* [Ocfs2-devel] [PATCH 01/17] ocfs2: Define data structures for discontiguous block groups.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 02/17] ocfs2: Allocate " Tao Ma
` (17 subsequent siblings)
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 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>
Signed-off-by: Tao Ma <tao.ma@oracle.com>
---
fs/ocfs2/localalloc.c | 2 +-
fs/ocfs2/ocfs2_fs.h | 53 ++++++++++++++++++++++++++++++++++++++++++------
fs/ocfs2/resize.c | 4 +-
fs/ocfs2/suballoc.c | 2 +-
fs/ocfs2/super.c | 2 +-
5 files changed, 51 insertions(+), 12 deletions(-)
diff --git a/fs/ocfs2/localalloc.c b/fs/ocfs2/localalloc.c
index 2882f79..cb06315 100644
--- a/fs/ocfs2/localalloc.c
+++ b/fs/ocfs2/localalloc.c
@@ -122,7 +122,7 @@ unsigned int ocfs2_la_default_mb(struct ocfs2_super *osb)
struct super_block *sb = osb->sb;
gd_mb = ocfs2_clusters_to_megabytes(osb->sb,
- 8 * ocfs2_group_bitmap_size(sb));
+ 8 * ocfs2_group_bitmap_size(sb, 0));
/*
* This takes care of files systems with very small group
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index d61a152..448aa8d 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.
@@ -832,6 +835,13 @@ 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
@@ -852,7 +862,29 @@ 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 {
@@ -1276,12 +1308,16 @@ 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;
}
@@ -1404,12 +1440,15 @@ 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 a821f66..5bbfc12 100644
--- a/fs/ocfs2/resize.c
+++ b/fs/ocfs2/resize.c
@@ -315,7 +315,7 @@ 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;
@@ -496,7 +496,7 @@ 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 f8aedb3..b82a667 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -364,7 +364,7 @@ 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 79d7d4c..9bbed28 100644
--- a/fs/ocfs2/super.c
+++ b/fs/ocfs2/super.c
@@ -2277,7 +2277,7 @@ static int ocfs2_initialize_super(struct super_block *sb,
osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
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] 30+ messages in thread* [Ocfs2-devel] [PATCH 02/17] ocfs2: Allocate discontiguous block groups.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 01/17] ocfs2: Define data structures for discontiguous block groups Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 03/17] ocfs2: Pass suballocation results back via a structure Tao Ma
` (16 subsequent siblings)
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 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>
Signed-off-by: Tao Ma <tao.ma@oracle.com>
---
fs/ocfs2/ocfs2.h | 7 +
fs/ocfs2/ocfs2_fs.h | 20 ++++
fs/ocfs2/suballoc.c | 304 ++++++++++++++++++++++++++++++++++++++++++++-------
3 files changed, 291 insertions(+), 40 deletions(-)
diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h
index b6d9110..79fe984 100644
--- a/fs/ocfs2/ocfs2.h
+++ b/fs/ocfs2/ocfs2.h
@@ -491,6 +491,13 @@ 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 448aa8d..888ba4e 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -1278,6 +1278,16 @@ 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;
@@ -1430,6 +1440,16 @@ 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 b82a667..9f985c5 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,
@@ -330,14 +331,36 @@ 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;
@@ -365,11 +388,16 @@ 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);
@@ -400,6 +428,218 @@ 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) {
+ }
+
+ ocfs2_journal_dirty(handle, bg_bh);
+
+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.
*/
@@ -415,16 +655,17 @@ 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),
@@ -450,44 +691,21 @@ 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),
@@ -497,10 +715,11 @@ 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);
@@ -529,6 +748,11 @@ 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] 30+ messages in thread* [Ocfs2-devel] [PATCH 03/17] ocfs2: Pass suballocation results back via a structure.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 01/17] ocfs2: Define data structures for discontiguous block groups Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 02/17] ocfs2: Allocate " Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 04/17] ocfs2: Add suballoc_loc to metadata blocks Tao Ma
` (15 subsequent siblings)
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 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>
Signed-off-by: Tao Ma <tao.ma@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 9f985c5..bc1d6b7 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,
@@ -1252,8 +1256,7 @@ 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;
@@ -1297,14 +1300,9 @@ 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
@@ -1460,14 +1458,13 @@ 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));
@@ -1494,15 +1491,15 @@ 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);
@@ -1514,16 +1511,14 @@ 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);
}
}
@@ -1534,7 +1529,7 @@ 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;
@@ -1547,10 +1542,10 @@ 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);
@@ -1593,20 +1588,17 @@ 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;
@@ -1614,17 +1606,15 @@ 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);
@@ -1632,7 +1622,7 @@ 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);
@@ -1648,13 +1638,11 @@ 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;
@@ -1683,8 +1671,8 @@ 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;
@@ -1709,11 +1697,11 @@ 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
@@ -1730,7 +1718,7 @@ 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);
@@ -1752,25 +1740,24 @@ 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);
ocfs2_journal_dirty(handle, ac->ac_bh);
status = ocfs2_block_group_set_bits(handle,
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);
@@ -1786,14 +1773,11 @@ 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;
@@ -1820,22 +1804,16 @@ 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;
@@ -1848,8 +1826,8 @@ 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) {
@@ -1873,8 +1851,7 @@ 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) {
@@ -1891,7 +1868,7 @@ 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:
@@ -1908,7 +1885,7 @@ 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));
@@ -1919,17 +1896,17 @@ 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);
@@ -1976,8 +1953,7 @@ 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();
@@ -1993,18 +1969,17 @@ 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;
@@ -2084,8 +2059,7 @@ 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();
@@ -2124,14 +2098,12 @@ 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);
}
}
@@ -2141,7 +2113,8 @@ 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 74e502e..26a5ad3 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] 30+ messages in thread* [Ocfs2-devel] [PATCH 04/17] ocfs2: Add suballoc_loc to metadata blocks.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (2 preceding siblings ...)
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 03/17] ocfs2: Pass suballocation results back via a structure Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 05/17] ocfs2: ocfs2_claim_suballoc_bits() doesn't need an osb argument Tao Ma
` (14 subsequent siblings)
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 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 888ba4e..a17bce5 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -507,7 +507,10 @@ 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 */
@@ -674,7 +677,11 @@ 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 */
@@ -809,7 +816,12 @@ 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
@@ -929,7 +941,11 @@ 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 */
@@ -1041,7 +1057,10 @@ 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] 30+ messages in thread* [Ocfs2-devel] [PATCH 05/17] ocfs2: ocfs2_claim_suballoc_bits() doesn't need an osb argument.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (3 preceding siblings ...)
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 04/17] ocfs2: Add suballoc_loc to metadata blocks Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 06/17] ocfs2: Trim suballocations if they cross discontiguous regions Tao Ma
` (13 subsequent siblings)
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 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 bc1d6b7..13b69ef 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,
@@ -1768,8 +1767,7 @@ 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,
@@ -1795,7 +1793,8 @@ 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),
@@ -1891,8 +1890,7 @@ 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,
@@ -1964,8 +1962,7 @@ 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,
@@ -2093,8 +2090,7 @@ 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] 30+ messages in thread* [Ocfs2-devel] [PATCH 06/17] ocfs2: Trim suballocations if they cross discontiguous regions
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (4 preceding siblings ...)
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 05/17] ocfs2: ocfs2_claim_suballoc_bits() doesn't need an osb argument Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 07/17] ocfs2: ocfs2_claim_*() don't need an ocfs2_super argument Tao Ma
` (12 subsequent siblings)
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 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 13b69ef..9e1c776 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -1583,6 +1583,48 @@ 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,
@@ -1612,6 +1654,9 @@ 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));
@@ -1701,6 +1746,9 @@ 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] 30+ messages in thread* [Ocfs2-devel] [PATCH 07/17] ocfs2: ocfs2_claim_*() don't need an ocfs2_super argument.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (5 preceding siblings ...)
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 06/17] ocfs2: Trim suballocations if they cross discontiguous regions Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 08/17] ocfs2: Return allocated metadata blknos on the ocfs2_suballoc_result Tao Ma
` (11 subsequent siblings)
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 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>
Signed-off-by: Tao Ma <tao.ma@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 a74ea70..e12c623 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,
@@ -4798,7 +4797,7 @@ 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)
@@ -7213,7 +7212,7 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode,
data_ac->ac_resv = &OCFS2_I(inode)->ip_la_data_resv;
- 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 6c9a28a..02c3f22 100644
--- a/fs/ocfs2/dir.c
+++ b/fs/ocfs2/dir.c
@@ -2402,7 +2402,7 @@ 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);
@@ -2544,7 +2544,7 @@ 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;
@@ -2979,7 +2979,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
*/
if (ocfs2_dir_resv_allowed(osb))
data_ac->ac_resv = &oi->ip_la_data_resv;
- 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;
@@ -3118,7 +3118,7 @@ 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 cb06315..4aa83df 100644
--- a/fs/ocfs2/localalloc.c
+++ b/fs/ocfs2/localalloc.c
@@ -1163,7 +1163,7 @@ 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:
@@ -1176,7 +1176,7 @@ 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 21d4a33..e5434a0 100644
--- a/fs/ocfs2/namei.c
+++ b/fs/ocfs2/namei.c
@@ -478,7 +478,7 @@ 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 44d3e1c..472db83 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) {
@@ -1297,7 +1297,7 @@ 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) {
@@ -1547,7 +1547,7 @@ 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) {
@@ -3271,7 +3271,7 @@ 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 9e1c776..34f3fbb 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -443,7 +443,7 @@ 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) {
@@ -485,7 +485,7 @@ 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;
@@ -1923,8 +1923,7 @@ 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,
@@ -1947,7 +1946,7 @@ 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);
@@ -1990,8 +1989,7 @@ 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,
@@ -2019,7 +2017,7 @@ 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);
@@ -2094,8 +2092,7 @@ 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,
@@ -2105,6 +2102,7 @@ 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();
@@ -2165,8 +2163,7 @@ 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,
@@ -2174,7 +2171,7 @@ 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 26a5ad3..2bc505a 100644
--- a/fs/ocfs2/suballoc.h
+++ b/fs/ocfs2/suballoc.h
@@ -83,22 +83,19 @@ 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,
@@ -107,8 +104,7 @@ 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 26f704c..49a15c4 100644
--- a/fs/ocfs2/xattr.c
+++ b/fs/ocfs2/xattr.c
@@ -2831,7 +2831,6 @@ 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;
@@ -2842,7 +2841,7 @@ 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) {
@@ -2867,7 +2866,8 @@ 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;
@@ -4232,7 +4232,6 @@ 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 =
@@ -4260,7 +4259,7 @@ 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);
@@ -5082,7 +5081,7 @@ 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)
@@ -6913,7 +6912,7 @@ 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] 30+ messages in thread* [Ocfs2-devel] [PATCH 08/17] ocfs2: Return allocated metadata blknos on the ocfs2_suballoc_result.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (6 preceding siblings ...)
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 07/17] ocfs2: ocfs2_claim_*() don't need an ocfs2_super argument Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 09/17] ocfs2: Set suballoc_loc on allocated metadata Tao Ma
` (10 subsequent siblings)
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 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 34f3fbb..424c62f 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 */
};
@@ -1583,9 +1584,9 @@ 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;
@@ -1595,32 +1596,35 @@ 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;
}
}
@@ -1655,7 +1659,7 @@ 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,
@@ -1747,7 +1751,7 @@ 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);
/*
@@ -1931,7 +1935,7 @@ 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));
@@ -1949,7 +1953,7 @@ 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;
@@ -1997,7 +2001,7 @@ 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();
@@ -2022,7 +2026,7 @@ 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;
@@ -2101,7 +2105,7 @@ 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();
@@ -2142,6 +2146,7 @@ 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] 30+ messages in thread* [Ocfs2-devel] [PATCH 09/17] ocfs2: Set suballoc_loc on allocated metadata.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (7 preceding siblings ...)
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 08/17] ocfs2: Return allocated metadata blknos on the ocfs2_suballoc_result Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 10/17] ocfs2: Grow discontig block groups in one transaction Tao Ma
` (9 subsequent siblings)
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 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 e12c623..dc029e0 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 02c3f22..341bb8f 100644
--- a/fs/ocfs2/dir.c
+++ b/fs/ocfs2/dir.c
@@ -2395,15 +2395,15 @@ 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;
@@ -2431,6 +2431,7 @@ 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 e5434a0..bfe4571 100644
--- a/fs/ocfs2/namei.c
+++ b/fs/ocfs2/namei.c
@@ -472,14 +472,15 @@ 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;
@@ -516,6 +517,7 @@ 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 472db83..4825ce1 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);
@@ -1283,7 +1284,7 @@ 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;
@@ -1297,7 +1298,7 @@ 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) {
@@ -1329,6 +1330,7 @@ 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);
@@ -1523,7 +1525,7 @@ 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;
@@ -1547,7 +1549,7 @@ 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) {
@@ -1575,6 +1577,7 @@ 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 424c62f..d939742 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 */
@@ -1608,6 +1610,7 @@ 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;
@@ -1618,14 +1621,17 @@ 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;
+ }
}
}
@@ -1930,6 +1936,7 @@ 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)
@@ -1952,6 +1959,7 @@ 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;
@@ -1997,11 +2005,12 @@ 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();
@@ -2025,6 +2034,7 @@ 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 2bc505a..2c41a2c 100644
--- a/fs/ocfs2/suballoc.h
+++ b/fs/ocfs2/suballoc.h
@@ -86,6 +86,7 @@ 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);
@@ -93,6 +94,7 @@ 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 49a15c4..0ac2a8a 100644
--- a/fs/ocfs2/xattr.c
+++ b/fs/ocfs2/xattr.c
@@ -2829,7 +2829,7 @@ 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;
@@ -2842,8 +2842,8 @@ 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;
@@ -2865,6 +2865,7 @@ 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] 30+ messages in thread* [Ocfs2-devel] [PATCH 10/17] ocfs2: Grow discontig block groups in one transaction.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (8 preceding siblings ...)
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 09/17] ocfs2: Set suballoc_loc on allocated metadata Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 11/17] ocfs2: Don't relink cluster groups when allocating discontig block groups Tao Ma
` (8 subsequent siblings)
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 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 7dc5656..b5baaa8 100644
--- a/fs/ocfs2/journal.h
+++ b/fs/ocfs2/journal.h
@@ -561,6 +561,18 @@ 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 d939742..9113da9 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -527,12 +527,6 @@ 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,
@@ -560,11 +554,12 @@ 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;
@@ -576,9 +571,13 @@ 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);
@@ -589,8 +588,7 @@ 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;
@@ -605,6 +603,13 @@ 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);
@@ -642,7 +647,7 @@ 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;
}
@@ -664,14 +669,11 @@ 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),
@@ -703,8 +705,7 @@ 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;
@@ -754,11 +755,6 @@ 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] 30+ messages in thread* [Ocfs2-devel] [PATCH 11/17] ocfs2: Don't relink cluster groups when allocating discontig block groups
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (9 preceding siblings ...)
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 10/17] ocfs2: Grow discontig block groups in one transaction Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 12/17] ocfs2: Some tiny bug fixes for discontiguous block allocation Tao Ma
` (7 subsequent siblings)
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 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 9113da9..b3053da 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -610,6 +610,14 @@ 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] 30+ messages in thread* [Ocfs2-devel] [PATCH 12/17] ocfs2: Some tiny bug fixes for discontiguous block allocation.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (10 preceding siblings ...)
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 11/17] ocfs2: Don't relink cluster groups when allocating discontig block groups Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 13/17] ocfs2: ocfs2_group_bitmap_size has to handle old volume Tao Ma
` (6 subsequent siblings)
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 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 | 38 ++++++++++++++++++++++++--------------
2 files changed, 25 insertions(+), 15 deletions(-)
diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h
index 79fe984..c60959a 100644
--- a/fs/ocfs2/ocfs2.h
+++ b/fs/ocfs2/ocfs2.h
@@ -491,7 +491,7 @@ 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 b3053da..ce92967 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -345,15 +345,17 @@ 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);
}
@@ -401,7 +403,7 @@ 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 */
@@ -510,8 +512,8 @@ 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;
@@ -542,10 +544,17 @@ 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);
+ needed = le16_to_cpu(cl->cl_cpg) -
+ le16_to_cpu(bg->bg_bits) / le16_to_cpu(cl->cl_bpc);
}
if (needed > 0) {
+ /*
+ * We have used up all the extent rec but can't fill up
+ * the cpg. So bail out.
+ */
+ status = -ENOSPC;
+ goto bail;
}
ocfs2_journal_dirty(handle, bg_bh);
@@ -598,7 +607,7 @@ 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;
}
@@ -674,7 +683,7 @@ 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;
@@ -730,11 +739,12 @@ 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);
@@ -1626,7 +1636,7 @@ 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;
@@ -2166,6 +2176,7 @@ 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) {
@@ -2174,8 +2185,7 @@ 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] 30+ messages in thread* [Ocfs2-devel] [PATCH 13/17] ocfs2: ocfs2_group_bitmap_size has to handle old volume.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (11 preceding siblings ...)
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 12/17] ocfs2: Some tiny bug fixes for discontiguous block allocation Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 14/17] ocfs2: Add ocfs2_gd_is_discontig Tao Ma
` (5 subsequent siblings)
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 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. So
pass the feature_incompat in and check it.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
---
fs/ocfs2/localalloc.c | 2 +-
fs/ocfs2/ocfs2_fs.h | 37 +++++++++++++++++++++++++------------
fs/ocfs2/resize.c | 6 ++++--
fs/ocfs2/suballoc.c | 3 ++-
fs/ocfs2/super.c | 3 ++-
5 files changed, 34 insertions(+), 17 deletions(-)
diff --git a/fs/ocfs2/localalloc.c b/fs/ocfs2/localalloc.c
index 4aa83df..6b15b24 100644
--- a/fs/ocfs2/localalloc.c
+++ b/fs/ocfs2/localalloc.c
@@ -122,7 +122,7 @@ unsigned int ocfs2_la_default_mb(struct ocfs2_super *osb)
struct super_block *sb = osb->sb;
gd_mb = ocfs2_clusters_to_megabytes(osb->sb,
- 8 * ocfs2_group_bitmap_size(sb, 0));
+ 8 * ocfs2_group_bitmap_size(sb, 0, osb->s_feature_incompat));
/*
* This takes care of files systems with very small group
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index a17bce5..67bb8a7 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -1338,15 +1338,21 @@ 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,
+ u32 feature_incompat)
{
- int size;
+ int size = sb->s_blocksize -
+ offsetof(struct ocfs2_group_desc, bg_bitmap);
- if (suballocator)
+ /*
+ * 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 &&
+ (feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG))
size = OCFS2_MAX_BG_BITMAP_SIZE;
- else
- size = sb->s_blocksize -
- offsetof(struct ocfs2_group_desc, bg_bitmap);
return size;
}
@@ -1479,15 +1485,22 @@ static inline int ocfs2_local_alloc_size(int blocksize)
return size;
}
-static inline int ocfs2_group_bitmap_size(int blocksize, int suballocator)
+static inline int ocfs2_group_bitmap_size(int blocksize,
+ int suballocator,
+ uint32_t feature_incompat)
{
- int size;
+ int size = sb->s_blocksize -
+ offsetof(struct ocfs2_group_desc, bg_bitmap);
- if (suballocator)
+ /*
+ * 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 &&
+ (feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG))
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 5bbfc12..dacd553 100644
--- a/fs/ocfs2/resize.c
+++ b/fs/ocfs2/resize.c
@@ -315,7 +315,8 @@ 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,
+ osb->s_feature_incompat) * 8) {
mlog(ML_ERROR, "The disk is too old and small. "
"Force to do offline resize.");
ret = -EINVAL;
@@ -496,7 +497,8 @@ 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,
+ osb->s_feature_incompat) * 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 ce92967..bfeb6ac 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -395,7 +395,8 @@ 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,
+ osb->s_feature_incompat));
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 9bbed28..fdea6ba 100644
--- a/fs/ocfs2/super.c
+++ b/fs/ocfs2/super.c
@@ -2277,7 +2277,8 @@ static int ocfs2_initialize_super(struct super_block *sb,
osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
iput(inode);
- osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0) * 8;
+ osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
+ osb->s_feature_incompat) * 8;
status = ocfs2_init_slot_info(osb);
if (status < 0) {
--
1.5.5
^ permalink raw reply related [flat|nested] 30+ messages in thread* [Ocfs2-devel] [PATCH 14/17] ocfs2: Add ocfs2_gd_is_discontig.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (12 preceding siblings ...)
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 13/17] ocfs2: ocfs2_group_bitmap_size has to handle old volume Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-26 21:49 ` Joel Becker
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 15/17] ocfs2: Free block to the right block group Tao Ma
` (4 subsequent siblings)
18 siblings, 1 reply; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 UTC (permalink / raw)
To: ocfs2-devel
Add ocfs2_gd_is_discontig so that we can test whether
a group descriptor is discontiguous or not.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
---
fs/ocfs2/ocfs2_fs.h | 6 ++++++
1 files changed, 6 insertions(+), 0 deletions(-)
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index 67bb8a7..b5b3cb7 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -1574,5 +1574,11 @@ static inline void ocfs2_set_de_type(struct ocfs2_dir_entry *de,
de->file_type = ocfs2_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
}
+static inline int ocfs2_gd_is_discontig(struct ocfs2_group_desc *gd)
+{
+ return ((offsetof(struct ocfs2_group_desc, bg_bitmap) + gd->bg_size ==
+ offsetof(struct ocfs2_group_desc, bg_list)) &&
+ gd->bg_list.l_next_free_rec);
+}
#endif /* _OCFS2_FS_H */
--
1.5.5
^ permalink raw reply related [flat|nested] 30+ messages in thread* [Ocfs2-devel] [PATCH 14/17] ocfs2: Add ocfs2_gd_is_discontig.
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 14/17] ocfs2: Add ocfs2_gd_is_discontig Tao Ma
@ 2010-04-26 21:49 ` Joel Becker
2010-04-27 0:47 ` Tao Ma
0 siblings, 1 reply; 30+ messages in thread
From: Joel Becker @ 2010-04-26 21:49 UTC (permalink / raw)
To: ocfs2-devel
On Thu, Apr 22, 2010 at 02:40:28PM +0800, Tao Ma wrote:
> Add ocfs2_gd_is_discontig so that we can test whether
> a group descriptor is discontiguous or not.
>
> Signed-off-by: Tao Ma <tao.ma@oracle.com>
> ---
> fs/ocfs2/ocfs2_fs.h | 6 ++++++
> 1 files changed, 6 insertions(+), 0 deletions(-)
>
> diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
> index 67bb8a7..b5b3cb7 100644
> --- a/fs/ocfs2/ocfs2_fs.h
> +++ b/fs/ocfs2/ocfs2_fs.h
> @@ -1574,5 +1574,11 @@ static inline void ocfs2_set_de_type(struct ocfs2_dir_entry *de,
> de->file_type = ocfs2_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
> }
>
> +static inline int ocfs2_gd_is_discontig(struct ocfs2_group_desc *gd)
> +{
> + return ((offsetof(struct ocfs2_group_desc, bg_bitmap) + gd->bg_size ==
> + offsetof(struct ocfs2_group_desc, bg_list)) &&
> + gd->bg_list.l_next_free_rec);
You're missing a paren around the addition:
return (((offsetof(struct ocfs2_group_desc, bg_bitmap) + gd->bg_size) ==
offsetof(struct ocfs2_group_desc, bg_list)) &&
gd->bg_list.l_next_free_rec);
Would this be more readable:
if ((offsetof(struct ocfs2_group_desc, bg_bitmap) + gd->bg_size) !=
offsetof(struct ocfs2_group_desc, bg_list))
return 0;
/*
* Only valid to check l_next_free_rec if
* bg_bitmap + bg_size == bg_list
*/
if (!gd->bg_list.l_next_free_rec)
return 0;
return 1;
Joel
--
Life's Little Instruction Book #267
"Lie on your back and look at the stars."
Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127
^ permalink raw reply [flat|nested] 30+ messages in thread* [Ocfs2-devel] [PATCH 14/17] ocfs2: Add ocfs2_gd_is_discontig.
2010-04-26 21:49 ` Joel Becker
@ 2010-04-27 0:47 ` Tao Ma
0 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-27 0:47 UTC (permalink / raw)
To: ocfs2-devel
Hi Joel,
Joel Becker wrote:
> On Thu, Apr 22, 2010 at 02:40:28PM +0800, Tao Ma wrote:
>> Add ocfs2_gd_is_discontig so that we can test whether
>> a group descriptor is discontiguous or not.
>>
>> Signed-off-by: Tao Ma <tao.ma@oracle.com>
>> ---
>> fs/ocfs2/ocfs2_fs.h | 6 ++++++
>> 1 files changed, 6 insertions(+), 0 deletions(-)
>>
>> diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
>> index 67bb8a7..b5b3cb7 100644
>> --- a/fs/ocfs2/ocfs2_fs.h
>> +++ b/fs/ocfs2/ocfs2_fs.h
>> @@ -1574,5 +1574,11 @@ static inline void ocfs2_set_de_type(struct ocfs2_dir_entry *de,
>> de->file_type = ocfs2_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
>> }
>>
>> +static inline int ocfs2_gd_is_discontig(struct ocfs2_group_desc *gd)
>> +{
>> + return ((offsetof(struct ocfs2_group_desc, bg_bitmap) + gd->bg_size ==
>> + offsetof(struct ocfs2_group_desc, bg_list)) &&
>> + gd->bg_list.l_next_free_rec);
>
> You're missing a paren around the addition:
>
> return (((offsetof(struct ocfs2_group_desc, bg_bitmap) + gd->bg_size) ==
> offsetof(struct ocfs2_group_desc, bg_list)) &&
> gd->bg_list.l_next_free_rec);
Actually '+' is more precedent than '=='. But yes, we have made a consensus
that we should use more paren to make it clear.
>
> Would this be more readable:
>
> if ((offsetof(struct ocfs2_group_desc, bg_bitmap) + gd->bg_size) !=
> offsetof(struct ocfs2_group_desc, bg_list))
> return 0;
> /*
> * Only valid to check l_next_free_rec if
> * bg_bitmap + bg_size == bg_list
> */
> if (!gd->bg_list.l_next_free_rec)
> return 0;
> return 1;
OK, I will use it. Here is the updated one.
Regards,
Tao
From 98733de3ad63dab2359d9643dda25e0fb4ddafc9 Mon Sep 17 00:00:00 2001
From: Tao Ma <tao.ma@oracle.com>
Date: Tue, 27 Apr 2010 08:30:28 +0800
Subject: [PATCH 14/17] ocfs2: Add ocfs2_gd_is_discontig.
Add ocfs2_gd_is_discontig so that we can test whether
a group descriptor is discontiguous or not.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
---
fs/ocfs2/ocfs2_fs.h | 13 +++++++++++++
1 files changed, 13 insertions(+), 0 deletions(-)
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index 67bb8a7..5f54a76 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -1574,5 +1574,18 @@ static inline void ocfs2_set_de_type(struct ocfs2_dir_entry *de,
de->file_type = ocfs2_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
}
+static inline int ocfs2_gd_is_discontig(struct ocfs2_group_desc *gd)
+{
+ if ((offsetof(struct ocfs2_group_desc, bg_bitmap) + gd->bg_size) !=
+ offsetof(struct ocfs2_group_desc, bg_list))
+ return 0;
+ /*
+ * Only valid to check l_next_free_rec if
+ * bg_bitmap + bg_size == bg_list.
+ */
+ if (!gd->bg_list.l_next_free_rec)
+ return 0;
+ return 1;
+}
#endif /* _OCFS2_FS_H */
--
1.5.5
^ permalink raw reply related [flat|nested] 30+ messages in thread
* [Ocfs2-devel] [PATCH 15/17] ocfs2: Free block to the right block group.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (13 preceding siblings ...)
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 14/17] ocfs2: Add ocfs2_gd_is_discontig Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 16/17] ocfs2: Set ac_last_group properly with discontig group Tao Ma
` (3 subsequent siblings)
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 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 dc029e0..008a11f 100644
--- a/fs/ocfs2/alloc.c
+++ b/fs/ocfs2/alloc.c
@@ -6215,6 +6215,7 @@ 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;
};
@@ -6261,8 +6262,11 @@ 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);
@@ -6310,7 +6314,7 @@ 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);
@@ -6450,8 +6454,8 @@ 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;
@@ -6464,7 +6468,7 @@ 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);
@@ -6474,6 +6478,7 @@ 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;
@@ -6490,6 +6495,7 @@ 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 341bb8f..3fea52d 100644
--- a/fs/ocfs2/dir.c
+++ b/fs/ocfs2/dir.c
@@ -4466,7 +4466,10 @@ 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 4825ce1..63dac01 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,
@@ -2108,6 +2111,7 @@ 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 bfeb6ac..a971ec6 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -2342,6 +2342,8 @@ 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 0ac2a8a..ff8a8ea 100644
--- a/fs/ocfs2/xattr.c
+++ b/fs/ocfs2/xattr.c
@@ -2466,7 +2466,10 @@ 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] 30+ messages in thread* [Ocfs2-devel] [PATCH 16/17] ocfs2: Set ac_last_group properly with discontig group.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (14 preceding siblings ...)
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 15/17] ocfs2: Free block to the right block group Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 17/17] ocfs2: enable discontig block group support Tao Ma
` (2 subsequent siblings)
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 UTC (permalink / raw)
To: ocfs2-devel
ac_last_group is used to record the last block group we
used during allocation. But the initialization process
only calls ocfs2_which_suballoc_group and fails to
use suballoc_loc properly. So let us do it.
Another function ocfs2_test_suballoc_bit also needs fix.
I have searched all the callers of ocfs2_which_suballoc_group,
and all the callers notices suballoc_loc now.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
---
fs/ocfs2/suballoc.c | 31 +++++++++++++++++++------------
1 files changed, 19 insertions(+), 12 deletions(-)
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index a971ec6..693656d 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -1986,10 +1986,10 @@ bail:
}
static void ocfs2_init_inode_ac_group(struct inode *dir,
- struct buffer_head *parent_fe_bh,
+ struct buffer_head *parent_di_bh,
struct ocfs2_alloc_context *ac)
{
- struct ocfs2_dinode *fe = (struct ocfs2_dinode *)parent_fe_bh->b_data;
+ struct ocfs2_dinode *di = (struct ocfs2_dinode *)parent_di_bh->b_data;
/*
* Try to allocate inodes from some specific group.
*
@@ -2003,10 +2003,14 @@ static void ocfs2_init_inode_ac_group(struct inode *dir,
if (OCFS2_I(dir)->ip_last_used_group &&
OCFS2_I(dir)->ip_last_used_slot == ac->ac_alloc_slot)
ac->ac_last_group = OCFS2_I(dir)->ip_last_used_group;
- else if (le16_to_cpu(fe->i_suballoc_slot) == ac->ac_alloc_slot)
- ac->ac_last_group = ocfs2_which_suballoc_group(
- le64_to_cpu(fe->i_blkno),
- le16_to_cpu(fe->i_suballoc_bit));
+ else if (le16_to_cpu(di->i_suballoc_slot) == ac->ac_alloc_slot) {
+ if (di->i_suballoc_loc)
+ ac->ac_last_group = le64_to_cpu(di->i_suballoc_loc);
+ else
+ ac->ac_last_group = ocfs2_which_suballoc_group(
+ le64_to_cpu(di->i_blkno),
+ le16_to_cpu(di->i_suballoc_bit));
+ }
}
static inline void ocfs2_save_inode_ac_group(struct inode *dir,
@@ -2585,7 +2589,7 @@ static int ocfs2_test_suballoc_bit(struct ocfs2_super *osb,
struct buffer_head *alloc_bh, u64 blkno,
u16 bit, int *res)
{
- struct ocfs2_dinode *alloc_fe;
+ struct ocfs2_dinode *alloc_di;
struct ocfs2_group_desc *group;
struct buffer_head *group_bh = NULL;
u64 bg_blkno;
@@ -2594,17 +2598,20 @@ static int ocfs2_test_suballoc_bit(struct ocfs2_super *osb,
mlog_entry("blkno: %llu bit: %u\n", (unsigned long long)blkno,
(unsigned int)bit);
- alloc_fe = (struct ocfs2_dinode *)alloc_bh->b_data;
- if ((bit + 1) > ocfs2_bits_per_group(&alloc_fe->id2.i_chain)) {
+ alloc_di = (struct ocfs2_dinode *)alloc_bh->b_data;
+ if ((bit + 1) > ocfs2_bits_per_group(&alloc_di->id2.i_chain)) {
mlog(ML_ERROR, "suballoc bit %u out of range of %u\n",
(unsigned int)bit,
- ocfs2_bits_per_group(&alloc_fe->id2.i_chain));
+ ocfs2_bits_per_group(&alloc_di->id2.i_chain));
status = -EINVAL;
goto bail;
}
- bg_blkno = ocfs2_which_suballoc_group(blkno, bit);
- status = ocfs2_read_group_descriptor(suballoc, alloc_fe, bg_blkno,
+ if (alloc_di->i_suballoc_loc)
+ bg_blkno = le64_to_cpu(alloc_di->i_suballoc_loc);
+ else
+ bg_blkno = ocfs2_which_suballoc_group(blkno, bit);
+ status = ocfs2_read_group_descriptor(suballoc, alloc_di, bg_blkno,
&group_bh);
if (status < 0) {
mlog(ML_ERROR, "read group %llu failed %d\n",
--
1.5.5
^ permalink raw reply related [flat|nested] 30+ messages in thread* [Ocfs2-devel] [PATCH 17/17] ocfs2: enable discontig block group support.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (15 preceding siblings ...)
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 16/17] ocfs2: Set ac_last_group properly with discontig group Tao Ma
@ 2010-04-22 6:40 ` Tao Ma
2010-05-06 1:36 ` [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Joel Becker
2010-05-17 19:21 ` Joel Becker
18 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-04-22 6:40 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 b5b3cb7..28a8618 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] 30+ messages in thread* [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (16 preceding siblings ...)
2010-04-22 6:40 ` [Ocfs2-devel] [PATCH 17/17] ocfs2: enable discontig block group support Tao Ma
@ 2010-05-06 1:36 ` Joel Becker
2010-05-06 6:20 ` Tao Ma
2010-05-17 19:21 ` Joel Becker
18 siblings, 1 reply; 30+ messages in thread
From: Joel Becker @ 2010-05-06 1:36 UTC (permalink / raw)
To: ocfs2-devel
On Thu, Apr 22, 2010 at 02:39:30PM +0800, Tao Ma wrote:
> Hi all,
> This is the v2 of discontiguous block group support in ocfs2.
Hey Tao,
Your current discontig-bg branch at oss doesn't pull cleanly
against merge-window. I was trying to pull it into ALL for visibility.
;-) I hope to have it in 2.6.35, so it will go atop the current
merge-window stuff anyway. Can you rebase atop that? I think the code
is in good shape aside from any bugs found in testing.
Joel
--
Life's Little Instruction Book #347
"Never waste the oppourtunity to tell someone you love them."
Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127
^ permalink raw reply [flat|nested] 30+ messages in thread* [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group.
2010-04-22 6:39 [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Tao Ma
` (17 preceding siblings ...)
2010-05-06 1:36 ` [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group Joel Becker
@ 2010-05-17 19:21 ` Joel Becker
2010-05-17 23:54 ` Tao Ma
2010-05-18 21:44 ` Mark Fasheh
18 siblings, 2 replies; 30+ messages in thread
From: Joel Becker @ 2010-05-17 19:21 UTC (permalink / raw)
To: ocfs2-devel
On Thu, Apr 22, 2010 at 02:39:30PM +0800, Tao Ma wrote:
> [View]
> http://oss.oracle.com/git/?p=tma/linux-2.6.git;a=shortlog;h=discontig-bg
> [Pull]
> git://oss.oracle.com/git/tma/linux-2.6.git discontig-bg
Hey everyone,
The merge window is open. I want to take this and push it to
Linus in the next few days. I want to clarify the state.
Tao,
Let me know if/when your discontig-bg branch has the code you
want to go upstream. It may already; I'm just confirming before I pull.
Tristan,
When we spoke last week, all tests had passed. I'm going to
assume that is still the case unless you tell me otherwise.
Mark,
Any chance you can give this a once over in the next two or so
days? I'm going to pull from Tao as soon as he confirms the latest code
is there, but then I'm going to give it a couple days in linux-next.
Joel
--
"If the human brain were so simple we could understand it, we would
be so simple that we could not."
- W. A. Clouston
Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127
^ permalink raw reply [flat|nested] 30+ messages in thread* [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group.
2010-05-17 19:21 ` Joel Becker
@ 2010-05-17 23:54 ` Tao Ma
2010-05-18 22:45 ` Joel Becker
2010-05-18 21:44 ` Mark Fasheh
1 sibling, 1 reply; 30+ messages in thread
From: Tao Ma @ 2010-05-17 23:54 UTC (permalink / raw)
To: ocfs2-devel
Joel Becker wrote:
> On Thu, Apr 22, 2010 at 02:39:30PM +0800, Tao Ma wrote:
>
>> [View]
>> http://oss.oracle.com/git/?p=tma/linux-2.6.git;a=shortlog;h=discontig-bg
>> [Pull]
>> git://oss.oracle.com/git/tma/linux-2.6.git discontig-bg
>>
>
> Hey everyone,
> The merge window is open. I want to take this and push it to
> Linus in the next few days. I want to clarify the state.
>
> Tao,
> Let me know if/when your discontig-bg branch has the code you
> want to go upstream. It may already; I'm just confirming before I pull.
>
It is ok there. please pull
git://oss.oracle.com/git/tma/linux-2.6.git discontig-bg
Regards,
Tao
^ permalink raw reply [flat|nested] 30+ messages in thread
* [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group.
2010-05-17 23:54 ` Tao Ma
@ 2010-05-18 22:45 ` Joel Becker
2010-05-18 22:49 ` Joel Becker
2010-05-18 23:00 ` Tao Ma
0 siblings, 2 replies; 30+ messages in thread
From: Joel Becker @ 2010-05-18 22:45 UTC (permalink / raw)
To: ocfs2-devel
On Tue, May 18, 2010 at 07:54:39AM +0800, Tao Ma wrote:
> > Let me know if/when your discontig-bg branch has the code you
> > want to go upstream. It may already; I'm just confirming before I pull.
> >
> It is ok there. please pull
>
> git://oss.oracle.com/git/tma/linux-2.6.git discontig-bg
Tao,
This branch has the bogus orig_value_size fix. The good one is
in the fixes branch, and they conflict. Also, you have three patches in
this branch without descriptions:
patch speedup.fix
patch unwritten.fix
patch DLM.fix
Please fix this.
--
Life's Little Instruction Book #197
"Don't forget, a person's greatest emotional need is to
feel appreciated."
Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127
^ permalink raw reply [flat|nested] 30+ messages in thread* [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group.
2010-05-18 22:45 ` Joel Becker
@ 2010-05-18 22:49 ` Joel Becker
2010-05-18 23:10 ` Tao Ma
2010-05-18 23:00 ` Tao Ma
1 sibling, 1 reply; 30+ messages in thread
From: Joel Becker @ 2010-05-18 22:49 UTC (permalink / raw)
To: ocfs2-devel
On Tue, May 18, 2010 at 03:45:56PM -0700, Joel Becker wrote:
> On Tue, May 18, 2010 at 07:54:39AM +0800, Tao Ma wrote:
> > > Let me know if/when your discontig-bg branch has the code you
> > > want to go upstream. It may already; I'm just confirming before I pull.
> > >
> > It is ok there. please pull
> >
> > git://oss.oracle.com/git/tma/linux-2.6.git discontig-bg
>
> Tao,
> This branch has the bogus orig_value_size fix. The good one is
> in the fixes branch, and they conflict. Also, you have three patches in
> this branch without descriptions:
>
> patch speedup.fix
> patch unwritten.fix
> patch DLM.fix
>
> Please fix this.
Actually, your discontig-bg branch doesn't look much like this
patch series at all.
Joel
--
"This is the end, beautiful friend.
This is the end, my only friend the end
Of our elaborate plans, the end
Of everything that stands, the end
No safety or surprise, the end
I'll never look into your eyes again."
Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127
^ permalink raw reply [flat|nested] 30+ messages in thread
* [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group.
2010-05-18 22:49 ` Joel Becker
@ 2010-05-18 23:10 ` Tao Ma
0 siblings, 0 replies; 30+ messages in thread
From: Tao Ma @ 2010-05-18 23:10 UTC (permalink / raw)
To: ocfs2-devel
Joel Becker wrote:
> On Tue, May 18, 2010 at 03:45:56PM -0700, Joel Becker wrote:
>
>> On Tue, May 18, 2010 at 07:54:39AM +0800, Tao Ma wrote:
>>
>>>> Let me know if/when your discontig-bg branch has the code you
>>>> want to go upstream. It may already; I'm just confirming before I pull.
>>>>
>>>>
>>> It is ok there. please pull
>>>
>>> git://oss.oracle.com/git/tma/linux-2.6.git discontig-bg
>>>
>> Tao,
>> This branch has the bogus orig_value_size fix. The good one is
>> in the fixes branch, and they conflict. Also, you have three patches in
>> this branch without descriptions:
>>
>> patch speedup.fix
>> patch unwritten.fix
>> patch DLM.fix
>>
>> Please fix this.
>>
>
> Actually, your discontig-bg branch doesn't look much like this
> patch series at all.
>
yes, there are 2 changes in this branch against the patch I sent to the
mailist(you can check the commit time for it).
1. for the patch "ocfs2_claim_*() don't need an ocfs2_super argument",
I rebased it since it has conflict with your new merge-window.
2. for the patch "Add ocfs2_gd_is_discontig", we need
le16_to_cpu(bg->bg_size).
Regards,
Tao
^ permalink raw reply [flat|nested] 30+ messages in thread
* [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group.
2010-05-18 22:45 ` Joel Becker
2010-05-18 22:49 ` Joel Becker
@ 2010-05-18 23:00 ` Tao Ma
2010-05-18 23:53 ` Joel Becker
1 sibling, 1 reply; 30+ messages in thread
From: Tao Ma @ 2010-05-18 23:00 UTC (permalink / raw)
To: ocfs2-devel
Joel Becker wrote:
> On Tue, May 18, 2010 at 07:54:39AM +0800, Tao Ma wrote:
>
>>> Let me know if/when your discontig-bg branch has the code you
>>> want to go upstream. It may already; I'm just confirming before I pull.
>>>
>>>
>> It is ok there. please pull
>>
>> git://oss.oracle.com/git/tma/linux-2.6.git discontig-bg
>>
>
> Tao,
> This branch has the bogus orig_value_size fix. The good one is
> in the fixes branch, and they conflict. Also, you have three patches in
> this branch without descriptions:
>
> patch speedup.fix
> patch unwritten.fix
> patch DLM.fix
>
> Please fix this.
>
done. Sorry for the inconvenience.
Regards,
Tao
^ permalink raw reply [flat|nested] 30+ messages in thread
* [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group.
2010-05-18 23:00 ` Tao Ma
@ 2010-05-18 23:53 ` Joel Becker
0 siblings, 0 replies; 30+ messages in thread
From: Joel Becker @ 2010-05-18 23:53 UTC (permalink / raw)
To: ocfs2-devel
On Wed, May 19, 2010 at 07:00:32AM +0800, Tao Ma wrote:
> Joel Becker wrote:
> > On Tue, May 18, 2010 at 07:54:39AM +0800, Tao Ma wrote:
> >
> >>> Let me know if/when your discontig-bg branch has the code you
> >>> want to go upstream. It may already; I'm just confirming before I pull.
> >>>
> >>>
> >> It is ok there. please pull
> >>
> >> git://oss.oracle.com/git/tma/linux-2.6.git discontig-bg
> >>
> >
> > Tao,
> > This branch has the bogus orig_value_size fix. The good one is
> > in the fixes branch, and they conflict. Also, you have three patches in
> > this branch without descriptions:
> >
> > patch speedup.fix
> > patch unwritten.fix
> > patch DLM.fix
> >
> > Please fix this.
> >
> done. Sorry for the inconvenience.
It's now in the merge-window branch of ocfs2.git.
Joel
--
"Egotist: a person more interested in himself than in me."
- Ambrose Bierce
Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127
^ permalink raw reply [flat|nested] 30+ messages in thread
* [Ocfs2-devel] [PATCH 00/17 V2] ocfs2: Support for discontiguous block group.
2010-05-17 19:21 ` Joel Becker
2010-05-17 23:54 ` Tao Ma
@ 2010-05-18 21:44 ` Mark Fasheh
1 sibling, 0 replies; 30+ messages in thread
From: Mark Fasheh @ 2010-05-18 21:44 UTC (permalink / raw)
To: ocfs2-devel
On Mon, May 17, 2010 at 12:21:47PM -0700, Joel Becker wrote:
> On Thu, Apr 22, 2010 at 02:39:30PM +0800, Tao Ma wrote:
> > [View]
> > http://oss.oracle.com/git/?p=tma/linux-2.6.git;a=shortlog;h=discontig-bg
> > [Pull]
> > git://oss.oracle.com/git/tma/linux-2.6.git discontig-bg
>
> Hey everyone,
> The merge window is open. I want to take this and push it to
> Linus in the next few days. I want to clarify the state.
>
> Tao,
> Let me know if/when your discontig-bg branch has the code you
> want to go upstream. It may already; I'm just confirming before I pull.
>
> Tristan,
> When we spoke last week, all tests had passed. I'm going to
> assume that is still the case unless you tell me otherwise.
>
> Mark,
> Any chance you can give this a once over in the next two or so
> days? I'm going to pull from Tao as soon as he confirms the latest code
> is there, but then I'm going to give it a couple days in linux-next.
Tristan's patches got my ack in mail. Tao's look good too (you can put my
ack on those too). Thanks,
--Mark
--
Mark Fasheh
^ permalink raw reply [flat|nested] 30+ messages in thread