#include "super.h"
#include "uptodate.h"
#include "xattr.h"
+#include "refcounttree.h"
#include "buffer_head_io.h"
.eo_fill_root_el = ocfs2_dx_root_fill_root_el,
};
+static void ocfs2_refcount_tree_fill_root_el(struct ocfs2_extent_tree *et)
+{
+ struct ocfs2_refcount_block *rb = et->et_object;
+
+ et->et_root_el = &rb->rf_list;
+}
+
+static void ocfs2_refcount_tree_set_last_eb_blk(struct ocfs2_extent_tree *et,
+ u64 blkno)
+{
+ struct ocfs2_refcount_block *rb = et->et_object;
+
+ rb->rf_last_eb_blk = cpu_to_le64(blkno);
+}
+
+static u64 ocfs2_refcount_tree_get_last_eb_blk(struct ocfs2_extent_tree *et)
+{
+ struct ocfs2_refcount_block *rb = et->et_object;
+
+ return le64_to_cpu(rb->rf_last_eb_blk);
+}
+
+static void ocfs2_refcount_tree_update_clusters(struct ocfs2_extent_tree *et,
+ u32 clusters)
+{
+ struct ocfs2_refcount_block *rb = et->et_object;
+
+ le32_add_cpu(&rb->rf_clusters, clusters);
+}
+
+static enum ocfs2_contig_type
+ocfs2_refcount_tree_extent_contig(struct ocfs2_extent_tree *et,
+ struct ocfs2_extent_rec *ext,
+ struct ocfs2_extent_rec *insert_rec)
+{
+ return CONTIG_NONE;
+}
+
+static struct ocfs2_extent_tree_operations ocfs2_refcount_tree_et_ops = {
+ .eo_set_last_eb_blk = ocfs2_refcount_tree_set_last_eb_blk,
+ .eo_get_last_eb_blk = ocfs2_refcount_tree_get_last_eb_blk,
+ .eo_update_clusters = ocfs2_refcount_tree_update_clusters,
+ .eo_fill_root_el = ocfs2_refcount_tree_fill_root_el,
+ .eo_extent_contig = ocfs2_refcount_tree_extent_contig,
+};
+
static void __ocfs2_init_extent_tree(struct ocfs2_extent_tree *et,
struct ocfs2_caching_info *ci,
struct buffer_head *bh,
NULL, &ocfs2_dx_root_et_ops);
}
+void ocfs2_init_refcount_extent_tree(struct ocfs2_extent_tree *et,
+ struct ocfs2_caching_info *ci,
+ struct buffer_head *bh)
+{
+ __ocfs2_init_extent_tree(et, ci, bh, ocfs2_journal_access_rb,
+ NULL, &ocfs2_refcount_tree_et_ops);
+}
+
static inline void ocfs2_et_set_last_eb_blk(struct ocfs2_extent_tree *et,
u64 new_last_eb_blk)
{
static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc);
static int ocfs2_cache_extent_block_free(struct ocfs2_cached_dealloc_ctxt *ctxt,
struct ocfs2_extent_block *eb);
-
-/*
- * Structures which describe a path through a btree, and functions to
- * manipulate them.
- *
- * The idea here is to be as generic as possible with the tree
- * manipulation code.
- */
-struct ocfs2_path_item {
- struct buffer_head *bh;
- struct ocfs2_extent_list *el;
-};
-
-#define OCFS2_MAX_PATH_DEPTH 5
-
-struct ocfs2_path {
- int p_tree_depth;
- ocfs2_journal_access_func p_root_access;
- struct ocfs2_path_item p_node[OCFS2_MAX_PATH_DEPTH];
-};
-
-#define path_root_bh(_path) ((_path)->p_node[0].bh)
-#define path_root_el(_path) ((_path)->p_node[0].el)
-#define path_root_access(_path)((_path)->p_root_access)
-#define path_leaf_bh(_path) ((_path)->p_node[(_path)->p_tree_depth].bh)
-#define path_leaf_el(_path) ((_path)->p_node[(_path)->p_tree_depth].el)
-#define path_num_items(_path) ((_path)->p_tree_depth + 1)
-
-static int ocfs2_find_path(struct ocfs2_caching_info *ci,
- struct ocfs2_path *path, u32 cpos);
static void ocfs2_adjust_rightmost_records(handle_t *handle,
struct ocfs2_extent_tree *et,
struct ocfs2_path *path,
* to build another path. Generally, this involves freeing the buffer
* heads.
*/
-static void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root)
+void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root)
{
int i, start = 0, depth = 0;
struct ocfs2_path_item *node;
path->p_tree_depth = depth;
}
-static void ocfs2_free_path(struct ocfs2_path *path)
+void ocfs2_free_path(struct ocfs2_path *path)
{
if (path) {
ocfs2_reinit_path(path, 0);
return path;
}
-static struct ocfs2_path *ocfs2_new_path_from_path(struct ocfs2_path *path)
+struct ocfs2_path *ocfs2_new_path_from_path(struct ocfs2_path *path)
{
return ocfs2_new_path(path_root_bh(path), path_root_el(path),
path_root_access(path));
}
-static struct ocfs2_path *ocfs2_new_path_from_et(struct ocfs2_extent_tree *et)
+struct ocfs2_path *ocfs2_new_path_from_et(struct ocfs2_extent_tree *et)
{
return ocfs2_new_path(et->et_root_bh, et->et_root_el,
et->et_root_journal_access);
* I don't like the way this function's name looks next to
* ocfs2_journal_access_path(), but I don't have a better one.
*/
-static int ocfs2_path_bh_journal_access(handle_t *handle,
- struct ocfs2_caching_info *ci,
- struct ocfs2_path *path,
- int idx)
+int ocfs2_path_bh_journal_access(handle_t *handle,
+ struct ocfs2_caching_info *ci,
+ struct ocfs2_path *path,
+ int idx)
{
ocfs2_journal_access_func access = path_root_access(path);
/*
* Convenience function to journal all components in a path.
*/
-static int ocfs2_journal_access_path(struct ocfs2_caching_info *ci,
- handle_t *handle,
- struct ocfs2_path *path)
+int ocfs2_journal_access_path(struct ocfs2_caching_info *ci,
+ handle_t *handle,
+ struct ocfs2_path *path)
{
int i, ret = 0;
ocfs2_path_insert_eb(fp->path, fp->index, bh);
fp->index++;
}
-static int ocfs2_find_path(struct ocfs2_caching_info *ci,
- struct ocfs2_path *path, u32 cpos)
+int ocfs2_find_path(struct ocfs2_caching_info *ci,
+ struct ocfs2_path *path, u32 cpos)
{
struct find_path_data data;
int op_credits,
struct ocfs2_path *path)
{
+ int ret;
int credits = (path->p_tree_depth - subtree_depth) * 2 + 1 + op_credits;
- if (handle->h_buffer_credits < credits)
- return ocfs2_extend_trans(handle, credits);
+ if (handle->h_buffer_credits < credits) {
+ ret = ocfs2_extend_trans(handle,
+ credits - handle->h_buffer_credits);
+ if (ret)
+ return ret;
+
+ if (unlikely(handle->h_buffer_credits < credits))
+ return ocfs2_extend_trans(handle, credits);
+ }
return 0;
}
*
* The array is assumed to be large enough to hold an entire path (tree depth).
*
- * Upon succesful return from this function:
+ * Upon successful return from this function:
*
* - The 'right_path' array will contain a path to the leaf block
* whose range contains e_cpos.
}
/*
- * Mark part or all of the extent record at split_index in the leaf
- * pointed to by path as written. This removes the unwritten
- * extent flag.
+ * Split part or all of the extent record at split_index in the leaf
+ * pointed to by path. Merge with the contiguous extent record if needed.
*
* Care is taken to handle contiguousness so as to not grow the tree.
*
* have been brought into cache (and pinned via the journal), so the
* extra overhead is not expressed in terms of disk reads.
*/
-static int __ocfs2_mark_extent_written(handle_t *handle,
- struct ocfs2_extent_tree *et,
- struct ocfs2_path *path,
- int split_index,
- struct ocfs2_extent_rec *split_rec,
- struct ocfs2_alloc_context *meta_ac,
- struct ocfs2_cached_dealloc_ctxt *dealloc)
+int ocfs2_split_extent(handle_t *handle,
+ struct ocfs2_extent_tree *et,
+ struct ocfs2_path *path,
+ int split_index,
+ struct ocfs2_extent_rec *split_rec,
+ struct ocfs2_alloc_context *meta_ac,
+ struct ocfs2_cached_dealloc_ctxt *dealloc)
{
int ret = 0;
struct ocfs2_extent_list *el = path_leaf_el(path);
struct ocfs2_merge_ctxt ctxt;
struct ocfs2_extent_list *rightmost_el;
- if (!(rec->e_flags & OCFS2_EXT_UNWRITTEN)) {
- ret = -EIO;
- mlog_errno(ret);
- goto out;
- }
-
if (le32_to_cpu(rec->e_cpos) > le32_to_cpu(split_rec->e_cpos) ||
((le32_to_cpu(rec->e_cpos) + le16_to_cpu(rec->e_leaf_clusters)) <
(le32_to_cpu(split_rec->e_cpos) + le16_to_cpu(split_rec->e_leaf_clusters)))) {
}
/*
- * Mark the already-existing extent at cpos as written for len clusters.
+ * Change the flags of the already-existing extent at cpos for len clusters.
+ *
+ * new_flags: the flags we want to set.
+ * clear_flags: the flags we want to clear.
+ * phys: the new physical offset we want this new extent starts from.
*
* If the existing extent is larger than the request, initiate a
* split. An attempt will be made at merging with adjacent extents.
*
* The caller is responsible for passing down meta_ac if we'll need it.
*/
-int ocfs2_mark_extent_written(struct inode *inode,
- struct ocfs2_extent_tree *et,
- handle_t *handle, u32 cpos, u32 len, u32 phys,
- struct ocfs2_alloc_context *meta_ac,
- struct ocfs2_cached_dealloc_ctxt *dealloc)
+int ocfs2_change_extent_flag(handle_t *handle,
+ struct ocfs2_extent_tree *et,
+ u32 cpos, u32 len, u32 phys,
+ struct ocfs2_alloc_context *meta_ac,
+ struct ocfs2_cached_dealloc_ctxt *dealloc,
+ int new_flags, int clear_flags)
{
int ret, index;
- u64 start_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys);
+ struct super_block *sb = ocfs2_metadata_cache_get_super(et->et_ci);
+ u64 start_blkno = ocfs2_clusters_to_blocks(sb, phys);
struct ocfs2_extent_rec split_rec;
struct ocfs2_path *left_path = NULL;
struct ocfs2_extent_list *el;
-
- mlog(0, "Inode %lu cpos %u, len %u, phys %u (%llu)\n",
- inode->i_ino, cpos, len, phys, (unsigned long long)start_blkno);
-
- if (!ocfs2_writes_unwritten_extents(OCFS2_SB(inode->i_sb))) {
- ocfs2_error(inode->i_sb, "Inode %llu has unwritten extents "
- "that are being written to, but the feature bit "
- "is not set in the super block.",
- (unsigned long long)OCFS2_I(inode)->ip_blkno);
- ret = -EROFS;
- goto out;
- }
-
- /*
- * XXX: This should be fixed up so that we just re-insert the
- * next extent records.
- */
- ocfs2_et_extent_map_truncate(et, 0);
+ struct ocfs2_extent_rec *rec;
left_path = ocfs2_new_path_from_et(et);
if (!left_path) {
index = ocfs2_search_extent_list(el, cpos);
if (index == -1 || index >= le16_to_cpu(el->l_next_free_rec)) {
- ocfs2_error(inode->i_sb,
- "Inode %llu has an extent at cpos %u which can no "
+ ocfs2_error(sb,
+ "Owner %llu has an extent at cpos %u which can no "
"longer be found.\n",
- (unsigned long long)OCFS2_I(inode)->ip_blkno, cpos);
+ (unsigned long long)
+ ocfs2_metadata_cache_owner(et->et_ci), cpos);
ret = -EROFS;
goto out;
}
+ ret = -EIO;
+ rec = &el->l_recs[index];
+ if (new_flags && (rec->e_flags & new_flags)) {
+ mlog(ML_ERROR, "Owner %llu tried to set %d flags on an "
+ "extent that already had them",
+ (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci),
+ new_flags);
+ goto out;
+ }
+
+ if (clear_flags && !(rec->e_flags & clear_flags)) {
+ mlog(ML_ERROR, "Owner %llu tried to clear %d flags on an "
+ "extent that didn't have them",
+ (unsigned long long)ocfs2_metadata_cache_owner(et->et_ci),
+ clear_flags);
+ goto out;
+ }
+
memset(&split_rec, 0, sizeof(struct ocfs2_extent_rec));
split_rec.e_cpos = cpu_to_le32(cpos);
split_rec.e_leaf_clusters = cpu_to_le16(len);
split_rec.e_blkno = cpu_to_le64(start_blkno);
- split_rec.e_flags = path_leaf_el(left_path)->l_recs[index].e_flags;
- split_rec.e_flags &= ~OCFS2_EXT_UNWRITTEN;
-
- ret = __ocfs2_mark_extent_written(handle, et, left_path,
- index, &split_rec, meta_ac,
- dealloc);
+ split_rec.e_flags = rec->e_flags;
+ if (new_flags)
+ split_rec.e_flags |= new_flags;
+ if (clear_flags)
+ split_rec.e_flags &= ~clear_flags;
+
+ ret = ocfs2_split_extent(handle, et, left_path,
+ index, &split_rec, meta_ac,
+ dealloc);
if (ret)
mlog_errno(ret);
out:
ocfs2_free_path(left_path);
return ret;
+
+}
+
+/*
+ * Mark the already-existing extent at cpos as written for len clusters.
+ * This removes the unwritten extent flag.
+ *
+ * If the existing extent is larger than the request, initiate a
+ * split. An attempt will be made at merging with adjacent extents.
+ *
+ * The caller is responsible for passing down meta_ac if we'll need it.
+ */
+int ocfs2_mark_extent_written(struct inode *inode,
+ struct ocfs2_extent_tree *et,
+ handle_t *handle, u32 cpos, u32 len, u32 phys,
+ struct ocfs2_alloc_context *meta_ac,
+ struct ocfs2_cached_dealloc_ctxt *dealloc)
+{
+ int ret;
+
+ mlog(0, "Inode %lu cpos %u, len %u, phys clusters %u\n",
+ inode->i_ino, cpos, len, phys);
+
+ if (!ocfs2_writes_unwritten_extents(OCFS2_SB(inode->i_sb))) {
+ ocfs2_error(inode->i_sb, "Inode %llu has unwritten extents "
+ "that are being written to, but the feature bit "
+ "is not set in the super block.",
+ (unsigned long long)OCFS2_I(inode)->ip_blkno);
+ ret = -EROFS;
+ goto out;
+ }
+
+ /*
+ * XXX: This should be fixed up so that we just re-insert the
+ * next extent records.
+ */
+ ocfs2_et_extent_map_truncate(et, 0);
+
+ ret = ocfs2_change_extent_flag(handle, et, cpos,
+ len, phys, meta_ac, dealloc,
+ 0, OCFS2_EXT_UNWRITTEN);
+ if (ret)
+ mlog_errno(ret);
+
+out:
+ return ret;
}
static int ocfs2_split_tree(handle_t *handle, struct ocfs2_extent_tree *et,
return fl;
}
-static int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
- int type, int slot, u64 blkno,
- unsigned int bit)
+int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
+ int type, int slot, u64 blkno,
+ unsigned int bit)
{
int ret;
struct ocfs2_per_slot_free_list *fl;
*/
static int ocfs2_trim_tree(struct inode *inode, struct ocfs2_path *path,
handle_t *handle, struct ocfs2_truncate_context *tc,
- u32 clusters_to_del, u64 *delete_start)
+ u32 clusters_to_del, u64 *delete_start, u8 *flags)
{
int ret, i, index = path->p_tree_depth;
u32 new_edge = 0;
struct ocfs2_extent_rec *rec;
*delete_start = 0;
+ *flags = 0;
while (index >= 0) {
bh = path->p_node[index].bh;
*delete_start = le64_to_cpu(rec->e_blkno)
+ ocfs2_clusters_to_blocks(inode->i_sb,
le16_to_cpu(rec->e_leaf_clusters));
+ *flags = rec->e_flags;
/*
* If it's now empty, remove this record.
struct buffer_head *fe_bh,
handle_t *handle,
struct ocfs2_truncate_context *tc,
- struct ocfs2_path *path)
+ struct ocfs2_path *path,
+ struct ocfs2_alloc_context *meta_ac)
{
int status;
struct ocfs2_dinode *fe;
struct ocfs2_extent_list *el;
struct buffer_head *last_eb_bh = NULL;
u64 delete_blk = 0;
+ u8 rec_flags;
fe = (struct ocfs2_dinode *) fe_bh->b_data;
inode->i_blocks = ocfs2_inode_sector_count(inode);
status = ocfs2_trim_tree(inode, path, handle, tc,
- clusters_to_del, &delete_blk);
+ clusters_to_del, &delete_blk, &rec_flags);
if (status) {
mlog_errno(status);
goto bail;
}
if (delete_blk) {
- status = ocfs2_truncate_log_append(osb, handle, delete_blk,
- clusters_to_del);
+ if (rec_flags & OCFS2_EXT_REFCOUNTED)
+ status = ocfs2_decrease_refcount(inode, handle,
+ ocfs2_blocks_to_clusters(osb->sb,
+ delete_blk),
+ clusters_to_del, meta_ac,
+ &tc->tc_dealloc, 1);
+ else
+ status = ocfs2_truncate_log_append(osb, handle,
+ delete_blk,
+ clusters_to_del);
if (status < 0) {
mlog_errno(status);
goto bail;
return 0;
}
-static void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle,
- unsigned int from, unsigned int to,
- struct page *page, int zero, u64 *phys)
+void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle,
+ unsigned int from, unsigned int to,
+ struct page *page, int zero, u64 *phys)
{
int ret, partial = 0;
ocfs2_unlock_and_free_pages(pages, numpages);
}
-static int ocfs2_grab_eof_pages(struct inode *inode, loff_t start, loff_t end,
- struct page **pages, int *num)
+int ocfs2_grab_pages(struct inode *inode, loff_t start, loff_t end,
+ struct page **pages, int *num)
{
int numpages, ret = 0;
- struct super_block *sb = inode->i_sb;
struct address_space *mapping = inode->i_mapping;
unsigned long index;
loff_t last_page_bytes;
BUG_ON(start > end);
- BUG_ON(start >> OCFS2_SB(sb)->s_clustersize_bits !=
- (end - 1) >> OCFS2_SB(sb)->s_clustersize_bits);
-
numpages = 0;
last_page_bytes = PAGE_ALIGN(end);
index = start >> PAGE_CACHE_SHIFT;
return ret;
}
+static int ocfs2_grab_eof_pages(struct inode *inode, loff_t start, loff_t end,
+ struct page **pages, int *num)
+{
+ struct super_block *sb = inode->i_sb;
+
+ BUG_ON(start >> OCFS2_SB(sb)->s_clustersize_bits !=
+ (end - 1) >> OCFS2_SB(sb)->s_clustersize_bits);
+
+ return ocfs2_grab_pages(inode, start, end, pages, num);
+}
+
/*
* Zero the area past i_size but still within an allocated
* cluster. This avoids exposing nonzero data on subsequent file
{
int status, i, credits, tl_sem = 0;
u32 clusters_to_del, new_highest_cpos, range;
+ u64 blkno = 0;
struct ocfs2_extent_list *el;
handle_t *handle = NULL;
struct inode *tl_inode = osb->osb_tl_inode;
struct ocfs2_path *path = NULL;
struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data;
+ struct ocfs2_alloc_context *meta_ac = NULL;
+ struct ocfs2_refcount_tree *ref_tree = NULL;
mlog_entry_void();
goto bail;
}
+ credits = 0;
+
/*
* Truncate always works against the rightmost tree branch.
*/
clusters_to_del = 0;
} else if (le32_to_cpu(el->l_recs[i].e_cpos) >= new_highest_cpos) {
clusters_to_del = ocfs2_rec_clusters(el, &el->l_recs[i]);
+ blkno = le64_to_cpu(el->l_recs[i].e_blkno);
} else if (range > new_highest_cpos) {
clusters_to_del = (ocfs2_rec_clusters(el, &el->l_recs[i]) +
le32_to_cpu(el->l_recs[i].e_cpos)) -
new_highest_cpos;
+ blkno = le64_to_cpu(el->l_recs[i].e_blkno) +
+ ocfs2_clusters_to_blocks(inode->i_sb,
+ ocfs2_rec_clusters(el, &el->l_recs[i]) -
+ clusters_to_del);
} else {
status = 0;
goto bail;
mlog(0, "clusters_to_del = %u in this pass, tail blk=%llu\n",
clusters_to_del, (unsigned long long)path_leaf_bh(path)->b_blocknr);
+ if (el->l_recs[i].e_flags & OCFS2_EXT_REFCOUNTED && clusters_to_del) {
+ BUG_ON(!(OCFS2_I(inode)->ip_dyn_features &
+ OCFS2_HAS_REFCOUNT_FL));
+
+ status = ocfs2_lock_refcount_tree(osb,
+ le64_to_cpu(di->i_refcount_loc),
+ 1, &ref_tree, NULL);
+ if (status) {
+ mlog_errno(status);
+ goto bail;
+ }
+
+ status = ocfs2_prepare_refcount_change_for_del(inode, fe_bh,
+ blkno,
+ clusters_to_del,
+ &credits,
+ &meta_ac);
+ if (status < 0) {
+ mlog_errno(status);
+ goto bail;
+ }
+ }
+
mutex_lock(&tl_inode->i_mutex);
tl_sem = 1;
/* ocfs2_truncate_log_needs_flush guarantees us at least one
}
}
- credits = ocfs2_calc_tree_trunc_credits(osb->sb, clusters_to_del,
+ credits += ocfs2_calc_tree_trunc_credits(osb->sb, clusters_to_del,
(struct ocfs2_dinode *)fe_bh->b_data,
el);
handle = ocfs2_start_trans(osb, credits);
}
status = ocfs2_do_truncate(osb, clusters_to_del, inode, fe_bh, handle,
- tc, path);
+ tc, path, meta_ac);
if (status < 0) {
mlog_errno(status);
goto bail;
ocfs2_reinit_path(path, 1);
+ if (meta_ac) {
+ ocfs2_free_alloc_context(meta_ac);
+ meta_ac = NULL;
+ }
+
+ if (ref_tree) {
+ ocfs2_unlock_refcount_tree(osb, ref_tree, 1);
+ ref_tree = NULL;
+ }
+
/*
* The check above will catch the case where we've truncated
* away all allocation.
if (handle)
ocfs2_commit_trans(osb, handle);
+ if (meta_ac)
+ ocfs2_free_alloc_context(meta_ac);
+
+ if (ref_tree)
+ ocfs2_unlock_refcount_tree(osb, ref_tree, 1);
+
ocfs2_run_deallocs(osb, &tc->tc_dealloc);
ocfs2_free_path(path);
git://ftp.safe.ca / safe / jmp / linux-2.6 / blobdiff