864 */
1865 goto flush_it;
1866 } else if ((nrblocks + (b_size >> mpd->inode->i_blkbits)) >
1867 EXT4_MAX_TRANS_DATA) {
1868 /*
1869 * Adding the new buffer_head would make it cross the
1870 * allowed limit for which we have journal credit
1871 * reserved. So limit the new bh->b_size
1872 */
1873 b_size = (EXT4_MAX_TRANS_DATA - nrblocks) <<
1874 mpd->inode->i_blkbits;
1875 /* we will do mpage_da_submit_io in the next loop */
1876 }
1877 }
1845 /* 1878 /*
1846 * First block in the extent 1879 * First block in the extent
1847 */ 1880 */
1848 if (lbh->b_size == 0) { 1881 if (lbh->b_size == 0) {
1849 lbh->b_blocknr = logical; 1882 lbh->b_blocknr = logical;
1850 lbh->b_size = bh->b_size; 1883 lbh->b_size = b_size;
1851 lbh->b_state = bh->b_state & BH_FLAGS; 1884 lbh->b_state = bh->b_state & BH_FLAGS;
1852 return; 1885 return;
1853 } 1886 }
1854 1887
1888 next = lbh->b_blocknr + nrblocks;
1855 /* 1889 /*
1856 * Can we merge the block to our big extent? 1890 * Can we merge the block to our big extent?
1857 */ 1891 */
1858 if (logical == next && (bh->b_state & BH_FLAGS) == lbh->b_state) { 1892 if (logical == next && (bh->b_state & BH_FLAGS) == lbh->b_state) {
1859 lbh->b_size += bh->b_size; 1893 lbh->b_size += b_size;
1860 return; 1894 return;
1861 } 1895 }
1862 1896
1897flush_it:
1863 /* 1898 /*
1864 * We couldn't merge the block to our extent, so we 1899 * We couldn't merge the block to our extent, so we
1865 * need to flush current extent and start new one 1900 * need to flush current extent and start new one
1866 */ 1901 */
1867 mpage_da_map_blocks(mpd); 1902 mpage_da_map_blocks(mpd);
1868 1903 mpage_da_submit_io(mpd);
1869 /* 1904 mpd->io_done = 1;
1870 * Now start a new extent 1905 return;
1871 */
1872 lbh->b_size = bh->b_size;
1873 lbh->b_state = bh->b_state & BH_FLAGS;
1874 lbh->b_blocknr = logical;
1875} 1906}
1876 1907
1877/* 1908/*
@@ -1891,17 +1922,35 @@ static int __mpage_da_writepage(struct page *page,
1891 struct buffer_head *bh, *head, fake; 1922 struct buffer_head *bh, *head, fake;
1892 sector_t logical; 1923 sector_t logical;
1893 1924
1925 if (mpd->io_done) {
1926 /*
1927 * Rest of the page in the page_vec
1928 * redirty then and skip then. We will
1929 * try to to write them again after
1930 * starting a new transaction
1931 */
1932 redirty_page_for_writepage(wbc, page);
1933 unlock_page(page);
1934 return MPAGE_DA_EXTENT_TAIL;
1935 }
1894 /* 1936 /*
1895 * Can we merge this page to current extent? 1937 * Can we merge this page to current extent?
1896 */ 1938 */
1897 if (mpd->next_page != page->index) { 1939 if (mpd->next_page != page->index) {
1898 /* 1940 /*
1899 * Nope, we can't. So, we map non-allocated blocks 1941 * Nope, we can't. So, we map non-allocated blocks
1900 * and start IO on them using __mpage_writepage() 1942 * and start IO on them using writepage()
1901 */ 1943 */
1902 if (mpd->next_page != mpd->first_page) { 1944 if (mpd->next_page != mpd->first_page) {
1903 mpage_da_map_blocks(mpd); 1945 mpage_da_map_blocks(mpd);
1904 mpage_da_submit_io(mpd); 1946 mpage_da_submit_io(mpd);
1947 /*
1948 * skip rest of the page in the page_vec
1949 */
1950 mpd->io_done = 1;
1951 redirty_page_for_writepage(wbc, page);
1952 unlock_page(page);
1953 return MPAGE_DA_EXTENT_TAIL;
1905 } 1954 }
1906 1955
1907 /* 1956 /*
@@ -1932,6 +1981,8 @@ static int __mpage_da_writepage(struct page *page,
1932 set_buffer_dirty(bh); 1981 set_buffer_dirty(bh);
1933 set_buffer_uptodate(bh); 1982 set_buffer_uptodate(bh);
1934 mpage_add_bh_to_extent(mpd, logical, bh); 1983 mpage_add_bh_to_extent(mpd, logical, bh);
1984 if (mpd->io_done)
1985 return MPAGE_DA_EXTENT_TAIL;
1935 } else { 1986 } else {
1936 /* 1987 /*
1937 * Page with regular buffer heads, just add all dirty ones 1988 * Page with regular buffer heads, just add all dirty ones
@@ -1940,8 +1991,12 @@ static int __mpage_da_writepage(struct page *page,
1940 bh = head; 1991 bh = head;
1941 do { 1992 do {
1942 BUG_ON(buffer_locked(bh)); 1993 BUG_ON(buffer_locked(bh));
1943 if (buffer_dirty(bh)) 1994 if (buffer_dirty(bh) &&
1995 (!buffer_mapped(bh) || buffer_delay(bh))) {
1944 mpage_add_bh_to_extent(mpd, logical, bh); 1996 mpage_add_bh_to_extent(mpd, logical, bh);
1997 if (mpd->io_done)
1998 return MPAGE_DA_EXTENT_TAIL;
1999 }
1945 logical++; 2000 logical++;
1946 } while ((bh = bh->b_this_page) != head); 2001 } while ((bh = bh->b_this_page) != head);
1947 } 2002 }
@@ -1960,22 +2015,13 @@ static int __mpage_da_writepage(struct page *page,
1960 * 2015 *
1961 * This is a library function, which implements the writepages() 2016 * This is a library function, which implements the writepages()
1962 * address_space_operation. 2017 * address_space_operation.
1963 *
1964 * In order to avoid duplication of logic that deals with partial pages,
1965 * multiple bio per page, etc, we find non-allocated blocks, allocate
1966 * them with minimal calls to ->get_block() and re-use __mpage_writepage()
1967 *
1968 * It's important that we call __mpage_writepage() only once for each
1969 * involved page, otherwise we'd have to implement more complicated logic
1970 * to deal with pages w/o PG_lock or w/ PG_writeback and so on.
1971 *
1972 * See comments to mpage_writepages()
1973 */ 2018 */
1974static int mpage_da_writepages(struct address_space *mapping, 2019static int mpage_da_writepages(struct address_space *mapping,
1975 struct writeback_control *wbc, 2020 struct writeback_control *wbc,
1976 get_block_t get_block) 2021 get_block_t get_block)
1977{ 2022{
1978 struct mpage_da_data mpd; 2023 struct mpage_da_data mpd;
2024 long to_write;
1979 int ret; 2025 int ret;
1980 2026
1981 if (!get_block) 2027 if (!get_block)
@@ -1989,17 +2035,22 @@ static int mpage_da_writepages(struct address_space *mapping,
1989 mpd.first_page = 0; 2035 mpd.first_page = 0;
1990 mpd.next_page = 0; 2036 mpd.next_page = 0;
1991 mpd.get_block = get_block; 2037 mpd.get_block = get_block;
2038 mpd.io_done = 0;
2039 mpd.pages_written = 0;
2040
2041 to_write = wbc->nr_to_write;
1992 2042
1993 ret = write_cache_pages(mapping, wbc, __mpage_da_writepage, &mpd); 2043 ret = write_cache_pages(mapping, wbc, __mpage_da_writepage, &mpd);
1994 2044
1995 /* 2045 /*
1996 * Handle last extent of pages 2046 * Handle last extent of pages
1997 */ 2047 */
1998 if (mpd.next_page != mpd.first_page) { 2048 if (!mpd.io_done && mpd.next_page != mpd.first_page) {
1999 mpage_da_map_blocks(&mpd); 2049 mpage_da_map_blocks(&mpd);
2000 mpage_da_submit_io(&mpd); 2050 mpage_da_submit_io(&mpd);
2001 } 2051 }
2002 2052
2053 wbc->nr_to_write = to_write - mpd.pages_written;
2003 return ret; 2054 return ret;
2004} 2055}
2005 2056
@@ -2204,63 +2255,95 @@ static int ext4_da_writepage(struct page *page,
2204} 2255}
2205 2256
2206/* 2257/*
2207 * For now just follow the DIO way to estimate the max credits 2258 * This is called via ext4_da_writepages() to
2208 * needed to write out EXT4_MAX_WRITEBACK_PAGES. 2259 * calulate the total number of credits to reserve to fit
2209 * todo: need to calculate the max credits need for 2260 * a single extent allocation into a single transaction,
2210 * extent based files, currently the DIO credits is based on 2261 * ext4_da_writpeages() will loop calling this before
2211 * indirect-blocks mapping way. 2262 * the block allocation.
2212 *
2213 * Probably should have a generic way to calculate credits
2214 * for DIO, writepages, and truncate
2215 */ 2263 */
2216#define EXT4_MAX_WRITEBACK_PAGES DIO_MAX_BLOCKS 2264
2217#define EXT4_MAX_WRITEBACK_CREDITS DIO_CREDITS 2265static int ext4_da_writepages_trans_blocks(struct inode *inode)
2266{
2267 int max_blocks = EXT4_I(inode)->i_reserved_data_blocks;
2268
2269 /*
2270 * With non-extent format the journal credit needed to
2271 * insert nrblocks contiguous block is dependent on
2272 * number of contiguous block. So we will limit
2273 * number of contiguous block to a sane value
2274 */
2275 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2276 (max_blocks > EXT4_MAX_TRANS_DATA))
2277 max_blocks = EXT4_MAX_TRANS_DATA;
2278
2279 return ext4_chunk_trans_blocks(inode, max_blocks);
2280}
2218 2281
2219static int ext4_da_writepages(struct address_space *mapping, 2282static int ext4_da_writepages(struct address_space *mapping,
2220 struct writeback_control *wbc) 2283 struct writeback_control *wbc)
2221{ 2284{
2222 struct inode *inode = mapping->host;
2223 handle_t *handle = NULL; 2285 handle_t *handle = NULL;
2224 int needed_blocks;
2225 int ret = 0;
2226 long to_write;
2227 loff_t range_start = 0; 2286 loff_t range_start = 0;
2287 struct inode *inode = mapping->host;
2288 int needed_blocks, ret = 0, nr_to_writebump = 0;
2289 long to_write, pages_skipped = 0;
2290 struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb);
2228 2291
2229 /* 2292 /*
2230 * No pages to write? This is mainly a kludge to avoid starting 2293 * No pages to write? This is mainly a kludge to avoid starting
2231 * a transaction for special inodes like journal inode on last iput() 2294 * a transaction for special inodes like journal inode on last iput()
2232 * because that could violate lock ordering on umount 2295 * because that could violate lock ordering on umount
2233 */ 2296 */
2234 if (!mapping->nrpages) 2297 if (!mapping->nrpages || !mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
2235 return 0; 2298 return 0;
2236
2237 /* 2299 /*
2238 * Estimate the worse case needed credits to write out 2300 * Make sure nr_to_write is >= sbi->s_mb_stream_request
2239 * EXT4_MAX_BUF_BLOCKS pages 2301 * This make sure small files blocks are allocated in
2302 * single attempt. This ensure that small files
2303 * get less fragmented.
2240 */ 2304 */
2241 needed_blocks = EXT4_MAX_WRITEBACK_CREDITS; 2305 if (wbc->nr_to_write < sbi->s_mb_stream_request) {
2306 nr_to_writebump = sbi->s_mb_stream_request - wbc->nr_to_write;
2307 wbc->nr_to_write = sbi->s_mb_stream_request;
2308 }
2242 2309
2243 to_write = wbc->nr_to_write; 2310 if (!wbc->range_cyclic)
2244 if (!wbc->range_cyclic) {
2245 /* 2311 /*
2246 * If range_cyclic is not set force range_cont 2312 * If range_cyclic is not set force range_cont
2247 * and save the old writeback_index 2313 * and save the old writeback_index
2248 */ 2314 */
2249 wbc->range_cont = 1; 2315 wbc->range_cont = 1;
2250 range_start = wbc->range_start;
2251 }
2252 2316
2253 while (!ret && to_write) { 2317 range_start = wbc->range_start;
2318 pages_skipped = wbc->pages_skipped;
2319
2320restart_loop:
2321 to_write = wbc->nr_to_write;
2322 while (!ret && to_write > 0) {
2323
2324 /*
2325 * we insert one extent at a time. So we need
2326 * credit needed for single extent allocation.
2327 * journalled mode is currently not supported
2328 * by delalloc
2329 */
2330 BUG_ON(ext4_should_journal_data(inode));
2331 needed_blocks = ext4_da_writepages_trans_blocks(inode);
2332
2254 /* start a new transaction*/ 2333 /* start a new transaction*/
2255 handle = ext4_journal_start(inode, needed_blocks); 2334 handle = ext4_journal_start(inode, needed_blocks);
2256 if (IS_ERR(handle)) { 2335 if (IS_ERR(handle)) {
2257 ret = PTR_ERR(handle); 2336 ret = PTR_ERR(handle);
2337 printk(KERN_EMERG "%s: jbd2_start: "
2338 "%ld pages, ino %lu; err %d\n", __func__,
2339 wbc->nr_to_write, inode->i_ino, ret);
2340 dump_stack();
2258 goto out_writepages; 2341 goto out_writepages;
2259 } 2342 }
2260 if (ext4_should_order_data(inode)) { 2343 if (ext4_should_order_data(inode)) {
2261 /* 2344 /*
2262 * With ordered mode we need to add 2345 * With ordered mode we need to add
2263 * the inode to the journal handle 2346 * the inode to the journal handl
2264 * when we do block allocation. 2347 * when we do block allocation.
2265 */ 2348 */
2266 ret = ext4_jbd2_file_inode(handle, inode); 2349 ret = ext4_jbd2_file_inode(handle, inode);
@@ -2268,20 +2351,20 @@ static int ext4_da_writepages(struct address_space *mapping,
2268 ext4_journal_stop(handle); 2351 ext4_journal_stop(handle);
2269 goto out_writepages; 2352 goto out_writepages;
2270 } 2353 }
2271
2272 } 2354 }
2273 /*
2274 * set the max dirty pages could be write at a time
2275 * to fit into the reserved transaction credits
2276 */
2277 if (wbc->nr_to_write > EXT4_MAX_WRITEBACK_PAGES)
2278 wbc->nr_to_write = EXT4_MAX_WRITEBACK_PAGES;
2279 2355
2280 to_write -= wbc->nr_to_write; 2356 to_write -= wbc->nr_to_write;
2281 ret = mpage_da_writepages(mapping, wbc, 2357 ret = mpage_da_writepages(mapping, wbc,
2282 ext4_da_get_block_write); 2358 ext4_da_get_block_write);
2283 ext4_journal_stop(handle); 2359 ext4_journal_stop(handle);
2284 if (wbc->nr_to_write) { 2360 if (ret == MPAGE_DA_EXTENT_TAIL) {
2361 /*
2362 * got one extent now try with
2363 * rest of the pages
2364 */
2365 to_write += wbc->nr_to_write;
2366 ret = 0;
2367 } else if (wbc->nr_to_write) {
2285 /* 2368 /*
2286 * There is no more writeout needed 2369 * There is no more writeout needed
2287 * or we requested for a noblocking writeout 2370 * or we requested for a noblocking writeout
@@ -2293,10 +2376,18 @@ static int ext4_da_writepages(struct address_space *mapping,
2293 wbc->nr_to_write = to_write; 2376 wbc->nr_to_write = to_write;
2294 } 2377 }
2295 2378
2296out_writepages: 2379 if (wbc->range_cont && (pages_skipped != wbc->pages_skipped)) {
2297 wbc->nr_to_write = to_write; 2380 /* We skipped pages in this loop */
2298 if (range_start)
2299 wbc->range_start = range_start; 2381 wbc->range_start = range_start;
2382 wbc->nr_to_write = to_write +
2383 wbc->pages_skipped - pages_skipped;
2384 wbc->pages_skipped = pages_skipped;
2385 goto restart_loop;
2386 }
2387
2388out_writepages:
2389 wbc->nr_to_write = to_write - nr_to_writebump;
2390 wbc->range_start = range_start;
2300 return ret; 2391 return ret;
2301} 2392}
2302 2393
@@ -3486,6 +3577,9 @@ void ext4_truncate(struct inode *inode)
3486 * modify the block allocation tree. 3577 * modify the block allocation tree.
3487 */ 3578 */
3488 down_write(&ei->i_data_sem); 3579 down_write(&ei->i_data_sem);
3580
3581 ext4_discard_reservation(inode);
3582
3489 /* 3583 /*
3490 * The orphan list entry will now protect us from any crash which 3584 * The orphan list entry will now protect us from any crash which
3491 * occurs before the truncate completes, so it is now safe to propagate 3585 * occurs before the truncate completes, so it is now safe to propagate
@@ -3555,8 +3649,6 @@ do_indirects:
3555 ; 3649 ;
3556 } 3650 }
3557 3651
3558 ext4_discard_reservation(inode);
3559
3560 up_write(&ei->i_data_sem); 3652 up_write(&ei->i_data_sem);
3561 inode->i_mtime = inode->i_ctime = ext4_current_time(inode); 3653 inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
3562 ext4_mark_inode_dirty(handle, inode); 3654 ext4_mark_inode_dirty(handle, inode);
@@ -4324,57 +4416,129 @@ int ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
4324 return 0; 4416 return 0;
4325} 4417}
4326 4418
4419static int ext4_indirect_trans_blocks(struct inode *inode, int nrblocks,
4420 int chunk)
4421{
4422 int indirects;
4423
4424 /* if nrblocks are contiguous */
4425 if (chunk) {
4426 /*
4427 * With N contiguous data blocks, it need at most
4428 * N/EXT4_ADDR_PER_BLOCK(inode->i_sb) indirect blocks
4429 * 2 dindirect blocks
4430 * 1 tindirect block
4431 */
4432 indirects = nrblocks / EXT4_ADDR_PER_BLOCK(inode->i_sb);
4433 return indirects + 3;
4434 }
4435 /*
4436 * if nrblocks are not contiguous, worse case, each block touch
4437 * a indirect block, and each indirect block touch a double indirect
4438 * block, plus a triple indirect block
4439 */
4440 indirects = nrblocks * 2 + 1;
4441 return indirects;
4442}
4443
4444static int ext4_index_trans_blocks(struct inode *inode, int nrblocks, int chunk)
4445{
4446 if (!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL))
4447 return ext4_indirect_trans_blocks(inode, nrblocks, 0);
4448 return ext4_ext_index_trans_blocks(inode, nrblocks, 0);
4449}
4327/* 4450/*
4328 * How many blocks doth make a writepage()? 4451 * Account for index blocks, block groups bitmaps and block group
4329 * 4452 * descriptor blocks if modify datablocks and index blocks
4330 * With N blocks per page, it may be: 4453 * worse case, the indexs blocks spread over different block groups
4331 * N data blocks
4332 * 2 indirect block
4333 * 2 dindirect
4334 * 1 tindirect
4335 * N+5 bitmap blocks (from the above)
4336 * N+5 group descriptor summary blocks
4337 * 1 inode block
4338 * 1 superblock.
4339 * 2 * EXT4_SINGLEDATA_TRANS_BLOCKS for the quote files
4340 * 4454 *
4341 * 3 * (N + 5) + 2 + 2 * EXT4_SINGLEDATA_TRANS_BLOCKS 4455 * If datablocks are discontiguous, they are possible to spread over
4456 * different block groups too. If they are contiugous, with flexbg,
4457 * they could still across block group boundary.
4342 * 4458 *
4343 * With ordered or writeback data it's the same, less the N data blocks. 4459 * Also account for superblock, inode, quota and xattr blocks
4460 */
4461int ext4_meta_trans_blocks(struct inode *inode, int nrblocks, int chunk)
4462{
4463 int groups, gdpblocks;
4464 int idxblocks;
4465 int ret = 0;
4466
4467 /*
4468 * How many index blocks need to touch to modify nrblocks?
4469 * The "Chunk" flag indicating whether the nrblocks is
4470 * physically contiguous on disk
4471 *
4472 * For Direct IO and fallocate, they calls get_block to allocate
4473 * one single extent at a time, so they could set the "Chunk" flag
4474 */
4475 idxblocks = ext4_index_trans_blocks(inode, nrblocks, chunk);
4476
4477 ret = idxblocks;
4478
4479 /*
4480 * Now let's see how many group bitmaps and group descriptors need
4481 * to account
4482 */
4483 groups = idxblocks;
4484 if (chunk)
4485 groups += 1;
4486 else
4487 groups += nrblocks;
4488
4489 gdpblocks = groups;
4490 if (groups > EXT4_SB(inode->i_sb)->s_groups_count)
4491 groups = EXT4_SB(inode->i_sb)->s_groups_count;
4492 if (groups > EXT4_SB(inode->i_sb)->s_gdb_count)
4493 gdpblocks = EXT4_SB(inode->i_sb)->s_gdb_count;
4494
4495 /* bitmaps and block group descriptor blocks */
4496 ret += groups + gdpblocks;
4497
4498 /* Blocks for super block, inode, quota and xattr blocks */
4499 ret += EXT4_META_TRANS_BLOCKS(inode->i_sb);
4500
4501 return ret;
4502}
4503
4504/*
4505 * Calulate the total number of credits to reserve to fit
4506 * the modification of a single pages into a single transaction,
4507 * which may include multiple chunks of block allocations.
4344 * 4508 *
4345 * If the inode's direct blocks can hold an integral number of pages then a 4509 * This could be called via ext4_write_begin()
4346 * page cannot straddle two indirect blocks, and we can only touch one indirect
4347 * and dindirect block, and the "5" above becomes "3".
4348 * 4510 *
4349 * This still overestimates under most circumstances. If we were to pass the 4511 * We need to consider the worse case, when
4350 * start and end offsets in here as well we could do block_to_path() on each 4512 * one new block per extent.
4351 * block and work out the exact number of indirects which are touched. Pah.
4352 */ 4513 */
4353
4354int ext4_writepage_trans_blocks(struct inode *inode) 4514int ext4_writepage_trans_blocks(struct inode *inode)
4355{ 4515{
4356 int bpp = ext4_journal_blocks_per_page(inode); 4516 int bpp = ext4_journal_blocks_per_page(inode);
4357 int indirects = (EXT4_NDIR_BLOCKS % bpp) ? 5 : 3;
4358 int ret; 4517 int ret;
4359 4518
4360 if (EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL) 4519 ret = ext4_meta_trans_blocks(inode, bpp, 0);
4361 return ext4_ext_writepage_trans_blocks(inode, bpp);
4362 4520
4521 /* Account for data blocks for journalled mode */
4363 if (ext4_should_journal_data(inode)) 4522 if (ext4_should_journal_data(inode))
4364 ret = 3 * (bpp + indirects) + 2; 4523 ret += bpp;
4365 else
4366 ret = 2 * (bpp + indirects) + 2;
4367
4368#ifdef CONFIG_QUOTA
4369 /* We know that structure was already allocated during DQUOT_INIT so
4370 * we will be updating only the data blocks + inodes */
4371 ret += 2*EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb);
4372#endif
4373
4374 return ret; 4524 return ret;
4375} 4525}
4376 4526
4377/* 4527/*
4528 * Calculate the journal credits for a chunk of data modification.
4529 *
4530 * This is called from DIO, fallocate or whoever calling
4531 * ext4_get_blocks_wrap() to map/allocate a chunk of contigous disk blocks.
4532 *
4533 * journal buffers for data blocks are not included here, as DIO
4534 * and fallocate do no need to journal data buffers.
4535 */
4536int ext4_chunk_trans_blocks(struct inode *inode, int nrblocks)
4537{
4538 return ext4_meta_trans_blocks(inode, nrblocks, 1);
4539}
4540
4541/*
4378 * The caller must have previously called ext4_reserve_inode_write(). 4542 * The caller must have previously called ext4_reserve_inode_write().
4379 * Give this, we know that the caller already has write access to iloc->bh. 4543 * Give this, we know that the caller already has write access to iloc->bh.
4380 */ 4544 */
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index 865e9ddb44d4..e0e3a5eb1ddb 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -3282,6 +3282,35 @@ static void ext4_mb_use_group_pa(struct ext4_allocation_context *ac,
3282} 3282}
3283 3283
3284/* 3284/*
3285 * Return the prealloc space that have minimal distance
3286 * from the goal block. @cpa is the prealloc
3287 * space that is having currently known minimal distance
3288 * from the goal block.
3289 */
3290static struct ext4_prealloc_space *
3291ext4_mb_check_group_pa(ext4_fsblk_t goal_block,
3292 struct ext4_prealloc_space *pa,
3293 struct ext4_prealloc_space *cpa)
3294{
3295 ext4_fsblk_t cur_distance, new_distance;
3296
3297 if (cpa == NULL) {
3298 atomic_inc(&pa->pa_count);
3299 return pa;
3300 }
3301 cur_distance = abs(goal_block - cpa->pa_pstart);
3302 new_distance = abs(goal_block - pa->pa_pstart);
3303
3304 if (cur_distance < new_distance)
3305 return cpa;
3306
3307 /* drop the previous reference */
3308 atomic_dec(&cpa->pa_count);
3309 atomic_inc(&pa->pa_count);
3310 return pa;
3311}
3312
3313/*
3285 * search goal blocks in preallocated space 3314 * search goal blocks in preallocated space
3286 */ 3315 */
3287static noinline_for_stack int 3316static noinline_for_stack int
@@ -3290,7 +3319,8 @@ ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
3290 int order, i; 3319 int order, i;
3291 struct ext4_inode_info *ei = EXT4_I(ac->ac_inode); 3320 struct ext4_inode_info *ei = EXT4_I(ac->ac_inode);
3292 struct ext4_locality_group *lg; 3321 struct ext4_locality_group *lg;
3293 struct ext4_prealloc_space *pa; 3322 struct ext4_prealloc_space *pa, *cpa = NULL;
3323 ext4_fsblk_t goal_block;
3294 3324
3295 /* only data can be preallocated */ 3325 /* only data can be preallocated */
3296 if (!(ac->ac_flags & EXT4_MB_HINT_DATA)) 3326 if (!(ac->ac_flags & EXT4_MB_HINT_DATA))
@@ -3333,6 +3363,13 @@ ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
3333 /* The max size of hash table is PREALLOC_TB_SIZE */ 3363 /* The max size of hash table is PREALLOC_TB_SIZE */
3334 order = PREALLOC_TB_SIZE - 1; 3364 order = PREALLOC_TB_SIZE - 1;
3335 3365
3366 goal_block = ac->ac_g_ex.fe_group * EXT4_BLOCKS_PER_GROUP(ac->ac_sb) +
3367 ac->ac_g_ex.fe_start +
3368 le32_to_cpu(EXT4_SB(ac->ac_sb)->s_es->s_first_data_block);
3369 /*
3370 * search for the prealloc space that is having
3371 * minimal distance from the goal block.
3372 */
3336 for (i = order; i < PREALLOC_TB_SIZE; i++) { 3373 for (i = order; i < PREALLOC_TB_SIZE; i++) {
3337 rcu_read_lock(); 3374 rcu_read_lock();
3338 list_for_each_entry_rcu(pa, &lg->lg_prealloc_list[i], 3375 list_for_each_entry_rcu(pa, &lg->lg_prealloc_list[i],
@@ -3340,17 +3377,19 @@ ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
3340 spin_lock(&pa->pa_lock); 3377 spin_lock(&pa->pa_lock);
3341 if (pa->pa_deleted == 0 && 3378 if (pa->pa_deleted == 0 &&
3342 pa->pa_free >= ac->ac_o_ex.fe_len) { 3379 pa->pa_free >= ac->ac_o_ex.fe_len) {
3343 atomic_inc(&pa->pa_count); 3380
3344 ext4_mb_use_group_pa(ac, pa); 3381 cpa = ext4_mb_check_group_pa(goal_block,
3345 spin_unlock(&pa->pa_lock); 3382 pa, cpa);
3346 ac->ac_criteria = 20;
3347 rcu_read_unlock();
3348 return 1;
3349 } 3383 }
3350 spin_unlock(&pa->pa_lock); 3384 spin_unlock(&pa->pa_lock);
3351 } 3385 }
3352 rcu_read_unlock(); 3386 rcu_read_unlock();
3353 } 3387 }
3388 if (cpa) {
3389 ext4_mb_use_group_pa(ac, cpa);
3390 ac->ac_criteria = 20;
3391 return 1;
3392 }
3354 return 0; 3393 return 0;
3355} 3394}
3356 3395
diff --git a/fs/ext4/migrate.c b/fs/ext4/migrate.c
index b9e077ba07e9..46fc0b5b12ba 100644
--- a/fs/ext4/migrate.c
+++ b/fs/ext4/migrate.c
@@ -53,7 +53,8 @@ static int finish_range(handle_t *handle, struct inode *inode,
53 * credit. But below we try to not accumalate too much 53 * credit. But below we try to not accumalate too much
54 * of them by restarting the journal. 54 * of them by restarting the journal.
55 */ 55 */
56 needed = ext4_ext_calc_credits_for_insert(inode, path); 56 needed = ext4_ext_calc_credits_for_single_extent(inode,
57 lb->last_block - lb->first_block + 1, path);
57 58
58 /* 59 /*
59 * Make sure the credit we accumalated is not really high 60 * Make sure the credit we accumalated is not really high
diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c
index 0a9265164265..b3d35604ea18 100644
--- a/fs/ext4/resize.c
+++ b/fs/ext4/resize.c
@@ -773,7 +773,8 @@ int ext4_group_add(struct super_block *sb, struct ext4_new_group_data *input)
773 773
774 if (reserved_gdb || gdb_off == 0) { 774 if (reserved_gdb || gdb_off == 0) {
775 if (!EXT4_HAS_COMPAT_FEATURE(sb, 775 if (!EXT4_HAS_COMPAT_FEATURE(sb,
776 EXT4_FEATURE_COMPAT_RESIZE_INODE)){ 776 EXT4_FEATURE_COMPAT_RESIZE_INODE)
777 || !le16_to_cpu(es->s_reserved_gdt_blocks)) {
777 ext4_warning(sb, __func__, 778 ext4_warning(sb, __func__,
778 "No reserved GDT blocks, can't resize"); 779 "No reserved GDT blocks, can't resize");
779 return -EPERM; 780 return -EPERM;
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index d5d77958b861..566344b926b7 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -568,6 +568,7 @@ static struct inode *ext4_alloc_inode(struct super_block *sb)
568#endif 568#endif
569 ei->i_block_alloc_info = NULL; 569 ei->i_block_alloc_info = NULL;
570 ei->vfs_inode.i_version = 1; 570 ei->vfs_inode.i_version = 1;
571 ei->vfs_inode.i_data.writeback_index = 0;
571 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache)); 572 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
572 INIT_LIST_HEAD(&ei->i_prealloc_list); 573 INIT_LIST_HEAD(&ei->i_prealloc_list);
573 spin_lock_init(&ei->i_prealloc_lock); 574 spin_lock_init(&ei->i_prealloc_lock);
diff --git a/fs/fat/inode.c b/fs/fat/inode.c
index 6d266d793e2c..80ff3381fa21 100644
--- a/fs/fat/inode.c
+++ b/fs/fat/inode.c
@@ -562,26 +562,23 @@ static int fat_write_inode(struct inode *inode, int wait)
562 struct buffer_head *bh; 562 struct buffer_head *bh;
563 struct msdos_dir_entry *raw_entry; 563 struct msdos_dir_entry *raw_entry;
564 loff_t i_pos; 564 loff_t i_pos;
565 int err = 0; 565 int err;
566 566
567retry: 567retry:
568 i_pos = MSDOS_I(inode)->i_pos; 568 i_pos = MSDOS_I(inode)->i_pos;
569 if (inode->i_ino == MSDOS_ROOT_INO || !i_pos) 569 if (inode->i_ino == MSDOS_ROOT_INO || !i_pos)
570 return 0; 570 return 0;
571 571
572 lock_super(sb);
573 bh = sb_bread(sb, i_pos >> sbi->dir_per_block_bits); 572 bh = sb_bread(sb, i_pos >> sbi->dir_per_block_bits);
574 if (!bh) { 573 if (!bh) {
575 printk(KERN_ERR "FAT: unable to read inode block " 574 printk(KERN_ERR "FAT: unable to read inode block "
576 "for updating (i_pos %lld)\n", i_pos); 575 "for updating (i_pos %lld)\n", i_pos);
577 err = -EIO; 576 return -EIO;
578 goto out;
579 } 577 }
580 spin_lock(&sbi->inode_hash_lock); 578 spin_lock(&sbi->inode_hash_lock);
581 if (i_pos != MSDOS_I(inode)->i_pos) { 579 if (i_pos != MSDOS_I(inode)->i_pos) {
582 spin_unlock(&sbi->inode_hash_lock); 580 spin_unlock(&sbi->inode_hash_lock);
583 brelse(bh); 581 brelse(bh);
584 unlock_super(sb);
585 goto retry; 582 goto retry;
586 } 583 }
587 584
@@ -607,11 +604,10 @@ retry:
607 } 604 }
608 spin_unlock(&sbi->inode_hash_lock); 605 spin_unlock(&sbi->inode_hash_lock);
609 mark_buffer_dirty(bh); 606 mark_buffer_dirty(bh);
607 err = 0;
610 if (wait) 608 if (wait)
611 err = sync_dirty_buffer(bh); 609 err = sync_dirty_buffer(bh);
612 brelse(bh); 610 brelse(bh);
613out:
614 unlock_super(sb);
615 return err; 611 return err;
616} 612}
617 613
diff --git a/fs/ioprio.c b/fs/ioprio.c
index c4a1c3c65aac..da3cc460d4df 100644
--- a/fs/ioprio.c
+++ b/fs/ioprio.c
@@ -115,11 +115,11 @@ asmlinkage long sys_ioprio_set(int which, int who, int ioprio)
115 pgrp = task_pgrp(current); 115 pgrp = task_pgrp(current);
116 else 116 else
117 pgrp = find_vpid(who); 117 pgrp = find_vpid(who);
118 do_each_pid_task(pgrp, PIDTYPE_PGID, p) { 118 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
119 ret = set_task_ioprio(p, ioprio); 119 ret = set_task_ioprio(p, ioprio);
120 if (ret) 120 if (ret)
121 break; 121 break;
122 } while_each_pid_task(pgrp, PIDTYPE_PGID, p); 122 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
123 break; 123 break;
124 case IOPRIO_WHO_USER: 124 case IOPRIO_WHO_USER:
125 if (!who) 125 if (!who)
@@ -204,7 +204,7 @@ asmlinkage long sys_ioprio_get(int which, int who)
204 pgrp = task_pgrp(current); 204 pgrp = task_pgrp(current);
205 else 205 else
206 pgrp = find_vpid(who); 206 pgrp = find_vpid(who);
207 do_each_pid_task(pgrp, PIDTYPE_PGID, p) { 207 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
208 tmpio = get_task_ioprio(p); 208 tmpio = get_task_ioprio(p);
209 if (tmpio < 0) 209 if (tmpio < 0)
210 continue; 210 continue;
@@ -212,7 +212,7 @@ asmlinkage long sys_ioprio_get(int which, int who)
212 ret = tmpio; 212 ret = tmpio;
213 else 213 else
214 ret = ioprio_best(ret, tmpio); 214 ret = ioprio_best(ret, tmpio);
215 } while_each_pid_task(pgrp, PIDTYPE_PGID, p); 215 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
216 break; 216 break;
217 case IOPRIO_WHO_USER: 217 case IOPRIO_WHO_USER:
218 if (!who) 218 if (!who)
diff --git a/fs/jffs2/jffs2_fs_i.h b/fs/jffs2/jffs2_fs_i.h
index 31559f45fdde..4c41db91eaa4 100644
--- a/fs/jffs2/jffs2_fs_i.h
+++ b/fs/jffs2/jffs2_fs_i.h
@@ -12,7 +12,6 @@
12#ifndef _JFFS2_FS_I 12#ifndef _JFFS2_FS_I
13#define _JFFS2_FS_I 13#define _JFFS2_FS_I
14 14
15#include <linux/version.h>
16#include <linux/rbtree.h> 15#include <linux/rbtree.h>
17#include <linux/posix_acl.h> 16#include <linux/posix_acl.h>
18#include <linux/mutex.h> 17#include <linux/mutex.h>
diff --git a/fs/proc/nommu.c b/fs/proc/nommu.c
index 79ecd281d2cb..3f87d2632947 100644
--- a/fs/proc/nommu.c
+++ b/fs/proc/nommu.c
@@ -52,14 +52,14 @@ int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma)
52 } 52 }
53 53
54 seq_printf(m, 54 seq_printf(m,
55 "%08lx-%08lx %c%c%c%c %08lx %02x:%02x %lu %n", 55 "%08lx-%08lx %c%c%c%c %08llx %02x:%02x %lu %n",
56 vma->vm_start, 56 vma->vm_start,
57 vma->vm_end, 57 vma->vm_end,
58 flags & VM_READ ? 'r' : '-', 58 flags & VM_READ ? 'r' : '-',
59 flags & VM_WRITE ? 'w' : '-', 59 flags & VM_WRITE ? 'w' : '-',
60 flags & VM_EXEC ? 'x' : '-', 60 flags & VM_EXEC ? 'x' : '-',
61 flags & VM_MAYSHARE ? flags & VM_SHARED ? 'S' : 's' : 'p', 61 flags & VM_MAYSHARE ? flags & VM_SHARED ? 'S' : 's' : 'p',
62 vma->vm_pgoff << PAGE_SHIFT, 62 ((loff_t)vma->vm_pgoff) << PAGE_SHIFT,
63 MAJOR(dev), MINOR(dev), ino, &len); 63 MAJOR(dev), MINOR(dev), ino, &len);
64 64
65 if (file) { 65 if (file) {
diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
index 7546a918f790..73d1891ee625 100644
--- a/fs/proc/task_mmu.c
+++ b/fs/proc/task_mmu.c
@@ -219,14 +219,14 @@ static int show_map(struct seq_file *m, void *v)
219 ino = inode->i_ino; 219 ino = inode->i_ino;
220 } 220 }
221 221
222 seq_printf(m, "%08lx-%08lx %c%c%c%c %08lx %02x:%02x %lu %n", 222 seq_printf(m, "%08lx-%08lx %c%c%c%c %08llx %02x:%02x %lu %n",
223 vma->vm_start, 223 vma->vm_start,
224 vma->vm_end, 224 vma->vm_end,
225 flags & VM_READ ? 'r' : '-', 225 flags & VM_READ ? 'r' : '-',
226 flags & VM_WRITE ? 'w' : '-', 226 flags & VM_WRITE ? 'w' : '-',
227 flags & VM_EXEC ? 'x' : '-', 227 flags & VM_EXEC ? 'x' : '-',
228 flags & VM_MAYSHARE ? 's' : 'p', 228 flags & VM_MAYSHARE ? 's' : 'p',
229 vma->vm_pgoff << PAGE_SHIFT, 229 ((loff_t)vma->vm_pgoff) << PAGE_SHIFT,
230 MAJOR(dev), MINOR(dev), ino, &len); 230 MAJOR(dev), MINOR(dev), ino, &len);
231 231
232 /* 232 /*
diff --git a/fs/xfs/xfs_dmapi.h b/fs/xfs/xfs_dmapi.h
index cdc2d3464a1a..2813cdd72375 100644
--- a/fs/xfs/xfs_dmapi.h
+++ b/fs/xfs/xfs_dmapi.h
@@ -18,7 +18,6 @@
18#ifndef __XFS_DMAPI_H__ 18#ifndef __XFS_DMAPI_H__
19#define __XFS_DMAPI_H__ 19#define __XFS_DMAPI_H__
20 20
21#include <linux/version.h>
22/* Values used to define the on-disk version of dm_attrname_t. All 21/* Values used to define the on-disk version of dm_attrname_t. All
23 * on-disk attribute names start with the 8-byte string "SGI_DMI_". 22 * on-disk attribute names start with the 8-byte string "SGI_DMI_".
24 * 23 *
generated by cgit v1.2.2 (git 2.25.0) at 2025-12-08 16:07:36 -0500