* [Ocfs2-devel] [PATCH 01/15] ocfs2: Define data structures for discontiguous block groups.
2010-04-01 2:53 [Ocfs2-devel] [PATCH 0/15] ocfs2: Support for discontiguous block group Tao Ma
@ 2010-04-01 2:58 ` Tao Ma
2010-04-01 2:58 ` [Ocfs2-devel] [PATCH 02/15] ocfs2: Allocate " Tao Ma
` (13 subsequent siblings)
14 siblings, 0 replies; 34+ messages in thread
From: Tao Ma @ 2010-04-01 2:58 UTC (permalink / raw)
To: ocfs2-devel
From: Joel Becker <joel.becker@oracle.com>
Defines the OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG feature bit and modifies
struct ocfs2_group_desc for the feature.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
fs/ocfs2/ocfs2_fs.h | 53 ++++++++++++++++++++++++++++++++++++++++++++------
fs/ocfs2/resize.c | 4 +-
fs/ocfs2/suballoc.c | 2 +-
fs/ocfs2/super.c | 2 +-
4 files changed, 50 insertions(+), 11 deletions(-)
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index bb37218..055cf07 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -165,6 +165,9 @@
/* Refcount tree support */
#define OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE 0x1000
+/* Discontigous block groups */
+#define OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG 0x2000
+
/*
* backup superblock flag is used to indicate that this volume
* has backup superblocks.
@@ -840,6 +843,13 @@ struct ocfs2_dx_leaf {
};
/*
+ * Largest bitmap for a block (suballocator) group in bytes. This limit
+ * does not affect cluster groups (global allocator). Cluster group
+ * bitmaps run to the end of the block.
+ */
+#define OCFS2_MAX_BG_BITMAP_SIZE 256
+
+/*
* On disk allocator group structure for OCFS2
*/
struct ocfs2_group_desc
@@ -860,7 +870,29 @@ struct ocfs2_group_desc
__le64 bg_blkno; /* Offset on disk, in blocks */
/*30*/ struct ocfs2_block_check bg_check; /* Error checking */
__le64 bg_reserved2;
-/*40*/ __u8 bg_bitmap[0];
+/*40*/ union {
+ __u8 bg_bitmap[0];
+ struct {
+ /*
+ * Block groups may be discontiguous when
+ * OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG is set.
+ * The extents of a discontigous block group are
+ * stored in bg_list. It is a flat list.
+ * l_tree_depth must always be zero. A
+ * discontiguous group is signified by a non-zero
+ * bg_list->l_next_free_rec. Only block groups
+ * can be discontiguous; Cluster groups cannot.
+ * We've never made a block group with more than
+ * 2048 blocks (256 bytes of bg_bitmap). This
+ * codifies that limit so that we can fit bg_list.
+ * bg_size of a discontiguous block group will
+ * be 256 to match bg_bitmap_filler.
+ */
+ __u8 bg_bitmap_filler[OCFS2_MAX_BG_BITMAP_SIZE];
+/*140*/ struct ocfs2_extent_list bg_list;
+ };
+ };
+/* Actual on-disk size is one block */
};
struct ocfs2_refcount_rec {
@@ -1284,12 +1316,16 @@ static inline u16 ocfs2_local_alloc_size(struct super_block *sb)
return size;
}
-static inline int ocfs2_group_bitmap_size(struct super_block *sb)
+static inline int ocfs2_group_bitmap_size(struct super_block *sb,
+ int suballocator)
{
int size;
- size = sb->s_blocksize -
- offsetof(struct ocfs2_group_desc, bg_bitmap);
+ if (suballocator)
+ size = OCFS2_MAX_BG_BITMAP_SIZE;
+ else
+ size = sb->s_blocksize -
+ offsetof(struct ocfs2_group_desc, bg_bitmap);
return size;
}
@@ -1412,12 +1448,15 @@ static inline int ocfs2_local_alloc_size(int blocksize)
return size;
}
-static inline int ocfs2_group_bitmap_size(int blocksize)
+static inline int ocfs2_group_bitmap_size(int blocksize, int suballocator)
{
int size;
- size = blocksize -
- offsetof(struct ocfs2_group_desc, bg_bitmap);
+ if (suballocator)
+ size = OCFS2_MAX_BG_BITMAP_SIZE;
+ else
+ size = blocksize -
+ offsetof(struct ocfs2_group_desc, bg_bitmap);
return size;
}
diff --git a/fs/ocfs2/resize.c b/fs/ocfs2/resize.c
index 3c3d673..0ff4835 100644
--- a/fs/ocfs2/resize.c
+++ b/fs/ocfs2/resize.c
@@ -319,7 +319,7 @@ int ocfs2_group_extend(struct inode * inode, int new_clusters)
BUG_ON(!OCFS2_IS_VALID_DINODE(fe));
if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
- ocfs2_group_bitmap_size(osb->sb) * 8) {
+ ocfs2_group_bitmap_size(osb->sb, 0) * 8) {
mlog(ML_ERROR, "The disk is too old and small. "
"Force to do offline resize.");
ret = -EINVAL;
@@ -500,7 +500,7 @@ int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input)
fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
- ocfs2_group_bitmap_size(osb->sb) * 8) {
+ ocfs2_group_bitmap_size(osb->sb, 0) * 8) {
mlog(ML_ERROR, "The disk is too old and small."
" Force to do offline resize.");
ret = -EINVAL;
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index c3c60bc..84125bf 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -363,7 +363,7 @@ static int ocfs2_block_group_fill(handle_t *handle,
memset(bg, 0, sb->s_blocksize);
strcpy(bg->bg_signature, OCFS2_GROUP_DESC_SIGNATURE);
bg->bg_generation = cpu_to_le32(OCFS2_SB(sb)->fs_generation);
- bg->bg_size = cpu_to_le16(ocfs2_group_bitmap_size(sb));
+ bg->bg_size = cpu_to_le16(ocfs2_group_bitmap_size(sb, 1));
bg->bg_bits = cpu_to_le16(ocfs2_bits_per_group(cl));
bg->bg_chain = cpu_to_le16(my_chain);
bg->bg_next_group = cl->cl_recs[my_chain].c_blkno;
diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c
index dee0319..de8547f 100644
--- a/fs/ocfs2/super.c
+++ b/fs/ocfs2/super.c
@@ -2226,7 +2226,7 @@ static int ocfs2_initialize_super(struct super_block *sb,
osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
iput(inode);
- osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
+ osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0) * 8;
status = ocfs2_init_slot_info(osb);
if (status < 0) {
--
1.5.5
^ permalink raw reply related [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 02/15] ocfs2: Allocate discontiguous block groups.
2010-04-01 2:53 [Ocfs2-devel] [PATCH 0/15] ocfs2: Support for discontiguous block group Tao Ma
2010-04-01 2:58 ` [Ocfs2-devel] [PATCH 01/15] ocfs2: Define data structures for discontiguous block groups Tao Ma
@ 2010-04-01 2:58 ` Tao Ma
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 03/15] ocfs2: Pass suballocation results back via a structure Tao Ma
` (12 subsequent siblings)
14 siblings, 0 replies; 34+ messages in thread
From: Tao Ma @ 2010-04-01 2:58 UTC (permalink / raw)
To: ocfs2-devel
From: Joel Becker <joel.becker@oracle.com>
If we cannot get a contiguous region for a block group, allocate a
discontiguous one when the filesystem supports it.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
fs/ocfs2/ocfs2.h | 7 +
fs/ocfs2/ocfs2_fs.h | 20 ++++
fs/ocfs2/suballoc.c | 306 ++++++++++++++++++++++++++++++++++++++++++++-------
3 files changed, 293 insertions(+), 40 deletions(-)
diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h
index 1238b49..b5fd76f 100644
--- a/fs/ocfs2/ocfs2.h
+++ b/fs/ocfs2/ocfs2.h
@@ -482,6 +482,13 @@ static inline int ocfs2_supports_indexed_dirs(struct ocfs2_super *osb)
return 0;
}
+static inline int ocfs2_supports_discontig_bh(struct ocfs2_super *osb)
+{
+ if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG)
+ return 1;
+ return 0;
+}
+
static inline unsigned int ocfs2_link_max(struct ocfs2_super *osb)
{
if (ocfs2_supports_indexed_dirs(osb))
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index 055cf07..0508e4e 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -1286,6 +1286,16 @@ static inline u16 ocfs2_extent_recs_per_eb(struct super_block *sb)
return size / sizeof(struct ocfs2_extent_rec);
}
+static inline u16 ocfs2_extent_recs_per_gd(struct super_block *sb)
+{
+ int size;
+
+ size = sb->s_blocksize -
+ offsetof(struct ocfs2_group_desc, bg_list.l_recs);
+
+ return size / sizeof(struct ocfs2_extent_rec);
+}
+
static inline int ocfs2_dx_entries_per_leaf(struct super_block *sb)
{
int size;
@@ -1438,6 +1448,16 @@ static inline int ocfs2_extent_recs_per_eb(int blocksize)
return size / sizeof(struct ocfs2_extent_rec);
}
+static inline int ocfs2_extent_recs_per_gd(int blocksize)
+{
+ int size;
+
+ size = blocksize -
+ offsetof(struct ocfs2_group_desc, bg_list.l_recs);
+
+ return size / sizeof(struct ocfs2_extent_rec);
+}
+
static inline int ocfs2_local_alloc_size(int blocksize)
{
int size;
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 84125bf..267d552 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -60,6 +60,7 @@ static int ocfs2_block_group_fill(handle_t *handle,
struct inode *alloc_inode,
struct buffer_head *bg_bh,
u64 group_blkno,
+ unsigned int group_clusters,
u16 my_chain,
struct ocfs2_chain_list *cl);
static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
@@ -329,14 +330,36 @@ out:
return rc;
}
+static void ocfs2_bg_discontig_add_extent(struct ocfs2_super *osb,
+ struct ocfs2_group_desc *bg,
+ struct ocfs2_chain_list *cl,
+ u64 p_blkno, u32 clusters)
+{
+ struct ocfs2_extent_list *el = &bg->bg_list;
+ struct ocfs2_extent_rec *rec;
+
+ BUG_ON(!ocfs2_supports_discontig_bh(osb));
+ if (!el->l_next_free_rec)
+ el->l_count = cpu_to_le16(ocfs2_extent_recs_per_gd(osb->sb));
+ rec = &el->l_recs[le16_to_cpu(el->l_next_free_rec)];
+ rec->e_blkno = p_blkno;
+ rec->e_cpos = cpu_to_le32(le16_to_cpu(bg->bg_bits) /
+ le16_to_cpu(cl->cl_bpc));
+ rec->e_leaf_clusters = cpu_to_le32(clusters);
+ le16_add_cpu(&bg->bg_bits, clusters * le16_to_cpu(cl->cl_bpc));
+ le16_add_cpu(&el->l_next_free_rec, 1);
+}
+
static int ocfs2_block_group_fill(handle_t *handle,
struct inode *alloc_inode,
struct buffer_head *bg_bh,
u64 group_blkno,
+ unsigned int group_clusters,
u16 my_chain,
struct ocfs2_chain_list *cl)
{
int status = 0;
+ struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb);
struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data;
struct super_block * sb = alloc_inode->i_sb;
@@ -364,11 +387,16 @@ static int ocfs2_block_group_fill(handle_t *handle,
strcpy(bg->bg_signature, OCFS2_GROUP_DESC_SIGNATURE);
bg->bg_generation = cpu_to_le32(OCFS2_SB(sb)->fs_generation);
bg->bg_size = cpu_to_le16(ocfs2_group_bitmap_size(sb, 1));
- bg->bg_bits = cpu_to_le16(ocfs2_bits_per_group(cl));
bg->bg_chain = cpu_to_le16(my_chain);
bg->bg_next_group = cl->cl_recs[my_chain].c_blkno;
bg->bg_parent_dinode = cpu_to_le64(OCFS2_I(alloc_inode)->ip_blkno);
bg->bg_blkno = cpu_to_le64(group_blkno);
+ if (group_clusters == le16_to_cpu(cl->cl_cpg))
+ bg->bg_bits = cpu_to_le16(ocfs2_bits_per_group(cl));
+ else
+ ocfs2_bg_discontig_add_extent(osb, bg, cl, bg->bg_blkno,
+ group_clusters);
+
/* set the 1st bit in the bitmap to account for the descriptor block */
ocfs2_set_bit(0, (unsigned long *)bg->bg_bitmap);
bg->bg_free_bits_count = cpu_to_le16(le16_to_cpu(bg->bg_bits) - 1);
@@ -401,6 +429,220 @@ static inline u16 ocfs2_find_smallest_chain(struct ocfs2_chain_list *cl)
return best;
}
+static struct buffer_head *
+ocfs2_block_group_alloc_contig(struct ocfs2_super *osb, handle_t *handle,
+ struct inode *alloc_inode,
+ struct ocfs2_alloc_context *ac,
+ struct ocfs2_chain_list *cl)
+{
+ int status;
+ u32 bit_off, num_bits;
+ u64 bg_blkno;
+ struct buffer_head *bg_bh;
+ unsigned int alloc_rec = ocfs2_find_smallest_chain(cl);
+
+ status = ocfs2_claim_clusters(osb, handle, ac,
+ le16_to_cpu(cl->cl_cpg), &bit_off,
+ &num_bits);
+ if (status < 0) {
+ if (status != -ENOSPC)
+ mlog_errno(status);
+ goto bail;
+ }
+
+ /* setup the group */
+ bg_blkno = ocfs2_clusters_to_blocks(osb->sb, bit_off);
+ mlog(0, "new descriptor, record %u, at block %llu\n",
+ alloc_rec, (unsigned long long)bg_blkno);
+
+ bg_bh = sb_getblk(osb->sb, bg_blkno);
+ if (!bg_bh) {
+ status = -EIO;
+ mlog_errno(status);
+ goto bail;
+ }
+ ocfs2_set_new_buffer_uptodate(INODE_CACHE(alloc_inode), bg_bh);
+
+ status = ocfs2_block_group_fill(handle, alloc_inode, bg_bh,
+ bg_blkno, num_bits, alloc_rec, cl);
+ if (status < 0) {
+ brelse(bg_bh);
+ mlog_errno(status);
+ }
+
+bail:
+ return status ? ERR_PTR(status) : bg_bh;
+}
+
+static int ocfs2_block_group_claim_bits(struct ocfs2_super *osb,
+ handle_t *handle,
+ struct ocfs2_alloc_context *ac,
+ unsigned int min_bits,
+ u32 *bit_off, u32 *num_bits)
+{
+ int status;
+
+ while (min_bits) {
+ status = ocfs2_claim_clusters(osb, handle, ac, min_bits,
+ bit_off, num_bits);
+ if (status != -ENOSPC)
+ break;
+
+ min_bits >>= 1;
+ }
+
+ return status;
+}
+
+static int ocfs2_block_group_grow_discontig(handle_t *handle,
+ struct inode *alloc_inode,
+ struct buffer_head *bg_bh,
+ struct ocfs2_alloc_context *ac,
+ struct ocfs2_chain_list *cl,
+ unsigned int min_bits)
+{
+ int status;
+ struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb);
+ struct ocfs2_group_desc *bg =
+ (struct ocfs2_group_desc *)bg_bh->b_data;
+ unsigned int needed =
+ ocfs2_bits_per_group(cl) - le16_to_cpu(bg->bg_bits);
+ u32 p_cpos, clusters;
+ u64 p_blkno;
+ struct ocfs2_extent_list *el = &bg->bg_list;
+
+ status = ocfs2_journal_access_gd(handle,
+ INODE_CACHE(alloc_inode),
+ bg_bh,
+ OCFS2_JOURNAL_ACCESS_CREATE);
+ if (status < 0) {
+ mlog_errno(status);
+ goto bail;
+ }
+
+ while ((needed > 0) && (le16_to_cpu(el->l_next_free_rec) <
+ le16_to_cpu(el->l_count))) {
+ status = ocfs2_extend_trans(handle, OCFS2_SUBALLOC_ALLOC);
+ if (status) {
+ mlog_errno(status);
+ goto bail;
+ }
+
+ if (min_bits > needed)
+ min_bits = needed;
+ status = ocfs2_block_group_claim_bits(osb, handle, ac,
+ min_bits, &p_cpos,
+ &clusters);
+ if (status < 0) {
+ if (status != -ENOSPC)
+ mlog_errno(status);
+ goto bail;
+ }
+ p_blkno = ocfs2_clusters_to_blocks(osb->sb, p_cpos);
+ ocfs2_bg_discontig_add_extent(osb, bg, cl, p_blkno,
+ clusters);
+
+ min_bits = clusters;
+ needed = ocfs2_bits_per_group(cl) - le16_to_cpu(bg->bg_bits);
+ }
+
+ if (needed > 0) {
+ }
+
+ status = ocfs2_journal_dirty(handle, bg_bh);
+ if (status < 0)
+ mlog_errno(status);
+
+bail:
+ return status;
+}
+
+static void ocfs2_bg_alloc_cleanup(struct inode *alloc_inode,
+ struct buffer_head *bg_bh,
+ struct ocfs2_cached_dealloc_ctxt *dealloc)
+{
+ int i;
+ struct ocfs2_group_desc *bg;
+ struct ocfs2_extent_list *el;
+ struct ocfs2_extent_rec *rec;
+
+ if (!bg_bh)
+ return;
+
+ bg = (struct ocfs2_group_desc *)bg_bh->b_data;
+ el = &bg->bg_list;
+ for (i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
+ rec = &el->l_recs[i];
+ ocfs2_cache_cluster_dealloc(dealloc,
+ le64_to_cpu(rec->e_blkno),
+ le32_to_cpu(rec->e_leaf_clusters));
+ }
+
+ ocfs2_remove_from_cache(INODE_CACHE(alloc_inode), bg_bh);
+ brelse(bg_bh);
+}
+
+static struct buffer_head *
+ocfs2_block_group_alloc_discontig(handle_t *handle,
+ struct inode *alloc_inode,
+ struct ocfs2_alloc_context *ac,
+ struct ocfs2_chain_list *cl,
+ struct ocfs2_cached_dealloc_ctxt *dealloc)
+{
+ int status;
+ u32 bit_off, num_bits;
+ u64 bg_blkno;
+ unsigned int min_bits = le16_to_cpu(cl->cl_cpg) >> 1;
+ struct buffer_head *bg_bh = NULL;
+ unsigned int alloc_rec = ocfs2_find_smallest_chain(cl);
+ struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb);
+
+ if (!ocfs2_supports_discontig_bh(osb)) {
+ status = -ENOSPC;
+ goto bail;
+ }
+
+ /* Claim the first region */
+ status = ocfs2_block_group_claim_bits(osb, handle, ac, min_bits,
+ &bit_off, &num_bits);
+ if (status < 0) {
+ if (status != -ENOSPC)
+ mlog_errno(status);
+ goto bail;
+ }
+ min_bits = num_bits;
+
+ /* setup the group */
+ bg_blkno = ocfs2_clusters_to_blocks(osb->sb, bit_off);
+ mlog(0, "new descriptor, record %u, at block %llu\n",
+ alloc_rec, (unsigned long long)bg_blkno);
+
+ bg_bh = sb_getblk(osb->sb, bg_blkno);
+ if (!bg_bh) {
+ status = -EIO;
+ mlog_errno(status);
+ goto bail;
+ }
+ ocfs2_set_new_buffer_uptodate(INODE_CACHE(alloc_inode), bg_bh);
+
+ status = ocfs2_block_group_fill(handle, alloc_inode, bg_bh,
+ bg_blkno, num_bits, alloc_rec, cl);
+ if (status < 0) {
+ mlog_errno(status);
+ goto bail;
+ }
+
+ status = ocfs2_block_group_grow_discontig(handle, alloc_inode,
+ bg_bh, ac, cl, min_bits);
+ if (status)
+ mlog_errno(status);
+
+bail:
+ if (status)
+ ocfs2_bg_alloc_cleanup(alloc_inode, bg_bh, dealloc);
+ return status ? ERR_PTR(status) : bg_bh;
+}
+
/*
* We expect the block group allocator to already be locked.
*/
@@ -416,16 +658,17 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
struct ocfs2_chain_list *cl;
struct ocfs2_alloc_context *ac = NULL;
handle_t *handle = NULL;
- u32 bit_off, num_bits;
- u16 alloc_rec;
u64 bg_blkno;
struct buffer_head *bg_bh = NULL;
struct ocfs2_group_desc *bg;
+ struct ocfs2_cached_dealloc_ctxt dealloc;
BUG_ON(ocfs2_is_cluster_bitmap(alloc_inode));
mlog_entry_void();
+ ocfs2_init_dealloc_ctxt(&dealloc);
+
cl = &fe->id2.i_chain;
status = ocfs2_reserve_clusters_with_limit(osb,
le16_to_cpu(cl->cl_cpg),
@@ -451,44 +694,21 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
(unsigned long long)*last_alloc_group);
ac->ac_last_group = *last_alloc_group;
}
- status = ocfs2_claim_clusters(osb,
- handle,
- ac,
- le16_to_cpu(cl->cl_cpg),
- &bit_off,
- &num_bits);
- if (status < 0) {
+
+ bg_bh = ocfs2_block_group_alloc_contig(osb, handle, alloc_inode,
+ ac, cl);
+ if (IS_ERR(bg_bh) && (PTR_ERR(bg_bh) == -ENOSPC))
+ bg_bh = ocfs2_block_group_alloc_discontig(handle,
+ alloc_inode,
+ ac, cl,
+ &dealloc);
+ if (IS_ERR(bg_bh)) {
+ status = PTR_ERR(bg_bh);
+ bg_bh = NULL;
if (status != -ENOSPC)
mlog_errno(status);
goto bail;
}
-
- alloc_rec = ocfs2_find_smallest_chain(cl);
-
- /* setup the group */
- bg_blkno = ocfs2_clusters_to_blocks(osb->sb, bit_off);
- mlog(0, "new descriptor, record %u, at block %llu\n",
- alloc_rec, (unsigned long long)bg_blkno);
-
- bg_bh = sb_getblk(osb->sb, bg_blkno);
- if (!bg_bh) {
- status = -EIO;
- mlog_errno(status);
- goto bail;
- }
- ocfs2_set_new_buffer_uptodate(INODE_CACHE(alloc_inode), bg_bh);
-
- status = ocfs2_block_group_fill(handle,
- alloc_inode,
- bg_bh,
- bg_blkno,
- alloc_rec,
- cl);
- if (status < 0) {
- mlog_errno(status);
- goto bail;
- }
-
bg = (struct ocfs2_group_desc *) bg_bh->b_data;
status = ocfs2_journal_access_di(handle, INODE_CACHE(alloc_inode),
@@ -498,10 +718,11 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
goto bail;
}
- le32_add_cpu(&cl->cl_recs[alloc_rec].c_free,
+ le32_add_cpu(&cl->cl_recs[bg->bg_chain].c_free,
le16_to_cpu(bg->bg_free_bits_count));
- le32_add_cpu(&cl->cl_recs[alloc_rec].c_total, le16_to_cpu(bg->bg_bits));
- cl->cl_recs[alloc_rec].c_blkno = cpu_to_le64(bg_blkno);
+ le32_add_cpu(&cl->cl_recs[bg->bg_chain].c_total,
+ le16_to_cpu(bg->bg_bits));
+ cl->cl_recs[bg->bg_chain].c_blkno = cpu_to_le64(bg_blkno);
if (le16_to_cpu(cl->cl_next_free_rec) < le16_to_cpu(cl->cl_count))
le16_add_cpu(&cl->cl_next_free_rec, 1);
@@ -534,6 +755,11 @@ bail:
if (handle)
ocfs2_commit_trans(osb, handle);
+ if (ocfs2_dealloc_has_cluster(&dealloc)) {
+ ocfs2_schedule_truncate_log_flush(osb, 1);
+ ocfs2_run_deallocs(osb, &dealloc);
+ }
+
if (ac)
ocfs2_free_alloc_context(ac);
--
1.5.5
^ permalink raw reply related [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 03/15] ocfs2: Pass suballocation results back via a structure.
2010-04-01 2:53 [Ocfs2-devel] [PATCH 0/15] ocfs2: Support for discontiguous block group Tao Ma
2010-04-01 2:58 ` [Ocfs2-devel] [PATCH 01/15] ocfs2: Define data structures for discontiguous block groups Tao Ma
2010-04-01 2:58 ` [Ocfs2-devel] [PATCH 02/15] ocfs2: Allocate " Tao Ma
@ 2010-04-01 2:59 ` Tao Ma
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 04/15] ocfs2: Add suballoc_loc to metadata blocks Tao Ma
` (11 subsequent siblings)
14 siblings, 0 replies; 34+ messages in thread
From: Tao Ma @ 2010-04-01 2:59 UTC (permalink / raw)
To: ocfs2-devel
From: Joel Becker <joel.becker@oracle.com>
We're going to be adding more info to a suballocator allocation. Rather
than growing every function in the chain, let's pass a result structure
around.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
fs/ocfs2/suballoc.c | 165 +++++++++++++++++++++-----------------------------
fs/ocfs2/suballoc.h | 5 +-
2 files changed, 72 insertions(+), 98 deletions(-)
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 267d552..8bb20bb 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -53,6 +53,12 @@
#define OCFS2_MAX_TO_STEAL 1024
+struct ocfs2_suballoc_result {
+ u64 sr_bg_blkno; /* The bg we allocated from */
+ unsigned int sr_bit_offset; /* The bit in the bg */
+ unsigned int sr_bits; /* How many bits we claimed */
+};
+
static inline void ocfs2_debug_bg(struct ocfs2_group_desc *bg);
static inline void ocfs2_debug_suballoc_inode(struct ocfs2_dinode *fe);
static inline u16 ocfs2_find_victim_chain(struct ocfs2_chain_list *cl);
@@ -74,20 +80,18 @@ static int ocfs2_cluster_group_search(struct inode *inode,
struct buffer_head *group_bh,
u32 bits_wanted, u32 min_bits,
u64 max_block,
- u16 *bit_off, u16 *bits_found);
+ struct ocfs2_suballoc_result *res);
static int ocfs2_block_group_search(struct inode *inode,
struct buffer_head *group_bh,
u32 bits_wanted, u32 min_bits,
u64 max_block,
- u16 *bit_off, u16 *bits_found);
+ struct ocfs2_suballoc_result *res);
static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
struct ocfs2_alloc_context *ac,
handle_t *handle,
u32 bits_wanted,
u32 min_bits,
- u16 *bit_off,
- unsigned int *num_bits,
- u64 *bg_blkno);
+ struct ocfs2_suballoc_result *res);
static int ocfs2_test_bg_bit_allocatable(struct buffer_head *bg_bh,
int nr);
static inline int ocfs2_block_group_set_bits(handle_t *handle,
@@ -1263,8 +1267,7 @@ static int ocfs2_block_group_find_clear_bits(struct ocfs2_super *osb,
struct buffer_head *bg_bh,
unsigned int bits_wanted,
unsigned int total_bits,
- u16 *bit_off,
- u16 *bits_found)
+ struct ocfs2_suballoc_result *res)
{
void *bitmap;
u16 best_offset, best_size;
@@ -1308,14 +1311,9 @@ static int ocfs2_block_group_find_clear_bits(struct ocfs2_super *osb,
}
}
- /* XXX: I think the first clause is equivalent to the second
- * - jlbec */
- if (found == bits_wanted) {
- *bit_off = start - found;
- *bits_found = found;
- } else if (best_size) {
- *bit_off = best_offset;
- *bits_found = best_size;
+ if (best_size) {
+ res->sr_bit_offset = best_offset;
+ res->sr_bits = best_size;
} else {
status = -ENOSPC;
/* No error log here -- see the comment above
@@ -1493,14 +1491,13 @@ static int ocfs2_cluster_group_search(struct inode *inode,
struct buffer_head *group_bh,
u32 bits_wanted, u32 min_bits,
u64 max_block,
- u16 *bit_off, u16 *bits_found)
+ struct ocfs2_suballoc_result *res)
{
int search = -ENOSPC;
int ret;
u64 blkoff;
struct ocfs2_group_desc *gd = (struct ocfs2_group_desc *) group_bh->b_data;
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
- u16 tmp_off, tmp_found;
unsigned int max_bits, gd_cluster_off;
BUG_ON(!ocfs2_is_cluster_bitmap(inode));
@@ -1527,15 +1524,15 @@ static int ocfs2_cluster_group_search(struct inode *inode,
ret = ocfs2_block_group_find_clear_bits(OCFS2_SB(inode->i_sb),
group_bh, bits_wanted,
- max_bits,
- &tmp_off, &tmp_found);
+ max_bits, res);
if (ret)
return ret;
if (max_block) {
blkoff = ocfs2_clusters_to_blocks(inode->i_sb,
gd_cluster_off +
- tmp_off + tmp_found);
+ res->sr_bit_offset +
+ res->sr_bits);
mlog(0, "Checking %llu against %llu\n",
(unsigned long long)blkoff,
(unsigned long long)max_block);
@@ -1547,16 +1544,14 @@ static int ocfs2_cluster_group_search(struct inode *inode,
* return success, but we still want to return
* -ENOSPC unless it found the minimum number
* of bits. */
- if (min_bits <= tmp_found) {
- *bit_off = tmp_off;
- *bits_found = tmp_found;
+ if (min_bits <= res->sr_bits)
search = 0; /* success */
- } else if (tmp_found) {
+ else if (res->sr_bits) {
/*
* Don't show bits which we'll be returning
* for allocation to the local alloc bitmap.
*/
- ocfs2_local_alloc_seen_free_bits(osb, tmp_found);
+ ocfs2_local_alloc_seen_free_bits(osb, res->sr_bits);
}
}
@@ -1567,7 +1562,7 @@ static int ocfs2_block_group_search(struct inode *inode,
struct buffer_head *group_bh,
u32 bits_wanted, u32 min_bits,
u64 max_block,
- u16 *bit_off, u16 *bits_found)
+ struct ocfs2_suballoc_result *res)
{
int ret = -ENOSPC;
u64 blkoff;
@@ -1580,10 +1575,10 @@ static int ocfs2_block_group_search(struct inode *inode,
ret = ocfs2_block_group_find_clear_bits(OCFS2_SB(inode->i_sb),
group_bh, bits_wanted,
le16_to_cpu(bg->bg_bits),
- bit_off, bits_found);
+ res);
if (!ret && max_block) {
- blkoff = le64_to_cpu(bg->bg_blkno) + *bit_off +
- *bits_found;
+ blkoff = le64_to_cpu(bg->bg_blkno) +
+ res->sr_bit_offset + res->sr_bits;
mlog(0, "Checking %llu against %llu\n",
(unsigned long long)blkoff,
(unsigned long long)max_block);
@@ -1629,20 +1624,17 @@ static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
handle_t *handle,
u32 bits_wanted,
u32 min_bits,
- u16 *bit_off,
- unsigned int *num_bits,
- u64 gd_blkno,
+ struct ocfs2_suballoc_result *res,
u16 *bits_left)
{
int ret;
- u16 found;
struct buffer_head *group_bh = NULL;
struct ocfs2_group_desc *gd;
struct ocfs2_dinode *di = (struct ocfs2_dinode *)ac->ac_bh->b_data;
struct inode *alloc_inode = ac->ac_inode;
- ret = ocfs2_read_group_descriptor(alloc_inode, di, gd_blkno,
- &group_bh);
+ ret = ocfs2_read_group_descriptor(alloc_inode, di,
+ res->sr_bg_blkno, &group_bh);
if (ret < 0) {
mlog_errno(ret);
return ret;
@@ -1650,17 +1642,15 @@ static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
gd = (struct ocfs2_group_desc *) group_bh->b_data;
ret = ac->ac_group_search(alloc_inode, group_bh, bits_wanted, min_bits,
- ac->ac_max_block, bit_off, &found);
+ ac->ac_max_block, res);
if (ret < 0) {
if (ret != -ENOSPC)
mlog_errno(ret);
goto out;
}
- *num_bits = found;
-
ret = ocfs2_alloc_dinode_update_counts(alloc_inode, handle, ac->ac_bh,
- *num_bits,
+ res->sr_bits,
le16_to_cpu(gd->bg_chain));
if (ret < 0) {
mlog_errno(ret);
@@ -1668,7 +1658,7 @@ static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
}
ret = ocfs2_block_group_set_bits(handle, alloc_inode, gd, group_bh,
- *bit_off, *num_bits);
+ res->sr_bit_offset, res->sr_bits);
if (ret < 0)
mlog_errno(ret);
@@ -1684,13 +1674,11 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
handle_t *handle,
u32 bits_wanted,
u32 min_bits,
- u16 *bit_off,
- unsigned int *num_bits,
- u64 *bg_blkno,
+ struct ocfs2_suballoc_result *res,
u16 *bits_left)
{
int status;
- u16 chain, tmp_bits;
+ u16 chain;
u32 tmp_used;
u64 next_group;
struct inode *alloc_inode = ac->ac_inode;
@@ -1719,8 +1707,8 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
* the 1st group with any empty bits. */
while ((status = ac->ac_group_search(alloc_inode, group_bh,
bits_wanted, min_bits,
- ac->ac_max_block, bit_off,
- &tmp_bits)) == -ENOSPC) {
+ ac->ac_max_block,
+ res)) == -ENOSPC) {
if (!bg->bg_next_group)
break;
@@ -1745,11 +1733,11 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
}
mlog(0, "alloc succeeds: we give %u bits from block group %llu\n",
- tmp_bits, (unsigned long long)le64_to_cpu(bg->bg_blkno));
+ res->sr_bits, (unsigned long long)le64_to_cpu(bg->bg_blkno));
- *num_bits = tmp_bits;
+ res->sr_bg_blkno = le64_to_cpu(bg->bg_blkno);
- BUG_ON(*num_bits == 0);
+ BUG_ON(res->sr_bits == 0);
/*
* Keep track of previous block descriptor read. When
@@ -1766,7 +1754,7 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
*/
if (ac->ac_allow_chain_relink &&
(prev_group_bh) &&
- (ocfs2_block_group_reasonably_empty(bg, *num_bits))) {
+ (ocfs2_block_group_reasonably_empty(bg, res->sr_bits))) {
status = ocfs2_relink_block_group(handle, alloc_inode,
ac->ac_bh, group_bh,
prev_group_bh, chain);
@@ -1788,8 +1776,8 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
}
tmp_used = le32_to_cpu(fe->id1.bitmap1.i_used);
- fe->id1.bitmap1.i_used = cpu_to_le32(*num_bits + tmp_used);
- le32_add_cpu(&cl->cl_recs[chain].c_free, -(*num_bits));
+ fe->id1.bitmap1.i_used = cpu_to_le32(res->sr_bits + tmp_used);
+ le32_add_cpu(&cl->cl_recs[chain].c_free, -res->sr_bits);
status = ocfs2_journal_dirty(handle,
ac->ac_bh);
@@ -1802,17 +1790,16 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
alloc_inode,
bg,
group_bh,
- *bit_off,
- *num_bits);
+ res->sr_bit_offset,
+ res->sr_bits);
if (status < 0) {
mlog_errno(status);
goto bail;
}
- mlog(0, "Allocated %u bits from suballocator %llu\n", *num_bits,
+ mlog(0, "Allocated %u bits from suballocator %llu\n", res->sr_bits,
(unsigned long long)le64_to_cpu(fe->i_blkno));
- *bg_blkno = le64_to_cpu(bg->bg_blkno);
*bits_left = le16_to_cpu(bg->bg_free_bits_count);
bail:
brelse(group_bh);
@@ -1828,14 +1815,11 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
handle_t *handle,
u32 bits_wanted,
u32 min_bits,
- u16 *bit_off,
- unsigned int *num_bits,
- u64 *bg_blkno)
+ struct ocfs2_suballoc_result *res)
{
int status;
u16 victim, i;
u16 bits_left = 0;
- u64 hint_blkno = ac->ac_last_group;
struct ocfs2_chain_list *cl;
struct ocfs2_dinode *fe;
@@ -1862,22 +1846,16 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
goto bail;
}
- if (hint_blkno) {
+ res->sr_bg_blkno = ac->ac_last_group;
+ if (res->sr_bg_blkno) {
/* Attempt to short-circuit the usual search mechanism
* by jumping straight to the most recently used
* allocation group. This helps us mantain some
* contiguousness across allocations. */
status = ocfs2_search_one_group(ac, handle, bits_wanted,
- min_bits, bit_off, num_bits,
- hint_blkno, &bits_left);
- if (!status) {
- /* Be careful to update *bg_blkno here as the
- * caller is expecting it to be filled in, and
- * ocfs2_search_one_group() won't do that for
- * us. */
- *bg_blkno = hint_blkno;
+ min_bits, res, &bits_left);
+ if (!status)
goto set_hint;
- }
if (status < 0 && status != -ENOSPC) {
mlog_errno(status);
goto bail;
@@ -1890,8 +1868,8 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
ac->ac_chain = victim;
ac->ac_allow_chain_relink = 1;
- status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits, bit_off,
- num_bits, bg_blkno, &bits_left);
+ status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits,
+ res, &bits_left);
if (!status)
goto set_hint;
if (status < 0 && status != -ENOSPC) {
@@ -1915,8 +1893,7 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
ac->ac_chain = i;
status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits,
- bit_off, num_bits, bg_blkno,
- &bits_left);
+ res, &bits_left);
if (!status)
break;
if (status < 0 && status != -ENOSPC) {
@@ -1933,7 +1910,7 @@ set_hint:
if (bits_left < min_bits)
ac->ac_last_group = 0;
else
- ac->ac_last_group = *bg_blkno;
+ ac->ac_last_group = res->sr_bg_blkno;
}
bail:
@@ -1950,7 +1927,7 @@ int ocfs2_claim_metadata(struct ocfs2_super *osb,
u64 *blkno_start)
{
int status;
- u64 bg_blkno;
+ struct ocfs2_suballoc_result res;
BUG_ON(!ac);
BUG_ON(ac->ac_bits_wanted < (ac->ac_bits_given + bits_wanted));
@@ -1961,17 +1938,17 @@ int ocfs2_claim_metadata(struct ocfs2_super *osb,
handle,
bits_wanted,
1,
- suballoc_bit_start,
- num_bits,
- &bg_blkno);
+ &res);
if (status < 0) {
mlog_errno(status);
goto bail;
}
atomic_inc(&osb->alloc_stats.bg_allocs);
- *blkno_start = bg_blkno + (u64) *suballoc_bit_start;
- ac->ac_bits_given += (*num_bits);
+ *suballoc_bit_start = res.sr_bit_offset;
+ *blkno_start = res.sr_bg_blkno + (u64)(res.sr_bit_offset);
+ ac->ac_bits_given += res.sr_bits;
+ *num_bits = res.sr_bits;
status = 0;
bail:
mlog_exit(status);
@@ -2018,8 +1995,7 @@ int ocfs2_claim_new_inode(struct ocfs2_super *osb,
u64 *fe_blkno)
{
int status;
- unsigned int num_bits;
- u64 bg_blkno;
+ struct ocfs2_suballoc_result res;
mlog_entry_void();
@@ -2035,18 +2011,17 @@ int ocfs2_claim_new_inode(struct ocfs2_super *osb,
handle,
1,
1,
- suballoc_bit,
- &num_bits,
- &bg_blkno);
+ &res);
if (status < 0) {
mlog_errno(status);
goto bail;
}
atomic_inc(&osb->alloc_stats.bg_allocs);
- BUG_ON(num_bits != 1);
+ BUG_ON(res.sr_bits != 1);
- *fe_blkno = bg_blkno + (u64) (*suballoc_bit);
+ *suballoc_bit = res.sr_bit_offset;
+ *fe_blkno = res.sr_bg_blkno + (u64)(res.sr_bit_offset);
ac->ac_bits_given++;
ocfs2_save_inode_ac_group(dir, ac);
status = 0;
@@ -2126,8 +2101,7 @@ int __ocfs2_claim_clusters(struct ocfs2_super *osb,
{
int status;
unsigned int bits_wanted = max_clusters;
- u64 bg_blkno = 0;
- u16 bg_bit_off;
+ struct ocfs2_suballoc_result res;
mlog_entry_void();
@@ -2164,14 +2138,12 @@ int __ocfs2_claim_clusters(struct ocfs2_super *osb,
handle,
bits_wanted,
min_clusters,
- &bg_bit_off,
- num_clusters,
- &bg_blkno);
+ &res);
if (!status) {
*cluster_start =
ocfs2_desc_bitmap_to_cluster_off(ac->ac_inode,
- bg_blkno,
- bg_bit_off);
+ res.sr_bg_blkno,
+ res.sr_bit_offset);
atomic_inc(&osb->alloc_stats.bitmap_data);
}
}
@@ -2181,7 +2153,8 @@ int __ocfs2_claim_clusters(struct ocfs2_super *osb,
goto bail;
}
- ac->ac_bits_given += *num_clusters;
+ ac->ac_bits_given += res.sr_bits;
+ *num_clusters = res.sr_bits;
bail:
mlog_exit(status);
diff --git a/fs/ocfs2/suballoc.h b/fs/ocfs2/suballoc.h
index fa60723..fbf7f17 100644
--- a/fs/ocfs2/suballoc.h
+++ b/fs/ocfs2/suballoc.h
@@ -26,13 +26,14 @@
#ifndef _CHAINALLOC_H_
#define _CHAINALLOC_H_
+struct ocfs2_suballoc_result;
typedef int (group_search_t)(struct inode *,
struct buffer_head *,
u32, /* bits_wanted */
u32, /* min_bits */
u64, /* max_block */
- u16 *, /* *bit_off */
- u16 *); /* *bits_found */
+ struct ocfs2_suballoc_result *);
+ /* found bits */
struct ocfs2_alloc_context {
struct inode *ac_inode; /* which bitmap are we allocating from? */
--
1.5.5
^ permalink raw reply related [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 04/15] ocfs2: Add suballoc_loc to metadata blocks.
2010-04-01 2:53 [Ocfs2-devel] [PATCH 0/15] ocfs2: Support for discontiguous block group Tao Ma
` (2 preceding siblings ...)
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 03/15] ocfs2: Pass suballocation results back via a structure Tao Ma
@ 2010-04-01 2:59 ` Tao Ma
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 05/15] ocfs2: ocfs2_claim_suballoc_bits() doesn't need an osb argument Tao Ma
` (10 subsequent siblings)
14 siblings, 0 replies; 34+ messages in thread
From: Tao Ma @ 2010-04-01 2:59 UTC (permalink / raw)
To: ocfs2-devel
From: Joel Becker <joel.becker@oracle.com>
We need a suballoc_loc field on any suballocated block. Define them.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
fs/ocfs2/ocfs2_fs.h | 29 ++++++++++++++++++++++++-----
1 files changed, 24 insertions(+), 5 deletions(-)
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index 0508e4e..42641f8 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -515,7 +515,10 @@ struct ocfs2_extent_block
block group */
__le32 h_fs_generation; /* Must match super block */
__le64 h_blkno; /* Offset on disk, in blocks */
-/*20*/ __le64 h_reserved3;
+/*20*/ __le64 h_suballoc_loc; /* Suballocator block group this
+ eb belongs to. Only valid
+ if allocated from a
+ discontiguous block group */
__le64 h_next_leaf_blk; /* Offset on disk, in blocks,
of next leaf header pointing
to data */
@@ -682,7 +685,11 @@ struct ocfs2_dinode {
/*80*/ struct ocfs2_block_check i_check; /* Error checking */
/*88*/ __le64 i_dx_root; /* Pointer to dir index root block */
/*90*/ __le64 i_refcount_loc;
- __le64 i_reserved2[4];
+ __le64 i_suballoc_loc; /* Suballocator block group this
+ inode belongs to. Only valid
+ if allocated from a
+ discontiguous block group */
+/*A0*/ __le64 i_reserved2[3];
/*B8*/ union {
__le64 i_pad1; /* Generic way to refer to this
64bit union */
@@ -817,7 +824,12 @@ struct ocfs2_dx_root_block {
__le32 dr_reserved2;
__le64 dr_free_blk; /* Pointer to head of free
* unindexed block list. */
- __le64 dr_reserved3[15];
+ __le64 dr_suballoc_loc; /* Suballocator block group
+ this root belongs to.
+ Only valid if allocated
+ from a discontiguous
+ block group */
+ __le64 dr_reserved3[14];
union {
struct ocfs2_extent_list dr_list; /* Keep this aligned to 128
* bits for maximum space
@@ -937,7 +949,11 @@ struct ocfs2_refcount_block {
/*40*/ __le32 rf_generation; /* generation number. all be the same
* for the same refcount tree. */
__le32 rf_reserved0;
- __le64 rf_reserved1[7];
+ __le64 rf_suballoc_loc; /* Suballocator block group this
+ refcount block belongs to. Only
+ valid if allocated from a
+ discontiguous block group */
+/*50*/ __le64 rf_reserved1[6];
/*80*/ union {
struct ocfs2_refcount_list rf_records; /* List of refcount
records */
@@ -1049,7 +1065,10 @@ struct ocfs2_xattr_block {
real xattr or a xattr tree. */
__le16 xb_reserved0;
__le32 xb_reserved1;
- __le64 xb_reserved2;
+ __le64 xb_suballoc_loc; /* Suballocator block group this
+ xattr block belongs to. Only
+ valid if allocated from a
+ discontiguous block group */
/*30*/ union {
struct ocfs2_xattr_header xb_header; /* xattr header if this
block contains xattr */
--
1.5.5
^ permalink raw reply related [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 05/15] ocfs2: ocfs2_claim_suballoc_bits() doesn't need an osb argument.
2010-04-01 2:53 [Ocfs2-devel] [PATCH 0/15] ocfs2: Support for discontiguous block group Tao Ma
` (3 preceding siblings ...)
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 04/15] ocfs2: Add suballoc_loc to metadata blocks Tao Ma
@ 2010-04-01 2:59 ` Tao Ma
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 06/15] ocfs2: Trim suballocations if they cross discontiguous regions Tao Ma
` (9 subsequent siblings)
14 siblings, 0 replies; 34+ messages in thread
From: Tao Ma @ 2010-04-01 2:59 UTC (permalink / raw)
To: ocfs2-devel
From: Joel Becker <joel.becker@oracle.com>
It's contained on ac->ac_inode->i_sb anyway.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
fs/ocfs2/suballoc.c | 18 +++++++-----------
1 files changed, 7 insertions(+), 11 deletions(-)
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 8bb20bb..298fb54 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -86,8 +86,7 @@ static int ocfs2_block_group_search(struct inode *inode,
u32 bits_wanted, u32 min_bits,
u64 max_block,
struct ocfs2_suballoc_result *res);
-static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
- struct ocfs2_alloc_context *ac,
+static int ocfs2_claim_suballoc_bits(struct ocfs2_alloc_context *ac,
handle_t *handle,
u32 bits_wanted,
u32 min_bits,
@@ -1810,8 +1809,7 @@ bail:
}
/* will give out up to bits_wanted contiguous bits. */
-static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
- struct ocfs2_alloc_context *ac,
+static int ocfs2_claim_suballoc_bits(struct ocfs2_alloc_context *ac,
handle_t *handle,
u32 bits_wanted,
u32 min_bits,
@@ -1837,7 +1835,8 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
if (le32_to_cpu(fe->id1.bitmap1.i_used) >=
le32_to_cpu(fe->id1.bitmap1.i_total)) {
- ocfs2_error(osb->sb, "Chain allocator dinode %llu has %u used "
+ ocfs2_error(ac->ac_inode->i_sb,
+ "Chain allocator dinode %llu has %u used "
"bits but only %u total.",
(unsigned long long)le64_to_cpu(fe->i_blkno),
le32_to_cpu(fe->id1.bitmap1.i_used),
@@ -1933,8 +1932,7 @@ int ocfs2_claim_metadata(struct ocfs2_super *osb,
BUG_ON(ac->ac_bits_wanted < (ac->ac_bits_given + bits_wanted));
BUG_ON(ac->ac_which != OCFS2_AC_USE_META);
- status = ocfs2_claim_suballoc_bits(osb,
- ac,
+ status = ocfs2_claim_suballoc_bits(ac,
handle,
bits_wanted,
1,
@@ -2006,8 +2004,7 @@ int ocfs2_claim_new_inode(struct ocfs2_super *osb,
ocfs2_init_inode_ac_group(dir, parent_fe_bh, ac);
- status = ocfs2_claim_suballoc_bits(osb,
- ac,
+ status = ocfs2_claim_suballoc_bits(ac,
handle,
1,
1,
@@ -2133,8 +2130,7 @@ int __ocfs2_claim_clusters(struct ocfs2_super *osb,
if (bits_wanted > (osb->bitmap_cpg - 1))
bits_wanted = osb->bitmap_cpg - 1;
- status = ocfs2_claim_suballoc_bits(osb,
- ac,
+ status = ocfs2_claim_suballoc_bits(ac,
handle,
bits_wanted,
min_clusters,
--
1.5.5
^ permalink raw reply related [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 06/15] ocfs2: Trim suballocations if they cross discontiguous regions
2010-04-01 2:53 [Ocfs2-devel] [PATCH 0/15] ocfs2: Support for discontiguous block group Tao Ma
` (4 preceding siblings ...)
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 05/15] ocfs2: ocfs2_claim_suballoc_bits() doesn't need an osb argument Tao Ma
@ 2010-04-01 2:59 ` Tao Ma
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 07/15] ocfs2: ocfs2_claim_*() don't need an ocfs2_super argument Tao Ma
` (8 subsequent siblings)
14 siblings, 0 replies; 34+ messages in thread
From: Tao Ma @ 2010-04-01 2:59 UTC (permalink / raw)
To: ocfs2-devel
From: Joel Becker <joel.becker@oracle.com>
A discontiguous block group can find a range of free bits that straddle
more than one region of its space. Callers can't handle that, so we
trim the returned bits until they fit within one region.
Only cluster allocations ask for min_bits>1. Discontiguous block groups
are only for block allocations. So min_bits doesn't matter here.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
fs/ocfs2/suballoc.c | 48 ++++++++++++++++++++++++++++++++++++++++++++++++
1 files changed, 48 insertions(+), 0 deletions(-)
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 298fb54..1910f09 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -1619,6 +1619,48 @@ out:
return ret;
}
+static int ocfs2_bg_discontig_trim_by_rec(struct ocfs2_suballoc_result *res,
+ struct ocfs2_extent_rec *rec,
+ struct ocfs2_chain_list *cl)
+{
+ unsigned int bpc = le16_to_cpu(cl->cl_bpc);
+ unsigned int bitoff = le32_to_cpu(rec->e_cpos) * bpc;
+ unsigned int bitcount = le32_to_cpu(rec->e_leaf_clusters) * bpc;
+
+ if (res->sr_bit_offset < bitoff)
+ return 0;
+ if (res->sr_bit_offset >= (bitoff + bitcount))
+ return 0;
+ if ((res->sr_bit_offset + res->sr_bits) > (bitoff + bitcount))
+ res->sr_bits = (bitoff + bitcount) - res->sr_bit_offset;
+ return 1;
+}
+
+static void ocfs2_bg_discontig_trim_result(struct ocfs2_alloc_context *ac,
+ struct ocfs2_group_desc *bg,
+ struct ocfs2_suballoc_result *res)
+{
+ int i;
+ struct ocfs2_extent_rec *rec;
+ struct ocfs2_dinode *di = (struct ocfs2_dinode *)ac->ac_bh->b_data;
+ struct ocfs2_chain_list *cl = &di->id2.i_chain;
+
+ if (!ocfs2_supports_discontig_bh(OCFS2_SB(ac->ac_inode->i_sb)))
+ return;
+
+ if (ocfs2_is_cluster_bitmap(ac->ac_inode))
+ return;
+
+ if (!bg->bg_list.l_next_free_rec)
+ return;
+
+ for (i = 0; i < le16_to_cpu(bg->bg_list.l_next_free_rec); i++) {
+ rec = &bg->bg_list.l_recs[i];
+ if (ocfs2_bg_discontig_trim_by_rec(res, rec, cl))
+ break;
+ }
+}
+
static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
handle_t *handle,
u32 bits_wanted,
@@ -1648,6 +1690,9 @@ static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
goto out;
}
+ if (!ret)
+ ocfs2_bg_discontig_trim_result(ac, gd, res);
+
ret = ocfs2_alloc_dinode_update_counts(alloc_inode, handle, ac->ac_bh,
res->sr_bits,
le16_to_cpu(gd->bg_chain));
@@ -1737,6 +1782,9 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
res->sr_bg_blkno = le64_to_cpu(bg->bg_blkno);
BUG_ON(res->sr_bits == 0);
+ if (!status)
+ ocfs2_bg_discontig_trim_result(ac, bg, res);
+
/*
* Keep track of previous block descriptor read. When
--
1.5.5
^ permalink raw reply related [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 07/15] ocfs2: ocfs2_claim_*() don't need an ocfs2_super argument.
2010-04-01 2:53 [Ocfs2-devel] [PATCH 0/15] ocfs2: Support for discontiguous block group Tao Ma
` (5 preceding siblings ...)
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 06/15] ocfs2: Trim suballocations if they cross discontiguous regions Tao Ma
@ 2010-04-01 2:59 ` Tao Ma
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 08/15] ocfs2: Return allocated metadata blknos on the ocfs2_suballoc_result Tao Ma
` (7 subsequent siblings)
14 siblings, 0 replies; 34+ messages in thread
From: Tao Ma @ 2010-04-01 2:59 UTC (permalink / raw)
To: ocfs2-devel
From: Joel Becker <joel.becker@oracle.com>
They all take an ocfs2_alloc_context, which has the allocation inode.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
fs/ocfs2/alloc.c | 7 +++----
fs/ocfs2/dir.c | 8 ++++----
fs/ocfs2/localalloc.c | 4 ++--
fs/ocfs2/namei.c | 2 +-
fs/ocfs2/refcounttree.c | 8 ++++----
fs/ocfs2/suballoc.c | 23 ++++++++++-------------
fs/ocfs2/suballoc.h | 12 ++++--------
fs/ocfs2/xattr.c | 13 ++++++-------
8 files changed, 34 insertions(+), 43 deletions(-)
diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
index 9f8bd91..57d7f7a 100644
--- a/fs/ocfs2/alloc.c
+++ b/fs/ocfs2/alloc.c
@@ -1015,8 +1015,7 @@ static int ocfs2_create_new_meta_bhs(handle_t *handle,
count = 0;
while (count < wanted) {
- status = ocfs2_claim_metadata(osb,
- handle,
+ status = ocfs2_claim_metadata(handle,
meta_ac,
wanted - count,
&suballoc_bit_start,
@@ -4866,7 +4865,7 @@ int ocfs2_add_clusters_in_btree(handle_t *handle,
goto leave;
}
- status = __ocfs2_claim_clusters(osb, handle, data_ac, 1,
+ status = __ocfs2_claim_clusters(handle, data_ac, 1,
clusters_to_add, &bit_off, &num_bits);
if (status < 0) {
if (status != -ENOSPC)
@@ -7307,7 +7306,7 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode,
goto out_commit;
did_quota = 1;
- ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off,
+ ret = ocfs2_claim_clusters(handle, data_ac, 1, &bit_off,
&num);
if (ret) {
mlog_errno(ret);
diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c
index efd77d0..06522b5 100644
--- a/fs/ocfs2/dir.c
+++ b/fs/ocfs2/dir.c
@@ -2411,7 +2411,7 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb,
struct ocfs2_dir_block_trailer *trailer =
ocfs2_trailer_from_bh(dirdata_bh, dir->i_sb);
- ret = ocfs2_claim_metadata(osb, handle, meta_ac, 1, &dr_suballoc_bit,
+ ret = ocfs2_claim_metadata(handle, meta_ac, 1, &dr_suballoc_bit,
&num_bits, &dr_blkno);
if (ret) {
mlog_errno(ret);
@@ -2558,7 +2558,7 @@ static int __ocfs2_dx_dir_new_cluster(struct inode *dir,
* chance of contiguousness as the directory grows in number
* of entries.
*/
- ret = __ocfs2_claim_clusters(osb, handle, data_ac, 1, 1, &phys, &num);
+ ret = __ocfs2_claim_clusters(handle, data_ac, 1, 1, &phys, &num);
if (ret) {
mlog_errno(ret);
goto out;
@@ -2991,7 +2991,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
* if we only get one now, that's enough to continue. The rest
* will be claimed after the conversion to extents.
*/
- ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off, &len);
+ ret = ocfs2_claim_clusters(handle, data_ac, 1, &bit_off, &len);
if (ret) {
mlog_errno(ret);
goto out_commit;
@@ -3138,7 +3138,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
* pass. Claim the 2nd cluster as a separate extent.
*/
if (alloc > len) {
- ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off,
+ ret = ocfs2_claim_clusters(handle, data_ac, 1, &bit_off,
&len);
if (ret) {
mlog_errno(ret);
diff --git a/fs/ocfs2/localalloc.c b/fs/ocfs2/localalloc.c
index ca992d9..e51d75c 100644
--- a/fs/ocfs2/localalloc.c
+++ b/fs/ocfs2/localalloc.c
@@ -1048,7 +1048,7 @@ static int ocfs2_local_alloc_new_window(struct ocfs2_super *osb,
/* we used the generic suballoc reserve function, but we set
* everything up nicely, so there's no reason why we can't use
* the more specific cluster api to claim bits. */
- status = ocfs2_claim_clusters(osb, handle, ac, osb->local_alloc_bits,
+ status = ocfs2_claim_clusters(handle, ac, osb->local_alloc_bits,
&cluster_off, &cluster_count);
if (status == -ENOSPC) {
retry_enospc:
@@ -1061,7 +1061,7 @@ retry_enospc:
OCFS2_LA_DISABLED)
goto bail;
- status = ocfs2_claim_clusters(osb, handle, ac,
+ status = ocfs2_claim_clusters(handle, ac,
osb->local_alloc_bits,
&cluster_off,
&cluster_count);
diff --git a/fs/ocfs2/namei.c b/fs/ocfs2/namei.c
index d9cd4e3..9c3e1fe 100644
--- a/fs/ocfs2/namei.c
+++ b/fs/ocfs2/namei.c
@@ -482,7 +482,7 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb,
*new_fe_bh = NULL;
- status = ocfs2_claim_new_inode(osb, handle, dir, parent_fe_bh,
+ status = ocfs2_claim_new_inode(handle, dir, parent_fe_bh,
inode_ac, &suballoc_bit, &fe_blkno);
if (status < 0) {
mlog_errno(status);
diff --git a/fs/ocfs2/refcounttree.c b/fs/ocfs2/refcounttree.c
index 9e96921..4fb18af 100644
--- a/fs/ocfs2/refcounttree.c
+++ b/fs/ocfs2/refcounttree.c
@@ -597,7 +597,7 @@ static int ocfs2_create_refcount_tree(struct inode *inode,
goto out_commit;
}
- ret = ocfs2_claim_metadata(osb, handle, meta_ac, 1,
+ ret = ocfs2_claim_metadata(handle, meta_ac, 1,
&suballoc_bit_start, &num_got,
&first_blkno);
if (ret) {
@@ -1299,7 +1299,7 @@ static int ocfs2_expand_inline_ref_root(handle_t *handle,
goto out;
}
- ret = ocfs2_claim_metadata(OCFS2_SB(sb), handle, meta_ac, 1,
+ ret = ocfs2_claim_metadata(handle, meta_ac, 1,
&suballoc_bit_start, &num_got,
&blkno);
if (ret) {
@@ -1549,7 +1549,7 @@ static int ocfs2_new_leaf_refcount_block(handle_t *handle,
goto out;
}
- ret = ocfs2_claim_metadata(OCFS2_SB(sb), handle, meta_ac, 1,
+ ret = ocfs2_claim_metadata(handle, meta_ac, 1,
&suballoc_bit_start, &num_got,
&blkno);
if (ret) {
@@ -3283,7 +3283,7 @@ static int ocfs2_make_clusters_writable(struct super_block *sb,
} else {
delete = 1;
- ret = __ocfs2_claim_clusters(osb, handle,
+ ret = __ocfs2_claim_clusters(handle,
context->data_ac,
1, set_len,
&new_bit, &new_len);
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 1910f09..0fd0357 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -444,7 +444,7 @@ ocfs2_block_group_alloc_contig(struct ocfs2_super *osb, handle_t *handle,
struct buffer_head *bg_bh;
unsigned int alloc_rec = ocfs2_find_smallest_chain(cl);
- status = ocfs2_claim_clusters(osb, handle, ac,
+ status = ocfs2_claim_clusters(handle, ac,
le16_to_cpu(cl->cl_cpg), &bit_off,
&num_bits);
if (status < 0) {
@@ -486,7 +486,7 @@ static int ocfs2_block_group_claim_bits(struct ocfs2_super *osb,
int status;
while (min_bits) {
- status = ocfs2_claim_clusters(osb, handle, ac, min_bits,
+ status = ocfs2_claim_clusters(handle, ac, min_bits,
bit_off, num_bits);
if (status != -ENOSPC)
break;
@@ -1965,8 +1965,7 @@ bail:
return status;
}
-int ocfs2_claim_metadata(struct ocfs2_super *osb,
- handle_t *handle,
+int ocfs2_claim_metadata(handle_t *handle,
struct ocfs2_alloc_context *ac,
u32 bits_wanted,
u16 *suballoc_bit_start,
@@ -1989,7 +1988,7 @@ int ocfs2_claim_metadata(struct ocfs2_super *osb,
mlog_errno(status);
goto bail;
}
- atomic_inc(&osb->alloc_stats.bg_allocs);
+ atomic_inc(&OCFS2_SB(ac->ac_inode->i_sb)->alloc_stats.bg_allocs);
*suballoc_bit_start = res.sr_bit_offset;
*blkno_start = res.sr_bg_blkno + (u64)(res.sr_bit_offset);
@@ -2032,8 +2031,7 @@ static inline void ocfs2_save_inode_ac_group(struct inode *dir,
OCFS2_I(dir)->ip_last_used_slot = ac->ac_alloc_slot;
}
-int ocfs2_claim_new_inode(struct ocfs2_super *osb,
- handle_t *handle,
+int ocfs2_claim_new_inode(handle_t *handle,
struct inode *dir,
struct buffer_head *parent_fe_bh,
struct ocfs2_alloc_context *ac,
@@ -2061,7 +2059,7 @@ int ocfs2_claim_new_inode(struct ocfs2_super *osb,
mlog_errno(status);
goto bail;
}
- atomic_inc(&osb->alloc_stats.bg_allocs);
+ atomic_inc(&OCFS2_SB(ac->ac_inode->i_sb)->alloc_stats.bg_allocs);
BUG_ON(res.sr_bits != 1);
@@ -2136,8 +2134,7 @@ static inline void ocfs2_block_to_cluster_group(struct inode *inode,
* contig. allocation, set to '1' to indicate we can deal with extents
* of any size.
*/
-int __ocfs2_claim_clusters(struct ocfs2_super *osb,
- handle_t *handle,
+int __ocfs2_claim_clusters(handle_t *handle,
struct ocfs2_alloc_context *ac,
u32 min_clusters,
u32 max_clusters,
@@ -2147,6 +2144,7 @@ int __ocfs2_claim_clusters(struct ocfs2_super *osb,
int status;
unsigned int bits_wanted = max_clusters;
struct ocfs2_suballoc_result res;
+ struct ocfs2_super *osb = OCFS2_SB(ac->ac_inode->i_sb);
mlog_entry_void();
@@ -2205,8 +2203,7 @@ bail:
return status;
}
-int ocfs2_claim_clusters(struct ocfs2_super *osb,
- handle_t *handle,
+int ocfs2_claim_clusters(handle_t *handle,
struct ocfs2_alloc_context *ac,
u32 min_clusters,
u32 *cluster_start,
@@ -2214,7 +2211,7 @@ int ocfs2_claim_clusters(struct ocfs2_super *osb,
{
unsigned int bits_wanted = ac->ac_bits_wanted - ac->ac_bits_given;
- return __ocfs2_claim_clusters(osb, handle, ac, min_clusters,
+ return __ocfs2_claim_clusters(handle, ac, min_clusters,
bits_wanted, cluster_start, num_clusters);
}
diff --git a/fs/ocfs2/suballoc.h b/fs/ocfs2/suballoc.h
index fbf7f17..28096dc 100644
--- a/fs/ocfs2/suballoc.h
+++ b/fs/ocfs2/suballoc.h
@@ -81,22 +81,19 @@ int ocfs2_reserve_clusters(struct ocfs2_super *osb,
u32 bits_wanted,
struct ocfs2_alloc_context **ac);
-int ocfs2_claim_metadata(struct ocfs2_super *osb,
- handle_t *handle,
+int ocfs2_claim_metadata(handle_t *handle,
struct ocfs2_alloc_context *ac,
u32 bits_wanted,
u16 *suballoc_bit_start,
u32 *num_bits,
u64 *blkno_start);
-int ocfs2_claim_new_inode(struct ocfs2_super *osb,
- handle_t *handle,
+int ocfs2_claim_new_inode(handle_t *handle,
struct inode *dir,
struct buffer_head *parent_fe_bh,
struct ocfs2_alloc_context *ac,
u16 *suballoc_bit,
u64 *fe_blkno);
-int ocfs2_claim_clusters(struct ocfs2_super *osb,
- handle_t *handle,
+int ocfs2_claim_clusters(handle_t *handle,
struct ocfs2_alloc_context *ac,
u32 min_clusters,
u32 *cluster_start,
@@ -105,8 +102,7 @@ int ocfs2_claim_clusters(struct ocfs2_super *osb,
* Use this variant of ocfs2_claim_clusters to specify a maxiumum
* number of clusters smaller than the allocation reserved.
*/
-int __ocfs2_claim_clusters(struct ocfs2_super *osb,
- handle_t *handle,
+int __ocfs2_claim_clusters(handle_t *handle,
struct ocfs2_alloc_context *ac,
u32 min_clusters,
u32 max_clusters,
diff --git a/fs/ocfs2/xattr.c b/fs/ocfs2/xattr.c
index d1b0d38..e8997ca 100644
--- a/fs/ocfs2/xattr.c
+++ b/fs/ocfs2/xattr.c
@@ -2848,7 +2848,6 @@ static int ocfs2_create_xattr_block(struct inode *inode,
u32 num_got;
u64 first_blkno;
struct ocfs2_dinode *di = (struct ocfs2_dinode *)inode_bh->b_data;
- struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
struct buffer_head *new_bh = NULL;
struct ocfs2_xattr_block *xblk;
@@ -2859,7 +2858,7 @@ static int ocfs2_create_xattr_block(struct inode *inode,
goto end;
}
- ret = ocfs2_claim_metadata(osb, ctxt->handle, ctxt->meta_ac, 1,
+ ret = ocfs2_claim_metadata(ctxt->handle, ctxt->meta_ac, 1,
&suballoc_bit_start, &num_got,
&first_blkno);
if (ret < 0) {
@@ -2884,7 +2883,8 @@ static int ocfs2_create_xattr_block(struct inode *inode,
strcpy((void *)xblk, OCFS2_XATTR_BLOCK_SIGNATURE);
xblk->xb_suballoc_slot = cpu_to_le16(ctxt->meta_ac->ac_alloc_slot);
xblk->xb_suballoc_bit = cpu_to_le16(suballoc_bit_start);
- xblk->xb_fs_generation = cpu_to_le32(osb->fs_generation);
+ xblk->xb_fs_generation =
+ cpu_to_le32(OCFS2_SB(inode->i_sb)->fs_generation);
xblk->xb_blkno = cpu_to_le64(first_blkno);
if (indexed) {
struct ocfs2_xattr_tree_root *xr = &xblk->xb_attrs.xb_root;
@@ -4249,7 +4249,6 @@ static int ocfs2_xattr_create_index_block(struct inode *inode,
u32 bit_off, len;
u64 blkno;
handle_t *handle = ctxt->handle;
- struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
struct ocfs2_inode_info *oi = OCFS2_I(inode);
struct buffer_head *xb_bh = xs->xattr_bh;
struct ocfs2_xattr_block *xb =
@@ -4277,7 +4276,7 @@ static int ocfs2_xattr_create_index_block(struct inode *inode,
goto out;
}
- ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac,
+ ret = __ocfs2_claim_clusters(handle, ctxt->data_ac,
1, 1, &bit_off, &len);
if (ret) {
mlog_errno(ret);
@@ -5099,7 +5098,7 @@ static int ocfs2_add_new_xattr_cluster(struct inode *inode,
goto leave;
}
- ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac, 1,
+ ret = __ocfs2_claim_clusters(handle, ctxt->data_ac, 1,
clusters_to_add, &bit_off, &num_bits);
if (ret < 0) {
if (ret != -ENOSPC)
@@ -6937,7 +6936,7 @@ static int ocfs2_reflink_xattr_rec(struct inode *inode,
goto out;
}
- ret = ocfs2_claim_clusters(osb, handle, data_ac,
+ ret = ocfs2_claim_clusters(handle, data_ac,
len, &p_cluster, &num_clusters);
if (ret) {
mlog_errno(ret);
--
1.5.5
^ permalink raw reply related [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 08/15] ocfs2: Return allocated metadata blknos on the ocfs2_suballoc_result.
2010-04-01 2:53 [Ocfs2-devel] [PATCH 0/15] ocfs2: Support for discontiguous block group Tao Ma
` (6 preceding siblings ...)
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 07/15] ocfs2: ocfs2_claim_*() don't need an ocfs2_super argument Tao Ma
@ 2010-04-01 2:59 ` Tao Ma
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 09/15] ocfs2: Set suballoc_loc on allocated metadata Tao Ma
` (6 subsequent siblings)
14 siblings, 0 replies; 34+ messages in thread
From: Tao Ma @ 2010-04-01 2:59 UTC (permalink / raw)
To: ocfs2-devel
From: Joel Becker <joel.becker@oracle.com>
Rather than calculating the resulting block number, return it on the
ocfs2_suballoc_result structure. This way we can calculate block
numbers for discontiguous block groups.
Cluster groups keep doing it the old way.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
fs/ocfs2/suballoc.c | 43 ++++++++++++++++++++++++-------------------
1 files changed, 24 insertions(+), 19 deletions(-)
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 0fd0357..6ac9884 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -55,6 +55,7 @@
struct ocfs2_suballoc_result {
u64 sr_bg_blkno; /* The bg we allocated from */
+ u64 sr_blkno; /* The first allocated block */
unsigned int sr_bit_offset; /* The bit in the bg */
unsigned int sr_bits; /* How many bits we claimed */
};
@@ -1619,9 +1620,9 @@ out:
return ret;
}
-static int ocfs2_bg_discontig_trim_by_rec(struct ocfs2_suballoc_result *res,
- struct ocfs2_extent_rec *rec,
- struct ocfs2_chain_list *cl)
+static int ocfs2_bg_discontig_fix_by_rec(struct ocfs2_suballoc_result *res,
+ struct ocfs2_extent_rec *rec,
+ struct ocfs2_chain_list *cl)
{
unsigned int bpc = le16_to_cpu(cl->cl_bpc);
unsigned int bitoff = le32_to_cpu(rec->e_cpos) * bpc;
@@ -1631,32 +1632,35 @@ static int ocfs2_bg_discontig_trim_by_rec(struct ocfs2_suballoc_result *res,
return 0;
if (res->sr_bit_offset >= (bitoff + bitcount))
return 0;
+ res->sr_blkno = le64_to_cpu(rec->e_blkno) +
+ (res->sr_bit_offset - bitoff);
if ((res->sr_bit_offset + res->sr_bits) > (bitoff + bitcount))
res->sr_bits = (bitoff + bitcount) - res->sr_bit_offset;
return 1;
}
-static void ocfs2_bg_discontig_trim_result(struct ocfs2_alloc_context *ac,
- struct ocfs2_group_desc *bg,
- struct ocfs2_suballoc_result *res)
+static void ocfs2_bg_discontig_fix_result(struct ocfs2_alloc_context *ac,
+ struct ocfs2_group_desc *bg,
+ struct ocfs2_suballoc_result *res)
{
int i;
struct ocfs2_extent_rec *rec;
struct ocfs2_dinode *di = (struct ocfs2_dinode *)ac->ac_bh->b_data;
struct ocfs2_chain_list *cl = &di->id2.i_chain;
- if (!ocfs2_supports_discontig_bh(OCFS2_SB(ac->ac_inode->i_sb)))
- return;
-
- if (ocfs2_is_cluster_bitmap(ac->ac_inode))
+ if (ocfs2_is_cluster_bitmap(ac->ac_inode)) {
+ res->sr_blkno = 0;
return;
+ }
- if (!bg->bg_list.l_next_free_rec)
+ res->sr_blkno = res->sr_bg_blkno + res->sr_bit_offset;
+ if (!ocfs2_supports_discontig_bh(OCFS2_SB(ac->ac_inode->i_sb)) ||
+ !bg->bg_list.l_next_free_rec)
return;
for (i = 0; i < le16_to_cpu(bg->bg_list.l_next_free_rec); i++) {
rec = &bg->bg_list.l_recs[i];
- if (ocfs2_bg_discontig_trim_by_rec(res, rec, cl))
+ if (ocfs2_bg_discontig_fix_by_rec(res, rec, cl))
break;
}
}
@@ -1691,7 +1695,7 @@ static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
}
if (!ret)
- ocfs2_bg_discontig_trim_result(ac, gd, res);
+ ocfs2_bg_discontig_fix_result(ac, gd, res);
ret = ocfs2_alloc_dinode_update_counts(alloc_inode, handle, ac->ac_bh,
res->sr_bits,
@@ -1783,7 +1787,7 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
BUG_ON(res->sr_bits == 0);
if (!status)
- ocfs2_bg_discontig_trim_result(ac, bg, res);
+ ocfs2_bg_discontig_fix_result(ac, bg, res);
/*
@@ -1973,7 +1977,7 @@ int ocfs2_claim_metadata(handle_t *handle,
u64 *blkno_start)
{
int status;
- struct ocfs2_suballoc_result res;
+ struct ocfs2_suballoc_result res = { .sr_blkno = 0, };
BUG_ON(!ac);
BUG_ON(ac->ac_bits_wanted < (ac->ac_bits_given + bits_wanted));
@@ -1991,7 +1995,7 @@ int ocfs2_claim_metadata(handle_t *handle,
atomic_inc(&OCFS2_SB(ac->ac_inode->i_sb)->alloc_stats.bg_allocs);
*suballoc_bit_start = res.sr_bit_offset;
- *blkno_start = res.sr_bg_blkno + (u64)(res.sr_bit_offset);
+ *blkno_start = res.sr_blkno;
ac->ac_bits_given += res.sr_bits;
*num_bits = res.sr_bits;
status = 0;
@@ -2039,7 +2043,7 @@ int ocfs2_claim_new_inode(handle_t *handle,
u64 *fe_blkno)
{
int status;
- struct ocfs2_suballoc_result res;
+ struct ocfs2_suballoc_result res = { .sr_blkno = 0, };
mlog_entry_void();
@@ -2064,7 +2068,7 @@ int ocfs2_claim_new_inode(handle_t *handle,
BUG_ON(res.sr_bits != 1);
*suballoc_bit = res.sr_bit_offset;
- *fe_blkno = res.sr_bg_blkno + (u64)(res.sr_bit_offset);
+ *fe_blkno = res.sr_blkno;
ac->ac_bits_given++;
ocfs2_save_inode_ac_group(dir, ac);
status = 0;
@@ -2143,7 +2147,7 @@ int __ocfs2_claim_clusters(handle_t *handle,
{
int status;
unsigned int bits_wanted = max_clusters;
- struct ocfs2_suballoc_result res;
+ struct ocfs2_suballoc_result res = { .sr_blkno = 0, };
struct ocfs2_super *osb = OCFS2_SB(ac->ac_inode->i_sb);
mlog_entry_void();
@@ -2182,6 +2186,7 @@ int __ocfs2_claim_clusters(handle_t *handle,
min_clusters,
&res);
if (!status) {
+ BUG_ON(res.sr_blkno); /* cluster alloc can't set */
*cluster_start =
ocfs2_desc_bitmap_to_cluster_off(ac->ac_inode,
res.sr_bg_blkno,
--
1.5.5
^ permalink raw reply related [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 09/15] ocfs2: Set suballoc_loc on allocated metadata.
2010-04-01 2:53 [Ocfs2-devel] [PATCH 0/15] ocfs2: Support for discontiguous block group Tao Ma
` (7 preceding siblings ...)
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 08/15] ocfs2: Return allocated metadata blknos on the ocfs2_suballoc_result Tao Ma
@ 2010-04-01 2:59 ` Tao Ma
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 10/15] ocfs2: Grow discontig block groups in one transaction Tao Ma
` (5 subsequent siblings)
14 siblings, 0 replies; 34+ messages in thread
From: Tao Ma @ 2010-04-01 2:59 UTC (permalink / raw)
To: ocfs2-devel
From: Joel Becker <joel.becker@oracle.com>
Get the suballoc_loc from ocfs2_claim_new_inode() or
ocfs2_claim_metadata(). Store it on the appropriate field of the block
we just allocated.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
fs/ocfs2/alloc.c | 4 +++-
fs/ocfs2/dir.c | 7 ++++---
fs/ocfs2/namei.c | 6 ++++--
fs/ocfs2/refcounttree.c | 15 +++++++++------
fs/ocfs2/suballoc.c | 16 +++++++++++++---
fs/ocfs2/suballoc.h | 2 ++
fs/ocfs2/xattr.c | 7 ++++---
7 files changed, 39 insertions(+), 18 deletions(-)
diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
index 57d7f7a..6b59b6f 100644
--- a/fs/ocfs2/alloc.c
+++ b/fs/ocfs2/alloc.c
@@ -1006,7 +1006,7 @@ static int ocfs2_create_new_meta_bhs(handle_t *handle,
int count, status, i;
u16 suballoc_bit_start;
u32 num_got;
- u64 first_blkno;
+ u64 suballoc_loc, first_blkno;
struct ocfs2_super *osb =
OCFS2_SB(ocfs2_metadata_cache_get_super(et->et_ci));
struct ocfs2_extent_block *eb;
@@ -1018,6 +1018,7 @@ static int ocfs2_create_new_meta_bhs(handle_t *handle,
status = ocfs2_claim_metadata(handle,
meta_ac,
wanted - count,
+ &suballoc_loc,
&suballoc_bit_start,
&num_got,
&first_blkno);
@@ -1051,6 +1052,7 @@ static int ocfs2_create_new_meta_bhs(handle_t *handle,
eb->h_fs_generation = cpu_to_le32(osb->fs_generation);
eb->h_suballoc_slot =
cpu_to_le16(meta_ac->ac_alloc_slot);
+ eb->h_suballoc_loc = cpu_to_le64(suballoc_loc);
eb->h_suballoc_bit = cpu_to_le16(suballoc_bit_start);
eb->h_list.l_count =
cpu_to_le16(ocfs2_extent_recs_per_eb(osb->sb));
diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c
index 06522b5..48b591d 100644
--- a/fs/ocfs2/dir.c
+++ b/fs/ocfs2/dir.c
@@ -2404,15 +2404,15 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb,
int ret;
struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data;
u16 dr_suballoc_bit;
- u64 dr_blkno;
+ u64 suballoc_loc, dr_blkno;
unsigned int num_bits;
struct buffer_head *dx_root_bh = NULL;
struct ocfs2_dx_root_block *dx_root;
struct ocfs2_dir_block_trailer *trailer =
ocfs2_trailer_from_bh(dirdata_bh, dir->i_sb);
- ret = ocfs2_claim_metadata(handle, meta_ac, 1, &dr_suballoc_bit,
- &num_bits, &dr_blkno);
+ ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc,
+ &dr_suballoc_bit, &num_bits, &dr_blkno);
if (ret) {
mlog_errno(ret);
goto out;
@@ -2440,6 +2440,7 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb,
memset(dx_root, 0, osb->sb->s_blocksize);
strcpy(dx_root->dr_signature, OCFS2_DX_ROOT_SIGNATURE);
dx_root->dr_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot);
+ dx_root->dr_suballoc_loc = cpu_to_le64(suballoc_loc);
dx_root->dr_suballoc_bit = cpu_to_le16(dr_suballoc_bit);
dx_root->dr_fs_generation = cpu_to_le32(osb->fs_generation);
dx_root->dr_blkno = cpu_to_le64(dr_blkno);
diff --git a/fs/ocfs2/namei.c b/fs/ocfs2/namei.c
index 9c3e1fe..6ec584e 100644
--- a/fs/ocfs2/namei.c
+++ b/fs/ocfs2/namei.c
@@ -476,14 +476,15 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb,
int status = 0;
struct ocfs2_dinode *fe = NULL;
struct ocfs2_extent_list *fel;
- u64 fe_blkno = 0;
+ u64 suballoc_loc, fe_blkno = 0;
u16 suballoc_bit;
u16 feat;
*new_fe_bh = NULL;
status = ocfs2_claim_new_inode(handle, dir, parent_fe_bh,
- inode_ac, &suballoc_bit, &fe_blkno);
+ inode_ac, &suballoc_loc,
+ &suballoc_bit, &fe_blkno);
if (status < 0) {
mlog_errno(status);
goto leave;
@@ -520,6 +521,7 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb,
fe->i_generation = cpu_to_le32(inode->i_generation);
fe->i_fs_generation = cpu_to_le32(osb->fs_generation);
fe->i_blkno = cpu_to_le64(fe_blkno);
+ fe->i_suballoc_loc = cpu_to_le64(suballoc_loc);
fe->i_suballoc_bit = cpu_to_le16(suballoc_bit);
fe->i_suballoc_slot = cpu_to_le16(inode_ac->ac_alloc_slot);
fe->i_uid = cpu_to_le32(inode->i_uid);
diff --git a/fs/ocfs2/refcounttree.c b/fs/ocfs2/refcounttree.c
index 4fb18af..3fc675f 100644
--- a/fs/ocfs2/refcounttree.c
+++ b/fs/ocfs2/refcounttree.c
@@ -571,7 +571,7 @@ static int ocfs2_create_refcount_tree(struct inode *inode,
struct ocfs2_refcount_tree *new_tree = NULL, *tree = NULL;
u16 suballoc_bit_start;
u32 num_got;
- u64 first_blkno;
+ u64 suballoc_loc, first_blkno;
BUG_ON(oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL);
@@ -597,7 +597,7 @@ static int ocfs2_create_refcount_tree(struct inode *inode,
goto out_commit;
}
- ret = ocfs2_claim_metadata(handle, meta_ac, 1,
+ ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc,
&suballoc_bit_start, &num_got,
&first_blkno);
if (ret) {
@@ -627,6 +627,7 @@ static int ocfs2_create_refcount_tree(struct inode *inode,
memset(rb, 0, inode->i_sb->s_blocksize);
strcpy((void *)rb, OCFS2_REFCOUNT_BLOCK_SIGNATURE);
rb->rf_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot);
+ rb->rf_suballoc_loc = cpu_to_le64(suballoc_loc);
rb->rf_suballoc_bit = cpu_to_le16(suballoc_bit_start);
rb->rf_fs_generation = cpu_to_le32(osb->fs_generation);
rb->rf_blkno = cpu_to_le64(first_blkno);
@@ -1285,7 +1286,7 @@ static int ocfs2_expand_inline_ref_root(handle_t *handle,
int ret;
u16 suballoc_bit_start;
u32 num_got;
- u64 blkno;
+ u64 suballoc_loc, blkno;
struct super_block *sb = ocfs2_metadata_cache_get_super(ci);
struct buffer_head *new_bh = NULL;
struct ocfs2_refcount_block *new_rb;
@@ -1299,7 +1300,7 @@ static int ocfs2_expand_inline_ref_root(handle_t *handle,
goto out;
}
- ret = ocfs2_claim_metadata(handle, meta_ac, 1,
+ ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc,
&suballoc_bit_start, &num_got,
&blkno);
if (ret) {
@@ -1331,6 +1332,7 @@ static int ocfs2_expand_inline_ref_root(handle_t *handle,
new_rb = (struct ocfs2_refcount_block *)new_bh->b_data;
new_rb->rf_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot);
+ new_rb->rf_suballoc_loc = cpu_to_le64(suballoc_loc);
new_rb->rf_suballoc_bit = cpu_to_le16(suballoc_bit_start);
new_rb->rf_blkno = cpu_to_le64(blkno);
new_rb->rf_cpos = cpu_to_le32(0);
@@ -1525,7 +1527,7 @@ static int ocfs2_new_leaf_refcount_block(handle_t *handle,
int ret;
u16 suballoc_bit_start;
u32 num_got, new_cpos;
- u64 blkno;
+ u64 suballoc_loc, blkno;
struct super_block *sb = ocfs2_metadata_cache_get_super(ci);
struct ocfs2_refcount_block *root_rb =
(struct ocfs2_refcount_block *)ref_root_bh->b_data;
@@ -1549,7 +1551,7 @@ static int ocfs2_new_leaf_refcount_block(handle_t *handle,
goto out;
}
- ret = ocfs2_claim_metadata(handle, meta_ac, 1,
+ ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc,
&suballoc_bit_start, &num_got,
&blkno);
if (ret) {
@@ -1577,6 +1579,7 @@ static int ocfs2_new_leaf_refcount_block(handle_t *handle,
memset(new_rb, 0, sb->s_blocksize);
strcpy((void *)new_rb, OCFS2_REFCOUNT_BLOCK_SIGNATURE);
new_rb->rf_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot);
+ new_rb->rf_suballoc_loc = cpu_to_le64(suballoc_loc);
new_rb->rf_suballoc_bit = cpu_to_le16(suballoc_bit_start);
new_rb->rf_fs_generation = cpu_to_le32(OCFS2_SB(sb)->fs_generation);
new_rb->rf_blkno = cpu_to_le64(blkno);
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 6ac9884..0ee6548 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -54,7 +54,9 @@
#define OCFS2_MAX_TO_STEAL 1024
struct ocfs2_suballoc_result {
- u64 sr_bg_blkno; /* The bg we allocated from */
+ u64 sr_bg_blkno; /* The bg we allocated from. Set
+ to 0 when a block group is
+ contiguous. */
u64 sr_blkno; /* The first allocated block */
unsigned int sr_bit_offset; /* The bit in the bg */
unsigned int sr_bits; /* How many bits we claimed */
@@ -1644,6 +1646,7 @@ static void ocfs2_bg_discontig_fix_result(struct ocfs2_alloc_context *ac,
struct ocfs2_suballoc_result *res)
{
int i;
+ u64 bg_blkno = res->sr_bg_blkno; /* Save off */
struct ocfs2_extent_rec *rec;
struct ocfs2_dinode *di = (struct ocfs2_dinode *)ac->ac_bh->b_data;
struct ocfs2_chain_list *cl = &di->id2.i_chain;
@@ -1654,14 +1657,17 @@ static void ocfs2_bg_discontig_fix_result(struct ocfs2_alloc_context *ac,
}
res->sr_blkno = res->sr_bg_blkno + res->sr_bit_offset;
+ res->sr_bg_blkno = 0; /* Clear it for contig block groups */
if (!ocfs2_supports_discontig_bh(OCFS2_SB(ac->ac_inode->i_sb)) ||
!bg->bg_list.l_next_free_rec)
return;
for (i = 0; i < le16_to_cpu(bg->bg_list.l_next_free_rec); i++) {
rec = &bg->bg_list.l_recs[i];
- if (ocfs2_bg_discontig_fix_by_rec(res, rec, cl))
+ if (ocfs2_bg_discontig_fix_by_rec(res, rec, cl)) {
+ res->sr_bg_blkno = bg_blkno; /* Restore */
break;
+ }
}
}
@@ -1972,6 +1978,7 @@ bail:
int ocfs2_claim_metadata(handle_t *handle,
struct ocfs2_alloc_context *ac,
u32 bits_wanted,
+ u64 *suballoc_loc,
u16 *suballoc_bit_start,
unsigned int *num_bits,
u64 *blkno_start)
@@ -1994,6 +2001,7 @@ int ocfs2_claim_metadata(handle_t *handle,
}
atomic_inc(&OCFS2_SB(ac->ac_inode->i_sb)->alloc_stats.bg_allocs);
+ *suballoc_loc = res.sr_bg_blkno;
*suballoc_bit_start = res.sr_bit_offset;
*blkno_start = res.sr_blkno;
ac->ac_bits_given += res.sr_bits;
@@ -2039,11 +2047,12 @@ int ocfs2_claim_new_inode(handle_t *handle,
struct inode *dir,
struct buffer_head *parent_fe_bh,
struct ocfs2_alloc_context *ac,
+ u64 *suballoc_loc,
u16 *suballoc_bit,
u64 *fe_blkno)
{
int status;
- struct ocfs2_suballoc_result res = { .sr_blkno = 0, };
+ struct ocfs2_suballoc_result res;
mlog_entry_void();
@@ -2067,6 +2076,7 @@ int ocfs2_claim_new_inode(handle_t *handle,
BUG_ON(res.sr_bits != 1);
+ *suballoc_loc = res.sr_bg_blkno;
*suballoc_bit = res.sr_bit_offset;
*fe_blkno = res.sr_blkno;
ac->ac_bits_given++;
diff --git a/fs/ocfs2/suballoc.h b/fs/ocfs2/suballoc.h
index 28096dc..c6e1ee7 100644
--- a/fs/ocfs2/suballoc.h
+++ b/fs/ocfs2/suballoc.h
@@ -84,6 +84,7 @@ int ocfs2_reserve_clusters(struct ocfs2_super *osb,
int ocfs2_claim_metadata(handle_t *handle,
struct ocfs2_alloc_context *ac,
u32 bits_wanted,
+ u64 *suballoc_loc,
u16 *suballoc_bit_start,
u32 *num_bits,
u64 *blkno_start);
@@ -91,6 +92,7 @@ int ocfs2_claim_new_inode(handle_t *handle,
struct inode *dir,
struct buffer_head *parent_fe_bh,
struct ocfs2_alloc_context *ac,
+ u64 *suballoc_loc,
u16 *suballoc_bit,
u64 *fe_blkno);
int ocfs2_claim_clusters(handle_t *handle,
diff --git a/fs/ocfs2/xattr.c b/fs/ocfs2/xattr.c
index e8997ca..863ad7a 100644
--- a/fs/ocfs2/xattr.c
+++ b/fs/ocfs2/xattr.c
@@ -2846,7 +2846,7 @@ static int ocfs2_create_xattr_block(struct inode *inode,
int ret;
u16 suballoc_bit_start;
u32 num_got;
- u64 first_blkno;
+ u64 suballoc_loc, first_blkno;
struct ocfs2_dinode *di = (struct ocfs2_dinode *)inode_bh->b_data;
struct buffer_head *new_bh = NULL;
struct ocfs2_xattr_block *xblk;
@@ -2859,8 +2859,8 @@ static int ocfs2_create_xattr_block(struct inode *inode,
}
ret = ocfs2_claim_metadata(ctxt->handle, ctxt->meta_ac, 1,
- &suballoc_bit_start, &num_got,
- &first_blkno);
+ &suballoc_loc, &suballoc_bit_start,
+ &num_got, &first_blkno);
if (ret < 0) {
mlog_errno(ret);
goto end;
@@ -2882,6 +2882,7 @@ static int ocfs2_create_xattr_block(struct inode *inode,
memset(xblk, 0, inode->i_sb->s_blocksize);
strcpy((void *)xblk, OCFS2_XATTR_BLOCK_SIGNATURE);
xblk->xb_suballoc_slot = cpu_to_le16(ctxt->meta_ac->ac_alloc_slot);
+ xblk->xb_suballoc_loc = cpu_to_le64(suballoc_loc);
xblk->xb_suballoc_bit = cpu_to_le16(suballoc_bit_start);
xblk->xb_fs_generation =
cpu_to_le32(OCFS2_SB(inode->i_sb)->fs_generation);
--
1.5.5
^ permalink raw reply related [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 10/15] ocfs2: Grow discontig block groups in one transaction.
2010-04-01 2:53 [Ocfs2-devel] [PATCH 0/15] ocfs2: Support for discontiguous block group Tao Ma
` (8 preceding siblings ...)
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 09/15] ocfs2: Set suballoc_loc on allocated metadata Tao Ma
@ 2010-04-01 2:59 ` Tao Ma
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 11/15] ocfs2: Don't relink cluster groups when allocating discontig block groups Tao Ma
` (4 subsequent siblings)
14 siblings, 0 replies; 34+ messages in thread
From: Tao Ma @ 2010-04-01 2:59 UTC (permalink / raw)
To: ocfs2-devel
From: Joel Becker <joel.becker@oracle.com>
Rather than extending the transaction every time we add an extent to a
discontiguous block group, we grab enough credits to fill the extent
list up front. This means we can free the bits in the same transaction
if we end up not getting enough space.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
fs/ocfs2/journal.h | 12 ++++++++++++
fs/ocfs2/suballoc.c | 48 ++++++++++++++++++++++--------------------------
2 files changed, 34 insertions(+), 26 deletions(-)
diff --git a/fs/ocfs2/journal.h b/fs/ocfs2/journal.h
index 3f74e09..45035f5 100644
--- a/fs/ocfs2/journal.h
+++ b/fs/ocfs2/journal.h
@@ -562,6 +562,18 @@ static inline int ocfs2_calc_group_alloc_credits(struct super_block *sb,
return blocks;
}
+/*
+ * Allocating a discontiguous block group requires the credits from
+ * ocfs2_calc_group_alloc_credits() as well as enough credits to fill
+ * the group descriptor's extent list. The caller already has started
+ * the transaction with ocfs2_calc_group_alloc_credits(). They extend
+ * it with these credits.
+ */
+static inline int ocfs2_calc_bg_discontig_credits(struct super_block *sb)
+{
+ return ocfs2_extent_recs_per_gd(sb);
+}
+
static inline int ocfs2_calc_tree_trunc_credits(struct super_block *sb,
unsigned int clusters_to_del,
struct ocfs2_dinode *fe,
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 0ee6548..679ac4d 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -528,12 +528,6 @@ static int ocfs2_block_group_grow_discontig(handle_t *handle,
while ((needed > 0) && (le16_to_cpu(el->l_next_free_rec) <
le16_to_cpu(el->l_count))) {
- status = ocfs2_extend_trans(handle, OCFS2_SUBALLOC_ALLOC);
- if (status) {
- mlog_errno(status);
- goto bail;
- }
-
if (min_bits > needed)
min_bits = needed;
status = ocfs2_block_group_claim_bits(osb, handle, ac,
@@ -563,11 +557,12 @@ bail:
return status;
}
-static void ocfs2_bg_alloc_cleanup(struct inode *alloc_inode,
- struct buffer_head *bg_bh,
- struct ocfs2_cached_dealloc_ctxt *dealloc)
+static void ocfs2_bg_alloc_cleanup(handle_t *handle,
+ struct ocfs2_alloc_context *cluster_ac,
+ struct inode *alloc_inode,
+ struct buffer_head *bg_bh)
{
- int i;
+ int i, ret;
struct ocfs2_group_desc *bg;
struct ocfs2_extent_list *el;
struct ocfs2_extent_rec *rec;
@@ -579,9 +574,13 @@ static void ocfs2_bg_alloc_cleanup(struct inode *alloc_inode,
el = &bg->bg_list;
for (i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
rec = &el->l_recs[i];
- ocfs2_cache_cluster_dealloc(dealloc,
- le64_to_cpu(rec->e_blkno),
- le32_to_cpu(rec->e_leaf_clusters));
+ ret = ocfs2_free_clusters(handle, cluster_ac->ac_inode,
+ cluster_ac->ac_bh,
+ le64_to_cpu(rec->e_blkno),
+ le32_to_cpu(rec->e_leaf_clusters));
+ if (ret)
+ mlog_errno(ret);
+ /* Try all the clusters to free */
}
ocfs2_remove_from_cache(INODE_CACHE(alloc_inode), bg_bh);
@@ -592,8 +591,7 @@ static struct buffer_head *
ocfs2_block_group_alloc_discontig(handle_t *handle,
struct inode *alloc_inode,
struct ocfs2_alloc_context *ac,
- struct ocfs2_chain_list *cl,
- struct ocfs2_cached_dealloc_ctxt *dealloc)
+ struct ocfs2_chain_list *cl)
{
int status;
u32 bit_off, num_bits;
@@ -608,6 +606,13 @@ ocfs2_block_group_alloc_discontig(handle_t *handle,
goto bail;
}
+ status = ocfs2_extend_trans(handle,
+ ocfs2_calc_bg_discontig_credits(osb->sb));
+ if (status) {
+ mlog_errno(status);
+ goto bail;
+ }
+
/* Claim the first region */
status = ocfs2_block_group_claim_bits(osb, handle, ac, min_bits,
&bit_off, &num_bits);
@@ -645,7 +650,7 @@ ocfs2_block_group_alloc_discontig(handle_t *handle,
bail:
if (status)
- ocfs2_bg_alloc_cleanup(alloc_inode, bg_bh, dealloc);
+ ocfs2_bg_alloc_cleanup(handle, ac, alloc_inode, bg_bh);
return status ? ERR_PTR(status) : bg_bh;
}
@@ -667,14 +672,11 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
u64 bg_blkno;
struct buffer_head *bg_bh = NULL;
struct ocfs2_group_desc *bg;
- struct ocfs2_cached_dealloc_ctxt dealloc;
BUG_ON(ocfs2_is_cluster_bitmap(alloc_inode));
mlog_entry_void();
- ocfs2_init_dealloc_ctxt(&dealloc);
-
cl = &fe->id2.i_chain;
status = ocfs2_reserve_clusters_with_limit(osb,
le16_to_cpu(cl->cl_cpg),
@@ -706,8 +708,7 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
if (IS_ERR(bg_bh) && (PTR_ERR(bg_bh) == -ENOSPC))
bg_bh = ocfs2_block_group_alloc_discontig(handle,
alloc_inode,
- ac, cl,
- &dealloc);
+ ac, cl);
if (IS_ERR(bg_bh)) {
status = PTR_ERR(bg_bh);
bg_bh = NULL;
@@ -761,11 +762,6 @@ bail:
if (handle)
ocfs2_commit_trans(osb, handle);
- if (ocfs2_dealloc_has_cluster(&dealloc)) {
- ocfs2_schedule_truncate_log_flush(osb, 1);
- ocfs2_run_deallocs(osb, &dealloc);
- }
-
if (ac)
ocfs2_free_alloc_context(ac);
--
1.5.5
^ permalink raw reply related [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 11/15] ocfs2: Don't relink cluster groups when allocating discontig block groups
2010-04-01 2:53 [Ocfs2-devel] [PATCH 0/15] ocfs2: Support for discontiguous block group Tao Ma
` (9 preceding siblings ...)
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 10/15] ocfs2: Grow discontig block groups in one transaction Tao Ma
@ 2010-04-01 2:59 ` Tao Ma
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 12/15] ocfs2: Some tiny bug fixes for discontiguous block allocation Tao Ma
` (3 subsequent siblings)
14 siblings, 0 replies; 34+ messages in thread
From: Tao Ma @ 2010-04-01 2:59 UTC (permalink / raw)
To: ocfs2-devel
From: Joel Becker <joel.becker@oracle.com>
We don't have enough credits, and the filesystem is in a full state
anyway.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
---
fs/ocfs2/suballoc.c | 8 ++++++++
1 files changed, 8 insertions(+), 0 deletions(-)
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 679ac4d..0d2cd75 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -613,6 +613,14 @@ ocfs2_block_group_alloc_discontig(handle_t *handle,
goto bail;
}
+ /*
+ * We're going to be grabbing from multiple cluster groups.
+ * We don't have enough credits to relink them all, and the
+ * cluster groups will be staying in cache for the duration of
+ * this operation.
+ */
+ ac->ac_allow_chain_relink = 0;
+
/* Claim the first region */
status = ocfs2_block_group_claim_bits(osb, handle, ac, min_bits,
&bit_off, &num_bits);
--
1.5.5
^ permalink raw reply related [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 12/15] ocfs2: Some tiny bug fixes for discontiguous block allocation.
2010-04-01 2:53 [Ocfs2-devel] [PATCH 0/15] ocfs2: Support for discontiguous block group Tao Ma
` (10 preceding siblings ...)
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 11/15] ocfs2: Don't relink cluster groups when allocating discontig block groups Tao Ma
@ 2010-04-01 2:59 ` Tao Ma
2010-04-01 3:34 ` Joel Becker
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume Tao Ma
` (2 subsequent siblings)
14 siblings, 1 reply; 34+ messages in thread
From: Tao Ma @ 2010-04-01 2:59 UTC (permalink / raw)
To: ocfs2-devel
The fixes include:
1. some endian problems.
2. we should use bit/bpc in ocfs2_block_group_grow_discontig to
allocate clusters.
3. set num_clusters properly in __ocfs2_claim_clusters.
4. change name from ocfs2_supports_discontig_bh to
ocfs2_supports_discontig_bg.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
---
fs/ocfs2/ocfs2.h | 2 +-
fs/ocfs2/suballoc.c | 41 ++++++++++++++++++++++-------------------
2 files changed, 23 insertions(+), 20 deletions(-)
diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h
index b5fd76f..85a0899 100644
--- a/fs/ocfs2/ocfs2.h
+++ b/fs/ocfs2/ocfs2.h
@@ -482,7 +482,7 @@ static inline int ocfs2_supports_indexed_dirs(struct ocfs2_super *osb)
return 0;
}
-static inline int ocfs2_supports_discontig_bh(struct ocfs2_super *osb)
+static inline int ocfs2_supports_discontig_bg(struct ocfs2_super *osb)
{
if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG)
return 1;
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 0d2cd75..0878b5d 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -344,15 +344,17 @@ static void ocfs2_bg_discontig_add_extent(struct ocfs2_super *osb,
struct ocfs2_extent_list *el = &bg->bg_list;
struct ocfs2_extent_rec *rec;
- BUG_ON(!ocfs2_supports_discontig_bh(osb));
+ BUG_ON(!ocfs2_supports_discontig_bg(osb));
if (!el->l_next_free_rec)
el->l_count = cpu_to_le16(ocfs2_extent_recs_per_gd(osb->sb));
rec = &el->l_recs[le16_to_cpu(el->l_next_free_rec)];
- rec->e_blkno = p_blkno;
+ rec->e_blkno = cpu_to_le64(p_blkno);
rec->e_cpos = cpu_to_le32(le16_to_cpu(bg->bg_bits) /
le16_to_cpu(cl->cl_bpc));
rec->e_leaf_clusters = cpu_to_le32(clusters);
le16_add_cpu(&bg->bg_bits, clusters * le16_to_cpu(cl->cl_bpc));
+ le16_add_cpu(&bg->bg_free_bits_count,
+ clusters * le16_to_cpu(cl->cl_bpc));
le16_add_cpu(&el->l_next_free_rec, 1);
}
@@ -400,7 +402,7 @@ static int ocfs2_block_group_fill(handle_t *handle,
if (group_clusters == le16_to_cpu(cl->cl_cpg))
bg->bg_bits = cpu_to_le16(ocfs2_bits_per_group(cl));
else
- ocfs2_bg_discontig_add_extent(osb, bg, cl, bg->bg_blkno,
+ ocfs2_bg_discontig_add_extent(osb, bg, cl, group_blkno,
group_clusters);
/* set the 1st bit in the bitmap to account for the descriptor block */
@@ -511,8 +513,8 @@ static int ocfs2_block_group_grow_discontig(handle_t *handle,
struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb);
struct ocfs2_group_desc *bg =
(struct ocfs2_group_desc *)bg_bh->b_data;
- unsigned int needed =
- ocfs2_bits_per_group(cl) - le16_to_cpu(bg->bg_bits);
+ unsigned int needed = le16_to_cpu(cl->cl_cpg) -
+ le16_to_cpu(bg->bg_bits) / le16_to_cpu(cl->cl_bpc);
u32 p_cpos, clusters;
u64 p_blkno;
struct ocfs2_extent_list *el = &bg->bg_list;
@@ -543,10 +545,8 @@ static int ocfs2_block_group_grow_discontig(handle_t *handle,
clusters);
min_bits = clusters;
- needed = ocfs2_bits_per_group(cl) - le16_to_cpu(bg->bg_bits);
- }
-
- if (needed > 0) {
+ needed = le16_to_cpu(cl->cl_cpg) -
+ le16_to_cpu(bg->bg_bits) / le16_to_cpu(cl->cl_bpc);
}
status = ocfs2_journal_dirty(handle, bg_bh);
@@ -601,13 +601,14 @@ ocfs2_block_group_alloc_discontig(handle_t *handle,
unsigned int alloc_rec = ocfs2_find_smallest_chain(cl);
struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb);
- if (!ocfs2_supports_discontig_bh(osb)) {
+ if (!ocfs2_supports_discontig_bg(osb)) {
status = -ENOSPC;
goto bail;
}
status = ocfs2_extend_trans(handle,
- ocfs2_calc_bg_discontig_credits(osb->sb));
+ ocfs2_calc_bg_discontig_credits(osb->sb) +
+ handle->h_buffer_credits);
if (status) {
mlog_errno(status);
goto bail;
@@ -677,7 +678,7 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
struct ocfs2_chain_list *cl;
struct ocfs2_alloc_context *ac = NULL;
handle_t *handle = NULL;
- u64 bg_blkno;
+ u16 alloc_rec;
struct buffer_head *bg_bh = NULL;
struct ocfs2_group_desc *bg;
@@ -733,18 +734,20 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
goto bail;
}
- le32_add_cpu(&cl->cl_recs[bg->bg_chain].c_free,
+ alloc_rec = le16_to_cpu(bg->bg_chain);
+ le32_add_cpu(&cl->cl_recs[alloc_rec].c_free,
le16_to_cpu(bg->bg_free_bits_count));
- le32_add_cpu(&cl->cl_recs[bg->bg_chain].c_total,
+ le32_add_cpu(&cl->cl_recs[alloc_rec].c_total,
le16_to_cpu(bg->bg_bits));
- cl->cl_recs[bg->bg_chain].c_blkno = cpu_to_le64(bg_blkno);
+ cl->cl_recs[alloc_rec].c_blkno = cpu_to_le64(bg->bg_blkno);
if (le16_to_cpu(cl->cl_next_free_rec) < le16_to_cpu(cl->cl_count))
le16_add_cpu(&cl->cl_next_free_rec, 1);
le32_add_cpu(&fe->id1.bitmap1.i_used, le16_to_cpu(bg->bg_bits) -
le16_to_cpu(bg->bg_free_bits_count));
le32_add_cpu(&fe->id1.bitmap1.i_total, le16_to_cpu(bg->bg_bits));
- le32_add_cpu(&fe->i_clusters, le16_to_cpu(cl->cl_cpg));
+ le32_add_cpu(&fe->i_clusters,
+ le16_to_cpu(bg->bg_bits) / le16_to_cpu(cl->cl_bpc));
status = ocfs2_journal_dirty(handle, bh);
if (status < 0) {
@@ -1662,7 +1665,7 @@ static void ocfs2_bg_discontig_fix_result(struct ocfs2_alloc_context *ac,
res->sr_blkno = res->sr_bg_blkno + res->sr_bit_offset;
res->sr_bg_blkno = 0; /* Clear it for contig block groups */
- if (!ocfs2_supports_discontig_bh(OCFS2_SB(ac->ac_inode->i_sb)) ||
+ if (!ocfs2_supports_discontig_bg(OCFS2_SB(ac->ac_inode->i_sb)) ||
!bg->bg_list.l_next_free_rec)
return;
@@ -2206,6 +2209,7 @@ int __ocfs2_claim_clusters(handle_t *handle,
res.sr_bg_blkno,
res.sr_bit_offset);
atomic_inc(&osb->alloc_stats.bitmap_data);
+ *num_clusters = res.sr_bits;
}
}
if (status < 0) {
@@ -2214,8 +2218,7 @@ int __ocfs2_claim_clusters(handle_t *handle,
goto bail;
}
- ac->ac_bits_given += res.sr_bits;
- *num_clusters = res.sr_bits;
+ ac->ac_bits_given += *num_clusters;
bail:
mlog_exit(status);
--
1.5.5
^ permalink raw reply related [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 12/15] ocfs2: Some tiny bug fixes for discontiguous block allocation.
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 12/15] ocfs2: Some tiny bug fixes for discontiguous block allocation Tao Ma
@ 2010-04-01 3:34 ` Joel Becker
2010-04-01 4:32 ` Tao Ma
0 siblings, 1 reply; 34+ messages in thread
From: Joel Becker @ 2010-04-01 3:34 UTC (permalink / raw)
To: ocfs2-devel
On Thu, Apr 01, 2010 at 10:59:09AM +0800, Tao Ma wrote:
> -static inline int ocfs2_supports_discontig_bh(struct ocfs2_super *osb)
> +static inline int ocfs2_supports_discontig_bg(struct ocfs2_super *osb)
D'oh!
> - rec->e_blkno = p_blkno;
> + rec->e_blkno = cpu_to_le64(p_blkno);
D'oh!!
> + le16_add_cpu(&bg->bg_free_bits_count,
> + clusters * le16_to_cpu(cl->cl_bpc));
D'oh!!!
> @@ -511,8 +513,8 @@ static int ocfs2_block_group_grow_discontig(handle_t *handle,
> struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb);
> struct ocfs2_group_desc *bg =
> (struct ocfs2_group_desc *)bg_bh->b_data;
> - unsigned int needed =
> - ocfs2_bits_per_group(cl) - le16_to_cpu(bg->bg_bits);
> + unsigned int needed = le16_to_cpu(cl->cl_cpg) -
> + le16_to_cpu(bg->bg_bits) / le16_to_cpu(cl->cl_bpc);
Ouch! But can you put parens around the divide? Make the
intention clear.
> @@ -733,18 +734,20 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
> goto bail;
> }
>
> - le32_add_cpu(&cl->cl_recs[bg->bg_chain].c_free,
> + alloc_rec = le16_to_cpu(bg->bg_chain);
> + le32_add_cpu(&cl->cl_recs[alloc_rec].c_free,
> le16_to_cpu(bg->bg_free_bits_count));
D'oh!!!!
> - le32_add_cpu(&cl->cl_recs[bg->bg_chain].c_total,
> + le32_add_cpu(&cl->cl_recs[alloc_rec].c_total,
> le16_to_cpu(bg->bg_bits));
> - cl->cl_recs[bg->bg_chain].c_blkno = cpu_to_le64(bg_blkno);
> + cl->cl_recs[alloc_rec].c_blkno = cpu_to_le64(bg->bg_blkno);
> if (le16_to_cpu(cl->cl_next_free_rec) < le16_to_cpu(cl->cl_count))
> le16_add_cpu(&cl->cl_next_free_rec, 1);
>
> le32_add_cpu(&fe->id1.bitmap1.i_used, le16_to_cpu(bg->bg_bits) -
> le16_to_cpu(bg->bg_free_bits_count));
> le32_add_cpu(&fe->id1.bitmap1.i_total, le16_to_cpu(bg->bg_bits));
> - le32_add_cpu(&fe->i_clusters, le16_to_cpu(cl->cl_cpg));
> + le32_add_cpu(&fe->i_clusters,
> + le16_to_cpu(bg->bg_bits) / le16_to_cpu(cl->cl_bpc));
You don't need to change this. Either we've allocated cpg or
we've failed out.
Joel
--
"There is shadow under this red rock.
(Come in under the shadow of this red rock)
And I will show you something different from either
Your shadow at morning striding behind you
Or your shadow at evening rising to meet you.
I will show you fear in a handful of dust."
Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127
^ permalink raw reply [flat|nested] 34+ messages in thread
* [Ocfs2-devel] [PATCH 12/15] ocfs2: Some tiny bug fixes for discontiguous block allocation.
2010-04-01 3:34 ` Joel Becker
@ 2010-04-01 4:32 ` Tao Ma
2010-04-01 17:44 ` Joel Becker
0 siblings, 1 reply; 34+ messages in thread
From: Tao Ma @ 2010-04-01 4:32 UTC (permalink / raw)
To: ocfs2-devel
Hi Joel,
Joel Becker wrote:
> On Thu, Apr 01, 2010 at 10:59:09AM +0800, Tao Ma wrote:
>> @@ -511,8 +513,8 @@ static int ocfs2_block_group_grow_discontig(handle_t *handle,
>> struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb);
>> struct ocfs2_group_desc *bg =
>> (struct ocfs2_group_desc *)bg_bh->b_data;
>> - unsigned int needed =
>> - ocfs2_bits_per_group(cl) - le16_to_cpu(bg->bg_bits);
>> + unsigned int needed = le16_to_cpu(cl->cl_cpg) -
>> + le16_to_cpu(bg->bg_bits) / le16_to_cpu(cl->cl_bpc);
>
> Ouch! But can you put parens around the divide? Make the
> intention clear.
no problem.
>
>> @@ -733,18 +734,20 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
>> goto bail;
>> }
>>
>> - le32_add_cpu(&cl->cl_recs[bg->bg_chain].c_free,
>> + alloc_rec = le16_to_cpu(bg->bg_chain);
>> + le32_add_cpu(&cl->cl_recs[alloc_rec].c_free,
>> le16_to_cpu(bg->bg_free_bits_count));
>
> D'oh!!!!
>
>> - le32_add_cpu(&cl->cl_recs[bg->bg_chain].c_total,
>> + le32_add_cpu(&cl->cl_recs[alloc_rec].c_total,
>> le16_to_cpu(bg->bg_bits));
>> - cl->cl_recs[bg->bg_chain].c_blkno = cpu_to_le64(bg_blkno);
>> + cl->cl_recs[alloc_rec].c_blkno = cpu_to_le64(bg->bg_blkno);
>> if (le16_to_cpu(cl->cl_next_free_rec) < le16_to_cpu(cl->cl_count))
>> le16_add_cpu(&cl->cl_next_free_rec, 1);
>>
>> le32_add_cpu(&fe->id1.bitmap1.i_used, le16_to_cpu(bg->bg_bits) -
>> le16_to_cpu(bg->bg_free_bits_count));
>> le32_add_cpu(&fe->id1.bitmap1.i_total, le16_to_cpu(bg->bg_bits));
>> - le32_add_cpu(&fe->i_clusters, le16_to_cpu(cl->cl_cpg));
>> + le32_add_cpu(&fe->i_clusters,
>> + le16_to_cpu(bg->bg_bits) / le16_to_cpu(cl->cl_bpc));
>
> You don't need to change this. Either we've allocated cpg or
> we've failed out.
Current we don't fail out in ocfs2_block_group_grow_discontig when we
used up all the extent record in the list. So actually our bitmap size
can be less than cpg. So you think we need to bail out?
I removed your
if (needed > 0) {
}
in this function. So maybe I am wrong? ;)
To bail out is simple, we just need to set the status to ENOSPC in this
case and it should be enough for us.
Regards,
Tao
^ permalink raw reply [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 12/15] ocfs2: Some tiny bug fixes for discontiguous block allocation.
2010-04-01 4:32 ` Tao Ma
@ 2010-04-01 17:44 ` Joel Becker
0 siblings, 0 replies; 34+ messages in thread
From: Joel Becker @ 2010-04-01 17:44 UTC (permalink / raw)
To: ocfs2-devel
On Thu, Apr 01, 2010 at 12:32:29PM +0800, Tao Ma wrote:
> Joel Becker wrote:
> >On Thu, Apr 01, 2010 at 10:59:09AM +0800, Tao Ma wrote:
> >> le32_add_cpu(&fe->id1.bitmap1.i_total, le16_to_cpu(bg->bg_bits));
> >>- le32_add_cpu(&fe->i_clusters, le16_to_cpu(cl->cl_cpg));
> >>+ le32_add_cpu(&fe->i_clusters,
> >>+ le16_to_cpu(bg->bg_bits) / le16_to_cpu(cl->cl_bpc));
> >
> > You don't need to change this. Either we've allocated cpg or
> >we've failed out.
> Current we don't fail out in ocfs2_block_group_grow_discontig when
> we used up all the extent record in the list. So actually our bitmap
> size can be less than cpg. So you think we need to bail out?
>
> I removed your
> if (needed > 0) {
> }
Um, there's supposed to be the -ENOSPC in there. Looks like
something I didn't complete.
We will NOT support block groups that are smaller than cpg.
Joel
--
"But then she looks me in the eye
And says, 'We're going to last forever,'
And man you know I can't begin to doubt it.
Cause it just feels so good and so free and so right,
I know we ain't never going to change our minds about it, Hey!
Here comes my girl."
Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127
^ permalink raw reply [flat|nested] 34+ messages in thread
* [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume.
2010-04-01 2:53 [Ocfs2-devel] [PATCH 0/15] ocfs2: Support for discontiguous block group Tao Ma
` (11 preceding siblings ...)
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 12/15] ocfs2: Some tiny bug fixes for discontiguous block allocation Tao Ma
@ 2010-04-01 2:59 ` Tao Ma
2010-04-01 3:38 ` Joel Becker
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 14/15] ocfs2: Free block to the right block group Tao Ma
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 15/15] ocfs2: enable discontig block group support Tao Ma
14 siblings, 1 reply; 34+ messages in thread
From: Tao Ma @ 2010-04-01 2:59 UTC (permalink / raw)
To: ocfs2-devel
ocfs2_group_bitmap_size has to handle the case when the
volume don't have discontiguous block group support.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
---
fs/ocfs2/ocfs2_fs.h | 5 +++--
fs/ocfs2/resize.c | 4 ++--
fs/ocfs2/suballoc.c | 3 ++-
fs/ocfs2/super.c | 2 +-
4 files changed, 8 insertions(+), 6 deletions(-)
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index 42641f8..29f32ad 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -1346,11 +1346,12 @@ static inline u16 ocfs2_local_alloc_size(struct super_block *sb)
}
static inline int ocfs2_group_bitmap_size(struct super_block *sb,
- int suballocator)
+ int suballocator,
+ int support_discontig_bg)
{
int size;
- if (suballocator)
+ if (suballocator && support_discontig_bg)
size = OCFS2_MAX_BG_BITMAP_SIZE;
else
size = sb->s_blocksize -
diff --git a/fs/ocfs2/resize.c b/fs/ocfs2/resize.c
index 0ff4835..e6eca60 100644
--- a/fs/ocfs2/resize.c
+++ b/fs/ocfs2/resize.c
@@ -319,7 +319,7 @@ int ocfs2_group_extend(struct inode * inode, int new_clusters)
BUG_ON(!OCFS2_IS_VALID_DINODE(fe));
if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
- ocfs2_group_bitmap_size(osb->sb, 0) * 8) {
+ ocfs2_group_bitmap_size(osb->sb, 0, 0) * 8) {
mlog(ML_ERROR, "The disk is too old and small. "
"Force to do offline resize.");
ret = -EINVAL;
@@ -500,7 +500,7 @@ int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input)
fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
- ocfs2_group_bitmap_size(osb->sb, 0) * 8) {
+ ocfs2_group_bitmap_size(osb->sb, 0, 0) * 8) {
mlog(ML_ERROR, "The disk is too old and small."
" Force to do offline resize.");
ret = -EINVAL;
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 0878b5d..0bf6151 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -394,7 +394,8 @@ static int ocfs2_block_group_fill(handle_t *handle,
memset(bg, 0, sb->s_blocksize);
strcpy(bg->bg_signature, OCFS2_GROUP_DESC_SIGNATURE);
bg->bg_generation = cpu_to_le32(OCFS2_SB(sb)->fs_generation);
- bg->bg_size = cpu_to_le16(ocfs2_group_bitmap_size(sb, 1));
+ bg->bg_size = cpu_to_le16(ocfs2_group_bitmap_size(sb, 1,
+ ocfs2_supports_discontig_bg(osb)));
bg->bg_chain = cpu_to_le16(my_chain);
bg->bg_next_group = cl->cl_recs[my_chain].c_blkno;
bg->bg_parent_dinode = cpu_to_le64(OCFS2_I(alloc_inode)->ip_blkno);
diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c
index de8547f..b6171df 100644
--- a/fs/ocfs2/super.c
+++ b/fs/ocfs2/super.c
@@ -2226,7 +2226,7 @@ static int ocfs2_initialize_super(struct super_block *sb,
osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
iput(inode);
- osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0) * 8;
+ osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0, 0) * 8;
status = ocfs2_init_slot_info(osb);
if (status < 0) {
--
1.5.5
^ permalink raw reply related [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume.
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume Tao Ma
@ 2010-04-01 3:38 ` Joel Becker
2010-04-01 4:22 ` Tao Ma
0 siblings, 1 reply; 34+ messages in thread
From: Joel Becker @ 2010-04-01 3:38 UTC (permalink / raw)
To: ocfs2-devel
On Thu, Apr 01, 2010 at 10:59:10AM +0800, Tao Ma wrote:
> ocfs2_group_bitmap_size has to handle the case when the
> volume don't have discontiguous block group support.
NAK. We specify 256 in all cases. This doesn't hurt old code
because we've never used more bits. Even though our old bg_size values
were much larger, we never used that space because bg_bits was set to
the smaller value.
Joel
--
"Get right to the heart of matters.
It's the heart that matters more."
Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127
^ permalink raw reply [flat|nested] 34+ messages in thread
* [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume.
2010-04-01 3:38 ` Joel Becker
@ 2010-04-01 4:22 ` Tao Ma
2010-04-01 4:34 ` Wengang Wang
2010-04-01 7:37 ` Joel Becker
0 siblings, 2 replies; 34+ messages in thread
From: Tao Ma @ 2010-04-01 4:22 UTC (permalink / raw)
To: ocfs2-devel
Hi Joel,
Joel Becker wrote:
> On Thu, Apr 01, 2010 at 10:59:10AM +0800, Tao Ma wrote:
>> ocfs2_group_bitmap_size has to handle the case when the
>> volume don't have discontiguous block group support.
>
> NAK. We specify 256 in all cases. This doesn't hurt old code
> because we've never used more bits. Even though our old bg_size values
> were much larger, we never used that space because bg_bits was set to
> the smaller value.
No, it hurts.
So with an old ocfs2-tools and a new kernel, the new kernel will set it
to 256 while the old ocfs2-tools refused to work by checking bg_size(see
chainalloc_process_group in libocfs2).
Regards,
Tao
^ permalink raw reply [flat|nested] 34+ messages in thread
* [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume.
2010-04-01 4:22 ` Tao Ma
@ 2010-04-01 4:34 ` Wengang Wang
2010-04-01 4:39 ` Tao Ma
2010-04-01 7:38 ` Joel Becker
2010-04-01 7:37 ` Joel Becker
1 sibling, 2 replies; 34+ messages in thread
From: Wengang Wang @ 2010-04-01 4:34 UTC (permalink / raw)
To: ocfs2-devel
On 10-04-01 12:22, Tao Ma wrote:
> Hi Joel,
>
> Joel Becker wrote:
> > On Thu, Apr 01, 2010 at 10:59:10AM +0800, Tao Ma wrote:
> >> ocfs2_group_bitmap_size has to handle the case when the
> >> volume don't have discontiguous block group support.
> >
> > NAK. We specify 256 in all cases. This doesn't hurt old code
> > because we've never used more bits. Even though our old bg_size values
> > were much larger, we never used that space because bg_bits was set to
> > the smaller value.
> No, it hurts.
Hi,
> So with an old ocfs2-tools and a new kernel, the new kernel will set it
> to 256 while the old ocfs2-tools refused to work by checking bg_size(see
> chainalloc_process_group in libocfs2).
Just for my info.
Is "old ocfs2-tools + new kernel" supported?
regards,
wengang.
^ permalink raw reply [flat|nested] 34+ messages in thread
* [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume.
2010-04-01 4:34 ` Wengang Wang
@ 2010-04-01 4:39 ` Tao Ma
2010-04-01 4:50 ` Tao Ma
2010-04-01 7:39 ` Joel Becker
2010-04-01 7:38 ` Joel Becker
1 sibling, 2 replies; 34+ messages in thread
From: Tao Ma @ 2010-04-01 4:39 UTC (permalink / raw)
To: ocfs2-devel
Hi wengang,
Wengang Wang wrote:
> On 10-04-01 12:22, Tao Ma wrote:
>> Hi Joel,
>>
>> Joel Becker wrote:
>>> On Thu, Apr 01, 2010 at 10:59:10AM +0800, Tao Ma wrote:
>>>> ocfs2_group_bitmap_size has to handle the case when the
>>>> volume don't have discontiguous block group support.
>>> NAK. We specify 256 in all cases. This doesn't hurt old code
>>> because we've never used more bits. Even though our old bg_size values
>>> were much larger, we never used that space because bg_bits was set to
>>> the smaller value.
>> No, it hurts.
>
> Hi,
>> So with an old ocfs2-tools and a new kernel, the new kernel will set it
>> to 256 while the old ocfs2-tools refused to work by checking bg_size(see
>> chainalloc_process_group in libocfs2).
>
> Just for my info.
> Is "old ocfs2-tools + new kernel" supported?
I guess yes? I am not sure whether a user can use a 2.6.35 kernel(the
next merge window) while using an ocfs2-tools 1.4.
Regards,
Tao
^ permalink raw reply [flat|nested] 34+ messages in thread
* [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume.
2010-04-01 4:39 ` Tao Ma
@ 2010-04-01 4:50 ` Tao Ma
2010-04-01 7:40 ` Joel Becker
2010-04-01 7:39 ` Joel Becker
1 sibling, 1 reply; 34+ messages in thread
From: Tao Ma @ 2010-04-01 4:50 UTC (permalink / raw)
To: ocfs2-devel
Tao Ma wrote:
> Hi wengang,
>
> Wengang Wang wrote:
>> On 10-04-01 12:22, Tao Ma wrote:
>>> Hi Joel,
>>>
>>> Joel Becker wrote:
>>>> On Thu, Apr 01, 2010 at 10:59:10AM +0800, Tao Ma wrote:
>>>>> ocfs2_group_bitmap_size has to handle the case when the
>>>>> volume don't have discontiguous block group support.
>>>> NAK. We specify 256 in all cases. This doesn't hurt old code
>>>> because we've never used more bits. Even though our old bg_size values
>>>> were much larger, we never used that space because bg_bits was set to
>>>> the smaller value.
>>> No, it hurts.
>> Hi,
>>> So with an old ocfs2-tools and a new kernel, the new kernel will set it
>>> to 256 while the old ocfs2-tools refused to work by checking bg_size(see
>>> chainalloc_process_group in libocfs2).
>> Just for my info.
>> Is "old ocfs2-tools + new kernel" supported?
> I guess yes? I am not sure whether a user can use a 2.6.35 kernel(the
> next merge window) while using an ocfs2-tools 1.4.
actually the most worse thing I am worried about it that: the old
fsck.ocfs2 refused to fsck the volume even in case we don't enable
discontig-bg because the new kernel set bg_size=256 so the new inode
group will be considered corrupted. I don't think it is endurable for a
user.
Regards,
Tao
^ permalink raw reply [flat|nested] 34+ messages in thread
* [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume.
2010-04-01 4:50 ` Tao Ma
@ 2010-04-01 7:40 ` Joel Becker
0 siblings, 0 replies; 34+ messages in thread
From: Joel Becker @ 2010-04-01 7:40 UTC (permalink / raw)
To: ocfs2-devel
On Thu, Apr 01, 2010 at 12:50:56PM +0800, Tao Ma wrote:
> actually the most worse thing I am worried about it that: the old
> fsck.ocfs2 refused to fsck the volume even in case we don't enable
> discontig-bg because the new kernel set bg_size=256 so the new inode
> group will be considered corrupted. I don't think it is endurable
> for a user.
You are absolutely right. We have to make that work.
Joel
--
"Reader, suppose you were and idiot. And suppose you were a member of
Congress. But I repeat myself."
- Mark Twain
Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127
^ permalink raw reply [flat|nested] 34+ messages in thread
* [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume.
2010-04-01 4:39 ` Tao Ma
2010-04-01 4:50 ` Tao Ma
@ 2010-04-01 7:39 ` Joel Becker
1 sibling, 0 replies; 34+ messages in thread
From: Joel Becker @ 2010-04-01 7:39 UTC (permalink / raw)
To: ocfs2-devel
On Thu, Apr 01, 2010 at 12:39:35PM +0800, Tao Ma wrote:
> Wengang Wang wrote:
> >Is "old ocfs2-tools + new kernel" supported?
> I guess yes? I am not sure whether a user can use a 2.6.35
> kernel(the next merge window) while using an ocfs2-tools 1.4.
Yes. Even ocfs2-tools 1.2 will work, as long as the filesystem
was formatted with --feature-level=max-compat.
Joel
--
"In the arms of the angel, fly away from here,
From this dark, cold hotel room and the endlessness that you fear.
You are pulled from the wreckage of your silent reverie.
In the arms of the angel, may you find some comfort here."
Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127
^ permalink raw reply [flat|nested] 34+ messages in thread
* [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume.
2010-04-01 4:34 ` Wengang Wang
2010-04-01 4:39 ` Tao Ma
@ 2010-04-01 7:38 ` Joel Becker
1 sibling, 0 replies; 34+ messages in thread
From: Joel Becker @ 2010-04-01 7:38 UTC (permalink / raw)
To: ocfs2-devel
On Thu, Apr 01, 2010 at 12:34:31PM +0800, Wengang Wang wrote:
> Just for my info.
> Is "old ocfs2-tools + new kernel" supported?
Absolutely. As is "old kernel + new ocfs2-tools". As long as
the features of the particular formatted filesystem are understood by
both tools and kernel, they should work.
Joel
--
Life's Little Instruction Book #30
"Never buy a house without a fireplace."
Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127
^ permalink raw reply [flat|nested] 34+ messages in thread
* [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume.
2010-04-01 4:22 ` Tao Ma
2010-04-01 4:34 ` Wengang Wang
@ 2010-04-01 7:37 ` Joel Becker
2010-04-01 7:46 ` Tao Ma
1 sibling, 1 reply; 34+ messages in thread
From: Joel Becker @ 2010-04-01 7:37 UTC (permalink / raw)
To: ocfs2-devel
On Thu, Apr 01, 2010 at 12:22:56PM +0800, Tao Ma wrote:
> Joel Becker wrote:
> >On Thu, Apr 01, 2010 at 10:59:10AM +0800, Tao Ma wrote:
> >>ocfs2_group_bitmap_size has to handle the case when the
> >>volume don't have discontiguous block group support.
> >
> > NAK. We specify 256 in all cases. This doesn't hurt old code
> >because we've never used more bits. Even though our old bg_size values
> >were much larger, we never used that space because bg_bits was set to
> >the smaller value.
> No, it hurts.
> So with an old ocfs2-tools and a new kernel, the new kernel will set
> it to 256 while the old ocfs2-tools refused to work by checking
> bg_size(see
> chainalloc_process_group in libocfs2).
*facepalm*
Well, we shot ourselves in the foot. The entire point of having
the size stored is so that we can vary it and the code still works.
Consider that "validation" of bg_size a bug. The correct validation is
((bg_bits + 7) / 8) <= bg_size.
So we do need the kernel to store the old-style bg_size in order
to keep compatibility. But we can do better. First, you don't need the
supports_discontig_bg argument. You already have the superblock, just
check ocfs2_supports_discontig_bg(OCFS2_SB(sb)).
Secondly, that function needs a big fat comment.
static inline int ocfs2_group_bitmap_size(struct super_block *sb,
int suballocator)
{
int size = sb->s_blocksize -
offsetof(struct ocfs2_group_desc, bg_bitmap);
/*
* The cluster allocator uses the entire block. Suballocators have
* never used more than OCFS2_MAX_BG_BITMAP_SIZE. Unfortunately, older
* code expects bg_size set to the maximum. Thus we must keep
* bg_size as-is unless discontig_bg is enabled.
*/
if (suballocator && ocfs2_supports_discontig_bg(OCFS2_SB(sb)))
size = OCFS2_MAX_BG_BITMAP_SIZE;
return size;
}
Joel
--
Life's Little Instruction Book #450
"Don't be afraid to say, 'I need help.'"
Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127
^ permalink raw reply [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume.
2010-04-01 7:37 ` Joel Becker
@ 2010-04-01 7:46 ` Tao Ma
2010-04-01 17:53 ` Joel Becker
0 siblings, 1 reply; 34+ messages in thread
From: Tao Ma @ 2010-04-01 7:46 UTC (permalink / raw)
To: ocfs2-devel
Hi Joel,
Joel Becker wrote:
> On Thu, Apr 01, 2010 at 12:22:56PM +0800, Tao Ma wrote:
>> Joel Becker wrote:
>>> On Thu, Apr 01, 2010 at 10:59:10AM +0800, Tao Ma wrote:
>>>> ocfs2_group_bitmap_size has to handle the case when the
>>>> volume don't have discontiguous block group support.
>>> NAK. We specify 256 in all cases. This doesn't hurt old code
>>> because we've never used more bits. Even though our old bg_size values
>>> were much larger, we never used that space because bg_bits was set to
>>> the smaller value.
>> No, it hurts.
>> So with an old ocfs2-tools and a new kernel, the new kernel will set
>> it to 256 while the old ocfs2-tools refused to work by checking
>> bg_size(see
>> chainalloc_process_group in libocfs2).
>
> *facepalm*
>
> Well, we shot ourselves in the foot. The entire point of having
> the size stored is so that we can vary it and the code still works.
> Consider that "validation" of bg_size a bug. The correct validation is
> ((bg_bits + 7) / 8) <= bg_size.
> So we do need the kernel to store the old-style bg_size in order
> to keep compatibility. But we can do better. First, you don't need the
> supports_discontig_bg argument. You already have the superblock, just
> check ocfs2_supports_discontig_bg(OCFS2_SB(sb)).
I have thought of this before, but finally I gave up. The reason is that
OCFS2_SB(sb) use ocfs2_super which is declared in ocfs2.h. So you see, I
don't want to overlap these 2 head files.
> Secondly, that function needs a big fat comment.
>
> static inline int ocfs2_group_bitmap_size(struct super_block *sb,
> int suballocator)
> {
> int size = sb->s_blocksize -
> offsetof(struct ocfs2_group_desc, bg_bitmap);
>
> /*
> * The cluster allocator uses the entire block. Suballocators have
> * never used more than OCFS2_MAX_BG_BITMAP_SIZE. Unfortunately, older
> * code expects bg_size set to the maximum. Thus we must keep
> * bg_size as-is unless discontig_bg is enabled.
> */
> if (suballocator && ocfs2_supports_discontig_bg(OCFS2_SB(sb)))
> size = OCFS2_MAX_BG_BITMAP_SIZE;
yes, I can add the comments.
Regards,
Tao
^ permalink raw reply [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume.
2010-04-01 7:46 ` Tao Ma
@ 2010-04-01 17:53 ` Joel Becker
2010-04-01 23:32 ` Tao Ma
0 siblings, 1 reply; 34+ messages in thread
From: Joel Becker @ 2010-04-01 17:53 UTC (permalink / raw)
To: ocfs2-devel
On Thu, Apr 01, 2010 at 03:46:43PM +0800, Tao Ma wrote:
> Joel Becker wrote:
> > So we do need the kernel to store the old-style bg_size in order
> >to keep compatibility. But we can do better. First, you don't need the
> >supports_discontig_bg argument. You already have the superblock, just
> >check ocfs2_supports_discontig_bg(OCFS2_SB(sb)).
> I have thought of this before, but finally I gave up. The reason is
> that OCFS2_SB(sb) use ocfs2_super which is declared in ocfs2.h. So
> you see, I don't want to overlap these 2 head files.
You're right about OCFS2_SB(). This comes back to your
userspace version of the interface too, where we don't even have the
superblock.
Perhaps the ocfs2_fs.h version of the function needs the awful
integer arguments. Maybe we can define a wrapper interface in ocfs2.h
that takes the smarter arguments. This would allow most call sites to
read better.
We could also define it as ocfs2_group_bitmap_size(struct
super_block *sb, int sysfile_type, u32 feature_incompat). This requires
we pass in arguments that are obvious on their face. We're going to
have to know the information when we call the function anyway.
Joel
--
Life's Little Instruction Book #450
"Don't be afraid to say, 'I need help.'"
Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127
^ permalink raw reply [flat|nested] 34+ messages in thread
* [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume.
2010-04-01 17:53 ` Joel Becker
@ 2010-04-01 23:32 ` Tao Ma
2010-04-01 23:43 ` Joel Becker
0 siblings, 1 reply; 34+ messages in thread
From: Tao Ma @ 2010-04-01 23:32 UTC (permalink / raw)
To: ocfs2-devel
Joel Becker wrote:
> On Thu, Apr 01, 2010 at 03:46:43PM +0800, Tao Ma wrote:
>
>> Joel Becker wrote:
>>
>>> So we do need the kernel to store the old-style bg_size in order
>>> to keep compatibility. But we can do better. First, you don't need the
>>> supports_discontig_bg argument. You already have the superblock, just
>>> check ocfs2_supports_discontig_bg(OCFS2_SB(sb)).
>>>
>> I have thought of this before, but finally I gave up. The reason is
>> that OCFS2_SB(sb) use ocfs2_super which is declared in ocfs2.h. So
>> you see, I don't want to overlap these 2 head files.
>>
>
> You're right about OCFS2_SB(). This comes back to your
> userspace version of the interface too, where we don't even have the
> superblock.
> Perhaps the ocfs2_fs.h version of the function needs the awful
> integer arguments. Maybe we can define a wrapper interface in ocfs2.h
> that takes the smarter arguments. This would allow most call sites to
> read better.
> We could also define it as ocfs2_group_bitmap_size(struct
> super_block *sb, int sysfile_type, u32 feature_incompat). This requires
> we pass in arguments that are obvious on their face. We're going to
> have to know the information when we call the function anyway.
>
I just went through all the callers of ocfs2_group_bitmap_size and it
seems that it is called in 2 scenarios:
1. set bg_size.
2. check some situation in resize.
So how about change it to
ocfs2_group_bitmap_size(struct super_block *sb, struct ocfs2_group_desc *gd)
so if gd is NULL, we returns the old value, if not, just return bg_size.
As for place of setting bg_size, we can set bg_bits first and then set
bg_size to (bg_bits +7)/8.
Regards,
Tao
> Joel
>
>
>
^ permalink raw reply [flat|nested] 34+ messages in thread
* [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume.
2010-04-01 23:32 ` Tao Ma
@ 2010-04-01 23:43 ` Joel Becker
0 siblings, 0 replies; 34+ messages in thread
From: Joel Becker @ 2010-04-01 23:43 UTC (permalink / raw)
To: ocfs2-devel
On Fri, Apr 02, 2010 at 07:32:52AM +0800, Tao Ma wrote:
> I just went through all the callers of ocfs2_group_bitmap_size and
> it seems that it is called in 2 scenarios:
> 1. set bg_size.
> 2. check some situation in resize.
>
> So how about change it to
> ocfs2_group_bitmap_size(struct super_block *sb, struct ocfs2_group_desc *gd)
> so if gd is NULL, we returns the old value, if not, just return bg_size.
That makes no sense, because we're asking for "what it should
be", not "what it is".
> As for place of setting bg_size, we can set bg_bits first and then
> set bg_size to (bg_bits +7)/8.
But when we grow a group, we'll set bg_size to 256, which blows
up tools.
Joel
--
You can use a screwdriver to screw in screws or to clean your ears,
however, the latter needs real skill, determination and a lack of fear
of injuring yourself. It is much the same with JavaScript.
- Chris Heilmann
Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127
^ permalink raw reply [flat|nested] 34+ messages in thread
* [Ocfs2-devel] [PATCH 14/15] ocfs2: Free block to the right block group.
2010-04-01 2:53 [Ocfs2-devel] [PATCH 0/15] ocfs2: Support for discontiguous block group Tao Ma
` (12 preceding siblings ...)
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 13/15] ocfs2: ocfs2_group_bitmap_size has to handle old volume Tao Ma
@ 2010-04-01 2:59 ` Tao Ma
2010-04-01 3:42 ` Joel Becker
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 15/15] ocfs2: enable discontig block group support Tao Ma
14 siblings, 1 reply; 34+ messages in thread
From: Tao Ma @ 2010-04-01 2:59 UTC (permalink / raw)
To: ocfs2-devel
In case the block we are going to free is allocated from
a discontiguous block group, we have to use suballoc_loc
to be the right group.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
---
fs/ocfs2/alloc.c | 18 ++++++++++++------
fs/ocfs2/alloc.h | 2 +-
fs/ocfs2/dir.c | 5 ++++-
fs/ocfs2/refcounttree.c | 6 +++++-
fs/ocfs2/suballoc.c | 2 ++
fs/ocfs2/xattr.c | 5 ++++-
6 files changed, 28 insertions(+), 10 deletions(-)
diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
index 6b59b6f..2f87ac0 100644
--- a/fs/ocfs2/alloc.c
+++ b/fs/ocfs2/alloc.c
@@ -6299,6 +6299,7 @@ int ocfs2_truncate_log_init(struct ocfs2_super *osb)
*/
struct ocfs2_cached_block_free {
struct ocfs2_cached_block_free *free_next;
+ u64 free_bg;
u64 free_blk;
unsigned int free_bit;
};
@@ -6345,8 +6346,11 @@ static int ocfs2_free_cached_blocks(struct ocfs2_super *osb,
}
while (head) {
- bg_blkno = ocfs2_which_suballoc_group(head->free_blk,
- head->free_bit);
+ if (head->free_bg)
+ bg_blkno = head->free_bg;
+ else
+ bg_blkno = ocfs2_which_suballoc_group(head->free_blk,
+ head->free_bit);
mlog(0, "Free bit: (bit %u, blkno %llu)\n",
head->free_bit, (unsigned long long)head->free_blk);
@@ -6394,7 +6398,7 @@ int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
int ret = 0;
struct ocfs2_cached_block_free *item;
- item = kmalloc(sizeof(*item), GFP_NOFS);
+ item = kzalloc(sizeof(*item), GFP_NOFS);
if (item == NULL) {
ret = -ENOMEM;
mlog_errno(ret);
@@ -6534,8 +6538,8 @@ ocfs2_find_per_slot_free_list(int type,
}
int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
- int type, int slot, u64 blkno,
- unsigned int bit)
+ int type, int slot, u64 suballoc,
+ u64 blkno, unsigned int bit)
{
int ret;
struct ocfs2_per_slot_free_list *fl;
@@ -6548,7 +6552,7 @@ int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
goto out;
}
- item = kmalloc(sizeof(*item), GFP_NOFS);
+ item = kzalloc(sizeof(*item), GFP_NOFS);
if (item == NULL) {
ret = -ENOMEM;
mlog_errno(ret);
@@ -6558,6 +6562,7 @@ int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
mlog(0, "Insert: (type %d, slot %u, bit %u, blk %llu)\n",
type, slot, bit, (unsigned long long)blkno);
+ item->free_bg = suballoc;
item->free_blk = blkno;
item->free_bit = bit;
item->free_next = fl->f_first;
@@ -6574,6 +6579,7 @@ static int ocfs2_cache_extent_block_free(struct ocfs2_cached_dealloc_ctxt *ctxt,
{
return ocfs2_cache_block_dealloc(ctxt, EXTENT_ALLOC_SYSTEM_INODE,
le16_to_cpu(eb->h_suballoc_slot),
+ le64_to_cpu(eb->h_suballoc_loc),
le64_to_cpu(eb->h_blkno),
le16_to_cpu(eb->h_suballoc_bit));
}
diff --git a/fs/ocfs2/alloc.h b/fs/ocfs2/alloc.h
index 1db4359..fc28d64 100644
--- a/fs/ocfs2/alloc.h
+++ b/fs/ocfs2/alloc.h
@@ -209,7 +209,7 @@ static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c)
int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
u64 blkno, unsigned int bit);
int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
- int type, int slot, u64 blkno,
+ int type, int slot, u64 suballoc, u64 blkno,
unsigned int bit);
static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c)
{
diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c
index 48b591d..9cb92d3 100644
--- a/fs/ocfs2/dir.c
+++ b/fs/ocfs2/dir.c
@@ -4491,7 +4491,10 @@ static int ocfs2_dx_dir_remove_index(struct inode *dir,
blk = le64_to_cpu(dx_root->dr_blkno);
bit = le16_to_cpu(dx_root->dr_suballoc_bit);
- bg_blkno = ocfs2_which_suballoc_group(blk, bit);
+ if (dx_root->dr_suballoc_loc)
+ bg_blkno = le64_to_cpu(dx_root->dr_suballoc_loc);
+ else
+ bg_blkno = ocfs2_which_suballoc_group(blk, bit);
ret = ocfs2_free_suballoc_bits(handle, dx_alloc_inode, dx_alloc_bh,
bit, bg_blkno, 1);
if (ret)
diff --git a/fs/ocfs2/refcounttree.c b/fs/ocfs2/refcounttree.c
index 3fc675f..668a018 100644
--- a/fs/ocfs2/refcounttree.c
+++ b/fs/ocfs2/refcounttree.c
@@ -792,7 +792,10 @@ int ocfs2_remove_refcount_tree(struct inode *inode, struct buffer_head *di_bh)
if (le32_to_cpu(rb->rf_count) == 1) {
blk = le64_to_cpu(rb->rf_blkno);
bit = le16_to_cpu(rb->rf_suballoc_bit);
- bg_blkno = ocfs2_which_suballoc_group(blk, bit);
+ if (rb->rf_suballoc_loc)
+ bg_blkno = le64_to_cpu(rb->rf_suballoc_loc);
+ else
+ bg_blkno = ocfs2_which_suballoc_group(blk, bit);
alloc_inode = ocfs2_get_system_file_inode(osb,
EXTENT_ALLOC_SYSTEM_INODE,
@@ -2116,6 +2119,7 @@ static int ocfs2_remove_refcount_extent(handle_t *handle,
*/
ret = ocfs2_cache_block_dealloc(dealloc, EXTENT_ALLOC_SYSTEM_INODE,
le16_to_cpu(rb->rf_suballoc_slot),
+ le64_to_cpu(rb->rf_suballoc_loc),
le64_to_cpu(rb->rf_blkno),
le16_to_cpu(rb->rf_suballoc_bit));
if (ret) {
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 0bf6151..bb0dbb9 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -2382,6 +2382,8 @@ int ocfs2_free_dinode(handle_t *handle,
u16 bit = le16_to_cpu(di->i_suballoc_bit);
u64 bg_blkno = ocfs2_which_suballoc_group(blk, bit);
+ if (di->i_suballoc_loc)
+ bg_blkno = le64_to_cpu(di->i_suballoc_loc);
return ocfs2_free_suballoc_bits(handle, inode_alloc_inode,
inode_alloc_bh, bit, bg_blkno, 1);
}
diff --git a/fs/ocfs2/xattr.c b/fs/ocfs2/xattr.c
index 863ad7a..2447e14 100644
--- a/fs/ocfs2/xattr.c
+++ b/fs/ocfs2/xattr.c
@@ -2479,7 +2479,10 @@ static int ocfs2_xattr_free_block(struct inode *inode,
xb = (struct ocfs2_xattr_block *)blk_bh->b_data;
blk = le64_to_cpu(xb->xb_blkno);
bit = le16_to_cpu(xb->xb_suballoc_bit);
- bg_blkno = ocfs2_which_suballoc_group(blk, bit);
+ if (xb->xb_suballoc_loc)
+ bg_blkno = le64_to_cpu(xb->xb_suballoc_loc);
+ else
+ bg_blkno = ocfs2_which_suballoc_group(blk, bit);
xb_alloc_inode = ocfs2_get_system_file_inode(osb,
EXTENT_ALLOC_SYSTEM_INODE,
--
1.5.5
^ permalink raw reply related [flat|nested] 34+ messages in thread* [Ocfs2-devel] [PATCH 15/15] ocfs2: enable discontig block group support.
2010-04-01 2:53 [Ocfs2-devel] [PATCH 0/15] ocfs2: Support for discontiguous block group Tao Ma
` (13 preceding siblings ...)
2010-04-01 2:59 ` [Ocfs2-devel] [PATCH 14/15] ocfs2: Free block to the right block group Tao Ma
@ 2010-04-01 2:59 ` Tao Ma
2010-04-01 3:43 ` Joel Becker
14 siblings, 1 reply; 34+ messages in thread
From: Tao Ma @ 2010-04-01 2:59 UTC (permalink / raw)
To: ocfs2-devel
Signed-off-by: Tao Ma <tao.ma@oracle.com>
---
fs/ocfs2/ocfs2_fs.h | 3 ++-
1 files changed, 2 insertions(+), 1 deletions(-)
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index 29f32ad..3e3f3b4 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -100,7 +100,8 @@
| OCFS2_FEATURE_INCOMPAT_XATTR \
| OCFS2_FEATURE_INCOMPAT_META_ECC \
| OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS \
- | OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE)
+ | OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE \
+ | OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG)
#define OCFS2_FEATURE_RO_COMPAT_SUPP (OCFS2_FEATURE_RO_COMPAT_UNWRITTEN \
| OCFS2_FEATURE_RO_COMPAT_USRQUOTA \
| OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
--
1.5.5
^ permalink raw reply related [flat|nested] 34+ messages in thread