[patch 24/36] ext4: dont use blocks freed but not yet committed inbuddy cache init

From: Greg KH
Date: Wed Feb 18 2009 - 17:41:26 EST



2.6.27-stable review patch. If anyone has any objections, please let us know.

------------------

From: Aneesh Kumar K.V <aneesh.kumar@xxxxxxxxxxxxxxxxxx>

(cherry picked from commit 7a2fcbf7f85737735fd44eb34b62315bccf6d6e4)

When we generate buddy cache (especially during resize) we need to
make sure we don't use the blocks freed but not yet comitted. This
makes sure we have the right value of free blocks count in the group
info and also in the bitmap. This also ensures the ordered mode
consistency

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@xxxxxxxxxxxxxxxxxx>
Signed-off-by: "Theodore Ts'o" <tytso@xxxxxxx>
Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxx>

---
fs/ext4/mballoc.c | 82 +++++++++++++++++++++++++++++++++++++++---------------
1 file changed, 60 insertions(+), 22 deletions(-)

--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -335,6 +335,8 @@ static struct kmem_cache *ext4_ac_cachep
static struct kmem_cache *ext4_free_ext_cachep;
static void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap,
ext4_group_t group);
+static void ext4_mb_generate_from_freelist(struct super_block *sb, void *bitmap,
+ ext4_group_t group);
static int ext4_mb_init_per_dev_proc(struct super_block *sb);
static int ext4_mb_destroy_per_dev_proc(struct super_block *sb);
static void ext4_mb_free_committed_blocks(struct super_block *);
@@ -858,7 +860,9 @@ static int ext4_mb_init_cache(struct pag
/*
* incore got set to the group block bitmap below
*/
+ ext4_lock_group(sb, group);
ext4_mb_generate_buddy(sb, data, incore, group);
+ ext4_unlock_group(sb, group);
incore = NULL;
} else {
/* this is block of bitmap */
@@ -872,6 +876,7 @@ static int ext4_mb_init_cache(struct pag

/* mark all preallocated blks used in in-core bitmap */
ext4_mb_generate_from_pa(sb, data, group);
+ ext4_mb_generate_from_freelist(sb, data, group);
ext4_unlock_group(sb, group);

/* set incore so that the buddy information can be
@@ -3579,6 +3584,32 @@ ext4_mb_use_preallocated(struct ext4_all
}

/*
+ * the function goes through all block freed in the group
+ * but not yet committed and marks them used in in-core bitmap.
+ * buddy must be generated from this bitmap
+ * Need to be called with ext4 group lock (ext4_lock_group)
+ */
+static void ext4_mb_generate_from_freelist(struct super_block *sb, void *bitmap,
+ ext4_group_t group)
+{
+ struct rb_node *n;
+ struct ext4_group_info *grp;
+ struct ext4_free_data *entry;
+
+ grp = ext4_get_group_info(sb, group);
+ n = rb_first(&(grp->bb_free_root));
+
+ while (n) {
+ entry = rb_entry(n, struct ext4_free_data, node);
+ mb_set_bits(sb_bgl_lock(EXT4_SB(sb), group),
+ bitmap, entry->start_blk,
+ entry->count);
+ n = rb_next(n);
+ }
+ return;
+}
+
+/*
* the function goes through all preallocation in this group and marks them
* used in in-core bitmap. buddy must be generated from this bitmap
* Need to be called with ext4 group lock (ext4_lock_group)
@@ -4709,27 +4740,22 @@ static int can_merge(struct ext4_free_da

static noinline_for_stack int
ext4_mb_free_metadata(handle_t *handle, struct ext4_buddy *e4b,
- ext4_group_t group, ext4_grpblk_t block, int count)
+ struct ext4_free_data *new_entry)
{
+ ext4_grpblk_t block;
+ struct ext4_free_data *entry;
struct ext4_group_info *db = e4b->bd_info;
struct super_block *sb = e4b->bd_sb;
struct ext4_sb_info *sbi = EXT4_SB(sb);
- struct ext4_free_data *entry, *new_entry;
struct rb_node **n = &db->bb_free_root.rb_node, *node;
struct rb_node *parent = NULL, *new_node;

-
BUG_ON(e4b->bd_bitmap_page == NULL);
BUG_ON(e4b->bd_buddy_page == NULL);

- new_entry = kmem_cache_alloc(ext4_free_ext_cachep, GFP_NOFS);
- new_entry->start_blk = block;
- new_entry->group = group;
- new_entry->count = count;
- new_entry->t_tid = handle->h_transaction->t_tid;
new_node = &new_entry->node;
+ block = new_entry->start_blk;

- ext4_lock_group(sb, group);
if (!*n) {
/* first free block exent. We need to
protect buddy cache from being freed,
@@ -4788,7 +4814,6 @@ ext4_mb_free_metadata(handle_t *handle,
spin_lock(&sbi->s_md_lock);
list_add(&new_entry->list, &sbi->s_active_transaction);
spin_unlock(&sbi->s_md_lock);
- ext4_unlock_group(sb, group);
return 0;
}

@@ -4895,15 +4920,6 @@ do_more:
BUG_ON(!mb_test_bit(bit + i, bitmap_bh->b_data));
}
#endif
- mb_clear_bits(sb_bgl_lock(sbi, block_group), bitmap_bh->b_data,
- bit, count);
-
- /* We dirtied the bitmap block */
- BUFFER_TRACE(bitmap_bh, "dirtied bitmap block");
- err = ext4_journal_dirty_metadata(handle, bitmap_bh);
- if (err)
- goto error_return;
-
if (ac) {
ac->ac_b_ex.fe_group = block_group;
ac->ac_b_ex.fe_start = bit;
@@ -4915,11 +4931,29 @@ do_more:
if (err)
goto error_return;
if (metadata) {
- /* blocks being freed are metadata. these blocks shouldn't
- * be used until this transaction is committed */
- ext4_mb_free_metadata(handle, &e4b, block_group, bit, count);
+ struct ext4_free_data *new_entry;
+ /*
+ * blocks being freed are metadata. these blocks shouldn't
+ * be used until this transaction is committed
+ */
+ new_entry = kmem_cache_alloc(ext4_free_ext_cachep, GFP_NOFS);
+ new_entry->start_blk = bit;
+ new_entry->group = block_group;
+ new_entry->count = count;
+ new_entry->t_tid = handle->h_transaction->t_tid;
+ ext4_lock_group(sb, block_group);
+ mb_clear_bits(sb_bgl_lock(sbi, block_group), bitmap_bh->b_data,
+ bit, count);
+ ext4_mb_free_metadata(handle, &e4b, new_entry);
+ ext4_unlock_group(sb, block_group);
} else {
ext4_lock_group(sb, block_group);
+ /* need to update group_info->bb_free and bitmap
+ * with group lock held. generate_buddy look at
+ * them with group lock_held
+ */
+ mb_clear_bits(sb_bgl_lock(sbi, block_group), bitmap_bh->b_data,
+ bit, count);
mb_free_blocks(inode, &e4b, bit, count);
ext4_mb_return_to_preallocation(inode, &e4b, block, count);
ext4_unlock_group(sb, block_group);
@@ -4942,6 +4976,10 @@ do_more:

*freed += count;

+ /* We dirtied the bitmap block */
+ BUFFER_TRACE(bitmap_bh, "dirtied bitmap block");
+ err = ext4_journal_dirty_metadata(handle, bitmap_bh);
+
/* And the group descriptor block */
BUFFER_TRACE(gd_bh, "dirtied group descriptor block");
ret = ext4_journal_dirty_metadata(handle, gd_bh);

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/