diff options
author | Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> | 2015-11-06 16:31:56 -0800 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2015-11-06 17:50:42 -0800 |
commit | d0c14a9ee79467cd6a04b281577e1e6f74806ab2 (patch) | |
tree | a757000613b5d690b972ead91eb5a5adefc8c4b5 /fs/nilfs2 | |
parent | da019954dd821682d6b2a8330c9c90acb943c456 (diff) | |
download | linux-d0c14a9ee79467cd6a04b281577e1e6f74806ab2.tar.bz2 |
nilfs2: free unused dat file blocks during garbage collection
As a nilfs2 volume ages, the amount of available disk space decreases
little by little due to bloat of DAT (disk address translation) metadata
file. Even if we delete all files in a file system and free their block
addresses from the DAT file through a garbage collection, empty DAT blocks
are not freed.
This fixes the issue by extending the deallocator of block addresses so
that empty data blocks and empty bitmap blocks of DAT are deleted.
The following comparison shows the effect of this patch. Each shows disk
amount information of a nilfs2 volume that we cleaned out by deleting all
files and running gc after having filled 90% of its capacity.
Before:
Filesystem 1K-blocks Used Available Use% Mounted on
/dev/sda1 500105212 3022844 472072192 1% /test
After:
Filesystem 1K-blocks Used Available Use% Mounted on
/dev/sda1 500105212 16380 475078656 1% /test
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'fs/nilfs2')
-rw-r--r-- | fs/nilfs2/alloc.c | 91 | ||||
-rw-r--r-- | fs/nilfs2/alloc.h | 1 |
2 files changed, 75 insertions, 17 deletions
diff --git a/fs/nilfs2/alloc.c b/fs/nilfs2/alloc.c index 225b79768865..b335a32e9561 100644 --- a/fs/nilfs2/alloc.c +++ b/fs/nilfs2/alloc.c @@ -154,13 +154,17 @@ nilfs_palloc_group_desc_nfrees(const struct nilfs_palloc_group_desc *desc, * @lock: spin lock protecting @desc * @n: delta to be added */ -static void +static u32 nilfs_palloc_group_desc_add_entries(struct nilfs_palloc_group_desc *desc, spinlock_t *lock, u32 n) { + u32 nfree; + spin_lock(lock); le32_add_cpu(&desc->pg_nfrees, n); + nfree = le32_to_cpu(desc->pg_nfrees); spin_unlock(lock); + return nfree; } /** @@ -735,12 +739,18 @@ int nilfs_palloc_freev(struct inode *inode, __u64 *entry_nrs, size_t nitems) unsigned char *bitmap; void *desc_kaddr, *bitmap_kaddr; unsigned long group, group_offset; - __u64 group_min_nr; + __u64 group_min_nr, last_nrs[8]; const unsigned long epg = nilfs_palloc_entries_per_group(inode); + const unsigned epb = NILFS_MDT(inode)->mi_entries_per_block; + unsigned entry_start, end, pos; spinlock_t *lock; - int i, j, n, ret; + int i, j, k, ret; + u32 nfree; for (i = 0; i < nitems; i = j) { + int change_group = false; + int nempties = 0, n = 0; + group = nilfs_palloc_group(inode, entry_nrs[i], &group_offset); ret = nilfs_palloc_get_desc_block(inode, group, 0, &desc_bh); if (ret < 0) @@ -755,17 +765,13 @@ int nilfs_palloc_freev(struct inode *inode, __u64 *entry_nrs, size_t nitems) /* Get the first entry number of the group */ group_min_nr = (__u64)group * epg; - desc_kaddr = kmap(desc_bh->b_page); - desc = nilfs_palloc_block_get_group_desc( - inode, group, desc_bh, desc_kaddr); bitmap_kaddr = kmap(bitmap_bh->b_page); bitmap = bitmap_kaddr + bh_offset(bitmap_bh); lock = nilfs_mdt_bgl_lock(inode, group); - for (j = i, n = 0; - j < nitems && entry_nrs[j] >= group_min_nr && - entry_nrs[j] < group_min_nr + epg; - j++) { - group_offset = entry_nrs[j] - group_min_nr; + + j = i; + entry_start = rounddown(group_offset, epb); + do { if (!nilfs_clear_bit_atomic(lock, group_offset, bitmap)) { nilfs_warning(inode->i_sb, __func__, @@ -775,18 +781,69 @@ int nilfs_palloc_freev(struct inode *inode, __u64 *entry_nrs, size_t nitems) } else { n++; } - } - nilfs_palloc_group_desc_add_entries(desc, lock, n); + + j++; + if (j >= nitems || entry_nrs[j] < group_min_nr || + entry_nrs[j] >= group_min_nr + epg) { + change_group = true; + } else { + group_offset = entry_nrs[j] - group_min_nr; + if (group_offset >= entry_start && + group_offset < entry_start + epb) { + /* This entry is in the same block */ + continue; + } + } + + /* Test if the entry block is empty or not */ + end = entry_start + epb; + pos = nilfs_find_next_bit(bitmap, end, entry_start); + if (pos >= end) { + last_nrs[nempties++] = entry_nrs[j - 1]; + if (nempties >= ARRAY_SIZE(last_nrs)) + break; + } + + if (change_group) + break; + + /* Go on to the next entry block */ + entry_start = rounddown(group_offset, epb); + } while (true); kunmap(bitmap_bh->b_page); - kunmap(desc_bh->b_page); + mark_buffer_dirty(bitmap_bh); + brelse(bitmap_bh); + for (k = 0; k < nempties; k++) { + ret = nilfs_palloc_delete_entry_block(inode, + last_nrs[k]); + if (ret && ret != -ENOENT) { + nilfs_warning(inode->i_sb, __func__, + "failed to delete block of entry %llu: ino=%lu, err=%d\n", + (unsigned long long)last_nrs[k], + (unsigned long)inode->i_ino, ret); + } + } + + desc_kaddr = kmap_atomic(desc_bh->b_page); + desc = nilfs_palloc_block_get_group_desc( + inode, group, desc_bh, desc_kaddr); + nfree = nilfs_palloc_group_desc_add_entries(desc, lock, n); + kunmap_atomic(desc_kaddr); mark_buffer_dirty(desc_bh); - mark_buffer_dirty(bitmap_bh); nilfs_mdt_mark_dirty(inode); - - brelse(bitmap_bh); brelse(desc_bh); + + if (nfree == nilfs_palloc_entries_per_group(inode)) { + ret = nilfs_palloc_delete_bitmap_block(inode, group); + if (ret && ret != -ENOENT) { + nilfs_warning(inode->i_sb, __func__, + "failed to delete bitmap block of group %lu: ino=%lu, err=%d\n", + group, + (unsigned long)inode->i_ino, ret); + } + } } return 0; } diff --git a/fs/nilfs2/alloc.h b/fs/nilfs2/alloc.h index 4bd6451b5703..6e6f49aa53df 100644 --- a/fs/nilfs2/alloc.h +++ b/fs/nilfs2/alloc.h @@ -77,6 +77,7 @@ int nilfs_palloc_freev(struct inode *, __u64 *, size_t); #define nilfs_set_bit_atomic ext2_set_bit_atomic #define nilfs_clear_bit_atomic ext2_clear_bit_atomic #define nilfs_find_next_zero_bit find_next_zero_bit_le +#define nilfs_find_next_bit find_next_bit_le /** * struct nilfs_bh_assoc - block offset and buffer head association |